| 1 | Updates from upstream's release/2.39/master branch, up to commit c7c3f5bf80ae86b34501f473f1a9fc545c911b7f. |
| 2 | |
| 3 | diff --git a/ADVISORIES b/ADVISORIES |
| 4 | new file mode 100644 |
| 5 | index 0000000000..d4e33f2df3 |
| 6 | --- /dev/null |
| 7 | +++ b/ADVISORIES |
| 8 | @@ -0,0 +1,2 @@ |
| 9 | +For the GNU C Library Security Advisories, see the git master branch: |
| 10 | +https://sourceware.org/git/?p=glibc.git;a=tree;f=advisories;hb=HEAD |
| 11 | diff --git a/Makeconfig b/Makeconfig |
| 12 | index 85e00cef94..522182abdc 100644 |
| 13 | --- a/Makeconfig |
| 14 | +++ b/Makeconfig |
| 15 | @@ -586,10 +586,13 @@ link-libc-rpath-link = -Wl,-rpath-link=$(rpath-link) |
| 16 | # before the expansion of LDLIBS-* variables). |
| 17 | |
| 18 | # Tests use -Wl,-rpath instead of -Wl,-rpath-link for |
| 19 | -# build-hardcoded-path-in-tests. |
| 20 | +# build-hardcoded-path-in-tests. Add -Wl,--disable-new-dtags to force |
| 21 | +# DT_RPATH instead of DT_RUNPATH which only applies to DT_NEEDED entries |
| 22 | +# in the executable and doesn't applies to DT_NEEDED entries in shared |
| 23 | +# libraries which are loaded via DT_NEEDED entries in the executable. |
| 24 | ifeq (yes,$(build-hardcoded-path-in-tests)) |
| 25 | -link-libc-tests-rpath-link = $(link-libc-rpath) |
| 26 | -link-test-modules-rpath-link = $(link-libc-rpath) |
| 27 | +link-libc-tests-rpath-link = $(link-libc-rpath) -Wl,--disable-new-dtags |
| 28 | +link-test-modules-rpath-link = $(link-libc-rpath) -Wl,--disable-new-dtags |
| 29 | else |
| 30 | link-libc-tests-rpath-link = $(link-libc-rpath-link) |
| 31 | link-test-modules-rpath-link = |
| 32 | diff --git a/Makefile b/Makefile |
| 33 | index 7052b46df8..2e351c0321 100644 |
| 34 | --- a/Makefile |
| 35 | +++ b/Makefile |
| 36 | @@ -577,6 +577,13 @@ $(objpfx)lint-makefiles.out: scripts/lint-makefiles.sh |
| 37 | $(SHELL) $< "$(PYTHON)" `pwd` > $@ ; \ |
| 38 | $(evaluate-test) |
| 39 | |
| 40 | +# Link libc.a as a whole to verify that it does not contain multiple |
| 41 | +# definitions of any symbols. |
| 42 | +tests-special += $(objpfx)link-static-libc.out |
| 43 | +$(objpfx)link-static-libc.out: |
| 44 | + $(LINK.o) $(whole-archive) -r $(objpfx)libc.a -o /dev/null > $@ 2>&1; \ |
| 45 | + $(evaluate-test) |
| 46 | + |
| 47 | # Print test summary for tests in $1 .sum file; |
| 48 | # $2 is optional test identifier. |
| 49 | # Fail if there are unexpected failures in the test results. |
| 50 | diff --git a/advisories/GLIBC-SA-2023-0001 b/advisories/GLIBC-SA-2023-0001 |
| 51 | deleted file mode 100644 |
| 52 | index 3d19c91b6a..0000000000 |
| 53 | --- a/advisories/GLIBC-SA-2023-0001 |
| 54 | +++ /dev/null |
| 55 | @@ -1,14 +0,0 @@ |
| 56 | -printf: incorrect output for integers with thousands separator and width field |
| 57 | - |
| 58 | -When the printf family of functions is called with a format specifier |
| 59 | -that uses an <apostrophe> (enable grouping) and a minimum width |
| 60 | -specifier, the resulting output could be larger than reasonably expected |
| 61 | -by a caller that computed a tight bound on the buffer size. The |
| 62 | -resulting larger than expected output could result in a buffer overflow |
| 63 | -in the printf family of functions. |
| 64 | - |
| 65 | -CVE-Id: CVE-2023-25139 |
| 66 | -Public-Date: 2023-02-02 |
| 67 | -Vulnerable-Commit: e88b9f0e5cc50cab57a299dc7efe1a4eb385161d (2.37) |
| 68 | -Fix-Commit: c980549cc6a1c03c23cc2fe3e7b0fe626a0364b0 (2.38) |
| 69 | -Fix-Commit: 07b9521fc6369d000216b96562ff7c0ed32a16c4 (2.37-4) |
| 70 | diff --git a/advisories/GLIBC-SA-2023-0002 b/advisories/GLIBC-SA-2023-0002 |
| 71 | deleted file mode 100644 |
| 72 | index 5122669a64..0000000000 |
| 73 | --- a/advisories/GLIBC-SA-2023-0002 |
| 74 | +++ /dev/null |
| 75 | @@ -1,15 +0,0 @@ |
| 76 | -getaddrinfo: Stack read overflow in no-aaaa mode |
| 77 | - |
| 78 | -If the system is configured in no-aaaa mode via /etc/resolv.conf, |
| 79 | -getaddrinfo is called for the AF_UNSPEC address family, and a DNS |
| 80 | -response is received over TCP that is larger than 2048 bytes, |
| 81 | -getaddrinfo may potentially disclose stack contents via the returned |
| 82 | -address data, or crash. |
| 83 | - |
| 84 | -CVE-Id: CVE-2023-4527 |
| 85 | -Public-Date: 2023-09-12 |
| 86 | -Vulnerable-Commit: f282cdbe7f436c75864e5640a409a10485e9abb2 (2.36) |
| 87 | -Fix-Commit: bd77dd7e73e3530203be1c52c8a29d08270cb25d (2.39) |
| 88 | -Fix-Commit: 4ea972b7edd7e36610e8cde18bf7a8149d7bac4f (2.36-113) |
| 89 | -Fix-Commit: b7529346025a130fee483d42178b5c118da971bb (2.37-38) |
| 90 | -Fix-Commit: b25508dd774b617f99419bdc3cf2ace4560cd2d6 (2.38-19) |
| 91 | diff --git a/advisories/GLIBC-SA-2023-0003 b/advisories/GLIBC-SA-2023-0003 |
| 92 | deleted file mode 100644 |
| 93 | index d3aef80348..0000000000 |
| 94 | --- a/advisories/GLIBC-SA-2023-0003 |
| 95 | +++ /dev/null |
| 96 | @@ -1,15 +0,0 @@ |
| 97 | -getaddrinfo: Potential use-after-free |
| 98 | - |
| 99 | -When an NSS plugin only implements the _gethostbyname2_r and |
| 100 | -_getcanonname_r callbacks, getaddrinfo could use memory that was freed |
| 101 | -during buffer resizing, potentially causing a crash or read or write to |
| 102 | -arbitrary memory. |
| 103 | - |
| 104 | -CVE-Id: CVE-2023-4806 |
| 105 | -Public-Date: 2023-09-12 |
| 106 | -Fix-Commit: 973fe93a5675c42798b2161c6f29c01b0e243994 (2.39) |
| 107 | -Fix-Commit: e09ee267c03e3150c2c9ba28625ab130705a485e (2.34-420) |
| 108 | -Fix-Commit: e3ccb230a961b4797510e6a1f5f21fd9021853e7 (2.35-270) |
| 109 | -Fix-Commit: a9728f798ec7f05454c95637ee6581afaa9b487d (2.36-115) |
| 110 | -Fix-Commit: 6529a7466c935f36e9006b854d6f4e1d4876f942 (2.37-39) |
| 111 | -Fix-Commit: 00ae4f10b504bc4564e9f22f00907093f1ab9338 (2.38-20) |
| 112 | diff --git a/advisories/GLIBC-SA-2023-0004 b/advisories/GLIBC-SA-2023-0004 |
| 113 | deleted file mode 100644 |
| 114 | index 5286a7aa54..0000000000 |
| 115 | --- a/advisories/GLIBC-SA-2023-0004 |
| 116 | +++ /dev/null |
| 117 | @@ -1,16 +0,0 @@ |
| 118 | -tunables: local privilege escalation through buffer overflow |
| 119 | - |
| 120 | -If a tunable of the form NAME=NAME=VAL is passed in the environment of a |
| 121 | -setuid program and NAME is valid, it may result in a buffer overflow, |
| 122 | -which could be exploited to achieve escalated privileges. This flaw was |
| 123 | -introduced in glibc 2.34. |
| 124 | - |
| 125 | -CVE-Id: CVE-2023-4911 |
| 126 | -Public-Date: 2023-10-03 |
| 127 | -Vulnerable-Commit: 2ed18c5b534d9e92fc006202a5af0df6b72e7aca (2.34) |
| 128 | -Fix-Commit: 1056e5b4c3f2d90ed2b4a55f96add28da2f4c8fa (2.39) |
| 129 | -Fix-Commit: dcc367f148bc92e7f3778a125f7a416b093964d9 (2.34-423) |
| 130 | -Fix-Commit: c84018a05aec80f5ee6f682db0da1130b0196aef (2.35-274) |
| 131 | -Fix-Commit: 22955ad85186ee05834e47e665056148ca07699c (2.36-118) |
| 132 | -Fix-Commit: b4e23c75aea756b4bddc4abcf27a1c6dca8b6bd3 (2.37-45) |
| 133 | -Fix-Commit: 750a45a783906a19591fb8ff6b7841470f1f5701 (2.38-27) |
| 134 | diff --git a/advisories/GLIBC-SA-2023-0005 b/advisories/GLIBC-SA-2023-0005 |
| 135 | deleted file mode 100644 |
| 136 | index cc4eb90b82..0000000000 |
| 137 | --- a/advisories/GLIBC-SA-2023-0005 |
| 138 | +++ /dev/null |
| 139 | @@ -1,18 +0,0 @@ |
| 140 | -getaddrinfo: DoS due to memory leak |
| 141 | - |
| 142 | -The fix for CVE-2023-4806 introduced a memory leak when an application |
| 143 | -calls getaddrinfo for AF_INET6 with AI_CANONNAME, AI_ALL and AI_V4MAPPED |
| 144 | -flags set. |
| 145 | - |
| 146 | -CVE-Id: CVE-2023-5156 |
| 147 | -Public-Date: 2023-09-25 |
| 148 | -Vulnerable-Commit: e09ee267c03e3150c2c9ba28625ab130705a485e (2.34-420) |
| 149 | -Vulnerable-Commit: e3ccb230a961b4797510e6a1f5f21fd9021853e7 (2.35-270) |
| 150 | -Vulnerable-Commit: a9728f798ec7f05454c95637ee6581afaa9b487d (2.36-115) |
| 151 | -Vulnerable-Commit: 6529a7466c935f36e9006b854d6f4e1d4876f942 (2.37-39) |
| 152 | -Vulnerable-Commit: 00ae4f10b504bc4564e9f22f00907093f1ab9338 (2.38-20) |
| 153 | -Fix-Commit: 8006457ab7e1cd556b919f477348a96fe88f2e49 (2.34-421) |
| 154 | -Fix-Commit: 17092c0311f954e6f3c010f73ce3a78c24ac279a (2.35-272) |
| 155 | -Fix-Commit: 856bac55f98dc840e7c27cfa82262b933385de90 (2.36-116) |
| 156 | -Fix-Commit: 4473d1b87d04b25cdd0e0354814eeaa421328268 (2.37-42) |
| 157 | -Fix-Commit: 5ee59ca371b99984232d7584fe2b1a758b4421d3 (2.38-24) |
| 158 | diff --git a/advisories/GLIBC-SA-2024-0001 b/advisories/GLIBC-SA-2024-0001 |
| 159 | deleted file mode 100644 |
| 160 | index 28931c75ae..0000000000 |
| 161 | --- a/advisories/GLIBC-SA-2024-0001 |
| 162 | +++ /dev/null |
| 163 | @@ -1,15 +0,0 @@ |
| 164 | -syslog: Heap buffer overflow in __vsyslog_internal |
| 165 | - |
| 166 | -__vsyslog_internal did not handle a case where printing a SYSLOG_HEADER |
| 167 | -containing a long program name failed to update the required buffer |
| 168 | -size, leading to the allocation and overflow of a too-small buffer on |
| 169 | -the heap. |
| 170 | - |
| 171 | -CVE-Id: CVE-2023-6246 |
| 172 | -Public-Date: 2024-01-30 |
| 173 | -Vulnerable-Commit: 52a5be0df411ef3ff45c10c7c308cb92993d15b1 (2.37) |
| 174 | -Fix-Commit: 6bd0e4efcc78f3c0115e5ea9739a1642807450da (2.39) |
| 175 | -Fix-Commit: 23514c72b780f3da097ecf33a793b7ba9c2070d2 (2.38-42) |
| 176 | -Fix-Commit: 97a4292aa4a2642e251472b878d0ec4c46a0e59a (2.37-57) |
| 177 | -Vulnerable-Commit: b0e7888d1fa2dbd2d9e1645ec8c796abf78880b9 (2.36-16) |
| 178 | -Fix-Commit: d1a83b6767f68b3cb5b4b4ea2617254acd040c82 (2.36-126) |
| 179 | diff --git a/advisories/GLIBC-SA-2024-0002 b/advisories/GLIBC-SA-2024-0002 |
| 180 | deleted file mode 100644 |
| 181 | index 940bfcf2fc..0000000000 |
| 182 | --- a/advisories/GLIBC-SA-2024-0002 |
| 183 | +++ /dev/null |
| 184 | @@ -1,15 +0,0 @@ |
| 185 | -syslog: Heap buffer overflow in __vsyslog_internal |
| 186 | - |
| 187 | -__vsyslog_internal used the return value of snprintf/vsnprintf to |
| 188 | -calculate buffer sizes for memory allocation. If these functions (for |
| 189 | -any reason) failed and returned -1, the resulting buffer would be too |
| 190 | -small to hold output. |
| 191 | - |
| 192 | -CVE-Id: CVE-2023-6779 |
| 193 | -Public-Date: 2024-01-30 |
| 194 | -Vulnerable-Commit: 52a5be0df411ef3ff45c10c7c308cb92993d15b1 (2.37) |
| 195 | -Fix-Commit: 7e5a0c286da33159d47d0122007aac016f3e02cd (2.39) |
| 196 | -Fix-Commit: d0338312aace5bbfef85e03055e1212dd0e49578 (2.38-43) |
| 197 | -Fix-Commit: 67062eccd9a65d7fda9976a56aeaaf6c25a80214 (2.37-58) |
| 198 | -Vulnerable-Commit: b0e7888d1fa2dbd2d9e1645ec8c796abf78880b9 (2.36-16) |
| 199 | -Fix-Commit: 2bc9d7c002bdac38b5c2a3f11b78e309d7765b83 (2.36-127) |
| 200 | diff --git a/advisories/GLIBC-SA-2024-0003 b/advisories/GLIBC-SA-2024-0003 |
| 201 | deleted file mode 100644 |
| 202 | index b43a5150ab..0000000000 |
| 203 | --- a/advisories/GLIBC-SA-2024-0003 |
| 204 | +++ /dev/null |
| 205 | @@ -1,13 +0,0 @@ |
| 206 | -syslog: Integer overflow in __vsyslog_internal |
| 207 | - |
| 208 | -__vsyslog_internal calculated a buffer size by adding two integers, but |
| 209 | -did not first check if the addition would overflow. |
| 210 | - |
| 211 | -CVE-Id: CVE-2023-6780 |
| 212 | -Public-Date: 2024-01-30 |
| 213 | -Vulnerable-Commit: 52a5be0df411ef3ff45c10c7c308cb92993d15b1 (2.37) |
| 214 | -Fix-Commit: ddf542da94caf97ff43cc2875c88749880b7259b (2.39) |
| 215 | -Fix-Commit: d37c2b20a4787463d192b32041c3406c2bd91de0 (2.38-44) |
| 216 | -Fix-Commit: 2b58cba076e912961ceaa5fa58588e4b10f791c0 (2.37-59) |
| 217 | -Vulnerable-Commit: b0e7888d1fa2dbd2d9e1645ec8c796abf78880b9 (2.36-16) |
| 218 | -Fix-Commit: b9b7d6a27aa0632f334352fa400771115b3c69b7 (2.36-128) |
| 219 | diff --git a/advisories/README b/advisories/README |
| 220 | deleted file mode 100644 |
| 221 | index 94e68b1350..0000000000 |
| 222 | --- a/advisories/README |
| 223 | +++ /dev/null |
| 224 | @@ -1,73 +0,0 @@ |
| 225 | -GNU C Library Security Advisory Format |
| 226 | -====================================== |
| 227 | - |
| 228 | -Security advisories in this directory follow a simple git commit log |
| 229 | -format, with a heading and free-format description augmented with tags |
| 230 | -to allow parsing key information. References to code changes are |
| 231 | -specific to the glibc repository and follow a specific format: |
| 232 | - |
| 233 | - Tag-name: <commit-ref> (release-version) |
| 234 | - |
| 235 | -The <commit-ref> indicates a specific commit in the repository. The |
| 236 | -release-version indicates the publicly consumable release in which this |
| 237 | -commit is known to exist. The release-version is derived from the |
| 238 | -git-describe format, (i.e. stripped out from glibc-2.34.NNN-gxxxx) and |
| 239 | -is of the form 2.34-NNN. If the -NNN suffix is absent, it means that |
| 240 | -the change is in that release tarball, otherwise the change is on the |
| 241 | -release/2.YY/master branch and not in any released tarball. |
| 242 | - |
| 243 | -The following tags are currently being used: |
| 244 | - |
| 245 | -CVE-Id: |
| 246 | -This is the CVE-Id assigned under the CVE Program |
| 247 | -(https://www.cve.org/). |
| 248 | - |
| 249 | -Public-Date: |
| 250 | -The date this issue became publicly known. |
| 251 | - |
| 252 | -Vulnerable-Commit: |
| 253 | -The commit that introduced this vulnerability. There could be multiple |
| 254 | -entries, one for each release branch in the glibc repository; the |
| 255 | -release-version portion of this tag should tell you which branch this is |
| 256 | -on. |
| 257 | - |
| 258 | -Fix-Commit: |
| 259 | -The commit that fixed this vulnerability. There could be multiple |
| 260 | -entries for each release branch in the glibc repository, indicating that |
| 261 | -all of those commits contributed to fixing that issue in each of those |
| 262 | -branches. |
| 263 | - |
| 264 | -Adding an Advisory |
| 265 | ------------------- |
| 266 | - |
| 267 | -An advisory for a CVE needs to be added on the master branch in two steps: |
| 268 | - |
| 269 | -1. Add the text of the advisory without any Fix-Commit tags along with |
| 270 | - the fix for the CVE. Add the Vulnerable-Commit tag, if applicable. |
| 271 | - The advisories directory does not exist in release branches, so keep |
| 272 | - the advisory text commit distinct from the code changes, to ease |
| 273 | - backports. Ask for the GLIBC-SA advisory number from the security |
| 274 | - team. |
| 275 | - |
| 276 | -2. Finish all backports on release branches and then back on the msater |
| 277 | - branch, add all commit refs to the advisory using the Fix-Commit |
| 278 | - tags. Don't bother adding the release-version subscript since the |
| 279 | - next step will overwrite it. |
| 280 | - |
| 281 | -3. Run the process-advisories.sh script in the scripts directory on the |
| 282 | - advisory: |
| 283 | - |
| 284 | - scripts/process-advisories.sh update GLIBC-SA-YYYY-NNNN |
| 285 | - |
| 286 | - (replace YYYY-NNNN with the actual advisory number). |
| 287 | - |
| 288 | -4. Verify the updated advisory and push the result. |
| 289 | - |
| 290 | -Getting a NEWS snippet from advisories |
| 291 | --------------------------------------- |
| 292 | - |
| 293 | -Run: |
| 294 | - |
| 295 | - scripts/process-advisories.sh news |
| 296 | - |
| 297 | -and copy the content into the NEWS file. |
| 298 | diff --git a/bits/wordsize.h b/bits/wordsize.h |
| 299 | index 14edae3a11..53013a9275 100644 |
| 300 | --- a/bits/wordsize.h |
| 301 | +++ b/bits/wordsize.h |
| 302 | @@ -21,7 +21,9 @@ |
| 303 | #define __WORDSIZE32_PTRDIFF_LONG |
| 304 | |
| 305 | /* Set to 1 in order to force time types to be 32 bits instead of 64 bits in |
| 306 | - struct lastlog and struct utmp{,x} on 64-bit ports. This may be done in |
| 307 | + struct lastlog and struct utmp{,x}. This may be done in |
| 308 | order to make 64-bit ports compatible with 32-bit ports. Set to 0 for |
| 309 | - 64-bit ports where the time types are 64-bits or for any 32-bit ports. */ |
| 310 | + 64-bit ports where the time types are 64-bits and new 32-bit ports |
| 311 | + where time_t is 64 bits, and there is no companion architecture with |
| 312 | + 32-bit time_t. */ |
| 313 | #define __WORDSIZE_TIME64_COMPAT32 |
| 314 | diff --git a/config.h.in b/config.h.in |
| 315 | index 44a34072a4..1e647de585 100644 |
| 316 | --- a/config.h.in |
| 317 | +++ b/config.h.in |
| 318 | @@ -141,6 +141,9 @@ |
| 319 | /* LOONGARCH floating-point ABI for ld.so. */ |
| 320 | #undef LOONGARCH_ABI_FRLEN |
| 321 | |
| 322 | +/* Define whether ARM used hard-float and support VFPvX-D32. */ |
| 323 | +#undef HAVE_ARM_PCS_VFP_D32 |
| 324 | + |
| 325 | /* Linux specific: minimum supported kernel version. */ |
| 326 | #undef __LINUX_KERNEL_VERSION |
| 327 | |
| 328 | @@ -283,6 +286,9 @@ |
| 329 | /* Define if x86 ISA level should be included in shared libraries. */ |
| 330 | #undef INCLUDE_X86_ISA_LEVEL |
| 331 | |
| 332 | +/* The x86 ISA level. 1 for baseline. Undefined on non-x86. */ |
| 333 | +#undef MINIMUM_X86_ISA_LEVEL |
| 334 | + |
| 335 | /* Define if -msahf is enabled by default on x86. */ |
| 336 | #undef HAVE_X86_LAHF_SAHF |
| 337 | |
| 338 | diff --git a/configure b/configure |
| 339 | index 59ff1e415d..432e40a592 100755 |
| 340 | --- a/configure |
| 341 | +++ b/configure |
| 342 | @@ -653,7 +653,7 @@ LIBGD |
| 343 | libc_cv_cc_loop_to_function |
| 344 | libc_cv_cc_submachine |
| 345 | libc_cv_cc_nofma |
| 346 | -libc_cv_mtls_dialect_gnu2 |
| 347 | +libc_cv_mtls_descriptor |
| 348 | libc_cv_has_glob_dat |
| 349 | libc_cv_fpie |
| 350 | libc_cv_z_execstack |
| 351 | @@ -4760,6 +4760,9 @@ libc_config_ok=no |
| 352 | # whether to use such directories. |
| 353 | with_fp_cond=1 |
| 354 | |
| 355 | +# A preconfigure script may define another name to TLS descriptor variant |
| 356 | +mtls_descriptor=gnu2 |
| 357 | + |
| 358 | if frags=`ls -d $srcdir/sysdeps/*/preconfigure 2> /dev/null` |
| 359 | then |
| 360 | { printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking for sysdeps preconfigure fragments" >&5 |
| 361 | @@ -7006,9 +7009,9 @@ fi |
| 362 | printf "%s\n" "$libc_cv_has_glob_dat" >&6; } |
| 363 | |
| 364 | |
| 365 | -{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking for -mtls-dialect=gnu2" >&5 |
| 366 | -printf %s "checking for -mtls-dialect=gnu2... " >&6; } |
| 367 | -if test ${libc_cv_mtls_dialect_gnu2+y} |
| 368 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking for tls descriptor support" >&5 |
| 369 | +printf %s "checking for tls descriptor support... " >&6; } |
| 370 | +if test ${libc_cv_mtls_descriptor+y} |
| 371 | then : |
| 372 | printf %s "(cached) " >&6 |
| 373 | else $as_nop |
| 374 | @@ -7019,25 +7022,25 @@ void foo (void) |
| 375 | i = 10; |
| 376 | } |
| 377 | EOF |
| 378 | -if { ac_try='${CC-cc} $CFLAGS $CPPFLAGS -fPIC -mtls-dialect=gnu2 -nostdlib -nostartfiles |
| 379 | - conftest.c -o conftest 1>&5' |
| 380 | +if { ac_try='${CC-cc} $CFLAGS $CPPFLAGS -fPIC -mtls-dialect=$mtls_descriptor -nostdlib -nostartfiles |
| 381 | + -shared conftest.c -o conftest 1>&5' |
| 382 | { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5 |
| 383 | (eval $ac_try) 2>&5 |
| 384 | ac_status=$? |
| 385 | printf "%s\n" "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5 |
| 386 | test $ac_status = 0; }; } |
| 387 | then |
| 388 | - libc_cv_mtls_dialect_gnu2=yes |
| 389 | + libc_cv_mtls_descriptor=$mtls_descriptor |
| 390 | else |
| 391 | - libc_cv_mtls_dialect_gnu2=no |
| 392 | + libc_cv_mtls_descriptor=no |
| 393 | fi |
| 394 | rm -f conftest* |
| 395 | fi |
| 396 | -{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: result: $libc_cv_mtls_dialect_gnu2" >&5 |
| 397 | -printf "%s\n" "$libc_cv_mtls_dialect_gnu2" >&6; } |
| 398 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: result: $libc_cv_mtls_descriptor" >&5 |
| 399 | +printf "%s\n" "$libc_cv_mtls_descriptor" >&6; } |
| 400 | |
| 401 | config_vars="$config_vars |
| 402 | -have-mtls-dialect-gnu2 = $libc_cv_mtls_dialect_gnu2" |
| 403 | +have-mtls-descriptor = $libc_cv_mtls_descriptor" |
| 404 | |
| 405 | { printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking if -Wno-ignored-attributes is required for aliases" >&5 |
| 406 | printf %s "checking if -Wno-ignored-attributes is required for aliases... " >&6; } |
| 407 | diff --git a/configure.ac b/configure.ac |
| 408 | index 65799e5685..bdc385d03c 100644 |
| 409 | --- a/configure.ac |
| 410 | +++ b/configure.ac |
| 411 | @@ -442,6 +442,9 @@ libc_config_ok=no |
| 412 | # whether to use such directories. |
| 413 | with_fp_cond=1 |
| 414 | |
| 415 | +# A preconfigure script may define another name to TLS descriptor variant |
| 416 | +mtls_descriptor=gnu2 |
| 417 | + |
| 418 | dnl Let sysdeps/*/preconfigure act here. |
| 419 | LIBC_PRECONFIGURE([$srcdir], [for sysdeps]) |
| 420 | |
| 421 | @@ -1287,7 +1290,7 @@ fi |
| 422 | rm -f conftest*]) |
| 423 | AC_SUBST(libc_cv_has_glob_dat) |
| 424 | |
| 425 | -AC_CACHE_CHECK([for -mtls-dialect=gnu2], libc_cv_mtls_dialect_gnu2, |
| 426 | +AC_CACHE_CHECK([for tls descriptor support], libc_cv_mtls_descriptor, |
| 427 | [dnl |
| 428 | cat > conftest.c <<EOF |
| 429 | __thread int i; |
| 430 | @@ -1296,16 +1299,16 @@ void foo (void) |
| 431 | i = 10; |
| 432 | } |
| 433 | EOF |
| 434 | -if AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS -fPIC -mtls-dialect=gnu2 -nostdlib -nostartfiles |
| 435 | - conftest.c -o conftest 1>&AS_MESSAGE_LOG_FD]) |
| 436 | +if AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS -fPIC -mtls-dialect=$mtls_descriptor -nostdlib -nostartfiles |
| 437 | + -shared conftest.c -o conftest 1>&AS_MESSAGE_LOG_FD]) |
| 438 | then |
| 439 | - libc_cv_mtls_dialect_gnu2=yes |
| 440 | + libc_cv_mtls_descriptor=$mtls_descriptor |
| 441 | else |
| 442 | - libc_cv_mtls_dialect_gnu2=no |
| 443 | + libc_cv_mtls_descriptor=no |
| 444 | fi |
| 445 | rm -f conftest*]) |
| 446 | -AC_SUBST(libc_cv_mtls_dialect_gnu2) |
| 447 | -LIBC_CONFIG_VAR([have-mtls-dialect-gnu2], [$libc_cv_mtls_dialect_gnu2]) |
| 448 | +AC_SUBST(libc_cv_mtls_descriptor) |
| 449 | +LIBC_CONFIG_VAR([have-mtls-descriptor], [$libc_cv_mtls_descriptor]) |
| 450 | |
| 451 | dnl clang emits an warning for a double alias redirection, to warn the |
| 452 | dnl original symbol is sed even when weak definition overrides it. |
| 453 | diff --git a/elf/Makefile b/elf/Makefile |
| 454 | index 5d78b659ce..a50a988e73 100644 |
| 455 | --- a/elf/Makefile |
| 456 | +++ b/elf/Makefile |
| 457 | @@ -170,6 +170,7 @@ CFLAGS-.op += $(call elide-stack-protector,.op,$(elide-routines.os)) |
| 458 | CFLAGS-.os += $(call elide-stack-protector,.os,$(all-rtld-routines)) |
| 459 | |
| 460 | # Add the requested compiler flags to the early startup code. |
| 461 | +CFLAGS-dl-misc.os += $(rtld-early-cflags) |
| 462 | CFLAGS-dl-printf.os += $(rtld-early-cflags) |
| 463 | CFLAGS-dl-setup_hash.os += $(rtld-early-cflags) |
| 464 | CFLAGS-dl-sysdep.os += $(rtld-early-cflags) |
| 465 | @@ -424,6 +425,7 @@ tests += \ |
| 466 | tst-glibc-hwcaps-prepend \ |
| 467 | tst-global1 \ |
| 468 | tst-global2 \ |
| 469 | + tst-gnu2-tls2 \ |
| 470 | tst-initfinilazyfail \ |
| 471 | tst-initorder \ |
| 472 | tst-initorder2 \ |
| 473 | @@ -846,6 +848,9 @@ modules-names += \ |
| 474 | tst-filterobj-flt \ |
| 475 | tst-finilazyfailmod \ |
| 476 | tst-globalmod2 \ |
| 477 | + tst-gnu2-tls2mod0 \ |
| 478 | + tst-gnu2-tls2mod1 \ |
| 479 | + tst-gnu2-tls2mod2 \ |
| 480 | tst-initlazyfailmod \ |
| 481 | tst-initorder2a \ |
| 482 | tst-initorder2b \ |
| 483 | @@ -995,13 +1000,13 @@ modules-names-tests = $(filter-out ifuncmod% tst-tlsmod%,\ |
| 484 | # For +depfiles in Makerules. |
| 485 | extra-test-objs += tst-auditmod17.os |
| 486 | |
| 487 | -ifeq (yes,$(have-mtls-dialect-gnu2)) |
| 488 | +ifneq (no,$(have-mtls-descriptor)) |
| 489 | tests += tst-gnu2-tls1 |
| 490 | modules-names += tst-gnu2-tls1mod |
| 491 | $(objpfx)tst-gnu2-tls1: $(objpfx)tst-gnu2-tls1mod.so |
| 492 | tst-gnu2-tls1mod.so-no-z-defs = yes |
| 493 | -CFLAGS-tst-gnu2-tls1mod.c += -mtls-dialect=gnu2 |
| 494 | -endif # $(have-mtls-dialect-gnu2) |
| 495 | +CFLAGS-tst-gnu2-tls1mod.c += -mtls-dialect=$(have-mtls-descriptor) |
| 496 | +endif # $(have-mtls-descriptor) |
| 497 | |
| 498 | ifeq (yes,$(have-protected-data)) |
| 499 | modules-names += tst-protected1moda tst-protected1modb |
| 500 | @@ -2968,11 +2973,11 @@ $(objpfx)tst-tls-allocation-failure-static-patched.out: \ |
| 501 | $(objpfx)tst-audit-tlsdesc: $(objpfx)tst-audit-tlsdesc-mod1.so \ |
| 502 | $(objpfx)tst-audit-tlsdesc-mod2.so \ |
| 503 | $(shared-thread-library) |
| 504 | -ifeq (yes,$(have-mtls-dialect-gnu2)) |
| 505 | +ifneq (no,$(have-mtls-descriptor)) |
| 506 | # The test is valid for all TLS types, but we want to exercise GNU2 |
| 507 | # TLS if possible. |
| 508 | -CFLAGS-tst-audit-tlsdesc-mod1.c += -mtls-dialect=gnu2 |
| 509 | -CFLAGS-tst-audit-tlsdesc-mod2.c += -mtls-dialect=gnu2 |
| 510 | +CFLAGS-tst-audit-tlsdesc-mod1.c += -mtls-dialect=$(have-mtls-descriptor) |
| 511 | +CFLAGS-tst-audit-tlsdesc-mod2.c += -mtls-dialect=$(have-mtls-descriptor) |
| 512 | endif |
| 513 | $(objpfx)tst-audit-tlsdesc-dlopen: $(shared-thread-library) |
| 514 | $(objpfx)tst-audit-tlsdesc-dlopen.out: $(objpfx)tst-audit-tlsdesc-mod1.so \ |
| 515 | @@ -3044,8 +3049,18 @@ $(objpfx)tst-tlsgap.out: \ |
| 516 | $(objpfx)tst-tlsgap-mod0.so \ |
| 517 | $(objpfx)tst-tlsgap-mod1.so \ |
| 518 | $(objpfx)tst-tlsgap-mod2.so |
| 519 | -ifeq (yes,$(have-mtls-dialect-gnu2)) |
| 520 | -CFLAGS-tst-tlsgap-mod0.c += -mtls-dialect=gnu2 |
| 521 | -CFLAGS-tst-tlsgap-mod1.c += -mtls-dialect=gnu2 |
| 522 | -CFLAGS-tst-tlsgap-mod2.c += -mtls-dialect=gnu2 |
| 523 | + |
| 524 | +$(objpfx)tst-gnu2-tls2: $(shared-thread-library) |
| 525 | +$(objpfx)tst-gnu2-tls2.out: \ |
| 526 | + $(objpfx)tst-gnu2-tls2mod0.so \ |
| 527 | + $(objpfx)tst-gnu2-tls2mod1.so \ |
| 528 | + $(objpfx)tst-gnu2-tls2mod2.so |
| 529 | + |
| 530 | +ifneq (no,$(have-mtls-descriptor)) |
| 531 | +CFLAGS-tst-tlsgap-mod0.c += -mtls-dialect=$(have-mtls-descriptor) |
| 532 | +CFLAGS-tst-tlsgap-mod1.c += -mtls-dialect=$(have-mtls-descriptor) |
| 533 | +CFLAGS-tst-tlsgap-mod2.c += -mtls-dialect=$(have-mtls-descriptor) |
| 534 | +CFLAGS-tst-gnu2-tls2mod0.c += -mtls-dialect=$(have-mtls-descriptor) |
| 535 | +CFLAGS-tst-gnu2-tls2mod1.c += -mtls-dialect=$(have-mtls-descriptor) |
| 536 | +CFLAGS-tst-gnu2-tls2mod2.c += -mtls-dialect=$(have-mtls-descriptor) |
| 537 | endif |
| 538 | diff --git a/elf/dl-diagnostics.c b/elf/dl-diagnostics.c |
| 539 | index 7345ebc4e5..aaf67b87e8 100644 |
| 540 | --- a/elf/dl-diagnostics.c |
| 541 | +++ b/elf/dl-diagnostics.c |
| 542 | @@ -235,6 +235,8 @@ _dl_print_diagnostics (char **environ) |
| 543 | _dl_diagnostics_print_labeled_value ("dl_hwcap", GLRO (dl_hwcap)); |
| 544 | _dl_diagnostics_print_labeled_value ("dl_hwcap_important", HWCAP_IMPORTANT); |
| 545 | _dl_diagnostics_print_labeled_value ("dl_hwcap2", GLRO (dl_hwcap2)); |
| 546 | + _dl_diagnostics_print_labeled_value ("dl_hwcap3", GLRO (dl_hwcap3)); |
| 547 | + _dl_diagnostics_print_labeled_value ("dl_hwcap4", GLRO (dl_hwcap4)); |
| 548 | _dl_diagnostics_print_labeled_string |
| 549 | ("dl_hwcaps_subdirs", _dl_hwcaps_subdirs); |
| 550 | _dl_diagnostics_print_labeled_value |
| 551 | diff --git a/elf/dl-support.c b/elf/dl-support.c |
| 552 | index 2f502c8b0d..451932dd03 100644 |
| 553 | --- a/elf/dl-support.c |
| 554 | +++ b/elf/dl-support.c |
| 555 | @@ -158,6 +158,8 @@ const ElfW(Phdr) *_dl_phdr; |
| 556 | size_t _dl_phnum; |
| 557 | uint64_t _dl_hwcap; |
| 558 | uint64_t _dl_hwcap2; |
| 559 | +uint64_t _dl_hwcap3; |
| 560 | +uint64_t _dl_hwcap4; |
| 561 | |
| 562 | enum dso_sort_algorithm _dl_dso_sort_algo; |
| 563 | |
| 564 | diff --git a/elf/dl-tunables.c b/elf/dl-tunables.c |
| 565 | index 03e1a68675..614ac9c047 100644 |
| 566 | --- a/elf/dl-tunables.c |
| 567 | +++ b/elf/dl-tunables.c |
| 568 | @@ -32,6 +32,7 @@ |
| 569 | #include <ldsodefs.h> |
| 570 | #include <array_length.h> |
| 571 | #include <dl-minimal-malloc.h> |
| 572 | +#include <dl-symbol-redir-ifunc.h> |
| 573 | |
| 574 | #define TUNABLES_INTERNAL 1 |
| 575 | #include "dl-tunables.h" |
| 576 | @@ -223,6 +224,7 @@ parse_tunables_string (const char *valstring, struct tunable_toset_t *tunables) |
| 577 | { |
| 578 | tunables[ntunables++] = |
| 579 | (struct tunable_toset_t) { cur, value, p - value }; |
| 580 | + |
| 581 | break; |
| 582 | } |
| 583 | } |
| 584 | @@ -234,23 +236,27 @@ parse_tunables_string (const char *valstring, struct tunable_toset_t *tunables) |
| 585 | static void |
| 586 | parse_tunables (const char *valstring) |
| 587 | { |
| 588 | - struct tunable_toset_t tunables[tunables_list_size]; |
| 589 | - int ntunables = parse_tunables_string (valstring, tunables); |
| 590 | - if (ntunables == -1) |
| 591 | + struct tunable_toset_t tunables[tunables_list_size] = { 0 }; |
| 592 | + if (parse_tunables_string (valstring, tunables) == -1) |
| 593 | { |
| 594 | _dl_error_printf ( |
| 595 | "WARNING: ld.so: invalid GLIBC_TUNABLES `%s': ignored.\n", valstring); |
| 596 | return; |
| 597 | } |
| 598 | |
| 599 | - for (int i = 0; i < ntunables; i++) |
| 600 | - if (!tunable_initialize (tunables[i].t, tunables[i].value, |
| 601 | - tunables[i].len)) |
| 602 | - _dl_error_printf ("WARNING: ld.so: invalid GLIBC_TUNABLES value `%.*s' " |
| 603 | - "for option `%s': ignored.\n", |
| 604 | - (int) tunables[i].len, |
| 605 | - tunables[i].value, |
| 606 | - tunables[i].t->name); |
| 607 | + for (int i = 0; i < tunables_list_size; i++) |
| 608 | + { |
| 609 | + if (tunables[i].t == NULL) |
| 610 | + continue; |
| 611 | + |
| 612 | + if (!tunable_initialize (tunables[i].t, tunables[i].value, |
| 613 | + tunables[i].len)) |
| 614 | + _dl_error_printf ("WARNING: ld.so: invalid GLIBC_TUNABLES value `%.*s' " |
| 615 | + "for option `%s': ignored.\n", |
| 616 | + (int) tunables[i].len, |
| 617 | + tunables[i].value, |
| 618 | + tunables[i].t->name); |
| 619 | + } |
| 620 | } |
| 621 | |
| 622 | /* Initialize the tunables list from the environment. For now we only use the |
| 623 | diff --git a/elf/elf.h b/elf/elf.h |
| 624 | index 455731663c..1c394c64cd 100644 |
| 625 | --- a/elf/elf.h |
| 626 | +++ b/elf/elf.h |
| 627 | @@ -1234,6 +1234,10 @@ typedef struct |
| 628 | #define AT_RSEQ_FEATURE_SIZE 27 /* rseq supported feature size. */ |
| 629 | #define AT_RSEQ_ALIGN 28 /* rseq allocation alignment. */ |
| 630 | |
| 631 | +/* More machine-dependent hints about processor capabilities. */ |
| 632 | +#define AT_HWCAP3 29 /* extension of AT_HWCAP. */ |
| 633 | +#define AT_HWCAP4 30 /* extension of AT_HWCAP. */ |
| 634 | + |
| 635 | #define AT_EXECFN 31 /* Filename of executable. */ |
| 636 | |
| 637 | /* Pointer to the global system page used for system calls and other |
| 638 | diff --git a/elf/tst-gnu2-tls2.c b/elf/tst-gnu2-tls2.c |
| 639 | new file mode 100644 |
| 640 | index 0000000000..7ac04d7f33 |
| 641 | --- /dev/null |
| 642 | +++ b/elf/tst-gnu2-tls2.c |
| 643 | @@ -0,0 +1,122 @@ |
| 644 | +/* Test TLSDESC relocation. |
| 645 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 646 | + This file is part of the GNU C Library. |
| 647 | + |
| 648 | + The GNU C Library is free software; you can redistribute it and/or |
| 649 | + modify it under the terms of the GNU Lesser General Public |
| 650 | + License as published by the Free Software Foundation; either |
| 651 | + version 2.1 of the License, or (at your option) any later version. |
| 652 | + |
| 653 | + The GNU C Library is distributed in the hope that it will be useful, |
| 654 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 655 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 656 | + Lesser General Public License for more details. |
| 657 | + |
| 658 | + You should have received a copy of the GNU Lesser General Public |
| 659 | + License along with the GNU C Library; if not, see |
| 660 | + <http://www.gnu.org/licenses/>. */ |
| 661 | + |
| 662 | +#include <stdio.h> |
| 663 | +#include <stdlib.h> |
| 664 | +#include <string.h> |
| 665 | +#include <dlfcn.h> |
| 666 | +#include <pthread.h> |
| 667 | +#include <support/xdlfcn.h> |
| 668 | +#include <support/xthread.h> |
| 669 | +#include <support/check.h> |
| 670 | +#include <support/test-driver.h> |
| 671 | +#include "tst-gnu2-tls2.h" |
| 672 | + |
| 673 | +#ifndef IS_SUPPORTED |
| 674 | +# define IS_SUPPORTED() true |
| 675 | +#endif |
| 676 | + |
| 677 | +/* An architecture can define it to clobber caller-saved registers in |
| 678 | + malloc below to verify that the implicit TLSDESC call won't change |
| 679 | + caller-saved registers. */ |
| 680 | +#ifndef PREPARE_MALLOC |
| 681 | +# define PREPARE_MALLOC() |
| 682 | +#endif |
| 683 | + |
| 684 | +extern void * __libc_malloc (size_t); |
| 685 | + |
| 686 | +size_t malloc_counter = 0; |
| 687 | + |
| 688 | +void * |
| 689 | +malloc (size_t n) |
| 690 | +{ |
| 691 | + PREPARE_MALLOC (); |
| 692 | + malloc_counter++; |
| 693 | + return __libc_malloc (n); |
| 694 | +} |
| 695 | + |
| 696 | +static void *mod[3]; |
| 697 | +#ifndef MOD |
| 698 | +# define MOD(i) "tst-gnu2-tls2mod" #i ".so" |
| 699 | +#endif |
| 700 | +static const char *modname[3] = { MOD(0), MOD(1), MOD(2) }; |
| 701 | +#undef MOD |
| 702 | + |
| 703 | +static void |
| 704 | +open_mod (int i) |
| 705 | +{ |
| 706 | + mod[i] = xdlopen (modname[i], RTLD_LAZY); |
| 707 | + printf ("open %s\n", modname[i]); |
| 708 | +} |
| 709 | + |
| 710 | +static void |
| 711 | +close_mod (int i) |
| 712 | +{ |
| 713 | + xdlclose (mod[i]); |
| 714 | + mod[i] = NULL; |
| 715 | + printf ("close %s\n", modname[i]); |
| 716 | +} |
| 717 | + |
| 718 | +static void |
| 719 | +access_mod (int i, const char *sym) |
| 720 | +{ |
| 721 | + struct tls var = { -1, -1, -1, -1 }; |
| 722 | + struct tls *(*f) (struct tls *) = xdlsym (mod[i], sym); |
| 723 | + /* Check that our malloc is called. */ |
| 724 | + malloc_counter = 0; |
| 725 | + struct tls *p = f (&var); |
| 726 | + TEST_VERIFY (malloc_counter != 0); |
| 727 | + printf ("access %s: %s() = %p\n", modname[i], sym, p); |
| 728 | + TEST_VERIFY_EXIT (memcmp (p, &var, sizeof (var)) == 0); |
| 729 | + ++(p->a); |
| 730 | +} |
| 731 | + |
| 732 | +static void * |
| 733 | +start (void *arg) |
| 734 | +{ |
| 735 | + /* The DTV generation is at the last dlopen of mod0 and the |
| 736 | + entry for mod1 is NULL. */ |
| 737 | + |
| 738 | + open_mod (1); /* Reuse modid of mod1. Uses dynamic TLS. */ |
| 739 | + |
| 740 | + /* Force the slow path in GNU2 TLS descriptor call. */ |
| 741 | + access_mod (1, "apply_tls"); |
| 742 | + |
| 743 | + return arg; |
| 744 | +} |
| 745 | + |
| 746 | +static int |
| 747 | +do_test (void) |
| 748 | +{ |
| 749 | + if (!IS_SUPPORTED ()) |
| 750 | + return EXIT_UNSUPPORTED; |
| 751 | + |
| 752 | + open_mod (0); |
| 753 | + open_mod (1); |
| 754 | + open_mod (2); |
| 755 | + close_mod (0); |
| 756 | + close_mod (1); /* Create modid gap at mod1. */ |
| 757 | + open_mod (0); /* Reuse modid of mod0, bump generation count. */ |
| 758 | + |
| 759 | + /* Create a thread where DTV of mod1 is NULL. */ |
| 760 | + pthread_t t = xpthread_create (NULL, start, NULL); |
| 761 | + xpthread_join (t); |
| 762 | + return 0; |
| 763 | +} |
| 764 | + |
| 765 | +#include <support/test-driver.c> |
| 766 | diff --git a/elf/tst-gnu2-tls2.h b/elf/tst-gnu2-tls2.h |
| 767 | new file mode 100644 |
| 768 | index 0000000000..1ade8151e2 |
| 769 | --- /dev/null |
| 770 | +++ b/elf/tst-gnu2-tls2.h |
| 771 | @@ -0,0 +1,40 @@ |
| 772 | +/* Test TLSDESC relocation. |
| 773 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 774 | + This file is part of the GNU C Library. |
| 775 | + |
| 776 | + The GNU C Library is free software; you can redistribute it and/or |
| 777 | + modify it under the terms of the GNU Lesser General Public |
| 778 | + License as published by the Free Software Foundation; either |
| 779 | + version 2.1 of the License, or (at your option) any later version. |
| 780 | + |
| 781 | + The GNU C Library is distributed in the hope that it will be useful, |
| 782 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 783 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 784 | + Lesser General Public License for more details. |
| 785 | + |
| 786 | + You should have received a copy of the GNU Lesser General Public |
| 787 | + License along with the GNU C Library; if not, see |
| 788 | + <https://www.gnu.org/licenses/>. */ |
| 789 | + |
| 790 | +#include <stdint.h> |
| 791 | + |
| 792 | +struct tls |
| 793 | +{ |
| 794 | + int64_t a, b, c, d; |
| 795 | +}; |
| 796 | + |
| 797 | +extern struct tls *apply_tls (struct tls *); |
| 798 | + |
| 799 | +/* An architecture can define them to verify that clobber caller-saved |
| 800 | + registers aren't changed by the implicit TLSDESC call. */ |
| 801 | +#ifndef INIT_TLSDESC_CALL |
| 802 | +# define INIT_TLSDESC_CALL() |
| 803 | +#endif |
| 804 | + |
| 805 | +#ifndef BEFORE_TLSDESC_CALL |
| 806 | +# define BEFORE_TLSDESC_CALL() |
| 807 | +#endif |
| 808 | + |
| 809 | +#ifndef AFTER_TLSDESC_CALL |
| 810 | +# define AFTER_TLSDESC_CALL() |
| 811 | +#endif |
| 812 | diff --git a/elf/tst-gnu2-tls2mod0.c b/elf/tst-gnu2-tls2mod0.c |
| 813 | new file mode 100644 |
| 814 | index 0000000000..3fe3c14277 |
| 815 | --- /dev/null |
| 816 | +++ b/elf/tst-gnu2-tls2mod0.c |
| 817 | @@ -0,0 +1,32 @@ |
| 818 | +/* DSO used by tst-gnu2-tls2. |
| 819 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 820 | + This file is part of the GNU C Library. |
| 821 | + |
| 822 | + The GNU C Library is free software; you can redistribute it and/or |
| 823 | + modify it under the terms of the GNU Lesser General Public |
| 824 | + License as published by the Free Software Foundation; either |
| 825 | + version 2.1 of the License, or (at your option) any later version. |
| 826 | + |
| 827 | + The GNU C Library is distributed in the hope that it will be useful, |
| 828 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 829 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 830 | + Lesser General Public License for more details. |
| 831 | + |
| 832 | + You should have received a copy of the GNU Lesser General Public |
| 833 | + License along with the GNU C Library; if not, see |
| 834 | + <https://www.gnu.org/licenses/>. */ |
| 835 | + |
| 836 | +#include <tst-gnu2-tls2.h> |
| 837 | + |
| 838 | +__thread struct tls tls_var0 __attribute__ ((visibility ("hidden"))); |
| 839 | + |
| 840 | +struct tls * |
| 841 | +apply_tls (struct tls *p) |
| 842 | +{ |
| 843 | + INIT_TLSDESC_CALL (); |
| 844 | + BEFORE_TLSDESC_CALL (); |
| 845 | + tls_var0 = *p; |
| 846 | + struct tls *ret = &tls_var0; |
| 847 | + AFTER_TLSDESC_CALL (); |
| 848 | + return ret; |
| 849 | +} |
| 850 | diff --git a/elf/tst-gnu2-tls2mod1.c b/elf/tst-gnu2-tls2mod1.c |
| 851 | new file mode 100644 |
| 852 | index 0000000000..e210538468 |
| 853 | --- /dev/null |
| 854 | +++ b/elf/tst-gnu2-tls2mod1.c |
| 855 | @@ -0,0 +1,32 @@ |
| 856 | +/* DSO used by tst-gnu2-tls2. |
| 857 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 858 | + This file is part of the GNU C Library. |
| 859 | + |
| 860 | + The GNU C Library is free software; you can redistribute it and/or |
| 861 | + modify it under the terms of the GNU Lesser General Public |
| 862 | + License as published by the Free Software Foundation; either |
| 863 | + version 2.1 of the License, or (at your option) any later version. |
| 864 | + |
| 865 | + The GNU C Library is distributed in the hope that it will be useful, |
| 866 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 867 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 868 | + Lesser General Public License for more details. |
| 869 | + |
| 870 | + You should have received a copy of the GNU Lesser General Public |
| 871 | + License along with the GNU C Library; if not, see |
| 872 | + <https://www.gnu.org/licenses/>. */ |
| 873 | + |
| 874 | +#include <tst-gnu2-tls2.h> |
| 875 | + |
| 876 | +__thread struct tls tls_var1[100] __attribute__ ((visibility ("hidden"))); |
| 877 | + |
| 878 | +struct tls * |
| 879 | +apply_tls (struct tls *p) |
| 880 | +{ |
| 881 | + INIT_TLSDESC_CALL (); |
| 882 | + BEFORE_TLSDESC_CALL (); |
| 883 | + tls_var1[1] = *p; |
| 884 | + struct tls *ret = &tls_var1[1]; |
| 885 | + AFTER_TLSDESC_CALL (); |
| 886 | + return ret; |
| 887 | +} |
| 888 | diff --git a/elf/tst-gnu2-tls2mod2.c b/elf/tst-gnu2-tls2mod2.c |
| 889 | new file mode 100644 |
| 890 | index 0000000000..6d3031dc5f |
| 891 | --- /dev/null |
| 892 | +++ b/elf/tst-gnu2-tls2mod2.c |
| 893 | @@ -0,0 +1,32 @@ |
| 894 | +/* DSO used by tst-gnu2-tls2. |
| 895 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 896 | + This file is part of the GNU C Library. |
| 897 | + |
| 898 | + The GNU C Library is free software; you can redistribute it and/or |
| 899 | + modify it under the terms of the GNU Lesser General Public |
| 900 | + License as published by the Free Software Foundation; either |
| 901 | + version 2.1 of the License, or (at your option) any later version. |
| 902 | + |
| 903 | + The GNU C Library is distributed in the hope that it will be useful, |
| 904 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 905 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 906 | + Lesser General Public License for more details. |
| 907 | + |
| 908 | + You should have received a copy of the GNU Lesser General Public |
| 909 | + License along with the GNU C Library; if not, see |
| 910 | + <https://www.gnu.org/licenses/>. */ |
| 911 | + |
| 912 | +#include <tst-gnu2-tls2.h> |
| 913 | + |
| 914 | +__thread struct tls tls_var2 __attribute__ ((visibility ("hidden"))); |
| 915 | + |
| 916 | +struct tls * |
| 917 | +apply_tls (struct tls *p) |
| 918 | +{ |
| 919 | + INIT_TLSDESC_CALL (); |
| 920 | + BEFORE_TLSDESC_CALL (); |
| 921 | + tls_var2 = *p; |
| 922 | + struct tls *ret = &tls_var2; |
| 923 | + AFTER_TLSDESC_CALL (); |
| 924 | + return ret; |
| 925 | +} |
| 926 | diff --git a/elf/tst-tunables.c b/elf/tst-tunables.c |
| 927 | index 095b5c81d9..dff34ed748 100644 |
| 928 | --- a/elf/tst-tunables.c |
| 929 | +++ b/elf/tst-tunables.c |
| 930 | @@ -17,6 +17,10 @@ |
| 931 | <https://www.gnu.org/licenses/>. */ |
| 932 | |
| 933 | #include <array_length.h> |
| 934 | +/* The test uses the tunable_list size, which is only exported for |
| 935 | + ld.so. This will result in a copy of tunable_list, which is ununsed by |
| 936 | + the test itself. */ |
| 937 | +#define TUNABLES_INTERNAL 1 |
| 938 | #include <dl-tunables.h> |
| 939 | #include <getopt.h> |
| 940 | #include <intprops.h> |
| 941 | @@ -24,12 +28,13 @@ |
| 942 | #include <stdlib.h> |
| 943 | #include <support/capture_subprocess.h> |
| 944 | #include <support/check.h> |
| 945 | +#include <support/support.h> |
| 946 | |
| 947 | static int restart; |
| 948 | #define CMDLINE_OPTIONS \ |
| 949 | { "restart", no_argument, &restart, 1 }, |
| 950 | |
| 951 | -static const struct test_t |
| 952 | +static struct test_t |
| 953 | { |
| 954 | const char *name; |
| 955 | const char *value; |
| 956 | @@ -284,6 +289,29 @@ static const struct test_t |
| 957 | 0, |
| 958 | 0, |
| 959 | }, |
| 960 | + /* Also check for repeated tunables with a count larger than the total number |
| 961 | + of tunables. */ |
| 962 | + { |
| 963 | + "GLIBC_TUNABLES", |
| 964 | + NULL, |
| 965 | + 2, |
| 966 | + 0, |
| 967 | + 0, |
| 968 | + }, |
| 969 | + { |
| 970 | + "GLIBC_TUNABLES", |
| 971 | + NULL, |
| 972 | + 1, |
| 973 | + 0, |
| 974 | + 0, |
| 975 | + }, |
| 976 | + { |
| 977 | + "GLIBC_TUNABLES", |
| 978 | + NULL, |
| 979 | + 0, |
| 980 | + 0, |
| 981 | + 0, |
| 982 | + }, |
| 983 | }; |
| 984 | |
| 985 | static int |
| 986 | @@ -327,6 +355,37 @@ do_test (int argc, char *argv[]) |
| 987 | spargv[i] = NULL; |
| 988 | } |
| 989 | |
| 990 | + /* Create a tunable line with the duplicate values with a total number |
| 991 | + larger than the different number of tunables. */ |
| 992 | + { |
| 993 | + enum { tunables_list_size = array_length (tunable_list) }; |
| 994 | + const char *value = ""; |
| 995 | + for (int i = 0; i < tunables_list_size; i++) |
| 996 | + value = xasprintf ("%sglibc.malloc.check=2%c", |
| 997 | + value, |
| 998 | + i == (tunables_list_size - 1) ? '\0' : ':'); |
| 999 | + tests[33].value = value; |
| 1000 | + } |
| 1001 | + /* Same as before, but the last tunable values is differen than the |
| 1002 | + rest. */ |
| 1003 | + { |
| 1004 | + enum { tunables_list_size = array_length (tunable_list) }; |
| 1005 | + const char *value = ""; |
| 1006 | + for (int i = 0; i < tunables_list_size - 1; i++) |
| 1007 | + value = xasprintf ("%sglibc.malloc.check=2:", value); |
| 1008 | + value = xasprintf ("%sglibc.malloc.check=1", value); |
| 1009 | + tests[34].value = value; |
| 1010 | + } |
| 1011 | + /* Same as before, but with an invalid last entry. */ |
| 1012 | + { |
| 1013 | + enum { tunables_list_size = array_length (tunable_list) }; |
| 1014 | + const char *value = ""; |
| 1015 | + for (int i = 0; i < tunables_list_size - 1; i++) |
| 1016 | + value = xasprintf ("%sglibc.malloc.check=2:", value); |
| 1017 | + value = xasprintf ("%sglibc.malloc.check=1=1", value); |
| 1018 | + tests[35].value = value; |
| 1019 | + } |
| 1020 | + |
| 1021 | for (int i = 0; i < array_length (tests); i++) |
| 1022 | { |
| 1023 | snprintf (nteststr, sizeof nteststr, "%d", i); |
| 1024 | diff --git a/iconvdata/Makefile b/iconvdata/Makefile |
| 1025 | index ea019ce5c0..7196a8744b 100644 |
| 1026 | --- a/iconvdata/Makefile |
| 1027 | +++ b/iconvdata/Makefile |
| 1028 | @@ -75,7 +75,8 @@ ifeq (yes,$(build-shared)) |
| 1029 | tests = bug-iconv1 bug-iconv2 tst-loading tst-e2big tst-iconv4 bug-iconv4 \ |
| 1030 | tst-iconv6 bug-iconv5 bug-iconv6 tst-iconv7 bug-iconv8 bug-iconv9 \ |
| 1031 | bug-iconv10 bug-iconv11 bug-iconv12 tst-iconv-big5-hkscs-to-2ucs4 \ |
| 1032 | - bug-iconv13 bug-iconv14 bug-iconv15 |
| 1033 | + bug-iconv13 bug-iconv14 bug-iconv15 \ |
| 1034 | + tst-iconv-iso-2022-cn-ext |
| 1035 | ifeq ($(have-thread-library),yes) |
| 1036 | tests += bug-iconv3 |
| 1037 | endif |
| 1038 | @@ -330,6 +331,8 @@ $(objpfx)bug-iconv14.out: $(addprefix $(objpfx), $(gconv-modules)) \ |
| 1039 | $(addprefix $(objpfx),$(modules.so)) |
| 1040 | $(objpfx)bug-iconv15.out: $(addprefix $(objpfx), $(gconv-modules)) \ |
| 1041 | $(addprefix $(objpfx),$(modules.so)) |
| 1042 | +$(objpfx)tst-iconv-iso-2022-cn-ext.out: $(addprefix $(objpfx), $(gconv-modules)) \ |
| 1043 | + $(addprefix $(objpfx),$(modules.so)) |
| 1044 | |
| 1045 | $(objpfx)iconv-test.out: run-iconv-test.sh \ |
| 1046 | $(addprefix $(objpfx), $(gconv-modules)) \ |
| 1047 | diff --git a/iconvdata/iso-2022-cn-ext.c b/iconvdata/iso-2022-cn-ext.c |
| 1048 | index b34c8a36f4..cce29b1969 100644 |
| 1049 | --- a/iconvdata/iso-2022-cn-ext.c |
| 1050 | +++ b/iconvdata/iso-2022-cn-ext.c |
| 1051 | @@ -574,6 +574,12 @@ DIAG_IGNORE_Os_NEEDS_COMMENT (5, "-Wmaybe-uninitialized"); |
| 1052 | { \ |
| 1053 | const char *escseq; \ |
| 1054 | \ |
| 1055 | + if (outptr + 4 > outend) \ |
| 1056 | + { \ |
| 1057 | + result = __GCONV_FULL_OUTPUT; \ |
| 1058 | + break; \ |
| 1059 | + } \ |
| 1060 | + \ |
| 1061 | assert (used == CNS11643_2_set); /* XXX */ \ |
| 1062 | escseq = "*H"; \ |
| 1063 | *outptr++ = ESC; \ |
| 1064 | @@ -587,6 +593,12 @@ DIAG_IGNORE_Os_NEEDS_COMMENT (5, "-Wmaybe-uninitialized"); |
| 1065 | { \ |
| 1066 | const char *escseq; \ |
| 1067 | \ |
| 1068 | + if (outptr + 4 > outend) \ |
| 1069 | + { \ |
| 1070 | + result = __GCONV_FULL_OUTPUT; \ |
| 1071 | + break; \ |
| 1072 | + } \ |
| 1073 | + \ |
| 1074 | assert ((used >> 5) >= 3 && (used >> 5) <= 7); \ |
| 1075 | escseq = "+I+J+K+L+M" + ((used >> 5) - 3) * 2; \ |
| 1076 | *outptr++ = ESC; \ |
| 1077 | diff --git a/iconvdata/tst-iconv-iso-2022-cn-ext.c b/iconvdata/tst-iconv-iso-2022-cn-ext.c |
| 1078 | new file mode 100644 |
| 1079 | index 0000000000..96a8765fd5 |
| 1080 | --- /dev/null |
| 1081 | +++ b/iconvdata/tst-iconv-iso-2022-cn-ext.c |
| 1082 | @@ -0,0 +1,128 @@ |
| 1083 | +/* Verify ISO-2022-CN-EXT does not write out of the bounds. |
| 1084 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 1085 | + This file is part of the GNU C Library. |
| 1086 | + |
| 1087 | + The GNU C Library is free software; you can redistribute it and/or |
| 1088 | + modify it under the terms of the GNU Lesser General Public |
| 1089 | + License as published by the Free Software Foundation; either |
| 1090 | + version 2.1 of the License, or (at your option) any later version. |
| 1091 | + |
| 1092 | + The GNU C Library is distributed in the hope that it will be useful, |
| 1093 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 1094 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 1095 | + Lesser General Public License for more details. |
| 1096 | + |
| 1097 | + You should have received a copy of the GNU Lesser General Public |
| 1098 | + License along with the GNU C Library; if not, see |
| 1099 | + <https://www.gnu.org/licenses/>. */ |
| 1100 | + |
| 1101 | +#include <stdio.h> |
| 1102 | +#include <string.h> |
| 1103 | + |
| 1104 | +#include <errno.h> |
| 1105 | +#include <iconv.h> |
| 1106 | +#include <sys/mman.h> |
| 1107 | + |
| 1108 | +#include <support/xunistd.h> |
| 1109 | +#include <support/check.h> |
| 1110 | +#include <support/support.h> |
| 1111 | + |
| 1112 | +/* The test sets up a two memory page buffer with the second page marked |
| 1113 | + PROT_NONE to trigger a fault if the conversion writes beyond the exact |
| 1114 | + expected amount. Then we carry out various conversions and precisely |
| 1115 | + place the start of the output buffer in order to trigger a SIGSEGV if the |
| 1116 | + process writes anywhere between 1 and page sized bytes more (only one |
| 1117 | + PROT_NONE page is setup as a canary) than expected. These tests exercise |
| 1118 | + all three of the cases in ISO-2022-CN-EXT where the converter must switch |
| 1119 | + character sets and may run out of buffer space while doing the |
| 1120 | + operation. */ |
| 1121 | + |
| 1122 | +static int |
| 1123 | +do_test (void) |
| 1124 | +{ |
| 1125 | + iconv_t cd = iconv_open ("ISO-2022-CN-EXT", "UTF-8"); |
| 1126 | + TEST_VERIFY_EXIT (cd != (iconv_t) -1); |
| 1127 | + |
| 1128 | + char *ntf; |
| 1129 | + size_t ntfsize; |
| 1130 | + char *outbufbase; |
| 1131 | + { |
| 1132 | + int pgz = getpagesize (); |
| 1133 | + TEST_VERIFY_EXIT (pgz > 0); |
| 1134 | + ntfsize = 2 * pgz; |
| 1135 | + |
| 1136 | + ntf = xmmap (NULL, ntfsize, PROT_READ | PROT_WRITE, MAP_PRIVATE |
| 1137 | + | MAP_ANONYMOUS, -1); |
| 1138 | + xmprotect (ntf + pgz, pgz, PROT_NONE); |
| 1139 | + |
| 1140 | + outbufbase = ntf + pgz; |
| 1141 | + } |
| 1142 | + |
| 1143 | + /* Check if SOdesignation escape sequence does not trigger an OOB write. */ |
| 1144 | + { |
| 1145 | + char inbuf[] = "\xe4\xba\xa4\xe6\x8d\xa2"; |
| 1146 | + |
| 1147 | + for (int i = 0; i < 9; i++) |
| 1148 | + { |
| 1149 | + char *inp = inbuf; |
| 1150 | + size_t inleft = sizeof (inbuf) - 1; |
| 1151 | + |
| 1152 | + char *outp = outbufbase - i; |
| 1153 | + size_t outleft = i; |
| 1154 | + |
| 1155 | + TEST_VERIFY_EXIT (iconv (cd, &inp, &inleft, &outp, &outleft) |
| 1156 | + == (size_t) -1); |
| 1157 | + TEST_COMPARE (errno, E2BIG); |
| 1158 | + |
| 1159 | + TEST_VERIFY_EXIT (iconv (cd, NULL, NULL, NULL, NULL) == 0); |
| 1160 | + } |
| 1161 | + } |
| 1162 | + |
| 1163 | + /* Same as before for SS2designation. */ |
| 1164 | + { |
| 1165 | + char inbuf[] = "㴽 \xe3\xb4\xbd"; |
| 1166 | + |
| 1167 | + for (int i = 0; i < 14; i++) |
| 1168 | + { |
| 1169 | + char *inp = inbuf; |
| 1170 | + size_t inleft = sizeof (inbuf) - 1; |
| 1171 | + |
| 1172 | + char *outp = outbufbase - i; |
| 1173 | + size_t outleft = i; |
| 1174 | + |
| 1175 | + TEST_VERIFY_EXIT (iconv (cd, &inp, &inleft, &outp, &outleft) |
| 1176 | + == (size_t) -1); |
| 1177 | + TEST_COMPARE (errno, E2BIG); |
| 1178 | + |
| 1179 | + TEST_VERIFY_EXIT (iconv (cd, NULL, NULL, NULL, NULL) == 0); |
| 1180 | + } |
| 1181 | + } |
| 1182 | + |
| 1183 | + /* Same as before for SS3designation. */ |
| 1184 | + { |
| 1185 | + char inbuf[] = "劄 \xe5\x8a\x84"; |
| 1186 | + |
| 1187 | + for (int i = 0; i < 14; i++) |
| 1188 | + { |
| 1189 | + char *inp = inbuf; |
| 1190 | + size_t inleft = sizeof (inbuf) - 1; |
| 1191 | + |
| 1192 | + char *outp = outbufbase - i; |
| 1193 | + size_t outleft = i; |
| 1194 | + |
| 1195 | + TEST_VERIFY_EXIT (iconv (cd, &inp, &inleft, &outp, &outleft) |
| 1196 | + == (size_t) -1); |
| 1197 | + TEST_COMPARE (errno, E2BIG); |
| 1198 | + |
| 1199 | + TEST_VERIFY_EXIT (iconv (cd, NULL, NULL, NULL, NULL) == 0); |
| 1200 | + } |
| 1201 | + } |
| 1202 | + |
| 1203 | + TEST_VERIFY_EXIT (iconv_close (cd) != -1); |
| 1204 | + |
| 1205 | + xmunmap (ntf, ntfsize); |
| 1206 | + |
| 1207 | + return 0; |
| 1208 | +} |
| 1209 | + |
| 1210 | +#include <support/test-driver.c> |
| 1211 | diff --git a/login/Makefile b/login/Makefile |
| 1212 | index 1e22008a61..f91190e3dc 100644 |
| 1213 | --- a/login/Makefile |
| 1214 | +++ b/login/Makefile |
| 1215 | @@ -44,7 +44,9 @@ subdir-dirs = programs |
| 1216 | vpath %.c programs |
| 1217 | |
| 1218 | tests := tst-utmp tst-utmpx tst-grantpt tst-ptsname tst-getlogin tst-updwtmpx \ |
| 1219 | - tst-pututxline-lockfail tst-pututxline-cache |
| 1220 | + tst-pututxline-lockfail tst-pututxline-cache tst-utmp-size tst-utmp-size-64 |
| 1221 | + |
| 1222 | +CFLAGS-tst-utmp-size-64.c += -D_FILE_OFFSET_BITS=64 -D_TIME_BITS=64 |
| 1223 | |
| 1224 | # Empty compatibility library for old binaries. |
| 1225 | extra-libs := libutil |
| 1226 | diff --git a/login/tst-utmp-size-64.c b/login/tst-utmp-size-64.c |
| 1227 | new file mode 100644 |
| 1228 | index 0000000000..7a581a4c12 |
| 1229 | --- /dev/null |
| 1230 | +++ b/login/tst-utmp-size-64.c |
| 1231 | @@ -0,0 +1,2 @@ |
| 1232 | +/* The on-disk layout must not change in time64 mode. */ |
| 1233 | +#include "tst-utmp-size.c" |
| 1234 | diff --git a/login/tst-utmp-size.c b/login/tst-utmp-size.c |
| 1235 | new file mode 100644 |
| 1236 | index 0000000000..1b7f7ff042 |
| 1237 | --- /dev/null |
| 1238 | +++ b/login/tst-utmp-size.c |
| 1239 | @@ -0,0 +1,33 @@ |
| 1240 | +/* Check expected sizes of struct utmp, struct utmpx, struct lastlog. |
| 1241 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 1242 | + This file is part of the GNU C Library. |
| 1243 | + |
| 1244 | + The GNU C Library is free software; you can redistribute it and/or |
| 1245 | + modify it under the terms of the GNU Lesser General Public |
| 1246 | + License as published by the Free Software Foundation; either |
| 1247 | + version 2.1 of the License, or (at your option) any later version. |
| 1248 | + |
| 1249 | + The GNU C Library is distributed in the hope that it will be useful, |
| 1250 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 1251 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 1252 | + Lesser General Public License for more details. |
| 1253 | + |
| 1254 | + You should have received a copy of the GNU Lesser General Public |
| 1255 | + License along with the GNU C Library; if not, see |
| 1256 | + <https://www.gnu.org/licenses/>. */ |
| 1257 | + |
| 1258 | +#include <utmp.h> |
| 1259 | +#include <utmpx.h> |
| 1260 | +#include <utmp-size.h> |
| 1261 | + |
| 1262 | +static int |
| 1263 | +do_test (void) |
| 1264 | +{ |
| 1265 | + _Static_assert (sizeof (struct utmp) == UTMP_SIZE, "struct utmp size"); |
| 1266 | + _Static_assert (sizeof (struct utmpx) == UTMP_SIZE, "struct utmpx size"); |
| 1267 | + _Static_assert (sizeof (struct lastlog) == LASTLOG_SIZE, |
| 1268 | + "struct lastlog size"); |
| 1269 | + return 0; |
| 1270 | +} |
| 1271 | + |
| 1272 | +#include <support/test-driver.c> |
| 1273 | diff --git a/manual/stdbit.texi b/manual/stdbit.texi |
| 1274 | index fe41c671d8..6c75ed9a20 100644 |
| 1275 | --- a/manual/stdbit.texi |
| 1276 | +++ b/manual/stdbit.texi |
| 1277 | @@ -32,7 +32,13 @@ and @code{unsigned long long int}. In addition, there is a |
| 1278 | corresponding type-generic macro (not listed below), named the same as |
| 1279 | the functions but without any suffix such as @samp{_uc}. The |
| 1280 | type-generic macro can only be used with an argument of an unsigned |
| 1281 | -integer type with a width of 8, 16, 32 or 64 bits. |
| 1282 | +integer type with a width of 8, 16, 32 or 64 bits, or when using |
| 1283 | +a compiler with support for |
| 1284 | +@uref{https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html,@code{__builtin_stdc_bit_ceil}}, |
| 1285 | +etc.@:, built-in functions such as GCC 14.1 or later |
| 1286 | +any unsigned integer type those built-in functions support. |
| 1287 | +In GCC 14.1 that includes support for @code{unsigned __int128} and |
| 1288 | +@code{unsigned _BitInt(@var{n})} if supported by the target. |
| 1289 | |
| 1290 | @deftypefun {unsigned int} stdc_leading_zeros_uc (unsigned char @var{x}) |
| 1291 | @deftypefunx {unsigned int} stdc_leading_zeros_us (unsigned short @var{x}) |
| 1292 | diff --git a/nscd/netgroupcache.c b/nscd/netgroupcache.c |
| 1293 | index 0c6e46f15c..01d554af9c 100644 |
| 1294 | --- a/nscd/netgroupcache.c |
| 1295 | +++ b/nscd/netgroupcache.c |
| 1296 | @@ -23,6 +23,7 @@ |
| 1297 | #include <stdlib.h> |
| 1298 | #include <unistd.h> |
| 1299 | #include <sys/mman.h> |
| 1300 | +#include <scratch_buffer.h> |
| 1301 | |
| 1302 | #include "../nss/netgroup.h" |
| 1303 | #include "nscd.h" |
| 1304 | @@ -65,6 +66,16 @@ struct dataset |
| 1305 | char strdata[0]; |
| 1306 | }; |
| 1307 | |
| 1308 | +/* Send a notfound response to FD. Always returns -1 to indicate an |
| 1309 | + ephemeral error. */ |
| 1310 | +static time_t |
| 1311 | +send_notfound (int fd) |
| 1312 | +{ |
| 1313 | + if (fd != -1) |
| 1314 | + TEMP_FAILURE_RETRY (send (fd, ¬found, sizeof (notfound), MSG_NOSIGNAL)); |
| 1315 | + return -1; |
| 1316 | +} |
| 1317 | + |
| 1318 | /* Sends a notfound message and prepares a notfound dataset to write to the |
| 1319 | cache. Returns true if there was enough memory to allocate the dataset and |
| 1320 | returns the dataset in DATASETP, total bytes to write in TOTALP and the |
| 1321 | @@ -83,8 +94,7 @@ do_notfound (struct database_dyn *db, int fd, request_header *req, |
| 1322 | total = sizeof (notfound); |
| 1323 | timeout = time (NULL) + db->negtimeout; |
| 1324 | |
| 1325 | - if (fd != -1) |
| 1326 | - TEMP_FAILURE_RETRY (send (fd, ¬found, total, MSG_NOSIGNAL)); |
| 1327 | + send_notfound (fd); |
| 1328 | |
| 1329 | dataset = mempool_alloc (db, sizeof (struct dataset) + req->key_len, 1); |
| 1330 | /* If we cannot permanently store the result, so be it. */ |
| 1331 | @@ -109,11 +119,78 @@ do_notfound (struct database_dyn *db, int fd, request_header *req, |
| 1332 | return cacheable; |
| 1333 | } |
| 1334 | |
| 1335 | +struct addgetnetgrentX_scratch |
| 1336 | +{ |
| 1337 | + /* This is the result that the caller should use. It can be NULL, |
| 1338 | + point into buffer, or it can be in the cache. */ |
| 1339 | + struct dataset *dataset; |
| 1340 | + |
| 1341 | + struct scratch_buffer buffer; |
| 1342 | + |
| 1343 | + /* Used internally in addgetnetgrentX as a staging area. */ |
| 1344 | + struct scratch_buffer tmp; |
| 1345 | + |
| 1346 | + /* Number of bytes in buffer that are actually used. */ |
| 1347 | + size_t buffer_used; |
| 1348 | +}; |
| 1349 | + |
| 1350 | +static void |
| 1351 | +addgetnetgrentX_scratch_init (struct addgetnetgrentX_scratch *scratch) |
| 1352 | +{ |
| 1353 | + scratch->dataset = NULL; |
| 1354 | + scratch_buffer_init (&scratch->buffer); |
| 1355 | + scratch_buffer_init (&scratch->tmp); |
| 1356 | + |
| 1357 | + /* Reserve space for the header. */ |
| 1358 | + scratch->buffer_used = sizeof (struct dataset); |
| 1359 | + static_assert (sizeof (struct dataset) < sizeof (scratch->tmp.__space), |
| 1360 | + "initial buffer space"); |
| 1361 | + memset (scratch->tmp.data, 0, sizeof (struct dataset)); |
| 1362 | +} |
| 1363 | + |
| 1364 | +static void |
| 1365 | +addgetnetgrentX_scratch_free (struct addgetnetgrentX_scratch *scratch) |
| 1366 | +{ |
| 1367 | + scratch_buffer_free (&scratch->buffer); |
| 1368 | + scratch_buffer_free (&scratch->tmp); |
| 1369 | +} |
| 1370 | + |
| 1371 | +/* Copy LENGTH bytes from S into SCRATCH. Returns NULL if SCRATCH |
| 1372 | + could not be resized, otherwise a pointer to the copy. */ |
| 1373 | +static char * |
| 1374 | +addgetnetgrentX_append_n (struct addgetnetgrentX_scratch *scratch, |
| 1375 | + const char *s, size_t length) |
| 1376 | +{ |
| 1377 | + while (true) |
| 1378 | + { |
| 1379 | + size_t remaining = scratch->buffer.length - scratch->buffer_used; |
| 1380 | + if (remaining >= length) |
| 1381 | + break; |
| 1382 | + if (!scratch_buffer_grow_preserve (&scratch->buffer)) |
| 1383 | + return NULL; |
| 1384 | + } |
| 1385 | + char *copy = scratch->buffer.data + scratch->buffer_used; |
| 1386 | + memcpy (copy, s, length); |
| 1387 | + scratch->buffer_used += length; |
| 1388 | + return copy; |
| 1389 | +} |
| 1390 | + |
| 1391 | +/* Copy S into SCRATCH, including its null terminator. Returns false |
| 1392 | + if SCRATCH could not be resized. */ |
| 1393 | +static bool |
| 1394 | +addgetnetgrentX_append (struct addgetnetgrentX_scratch *scratch, const char *s) |
| 1395 | +{ |
| 1396 | + if (s == NULL) |
| 1397 | + s = ""; |
| 1398 | + return addgetnetgrentX_append_n (scratch, s, strlen (s) + 1) != NULL; |
| 1399 | +} |
| 1400 | + |
| 1401 | +/* Caller must initialize and free *SCRATCH. If the return value is |
| 1402 | + negative, this function has sent a notfound response. */ |
| 1403 | static time_t |
| 1404 | addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1405 | const char *key, uid_t uid, struct hashentry *he, |
| 1406 | - struct datahead *dh, struct dataset **resultp, |
| 1407 | - void **tofreep) |
| 1408 | + struct datahead *dh, struct addgetnetgrentX_scratch *scratch) |
| 1409 | { |
| 1410 | if (__glibc_unlikely (debug_level > 0)) |
| 1411 | { |
| 1412 | @@ -132,14 +209,10 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1413 | |
| 1414 | char *key_copy = NULL; |
| 1415 | struct __netgrent data; |
| 1416 | - size_t buflen = MAX (1024, sizeof (*dataset) + req->key_len); |
| 1417 | - size_t buffilled = sizeof (*dataset); |
| 1418 | - char *buffer = NULL; |
| 1419 | size_t nentries = 0; |
| 1420 | size_t group_len = strlen (key) + 1; |
| 1421 | struct name_list *first_needed |
| 1422 | = alloca (sizeof (struct name_list) + group_len); |
| 1423 | - *tofreep = NULL; |
| 1424 | |
| 1425 | if (netgroup_database == NULL |
| 1426 | && !__nss_database_get (nss_database_netgroup, &netgroup_database)) |
| 1427 | @@ -147,12 +220,10 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1428 | /* No such service. */ |
| 1429 | cacheable = do_notfound (db, fd, req, key, &dataset, &total, &timeout, |
| 1430 | &key_copy); |
| 1431 | - goto writeout; |
| 1432 | + goto maybe_cache_add; |
| 1433 | } |
| 1434 | |
| 1435 | memset (&data, '\0', sizeof (data)); |
| 1436 | - buffer = xmalloc (buflen); |
| 1437 | - *tofreep = buffer; |
| 1438 | first_needed->next = first_needed; |
| 1439 | memcpy (first_needed->name, key, group_len); |
| 1440 | data.needed_groups = first_needed; |
| 1441 | @@ -195,8 +266,8 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1442 | while (1) |
| 1443 | { |
| 1444 | int e; |
| 1445 | - status = getfct.f (&data, buffer + buffilled, |
| 1446 | - buflen - buffilled - req->key_len, &e); |
| 1447 | + status = getfct.f (&data, scratch->tmp.data, |
| 1448 | + scratch->tmp.length, &e); |
| 1449 | if (status == NSS_STATUS_SUCCESS) |
| 1450 | { |
| 1451 | if (data.type == triple_val) |
| 1452 | @@ -204,68 +275,10 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1453 | const char *nhost = data.val.triple.host; |
| 1454 | const char *nuser = data.val.triple.user; |
| 1455 | const char *ndomain = data.val.triple.domain; |
| 1456 | - |
| 1457 | - size_t hostlen = strlen (nhost ?: "") + 1; |
| 1458 | - size_t userlen = strlen (nuser ?: "") + 1; |
| 1459 | - size_t domainlen = strlen (ndomain ?: "") + 1; |
| 1460 | - |
| 1461 | - if (nhost == NULL || nuser == NULL || ndomain == NULL |
| 1462 | - || nhost > nuser || nuser > ndomain) |
| 1463 | - { |
| 1464 | - const char *last = nhost; |
| 1465 | - if (last == NULL |
| 1466 | - || (nuser != NULL && nuser > last)) |
| 1467 | - last = nuser; |
| 1468 | - if (last == NULL |
| 1469 | - || (ndomain != NULL && ndomain > last)) |
| 1470 | - last = ndomain; |
| 1471 | - |
| 1472 | - size_t bufused |
| 1473 | - = (last == NULL |
| 1474 | - ? buffilled |
| 1475 | - : last + strlen (last) + 1 - buffer); |
| 1476 | - |
| 1477 | - /* We have to make temporary copies. */ |
| 1478 | - size_t needed = hostlen + userlen + domainlen; |
| 1479 | - |
| 1480 | - if (buflen - req->key_len - bufused < needed) |
| 1481 | - { |
| 1482 | - buflen += MAX (buflen, 2 * needed); |
| 1483 | - /* Save offset in the old buffer. We don't |
| 1484 | - bother with the NULL check here since |
| 1485 | - we'll do that later anyway. */ |
| 1486 | - size_t nhostdiff = nhost - buffer; |
| 1487 | - size_t nuserdiff = nuser - buffer; |
| 1488 | - size_t ndomaindiff = ndomain - buffer; |
| 1489 | - |
| 1490 | - char *newbuf = xrealloc (buffer, buflen); |
| 1491 | - /* Fix up the triplet pointers into the new |
| 1492 | - buffer. */ |
| 1493 | - nhost = (nhost ? newbuf + nhostdiff |
| 1494 | - : NULL); |
| 1495 | - nuser = (nuser ? newbuf + nuserdiff |
| 1496 | - : NULL); |
| 1497 | - ndomain = (ndomain ? newbuf + ndomaindiff |
| 1498 | - : NULL); |
| 1499 | - *tofreep = buffer = newbuf; |
| 1500 | - } |
| 1501 | - |
| 1502 | - nhost = memcpy (buffer + bufused, |
| 1503 | - nhost ?: "", hostlen); |
| 1504 | - nuser = memcpy ((char *) nhost + hostlen, |
| 1505 | - nuser ?: "", userlen); |
| 1506 | - ndomain = memcpy ((char *) nuser + userlen, |
| 1507 | - ndomain ?: "", domainlen); |
| 1508 | - } |
| 1509 | - |
| 1510 | - char *wp = buffer + buffilled; |
| 1511 | - wp = memmove (wp, nhost ?: "", hostlen); |
| 1512 | - wp += hostlen; |
| 1513 | - wp = memmove (wp, nuser ?: "", userlen); |
| 1514 | - wp += userlen; |
| 1515 | - wp = memmove (wp, ndomain ?: "", domainlen); |
| 1516 | - wp += domainlen; |
| 1517 | - buffilled = wp - buffer; |
| 1518 | + if (!(addgetnetgrentX_append (scratch, nhost) |
| 1519 | + && addgetnetgrentX_append (scratch, nuser) |
| 1520 | + && addgetnetgrentX_append (scratch, ndomain))) |
| 1521 | + return send_notfound (fd); |
| 1522 | ++nentries; |
| 1523 | } |
| 1524 | else |
| 1525 | @@ -317,8 +330,8 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1526 | } |
| 1527 | else if (status == NSS_STATUS_TRYAGAIN && e == ERANGE) |
| 1528 | { |
| 1529 | - buflen *= 2; |
| 1530 | - *tofreep = buffer = xrealloc (buffer, buflen); |
| 1531 | + if (!scratch_buffer_grow (&scratch->tmp)) |
| 1532 | + return send_notfound (fd); |
| 1533 | } |
| 1534 | else if (status == NSS_STATUS_RETURN |
| 1535 | || status == NSS_STATUS_NOTFOUND |
| 1536 | @@ -348,13 +361,20 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1537 | { |
| 1538 | cacheable = do_notfound (db, fd, req, key, &dataset, &total, &timeout, |
| 1539 | &key_copy); |
| 1540 | - goto writeout; |
| 1541 | + goto maybe_cache_add; |
| 1542 | } |
| 1543 | |
| 1544 | - total = buffilled; |
| 1545 | + /* Capture the result size without the key appended. */ |
| 1546 | + total = scratch->buffer_used; |
| 1547 | + |
| 1548 | + /* Make a copy of the key. The scratch buffer must not move after |
| 1549 | + this point. */ |
| 1550 | + key_copy = addgetnetgrentX_append_n (scratch, key, req->key_len); |
| 1551 | + if (key_copy == NULL) |
| 1552 | + return send_notfound (fd); |
| 1553 | |
| 1554 | /* Fill in the dataset. */ |
| 1555 | - dataset = (struct dataset *) buffer; |
| 1556 | + dataset = scratch->buffer.data; |
| 1557 | timeout = datahead_init_pos (&dataset->head, total + req->key_len, |
| 1558 | total - offsetof (struct dataset, resp), |
| 1559 | he == NULL ? 0 : dh->nreloads + 1, |
| 1560 | @@ -363,11 +383,7 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1561 | dataset->resp.version = NSCD_VERSION; |
| 1562 | dataset->resp.found = 1; |
| 1563 | dataset->resp.nresults = nentries; |
| 1564 | - dataset->resp.result_len = buffilled - sizeof (*dataset); |
| 1565 | - |
| 1566 | - assert (buflen - buffilled >= req->key_len); |
| 1567 | - key_copy = memcpy (buffer + buffilled, key, req->key_len); |
| 1568 | - buffilled += req->key_len; |
| 1569 | + dataset->resp.result_len = total - sizeof (*dataset); |
| 1570 | |
| 1571 | /* Now we can determine whether on refill we have to create a new |
| 1572 | record or not. */ |
| 1573 | @@ -398,7 +414,7 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1574 | if (__glibc_likely (newp != NULL)) |
| 1575 | { |
| 1576 | /* Adjust pointer into the memory block. */ |
| 1577 | - key_copy = (char *) newp + (key_copy - buffer); |
| 1578 | + key_copy = (char *) newp + (key_copy - (char *) dataset); |
| 1579 | |
| 1580 | dataset = memcpy (newp, dataset, total + req->key_len); |
| 1581 | cacheable = true; |
| 1582 | @@ -410,14 +426,12 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1583 | } |
| 1584 | |
| 1585 | if (he == NULL && fd != -1) |
| 1586 | - { |
| 1587 | - /* We write the dataset before inserting it to the database |
| 1588 | - since while inserting this thread might block and so would |
| 1589 | - unnecessarily let the receiver wait. */ |
| 1590 | - writeout: |
| 1591 | + /* We write the dataset before inserting it to the database since |
| 1592 | + while inserting this thread might block and so would |
| 1593 | + unnecessarily let the receiver wait. */ |
| 1594 | writeall (fd, &dataset->resp, dataset->head.recsize); |
| 1595 | - } |
| 1596 | |
| 1597 | + maybe_cache_add: |
| 1598 | if (cacheable) |
| 1599 | { |
| 1600 | /* If necessary, we also propagate the data to disk. */ |
| 1601 | @@ -441,7 +455,7 @@ addgetnetgrentX (struct database_dyn *db, int fd, request_header *req, |
| 1602 | } |
| 1603 | |
| 1604 | out: |
| 1605 | - *resultp = dataset; |
| 1606 | + scratch->dataset = dataset; |
| 1607 | |
| 1608 | return timeout; |
| 1609 | } |
| 1610 | @@ -462,6 +476,9 @@ addinnetgrX (struct database_dyn *db, int fd, request_header *req, |
| 1611 | if (user != NULL) |
| 1612 | key = strchr (key, '\0') + 1; |
| 1613 | const char *domain = *key++ ? key : NULL; |
| 1614 | + struct addgetnetgrentX_scratch scratch; |
| 1615 | + |
| 1616 | + addgetnetgrentX_scratch_init (&scratch); |
| 1617 | |
| 1618 | if (__glibc_unlikely (debug_level > 0)) |
| 1619 | { |
| 1620 | @@ -477,12 +494,8 @@ addinnetgrX (struct database_dyn *db, int fd, request_header *req, |
| 1621 | group, group_len, |
| 1622 | db, uid); |
| 1623 | time_t timeout; |
| 1624 | - void *tofree; |
| 1625 | if (result != NULL) |
| 1626 | - { |
| 1627 | - timeout = result->head.timeout; |
| 1628 | - tofree = NULL; |
| 1629 | - } |
| 1630 | + timeout = result->head.timeout; |
| 1631 | else |
| 1632 | { |
| 1633 | request_header req_get = |
| 1634 | @@ -491,7 +504,10 @@ addinnetgrX (struct database_dyn *db, int fd, request_header *req, |
| 1635 | .key_len = group_len |
| 1636 | }; |
| 1637 | timeout = addgetnetgrentX (db, -1, &req_get, group, uid, NULL, NULL, |
| 1638 | - &result, &tofree); |
| 1639 | + &scratch); |
| 1640 | + result = scratch.dataset; |
| 1641 | + if (timeout < 0) |
| 1642 | + goto out; |
| 1643 | } |
| 1644 | |
| 1645 | struct indataset |
| 1646 | @@ -502,24 +518,26 @@ addinnetgrX (struct database_dyn *db, int fd, request_header *req, |
| 1647 | = (struct indataset *) mempool_alloc (db, |
| 1648 | sizeof (*dataset) + req->key_len, |
| 1649 | 1); |
| 1650 | - struct indataset dataset_mem; |
| 1651 | bool cacheable = true; |
| 1652 | if (__glibc_unlikely (dataset == NULL)) |
| 1653 | { |
| 1654 | cacheable = false; |
| 1655 | - dataset = &dataset_mem; |
| 1656 | + /* The alloca is safe because nscd_run_worker verfies that |
| 1657 | + key_len is not larger than MAXKEYLEN. */ |
| 1658 | + dataset = alloca (sizeof (*dataset) + req->key_len); |
| 1659 | } |
| 1660 | |
| 1661 | datahead_init_pos (&dataset->head, sizeof (*dataset) + req->key_len, |
| 1662 | sizeof (innetgroup_response_header), |
| 1663 | - he == NULL ? 0 : dh->nreloads + 1, result->head.ttl); |
| 1664 | + he == NULL ? 0 : dh->nreloads + 1, |
| 1665 | + result == NULL ? db->negtimeout : result->head.ttl); |
| 1666 | /* Set the notfound status and timeout based on the result from |
| 1667 | getnetgrent. */ |
| 1668 | - dataset->head.notfound = result->head.notfound; |
| 1669 | + dataset->head.notfound = result == NULL || result->head.notfound; |
| 1670 | dataset->head.timeout = timeout; |
| 1671 | |
| 1672 | dataset->resp.version = NSCD_VERSION; |
| 1673 | - dataset->resp.found = result->resp.found; |
| 1674 | + dataset->resp.found = result != NULL && result->resp.found; |
| 1675 | /* Until we find a matching entry the result is 0. */ |
| 1676 | dataset->resp.result = 0; |
| 1677 | |
| 1678 | @@ -567,7 +585,9 @@ addinnetgrX (struct database_dyn *db, int fd, request_header *req, |
| 1679 | goto out; |
| 1680 | } |
| 1681 | |
| 1682 | - if (he == NULL) |
| 1683 | + /* addgetnetgrentX may have already sent a notfound response. Do |
| 1684 | + not send another one. */ |
| 1685 | + if (he == NULL && dataset->resp.found) |
| 1686 | { |
| 1687 | /* We write the dataset before inserting it to the database |
| 1688 | since while inserting this thread might block and so would |
| 1689 | @@ -601,7 +621,7 @@ addinnetgrX (struct database_dyn *db, int fd, request_header *req, |
| 1690 | } |
| 1691 | |
| 1692 | out: |
| 1693 | - free (tofree); |
| 1694 | + addgetnetgrentX_scratch_free (&scratch); |
| 1695 | return timeout; |
| 1696 | } |
| 1697 | |
| 1698 | @@ -611,11 +631,12 @@ addgetnetgrentX_ignore (struct database_dyn *db, int fd, request_header *req, |
| 1699 | const char *key, uid_t uid, struct hashentry *he, |
| 1700 | struct datahead *dh) |
| 1701 | { |
| 1702 | - struct dataset *ignore; |
| 1703 | - void *tofree; |
| 1704 | - time_t timeout = addgetnetgrentX (db, fd, req, key, uid, he, dh, |
| 1705 | - &ignore, &tofree); |
| 1706 | - free (tofree); |
| 1707 | + struct addgetnetgrentX_scratch scratch; |
| 1708 | + addgetnetgrentX_scratch_init (&scratch); |
| 1709 | + time_t timeout = addgetnetgrentX (db, fd, req, key, uid, he, dh, &scratch); |
| 1710 | + addgetnetgrentX_scratch_free (&scratch); |
| 1711 | + if (timeout < 0) |
| 1712 | + timeout = 0; |
| 1713 | return timeout; |
| 1714 | } |
| 1715 | |
| 1716 | @@ -659,5 +680,9 @@ readdinnetgr (struct database_dyn *db, struct hashentry *he, |
| 1717 | .key_len = he->len |
| 1718 | }; |
| 1719 | |
| 1720 | - return addinnetgrX (db, -1, &req, db->data + he->key, he->owner, he, dh); |
| 1721 | + time_t timeout = addinnetgrX (db, -1, &req, db->data + he->key, he->owner, |
| 1722 | + he, dh); |
| 1723 | + if (timeout < 0) |
| 1724 | + timeout = 0; |
| 1725 | + return timeout; |
| 1726 | } |
| 1727 | diff --git a/socket/Makefile b/socket/Makefile |
| 1728 | index 74ca5b8452..fc1bd0a260 100644 |
| 1729 | --- a/socket/Makefile |
| 1730 | +++ b/socket/Makefile |
| 1731 | @@ -70,6 +70,7 @@ tests := \ |
| 1732 | tst-accept4 \ |
| 1733 | tst-cmsg_cloexec \ |
| 1734 | tst-cmsghdr \ |
| 1735 | + tst-connect \ |
| 1736 | tst-sockopt \ |
| 1737 | # tests |
| 1738 | |
| 1739 | diff --git a/socket/tst-connect.c b/socket/tst-connect.c |
| 1740 | new file mode 100644 |
| 1741 | index 0000000000..ec2fdd92c0 |
| 1742 | --- /dev/null |
| 1743 | +++ b/socket/tst-connect.c |
| 1744 | @@ -0,0 +1,113 @@ |
| 1745 | +/* Test the connect function. |
| 1746 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 1747 | + This file is part of the GNU C Library. |
| 1748 | + |
| 1749 | + The GNU C Library is free software; you can redistribute it and/or |
| 1750 | + modify it under the terms of the GNU Lesser General Public |
| 1751 | + License as published by the Free Software Foundation; either |
| 1752 | + version 2.1 of the License, or (at your option) any later version. |
| 1753 | + |
| 1754 | + The GNU C Library is distributed in the hope that it will be useful, |
| 1755 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 1756 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 1757 | + Lesser General Public License for more details. |
| 1758 | + |
| 1759 | + You should have received a copy of the GNU Lesser General Public |
| 1760 | + License along with the GNU C Library; if not, see |
| 1761 | + <https://www.gnu.org/licenses/>. */ |
| 1762 | + |
| 1763 | +#include <arpa/inet.h> |
| 1764 | +#include <errno.h> |
| 1765 | +#include <fcntl.h> |
| 1766 | +#include <signal.h> |
| 1767 | +#include <stdbool.h> |
| 1768 | +#include <support/check.h> |
| 1769 | +#include <support/xsocket.h> |
| 1770 | +#include <support/xunistd.h> |
| 1771 | +#include <sys/socket.h> |
| 1772 | +#include <stdio.h> |
| 1773 | + |
| 1774 | +static struct sockaddr_in server_address; |
| 1775 | + |
| 1776 | +int |
| 1777 | +open_socket_inet_tcp (void) |
| 1778 | +{ |
| 1779 | + int fd = socket (AF_INET, SOCK_STREAM, IPPROTO_TCP); |
| 1780 | + if (fd < 0) |
| 1781 | + { |
| 1782 | + if (errno == EAFNOSUPPORT) |
| 1783 | + FAIL_UNSUPPORTED ("The host does not support IPv4"); |
| 1784 | + else |
| 1785 | + FAIL_EXIT1 ("socket (AF_INET, SOCK_STREAM, IPPROTO_TCP): %m\n"); |
| 1786 | + } |
| 1787 | + return fd; |
| 1788 | +} |
| 1789 | + |
| 1790 | +static pid_t |
| 1791 | +start_server (void) |
| 1792 | +{ |
| 1793 | + server_address.sin_family = AF_INET; |
| 1794 | + server_address.sin_port = 0; |
| 1795 | + server_address.sin_addr.s_addr = htonl (INADDR_LOOPBACK); |
| 1796 | + |
| 1797 | + int server_sock = open_socket_inet_tcp (); |
| 1798 | + |
| 1799 | + xbind (server_sock, (struct sockaddr *) &server_address, |
| 1800 | + sizeof (server_address)); |
| 1801 | + |
| 1802 | + socklen_t sa_len = sizeof (server_address); |
| 1803 | + xgetsockname (server_sock, (struct sockaddr *) &server_address, &sa_len); |
| 1804 | + xlisten (server_sock, 5); |
| 1805 | + |
| 1806 | + pid_t my_pid = xfork (); |
| 1807 | + if (my_pid > 0) |
| 1808 | + { |
| 1809 | + xclose (server_sock); |
| 1810 | + return my_pid; |
| 1811 | + } |
| 1812 | + |
| 1813 | + struct sockaddr_in client_address; |
| 1814 | + socklen_t ca_len = sizeof (server_address); |
| 1815 | + int client_sock = xaccept (server_sock, (struct sockaddr *) &client_address, |
| 1816 | + &ca_len); |
| 1817 | + printf ("socket accepted %d\n", client_sock); |
| 1818 | + |
| 1819 | + _exit (0); |
| 1820 | +} |
| 1821 | + |
| 1822 | +static int |
| 1823 | +do_test (void) |
| 1824 | +{ |
| 1825 | + pid_t serv_pid; |
| 1826 | + struct sockaddr_in peer; |
| 1827 | + socklen_t peer_len; |
| 1828 | + |
| 1829 | + serv_pid = start_server (); |
| 1830 | + int client_sock = open_socket_inet_tcp (); |
| 1831 | + xconnect (client_sock, (const struct sockaddr *) &server_address, |
| 1832 | + sizeof (server_address)); |
| 1833 | + |
| 1834 | + /* A second connect with same arguments should fail with EISCONN. */ |
| 1835 | + int result = connect (client_sock, |
| 1836 | + (const struct sockaddr *) &server_address, |
| 1837 | + sizeof (server_address)); |
| 1838 | + if (result == 0 || errno != EISCONN) |
| 1839 | + FAIL_EXIT1 ("Second connect (%d), should fail with EISCONN: %m", |
| 1840 | + client_sock); |
| 1841 | + |
| 1842 | + peer_len = sizeof (peer); |
| 1843 | + xgetpeername (client_sock, (struct sockaddr *) &peer, &peer_len); |
| 1844 | + TEST_COMPARE (peer_len, sizeof (peer)); |
| 1845 | + TEST_COMPARE (peer.sin_port, server_address.sin_port); |
| 1846 | + TEST_COMPARE_BLOB (&peer.sin_addr, sizeof (peer.sin_addr), |
| 1847 | + &server_address.sin_addr, |
| 1848 | + sizeof (server_address.sin_addr)); |
| 1849 | + |
| 1850 | + int status; |
| 1851 | + xwaitpid (serv_pid, &status, 0); |
| 1852 | + TEST_COMPARE (status, 0); |
| 1853 | + |
| 1854 | + return 0; |
| 1855 | +} |
| 1856 | + |
| 1857 | +#include <support/test-driver.c> |
| 1858 | diff --git a/stdlib/Makefile b/stdlib/Makefile |
| 1859 | index d587f054d1..9898cc5d8a 100644 |
| 1860 | --- a/stdlib/Makefile |
| 1861 | +++ b/stdlib/Makefile |
| 1862 | @@ -308,6 +308,7 @@ tests := \ |
| 1863 | tst-setcontext10 \ |
| 1864 | tst-setcontext11 \ |
| 1865 | tst-stdbit-Wconversion \ |
| 1866 | + tst-stdbit-builtins \ |
| 1867 | tst-stdc_bit_ceil \ |
| 1868 | tst-stdc_bit_floor \ |
| 1869 | tst-stdc_bit_width \ |
| 1870 | diff --git a/stdlib/stdbit.h b/stdlib/stdbit.h |
| 1871 | index f334eb174d..2801590c63 100644 |
| 1872 | --- a/stdlib/stdbit.h |
| 1873 | +++ b/stdlib/stdbit.h |
| 1874 | @@ -64,9 +64,13 @@ extern unsigned int stdc_leading_zeros_ul (unsigned long int __x) |
| 1875 | __extension__ |
| 1876 | extern unsigned int stdc_leading_zeros_ull (unsigned long long int __x) |
| 1877 | __THROW __attribute_const__; |
| 1878 | -#define stdc_leading_zeros(x) \ |
| 1879 | +#if __glibc_has_builtin (__builtin_stdc_leading_zeros) |
| 1880 | +# define stdc_leading_zeros(x) (__builtin_stdc_leading_zeros (x)) |
| 1881 | +#else |
| 1882 | +# define stdc_leading_zeros(x) \ |
| 1883 | (stdc_leading_zeros_ull (x) \ |
| 1884 | - (unsigned int) (8 * (sizeof (0ULL) - sizeof (x)))) |
| 1885 | +#endif |
| 1886 | |
| 1887 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 1888 | static __always_inline unsigned int |
| 1889 | @@ -116,9 +120,13 @@ extern unsigned int stdc_leading_ones_ul (unsigned long int __x) |
| 1890 | __extension__ |
| 1891 | extern unsigned int stdc_leading_ones_ull (unsigned long long int __x) |
| 1892 | __THROW __attribute_const__; |
| 1893 | -#define stdc_leading_ones(x) \ |
| 1894 | +#if __glibc_has_builtin (__builtin_stdc_leading_ones) |
| 1895 | +# define stdc_leading_ones(x) (__builtin_stdc_leading_ones (x)) |
| 1896 | +#else |
| 1897 | +# define stdc_leading_ones(x) \ |
| 1898 | (stdc_leading_ones_ull ((unsigned long long int) (x) \ |
| 1899 | << 8 * (sizeof (0ULL) - sizeof (x)))) |
| 1900 | +#endif |
| 1901 | |
| 1902 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 1903 | static __always_inline unsigned int |
| 1904 | @@ -168,11 +176,15 @@ extern unsigned int stdc_trailing_zeros_ul (unsigned long int __x) |
| 1905 | __extension__ |
| 1906 | extern unsigned int stdc_trailing_zeros_ull (unsigned long long int __x) |
| 1907 | __THROW __attribute_const__; |
| 1908 | -#define stdc_trailing_zeros(x) \ |
| 1909 | +#if __glibc_has_builtin (__builtin_stdc_trailing_zeros) |
| 1910 | +# define stdc_trailing_zeros(x) (__builtin_stdc_trailing_zeros (x)) |
| 1911 | +#else |
| 1912 | +# define stdc_trailing_zeros(x) \ |
| 1913 | (sizeof (x) == 8 ? stdc_trailing_zeros_ull (x) \ |
| 1914 | : sizeof (x) == 4 ? stdc_trailing_zeros_ui (x) \ |
| 1915 | : sizeof (x) == 2 ? stdc_trailing_zeros_us (__pacify_uint16 (x)) \ |
| 1916 | : stdc_trailing_zeros_uc (__pacify_uint8 (x))) |
| 1917 | +#endif |
| 1918 | |
| 1919 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_ctzll) |
| 1920 | static __always_inline unsigned int |
| 1921 | @@ -222,7 +234,11 @@ extern unsigned int stdc_trailing_ones_ul (unsigned long int __x) |
| 1922 | __extension__ |
| 1923 | extern unsigned int stdc_trailing_ones_ull (unsigned long long int __x) |
| 1924 | __THROW __attribute_const__; |
| 1925 | -#define stdc_trailing_ones(x) (stdc_trailing_ones_ull (x)) |
| 1926 | +#if __glibc_has_builtin (__builtin_stdc_trailing_ones) |
| 1927 | +# define stdc_trailing_ones(x) (__builtin_stdc_trailing_ones (x)) |
| 1928 | +#else |
| 1929 | +# define stdc_trailing_ones(x) (stdc_trailing_ones_ull (x)) |
| 1930 | +#endif |
| 1931 | |
| 1932 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_ctzll) |
| 1933 | static __always_inline unsigned int |
| 1934 | @@ -272,11 +288,15 @@ extern unsigned int stdc_first_leading_zero_ul (unsigned long int __x) |
| 1935 | __extension__ |
| 1936 | extern unsigned int stdc_first_leading_zero_ull (unsigned long long int __x) |
| 1937 | __THROW __attribute_const__; |
| 1938 | -#define stdc_first_leading_zero(x) \ |
| 1939 | +#if __glibc_has_builtin (__builtin_stdc_first_leading_zero) |
| 1940 | +# define stdc_first_leading_zero(x) (__builtin_stdc_first_leading_zero (x)) |
| 1941 | +#else |
| 1942 | +# define stdc_first_leading_zero(x) \ |
| 1943 | (sizeof (x) == 8 ? stdc_first_leading_zero_ull (x) \ |
| 1944 | : sizeof (x) == 4 ? stdc_first_leading_zero_ui (x) \ |
| 1945 | : sizeof (x) == 2 ? stdc_first_leading_zero_us (__pacify_uint16 (x)) \ |
| 1946 | : stdc_first_leading_zero_uc (__pacify_uint8 (x))) |
| 1947 | +#endif |
| 1948 | |
| 1949 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 1950 | static __always_inline unsigned int |
| 1951 | @@ -326,11 +346,15 @@ extern unsigned int stdc_first_leading_one_ul (unsigned long int __x) |
| 1952 | __extension__ |
| 1953 | extern unsigned int stdc_first_leading_one_ull (unsigned long long int __x) |
| 1954 | __THROW __attribute_const__; |
| 1955 | -#define stdc_first_leading_one(x) \ |
| 1956 | +#if __glibc_has_builtin (__builtin_stdc_first_leading_one) |
| 1957 | +# define stdc_first_leading_one(x) (__builtin_stdc_first_leading_one (x)) |
| 1958 | +#else |
| 1959 | +# define stdc_first_leading_one(x) \ |
| 1960 | (sizeof (x) == 8 ? stdc_first_leading_one_ull (x) \ |
| 1961 | : sizeof (x) == 4 ? stdc_first_leading_one_ui (x) \ |
| 1962 | : sizeof (x) == 2 ? stdc_first_leading_one_us (__pacify_uint16 (x)) \ |
| 1963 | : stdc_first_leading_one_uc (__pacify_uint8 (x))) |
| 1964 | +#endif |
| 1965 | |
| 1966 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 1967 | static __always_inline unsigned int |
| 1968 | @@ -380,11 +404,15 @@ extern unsigned int stdc_first_trailing_zero_ul (unsigned long int __x) |
| 1969 | __extension__ |
| 1970 | extern unsigned int stdc_first_trailing_zero_ull (unsigned long long int __x) |
| 1971 | __THROW __attribute_const__; |
| 1972 | -#define stdc_first_trailing_zero(x) \ |
| 1973 | +#if __glibc_has_builtin (__builtin_stdc_first_trailing_zero) |
| 1974 | +# define stdc_first_trailing_zero(x) (__builtin_stdc_first_trailing_zero (x)) |
| 1975 | +#else |
| 1976 | +# define stdc_first_trailing_zero(x) \ |
| 1977 | (sizeof (x) == 8 ? stdc_first_trailing_zero_ull (x) \ |
| 1978 | : sizeof (x) == 4 ? stdc_first_trailing_zero_ui (x) \ |
| 1979 | : sizeof (x) == 2 ? stdc_first_trailing_zero_us (__pacify_uint16 (x)) \ |
| 1980 | : stdc_first_trailing_zero_uc (__pacify_uint8 (x))) |
| 1981 | +#endif |
| 1982 | |
| 1983 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_ctzll) |
| 1984 | static __always_inline unsigned int |
| 1985 | @@ -434,11 +462,15 @@ extern unsigned int stdc_first_trailing_one_ul (unsigned long int __x) |
| 1986 | __extension__ |
| 1987 | extern unsigned int stdc_first_trailing_one_ull (unsigned long long int __x) |
| 1988 | __THROW __attribute_const__; |
| 1989 | -#define stdc_first_trailing_one(x) \ |
| 1990 | +#if __glibc_has_builtin (__builtin_stdc_first_trailing_one) |
| 1991 | +# define stdc_first_trailing_one(x) (__builtin_stdc_first_trailing_one (x)) |
| 1992 | +#else |
| 1993 | +# define stdc_first_trailing_one(x) \ |
| 1994 | (sizeof (x) == 8 ? stdc_first_trailing_one_ull (x) \ |
| 1995 | : sizeof (x) == 4 ? stdc_first_trailing_one_ui (x) \ |
| 1996 | : sizeof (x) == 2 ? stdc_first_trailing_one_us (__pacify_uint16 (x)) \ |
| 1997 | : stdc_first_trailing_one_uc (__pacify_uint8 (x))) |
| 1998 | +#endif |
| 1999 | |
| 2000 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_ctzll) |
| 2001 | static __always_inline unsigned int |
| 2002 | @@ -488,9 +520,13 @@ extern unsigned int stdc_count_zeros_ul (unsigned long int __x) |
| 2003 | __extension__ |
| 2004 | extern unsigned int stdc_count_zeros_ull (unsigned long long int __x) |
| 2005 | __THROW __attribute_const__; |
| 2006 | -#define stdc_count_zeros(x) \ |
| 2007 | +#if __glibc_has_builtin (__builtin_stdc_count_zeros) |
| 2008 | +# define stdc_count_zeros(x) (__builtin_stdc_count_zeros (x)) |
| 2009 | +#else |
| 2010 | +# define stdc_count_zeros(x) \ |
| 2011 | (stdc_count_zeros_ull (x) \ |
| 2012 | - (unsigned int) (8 * (sizeof (0ULL) - sizeof (x)))) |
| 2013 | +#endif |
| 2014 | |
| 2015 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_popcountll) |
| 2016 | static __always_inline unsigned int |
| 2017 | @@ -540,7 +576,11 @@ extern unsigned int stdc_count_ones_ul (unsigned long int __x) |
| 2018 | __extension__ |
| 2019 | extern unsigned int stdc_count_ones_ull (unsigned long long int __x) |
| 2020 | __THROW __attribute_const__; |
| 2021 | -#define stdc_count_ones(x) (stdc_count_ones_ull (x)) |
| 2022 | +#if __glibc_has_builtin (__builtin_stdc_count_ones) |
| 2023 | +# define stdc_count_ones(x) (__builtin_stdc_count_ones (x)) |
| 2024 | +#else |
| 2025 | +# define stdc_count_ones(x) (stdc_count_ones_ull (x)) |
| 2026 | +#endif |
| 2027 | |
| 2028 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_popcountll) |
| 2029 | static __always_inline unsigned int |
| 2030 | @@ -590,10 +630,14 @@ extern bool stdc_has_single_bit_ul (unsigned long int __x) |
| 2031 | __extension__ |
| 2032 | extern bool stdc_has_single_bit_ull (unsigned long long int __x) |
| 2033 | __THROW __attribute_const__; |
| 2034 | -#define stdc_has_single_bit(x) \ |
| 2035 | +#if __glibc_has_builtin (__builtin_stdc_has_single_bit) |
| 2036 | +# define stdc_has_single_bit(x) (__builtin_stdc_has_single_bit (x)) |
| 2037 | +#else |
| 2038 | +# define stdc_has_single_bit(x) \ |
| 2039 | ((bool) (sizeof (x) <= sizeof (unsigned int) \ |
| 2040 | ? stdc_has_single_bit_ui (x) \ |
| 2041 | : stdc_has_single_bit_ull (x))) |
| 2042 | +#endif |
| 2043 | |
| 2044 | static __always_inline bool |
| 2045 | __hsb64_inline (uint64_t __x) |
| 2046 | @@ -641,7 +685,11 @@ extern unsigned int stdc_bit_width_ul (unsigned long int __x) |
| 2047 | __extension__ |
| 2048 | extern unsigned int stdc_bit_width_ull (unsigned long long int __x) |
| 2049 | __THROW __attribute_const__; |
| 2050 | -#define stdc_bit_width(x) (stdc_bit_width_ull (x)) |
| 2051 | +#if __glibc_has_builtin (__builtin_stdc_bit_width) |
| 2052 | +# define stdc_bit_width(x) (__builtin_stdc_bit_width (x)) |
| 2053 | +#else |
| 2054 | +# define stdc_bit_width(x) (stdc_bit_width_ull (x)) |
| 2055 | +#endif |
| 2056 | |
| 2057 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 2058 | static __always_inline unsigned int |
| 2059 | @@ -691,7 +739,11 @@ extern unsigned long int stdc_bit_floor_ul (unsigned long int __x) |
| 2060 | __extension__ |
| 2061 | extern unsigned long long int stdc_bit_floor_ull (unsigned long long int __x) |
| 2062 | __THROW __attribute_const__; |
| 2063 | -#define stdc_bit_floor(x) ((__typeof (x)) stdc_bit_floor_ull (x)) |
| 2064 | +#if __glibc_has_builtin (__builtin_stdc_bit_floor) |
| 2065 | +# define stdc_bit_floor(x) (__builtin_stdc_bit_floor (x)) |
| 2066 | +#else |
| 2067 | +# define stdc_bit_floor(x) ((__typeof (x)) stdc_bit_floor_ull (x)) |
| 2068 | +#endif |
| 2069 | |
| 2070 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 2071 | static __always_inline uint64_t |
| 2072 | @@ -743,7 +795,11 @@ extern unsigned long int stdc_bit_ceil_ul (unsigned long int __x) |
| 2073 | __extension__ |
| 2074 | extern unsigned long long int stdc_bit_ceil_ull (unsigned long long int __x) |
| 2075 | __THROW __attribute_const__; |
| 2076 | -#define stdc_bit_ceil(x) ((__typeof (x)) stdc_bit_ceil_ull (x)) |
| 2077 | +#if __glibc_has_builtin (__builtin_stdc_bit_ceil) |
| 2078 | +# define stdc_bit_ceil(x) (__builtin_stdc_bit_ceil (x)) |
| 2079 | +#else |
| 2080 | +# define stdc_bit_ceil(x) ((__typeof (x)) stdc_bit_ceil_ull (x)) |
| 2081 | +#endif |
| 2082 | |
| 2083 | #if __GNUC_PREREQ (3, 4) || __glibc_has_builtin (__builtin_clzll) |
| 2084 | static __always_inline uint64_t |
| 2085 | diff --git a/stdlib/tst-stdbit-builtins.c b/stdlib/tst-stdbit-builtins.c |
| 2086 | new file mode 100644 |
| 2087 | index 0000000000..536841ca8a |
| 2088 | --- /dev/null |
| 2089 | +++ b/stdlib/tst-stdbit-builtins.c |
| 2090 | @@ -0,0 +1,778 @@ |
| 2091 | +/* Test <stdbit.h> type-generic macros with compiler __builtin_stdc_* support. |
| 2092 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 2093 | + This file is part of the GNU C Library. |
| 2094 | + |
| 2095 | + The GNU C Library is free software; you can redistribute it and/or |
| 2096 | + modify it under the terms of the GNU Lesser General Public |
| 2097 | + License as published by the Free Software Foundation; either |
| 2098 | + version 2.1 of the License, or (at your option) any later version. |
| 2099 | + |
| 2100 | + The GNU C Library is distributed in the hope that it will be useful, |
| 2101 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 2102 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 2103 | + Lesser General Public License for more details. |
| 2104 | + |
| 2105 | + You should have received a copy of the GNU Lesser General Public |
| 2106 | + License along with the GNU C Library; if not, see |
| 2107 | + <https://www.gnu.org/licenses/>. */ |
| 2108 | + |
| 2109 | +#include <stdbit.h> |
| 2110 | +#include <limits.h> |
| 2111 | +#include <support/check.h> |
| 2112 | + |
| 2113 | +#if __glibc_has_builtin (__builtin_stdc_leading_zeros) \ |
| 2114 | + && __glibc_has_builtin (__builtin_stdc_leading_ones) \ |
| 2115 | + && __glibc_has_builtin (__builtin_stdc_trailing_zeros) \ |
| 2116 | + && __glibc_has_builtin (__builtin_stdc_trailing_ones) \ |
| 2117 | + && __glibc_has_builtin (__builtin_stdc_first_leading_zero) \ |
| 2118 | + && __glibc_has_builtin (__builtin_stdc_first_leading_one) \ |
| 2119 | + && __glibc_has_builtin (__builtin_stdc_first_trailing_zero) \ |
| 2120 | + && __glibc_has_builtin (__builtin_stdc_first_trailing_one) \ |
| 2121 | + && __glibc_has_builtin (__builtin_stdc_count_zeros) \ |
| 2122 | + && __glibc_has_builtin (__builtin_stdc_count_ones) \ |
| 2123 | + && __glibc_has_builtin (__builtin_stdc_has_single_bit) \ |
| 2124 | + && __glibc_has_builtin (__builtin_stdc_bit_width) \ |
| 2125 | + && __glibc_has_builtin (__builtin_stdc_bit_floor) \ |
| 2126 | + && __glibc_has_builtin (__builtin_stdc_bit_ceil) |
| 2127 | + |
| 2128 | +# if !defined (BITINT_MAXWIDTH) && defined (__BITINT_MAXWIDTH__) |
| 2129 | +# define BITINT_MAXWIDTH __BITINT_MAXWIDTH__ |
| 2130 | +# endif |
| 2131 | + |
| 2132 | +typedef unsigned char uc; |
| 2133 | +typedef unsigned short us; |
| 2134 | +typedef unsigned int ui; |
| 2135 | +typedef unsigned long int ul; |
| 2136 | +typedef unsigned long long int ull; |
| 2137 | + |
| 2138 | +# define expr_has_type(e, t) _Generic (e, default : 0, t : 1) |
| 2139 | + |
| 2140 | +static int |
| 2141 | +do_test (void) |
| 2142 | +{ |
| 2143 | + TEST_COMPARE (stdc_leading_zeros ((uc) 0), CHAR_BIT); |
| 2144 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros ((uc) 0), ui), 1); |
| 2145 | + TEST_COMPARE (stdc_leading_zeros ((us) 0), sizeof (short) * CHAR_BIT); |
| 2146 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros ((us) 0), ui), 1); |
| 2147 | + TEST_COMPARE (stdc_leading_zeros (0U), sizeof (int) * CHAR_BIT); |
| 2148 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros (0U), ui), 1); |
| 2149 | + TEST_COMPARE (stdc_leading_zeros (0UL), sizeof (long int) * CHAR_BIT); |
| 2150 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros (0UL), ui), 1); |
| 2151 | + TEST_COMPARE (stdc_leading_zeros (0ULL), sizeof (long long int) * CHAR_BIT); |
| 2152 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros (0ULL), ui), 1); |
| 2153 | + TEST_COMPARE (stdc_leading_zeros ((uc) ~0U), 0); |
| 2154 | + TEST_COMPARE (stdc_leading_zeros ((us) ~0U), 0); |
| 2155 | + TEST_COMPARE (stdc_leading_zeros (~0U), 0); |
| 2156 | + TEST_COMPARE (stdc_leading_zeros (~0UL), 0); |
| 2157 | + TEST_COMPARE (stdc_leading_zeros (~0ULL), 0); |
| 2158 | + TEST_COMPARE (stdc_leading_zeros ((uc) 3), CHAR_BIT - 2); |
| 2159 | + TEST_COMPARE (stdc_leading_zeros ((us) 9), sizeof (short) * CHAR_BIT - 4); |
| 2160 | + TEST_COMPARE (stdc_leading_zeros (34U), sizeof (int) * CHAR_BIT - 6); |
| 2161 | + TEST_COMPARE (stdc_leading_zeros (130UL), sizeof (long int) * CHAR_BIT - 8); |
| 2162 | + TEST_COMPARE (stdc_leading_zeros (512ULL), |
| 2163 | + sizeof (long long int) * CHAR_BIT - 10); |
| 2164 | + TEST_COMPARE (stdc_leading_ones ((uc) 0), 0); |
| 2165 | + TEST_COMPARE (expr_has_type (stdc_leading_ones ((uc) 0), ui), 1); |
| 2166 | + TEST_COMPARE (stdc_leading_ones ((us) 0), 0); |
| 2167 | + TEST_COMPARE (expr_has_type (stdc_leading_ones ((us) 0), ui), 1); |
| 2168 | + TEST_COMPARE (stdc_leading_ones (0U), 0); |
| 2169 | + TEST_COMPARE (expr_has_type (stdc_leading_ones (0U), ui), 1); |
| 2170 | + TEST_COMPARE (stdc_leading_ones (0UL), 0); |
| 2171 | + TEST_COMPARE (expr_has_type (stdc_leading_ones (0UL), ui), 1); |
| 2172 | + TEST_COMPARE (stdc_leading_ones (0ULL), 0); |
| 2173 | + TEST_COMPARE (expr_has_type (stdc_leading_ones (0ULL), ui), 1); |
| 2174 | + TEST_COMPARE (stdc_leading_ones ((uc) ~0U), CHAR_BIT); |
| 2175 | + TEST_COMPARE (stdc_leading_ones ((us) ~0U), sizeof (short) * CHAR_BIT); |
| 2176 | + TEST_COMPARE (stdc_leading_ones (~0U), sizeof (int) * CHAR_BIT); |
| 2177 | + TEST_COMPARE (stdc_leading_ones (~0UL), sizeof (long int) * CHAR_BIT); |
| 2178 | + TEST_COMPARE (stdc_leading_ones (~0ULL), sizeof (long long int) * CHAR_BIT); |
| 2179 | + TEST_COMPARE (stdc_leading_ones ((uc) ~3), CHAR_BIT - 2); |
| 2180 | + TEST_COMPARE (stdc_leading_ones ((us) ~9), sizeof (short) * CHAR_BIT - 4); |
| 2181 | + TEST_COMPARE (stdc_leading_ones (~34U), sizeof (int) * CHAR_BIT - 6); |
| 2182 | + TEST_COMPARE (stdc_leading_ones (~130UL), sizeof (long int) * CHAR_BIT - 8); |
| 2183 | + TEST_COMPARE (stdc_leading_ones (~512ULL), |
| 2184 | + sizeof (long long int) * CHAR_BIT - 10); |
| 2185 | + TEST_COMPARE (stdc_trailing_zeros ((uc) 0), CHAR_BIT); |
| 2186 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros ((uc) 0), ui), 1); |
| 2187 | + TEST_COMPARE (stdc_trailing_zeros ((us) 0), sizeof (short) * CHAR_BIT); |
| 2188 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros ((us) 0), ui), 1); |
| 2189 | + TEST_COMPARE (stdc_trailing_zeros (0U), sizeof (int) * CHAR_BIT); |
| 2190 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros (0U), ui), 1); |
| 2191 | + TEST_COMPARE (stdc_trailing_zeros (0UL), sizeof (long int) * CHAR_BIT); |
| 2192 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros (0UL), ui), 1); |
| 2193 | + TEST_COMPARE (stdc_trailing_zeros (0ULL), sizeof (long long int) * CHAR_BIT); |
| 2194 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros (0ULL), ui), 1); |
| 2195 | + TEST_COMPARE (stdc_trailing_zeros ((uc) ~0U), 0); |
| 2196 | + TEST_COMPARE (stdc_trailing_zeros ((us) ~0U), 0); |
| 2197 | + TEST_COMPARE (stdc_trailing_zeros (~0U), 0); |
| 2198 | + TEST_COMPARE (stdc_trailing_zeros (~0UL), 0); |
| 2199 | + TEST_COMPARE (stdc_trailing_zeros (~0ULL), 0); |
| 2200 | + TEST_COMPARE (stdc_trailing_zeros ((uc) 2), 1); |
| 2201 | + TEST_COMPARE (stdc_trailing_zeros ((us) 24), 3); |
| 2202 | + TEST_COMPARE (stdc_trailing_zeros (32U), 5); |
| 2203 | + TEST_COMPARE (stdc_trailing_zeros (128UL), 7); |
| 2204 | + TEST_COMPARE (stdc_trailing_zeros (512ULL), 9); |
| 2205 | + TEST_COMPARE (stdc_trailing_ones ((uc) 0), 0); |
| 2206 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones ((uc) 0), ui), 1); |
| 2207 | + TEST_COMPARE (stdc_trailing_ones ((us) 0), 0); |
| 2208 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones ((us) 0), ui), 1); |
| 2209 | + TEST_COMPARE (stdc_trailing_ones (0U), 0); |
| 2210 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones (0U), ui), 1); |
| 2211 | + TEST_COMPARE (stdc_trailing_ones (0UL), 0); |
| 2212 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones (0UL), ui), 1); |
| 2213 | + TEST_COMPARE (stdc_trailing_ones (0ULL), 0); |
| 2214 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones (0ULL), ui), 1); |
| 2215 | + TEST_COMPARE (stdc_trailing_ones ((uc) ~0U), CHAR_BIT); |
| 2216 | + TEST_COMPARE (stdc_trailing_ones ((us) ~0U), sizeof (short) * CHAR_BIT); |
| 2217 | + TEST_COMPARE (stdc_trailing_ones (~0U), sizeof (int) * CHAR_BIT); |
| 2218 | + TEST_COMPARE (stdc_trailing_ones (~0UL), sizeof (long int) * CHAR_BIT); |
| 2219 | + TEST_COMPARE (stdc_trailing_ones (~0ULL), sizeof (long long int) * CHAR_BIT); |
| 2220 | + TEST_COMPARE (stdc_trailing_ones ((uc) 5), 1); |
| 2221 | + TEST_COMPARE (stdc_trailing_ones ((us) 15), 4); |
| 2222 | + TEST_COMPARE (stdc_trailing_ones (127U), 7); |
| 2223 | + TEST_COMPARE (stdc_trailing_ones (511UL), 9); |
| 2224 | + TEST_COMPARE (stdc_trailing_ones (~0ULL >> 2), |
| 2225 | + sizeof (long long int) * CHAR_BIT - 2); |
| 2226 | + TEST_COMPARE (stdc_first_leading_zero ((uc) 0), 1); |
| 2227 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero ((uc) 0), ui), 1); |
| 2228 | + TEST_COMPARE (stdc_first_leading_zero ((us) 0), 1); |
| 2229 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero ((us) 0), ui), 1); |
| 2230 | + TEST_COMPARE (stdc_first_leading_zero (0U), 1); |
| 2231 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero (0U), ui), 1); |
| 2232 | + TEST_COMPARE (stdc_first_leading_zero (0UL), 1); |
| 2233 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero (0UL), ui), 1); |
| 2234 | + TEST_COMPARE (stdc_first_leading_zero (0ULL), 1); |
| 2235 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero (0ULL), ui), 1); |
| 2236 | + TEST_COMPARE (stdc_first_leading_zero ((uc) ~0U), 0); |
| 2237 | + TEST_COMPARE (stdc_first_leading_zero ((us) ~0U), 0); |
| 2238 | + TEST_COMPARE (stdc_first_leading_zero (~0U), 0); |
| 2239 | + TEST_COMPARE (stdc_first_leading_zero (~0UL), 0); |
| 2240 | + TEST_COMPARE (stdc_first_leading_zero (~0ULL), 0); |
| 2241 | + TEST_COMPARE (stdc_first_leading_zero ((uc) ~3U), CHAR_BIT - 1); |
| 2242 | + TEST_COMPARE (stdc_first_leading_zero ((us) ~15U), |
| 2243 | + sizeof (short) * CHAR_BIT - 3); |
| 2244 | + TEST_COMPARE (stdc_first_leading_zero (~63U), sizeof (int) * CHAR_BIT - 5); |
| 2245 | + TEST_COMPARE (stdc_first_leading_zero (~255UL), |
| 2246 | + sizeof (long int) * CHAR_BIT - 7); |
| 2247 | + TEST_COMPARE (stdc_first_leading_zero (~1023ULL), |
| 2248 | + sizeof (long long int) * CHAR_BIT - 9); |
| 2249 | + TEST_COMPARE (stdc_first_leading_one ((uc) 0), 0); |
| 2250 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one ((uc) 0), ui), 1); |
| 2251 | + TEST_COMPARE (stdc_first_leading_one ((us) 0), 0); |
| 2252 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one ((us) 0), ui), 1); |
| 2253 | + TEST_COMPARE (stdc_first_leading_one (0U), 0); |
| 2254 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one (0U), ui), 1); |
| 2255 | + TEST_COMPARE (stdc_first_leading_one (0UL), 0); |
| 2256 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one (0UL), ui), 1); |
| 2257 | + TEST_COMPARE (stdc_first_leading_one (0ULL), 0); |
| 2258 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one (0ULL), ui), 1); |
| 2259 | + TEST_COMPARE (stdc_first_leading_one ((uc) ~0U), 1); |
| 2260 | + TEST_COMPARE (stdc_first_leading_one ((us) ~0U), 1); |
| 2261 | + TEST_COMPARE (stdc_first_leading_one (~0U), 1); |
| 2262 | + TEST_COMPARE (stdc_first_leading_one (~0UL), 1); |
| 2263 | + TEST_COMPARE (stdc_first_leading_one (~0ULL), 1); |
| 2264 | + TEST_COMPARE (stdc_first_leading_one ((uc) 3), CHAR_BIT - 1); |
| 2265 | + TEST_COMPARE (stdc_first_leading_one ((us) 9), |
| 2266 | + sizeof (short) * CHAR_BIT - 3); |
| 2267 | + TEST_COMPARE (stdc_first_leading_one (34U), sizeof (int) * CHAR_BIT - 5); |
| 2268 | + TEST_COMPARE (stdc_first_leading_one (130UL), |
| 2269 | + sizeof (long int) * CHAR_BIT - 7); |
| 2270 | + TEST_COMPARE (stdc_first_leading_one (512ULL), |
| 2271 | + sizeof (long long int) * CHAR_BIT - 9); |
| 2272 | + TEST_COMPARE (stdc_first_trailing_zero ((uc) 0), 1); |
| 2273 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero ((uc) 0), ui), 1); |
| 2274 | + TEST_COMPARE (stdc_first_trailing_zero ((us) 0), 1); |
| 2275 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero ((us) 0), ui), 1); |
| 2276 | + TEST_COMPARE (stdc_first_trailing_zero (0U), 1); |
| 2277 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero (0U), ui), 1); |
| 2278 | + TEST_COMPARE (stdc_first_trailing_zero (0UL), 1); |
| 2279 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero (0UL), ui), 1); |
| 2280 | + TEST_COMPARE (stdc_first_trailing_zero (0ULL), 1); |
| 2281 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero (0ULL), ui), 1); |
| 2282 | + TEST_COMPARE (stdc_first_trailing_zero ((uc) ~0U), 0); |
| 2283 | + TEST_COMPARE (stdc_first_trailing_zero ((us) ~0U), 0); |
| 2284 | + TEST_COMPARE (stdc_first_trailing_zero (~0U), 0); |
| 2285 | + TEST_COMPARE (stdc_first_trailing_zero (~0UL), 0); |
| 2286 | + TEST_COMPARE (stdc_first_trailing_zero (~0ULL), 0); |
| 2287 | + TEST_COMPARE (stdc_first_trailing_zero ((uc) 2), 1); |
| 2288 | + TEST_COMPARE (stdc_first_trailing_zero ((us) 15), 5); |
| 2289 | + TEST_COMPARE (stdc_first_trailing_zero (63U), 7); |
| 2290 | + TEST_COMPARE (stdc_first_trailing_zero (128UL), 1); |
| 2291 | + TEST_COMPARE (stdc_first_trailing_zero (511ULL), 10); |
| 2292 | + TEST_COMPARE (stdc_first_trailing_one ((uc) 0), 0); |
| 2293 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one ((uc) 0), ui), 1); |
| 2294 | + TEST_COMPARE (stdc_first_trailing_one ((us) 0), 0); |
| 2295 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one ((us) 0), ui), 1); |
| 2296 | + TEST_COMPARE (stdc_first_trailing_one (0U), 0); |
| 2297 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one (0U), ui), 1); |
| 2298 | + TEST_COMPARE (stdc_first_trailing_one (0UL), 0); |
| 2299 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one (0UL), ui), 1); |
| 2300 | + TEST_COMPARE (stdc_first_trailing_one (0ULL), 0); |
| 2301 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one (0ULL), ui), 1); |
| 2302 | + TEST_COMPARE (stdc_first_trailing_one ((uc) ~0U), 1); |
| 2303 | + TEST_COMPARE (stdc_first_trailing_one ((us) ~0U), 1); |
| 2304 | + TEST_COMPARE (stdc_first_trailing_one (~0U), 1); |
| 2305 | + TEST_COMPARE (stdc_first_trailing_one (~0UL), 1); |
| 2306 | + TEST_COMPARE (stdc_first_trailing_one (~0ULL), 1); |
| 2307 | + TEST_COMPARE (stdc_first_trailing_one ((uc) 4), 3); |
| 2308 | + TEST_COMPARE (stdc_first_trailing_one ((us) 96), 6); |
| 2309 | + TEST_COMPARE (stdc_first_trailing_one (127U), 1); |
| 2310 | + TEST_COMPARE (stdc_first_trailing_one (511UL), 1); |
| 2311 | + TEST_COMPARE (stdc_first_trailing_one (~0ULL << 12), 13); |
| 2312 | + TEST_COMPARE (stdc_count_zeros ((uc) 0), CHAR_BIT); |
| 2313 | + TEST_COMPARE (expr_has_type (stdc_count_zeros ((uc) 0), ui), 1); |
| 2314 | + TEST_COMPARE (stdc_count_zeros ((us) 0), sizeof (short) * CHAR_BIT); |
| 2315 | + TEST_COMPARE (expr_has_type (stdc_count_zeros ((us) 0), ui), 1); |
| 2316 | + TEST_COMPARE (stdc_count_zeros (0U), sizeof (int) * CHAR_BIT); |
| 2317 | + TEST_COMPARE (expr_has_type (stdc_count_zeros (0U), ui), 1); |
| 2318 | + TEST_COMPARE (stdc_count_zeros (0UL), sizeof (long int) * CHAR_BIT); |
| 2319 | + TEST_COMPARE (expr_has_type (stdc_count_zeros (0UL), ui), 1); |
| 2320 | + TEST_COMPARE (stdc_count_zeros (0ULL), sizeof (long long int) * CHAR_BIT); |
| 2321 | + TEST_COMPARE (expr_has_type (stdc_count_zeros (0ULL), ui), 1); |
| 2322 | + TEST_COMPARE (stdc_count_zeros ((uc) ~0U), 0); |
| 2323 | + TEST_COMPARE (stdc_count_zeros ((us) ~0U), 0); |
| 2324 | + TEST_COMPARE (stdc_count_zeros (~0U), 0); |
| 2325 | + TEST_COMPARE (stdc_count_zeros (~0UL), 0); |
| 2326 | + TEST_COMPARE (stdc_count_zeros (~0ULL), 0); |
| 2327 | + TEST_COMPARE (stdc_count_zeros ((uc) 1U), CHAR_BIT - 1); |
| 2328 | + TEST_COMPARE (stdc_count_zeros ((us) 42), sizeof (short) * CHAR_BIT - 3); |
| 2329 | + TEST_COMPARE (stdc_count_zeros (291U), sizeof (int) * CHAR_BIT - 4); |
| 2330 | + TEST_COMPARE (stdc_count_zeros (~1315UL), 5); |
| 2331 | + TEST_COMPARE (stdc_count_zeros (3363ULL), |
| 2332 | + sizeof (long long int) * CHAR_BIT - 6); |
| 2333 | + TEST_COMPARE (stdc_count_ones ((uc) 0), 0); |
| 2334 | + TEST_COMPARE (expr_has_type (stdc_count_ones ((uc) 0), ui), 1); |
| 2335 | + TEST_COMPARE (stdc_count_ones ((us) 0), 0); |
| 2336 | + TEST_COMPARE (expr_has_type (stdc_count_ones ((us) 0), ui), 1); |
| 2337 | + TEST_COMPARE (stdc_count_ones (0U), 0); |
| 2338 | + TEST_COMPARE (expr_has_type (stdc_count_ones (0U), ui), 1); |
| 2339 | + TEST_COMPARE (stdc_count_ones (0UL), 0); |
| 2340 | + TEST_COMPARE (expr_has_type (stdc_count_ones (0UL), ui), 1); |
| 2341 | + TEST_COMPARE (stdc_count_ones (0ULL), 0); |
| 2342 | + TEST_COMPARE (expr_has_type (stdc_count_ones (0ULL), ui), 1); |
| 2343 | + TEST_COMPARE (stdc_count_ones ((uc) ~0U), CHAR_BIT); |
| 2344 | + TEST_COMPARE (stdc_count_ones ((us) ~0U), sizeof (short) * CHAR_BIT); |
| 2345 | + TEST_COMPARE (stdc_count_ones (~0U), sizeof (int) * CHAR_BIT); |
| 2346 | + TEST_COMPARE (stdc_count_ones (~0UL), sizeof (long int) * CHAR_BIT); |
| 2347 | + TEST_COMPARE (stdc_count_ones (~0ULL), sizeof (long long int) * CHAR_BIT); |
| 2348 | + TEST_COMPARE (stdc_count_ones ((uc) ~1U), CHAR_BIT - 1); |
| 2349 | + TEST_COMPARE (stdc_count_ones ((us) ~42), sizeof (short) * CHAR_BIT - 3); |
| 2350 | + TEST_COMPARE (stdc_count_ones (~291U), sizeof (int) * CHAR_BIT - 4); |
| 2351 | + TEST_COMPARE (stdc_count_ones (1315UL), 5); |
| 2352 | + TEST_COMPARE (stdc_count_ones (~3363ULL), |
| 2353 | + sizeof (long long int) * CHAR_BIT - 6); |
| 2354 | + TEST_COMPARE (stdc_has_single_bit ((uc) 0), 0); |
| 2355 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit ((uc) 0), _Bool), 1); |
| 2356 | + TEST_COMPARE (stdc_has_single_bit ((us) 0), 0); |
| 2357 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit ((us) 0), _Bool), 1); |
| 2358 | + TEST_COMPARE (stdc_has_single_bit (0U), 0); |
| 2359 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit (0U), _Bool), 1); |
| 2360 | + TEST_COMPARE (stdc_has_single_bit (0UL), 0); |
| 2361 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit (0UL), _Bool), 1); |
| 2362 | + TEST_COMPARE (stdc_has_single_bit (0ULL), 0); |
| 2363 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit (0ULL), _Bool), 1); |
| 2364 | + TEST_COMPARE (stdc_has_single_bit ((uc) 2), 1); |
| 2365 | + TEST_COMPARE (stdc_has_single_bit ((us) 8), 1); |
| 2366 | + TEST_COMPARE (stdc_has_single_bit (32U), 1); |
| 2367 | + TEST_COMPARE (stdc_has_single_bit (128UL), 1); |
| 2368 | + TEST_COMPARE (stdc_has_single_bit (512ULL), 1); |
| 2369 | + TEST_COMPARE (stdc_has_single_bit ((uc) 7), 0); |
| 2370 | + TEST_COMPARE (stdc_has_single_bit ((us) 96), 0); |
| 2371 | + TEST_COMPARE (stdc_has_single_bit (513U), 0); |
| 2372 | + TEST_COMPARE (stdc_has_single_bit (1022UL), 0); |
| 2373 | + TEST_COMPARE (stdc_has_single_bit (12ULL), 0); |
| 2374 | + TEST_COMPARE (stdc_bit_width ((uc) 0), 0); |
| 2375 | + TEST_COMPARE (expr_has_type (stdc_bit_width ((uc) 0), ui), 1); |
| 2376 | + TEST_COMPARE (stdc_bit_width ((us) 0), 0); |
| 2377 | + TEST_COMPARE (expr_has_type (stdc_bit_width ((us) 0), ui), 1); |
| 2378 | + TEST_COMPARE (stdc_bit_width (0U), 0); |
| 2379 | + TEST_COMPARE (expr_has_type (stdc_bit_width (0U), ui), 1); |
| 2380 | + TEST_COMPARE (stdc_bit_width (0UL), 0); |
| 2381 | + TEST_COMPARE (expr_has_type (stdc_bit_width (0UL), ui), 1); |
| 2382 | + TEST_COMPARE (stdc_bit_width (0ULL), 0); |
| 2383 | + TEST_COMPARE (expr_has_type (stdc_bit_width (0ULL), ui), 1); |
| 2384 | + TEST_COMPARE (stdc_bit_width ((uc) ~0U), CHAR_BIT); |
| 2385 | + TEST_COMPARE (stdc_bit_width ((us) ~0U), sizeof (short) * CHAR_BIT); |
| 2386 | + TEST_COMPARE (stdc_bit_width (~0U), sizeof (int) * CHAR_BIT); |
| 2387 | + TEST_COMPARE (stdc_bit_width (~0UL), sizeof (long int) * CHAR_BIT); |
| 2388 | + TEST_COMPARE (stdc_bit_width (~0ULL), sizeof (long long int) * CHAR_BIT); |
| 2389 | + TEST_COMPARE (stdc_bit_width ((uc) ((uc) ~0U >> 1)), CHAR_BIT - 1); |
| 2390 | + TEST_COMPARE (stdc_bit_width ((uc) 6), 3); |
| 2391 | + TEST_COMPARE (stdc_bit_width ((us) 12U), 4); |
| 2392 | + TEST_COMPARE (stdc_bit_width ((us) ((us) ~0U >> 5)), |
| 2393 | + sizeof (short) * CHAR_BIT - 5); |
| 2394 | + TEST_COMPARE (stdc_bit_width (137U), 8); |
| 2395 | + TEST_COMPARE (stdc_bit_width (269U), 9); |
| 2396 | + TEST_COMPARE (stdc_bit_width (39UL), 6); |
| 2397 | + TEST_COMPARE (stdc_bit_width (~0UL >> 2), sizeof (long int) * CHAR_BIT - 2); |
| 2398 | + TEST_COMPARE (stdc_bit_width (1023ULL), 10); |
| 2399 | + TEST_COMPARE (stdc_bit_width (1024ULL), 11); |
| 2400 | + TEST_COMPARE (stdc_bit_floor ((uc) 0), 0); |
| 2401 | + TEST_COMPARE (expr_has_type (stdc_bit_floor ((uc) 0), uc), 1); |
| 2402 | + TEST_COMPARE (stdc_bit_floor ((us) 0), 0); |
| 2403 | + TEST_COMPARE (expr_has_type (stdc_bit_floor ((us) 0), us), 1); |
| 2404 | + TEST_COMPARE (stdc_bit_floor (0U), 0U); |
| 2405 | + TEST_COMPARE (expr_has_type (stdc_bit_floor (0U), ui), 1); |
| 2406 | + TEST_COMPARE (stdc_bit_floor (0UL), 0UL); |
| 2407 | + TEST_COMPARE (expr_has_type (stdc_bit_floor (0UL), ul), 1); |
| 2408 | + TEST_COMPARE (stdc_bit_floor (0ULL), 0ULL); |
| 2409 | + TEST_COMPARE (expr_has_type (stdc_bit_floor (0ULL), ull), 1); |
| 2410 | + TEST_COMPARE (stdc_bit_floor ((uc) ~0U), (1U << (CHAR_BIT - 1))); |
| 2411 | + TEST_COMPARE (stdc_bit_floor ((us) ~0U), |
| 2412 | + (1U << (sizeof (short) * CHAR_BIT - 1))); |
| 2413 | + TEST_COMPARE (stdc_bit_floor (~0U), (1U << (sizeof (int) * CHAR_BIT - 1))); |
| 2414 | + TEST_COMPARE (stdc_bit_floor (~0UL), |
| 2415 | + (1UL << (sizeof (long int) * CHAR_BIT - 1))); |
| 2416 | + TEST_COMPARE (stdc_bit_floor (~0ULL), |
| 2417 | + (1ULL << (sizeof (long long int) * CHAR_BIT - 1))); |
| 2418 | + TEST_COMPARE (stdc_bit_floor ((uc) 4), 4); |
| 2419 | + TEST_COMPARE (stdc_bit_floor ((uc) 7), 4); |
| 2420 | + TEST_COMPARE (stdc_bit_floor ((us) 8U), 8); |
| 2421 | + TEST_COMPARE (stdc_bit_floor ((us) 31U), 16); |
| 2422 | + TEST_COMPARE (stdc_bit_floor (137U), 128U); |
| 2423 | + TEST_COMPARE (stdc_bit_floor (269U), 256U); |
| 2424 | + TEST_COMPARE (stdc_bit_floor (511UL), 256UL); |
| 2425 | + TEST_COMPARE (stdc_bit_floor (512UL), 512UL); |
| 2426 | + TEST_COMPARE (stdc_bit_floor (513UL), 512ULL); |
| 2427 | + TEST_COMPARE (stdc_bit_floor (1024ULL), 1024ULL); |
| 2428 | + TEST_COMPARE (stdc_bit_ceil ((uc) 0), 1); |
| 2429 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil ((uc) 0), uc), 1); |
| 2430 | + TEST_COMPARE (stdc_bit_ceil ((us) 0), 1); |
| 2431 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil ((us) 0), us), 1); |
| 2432 | + TEST_COMPARE (stdc_bit_ceil (0U), 1U); |
| 2433 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil (0U), ui), 1); |
| 2434 | + TEST_COMPARE (stdc_bit_ceil (0UL), 1UL); |
| 2435 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil (0UL), ul), 1); |
| 2436 | + TEST_COMPARE (stdc_bit_ceil (0ULL), 1ULL); |
| 2437 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil (0ULL), ull), 1); |
| 2438 | + TEST_COMPARE (stdc_bit_ceil ((uc) ~0U), 0); |
| 2439 | + TEST_COMPARE (stdc_bit_ceil ((us) ~0U), 0); |
| 2440 | + TEST_COMPARE (stdc_bit_ceil (~0U), 0U); |
| 2441 | + TEST_COMPARE (stdc_bit_ceil (~0UL), 0UL); |
| 2442 | + TEST_COMPARE (stdc_bit_ceil (~0ULL), 0ULL); |
| 2443 | + TEST_COMPARE (stdc_bit_ceil ((uc) ((uc) ~0U >> 1)), (1U << (CHAR_BIT - 1))); |
| 2444 | + TEST_COMPARE (stdc_bit_ceil ((uc) ((uc) ~0U >> 1)), (1U << (CHAR_BIT - 1))); |
| 2445 | + TEST_COMPARE (stdc_bit_ceil ((us) ((us) ~0U >> 1)), |
| 2446 | + (1U << (sizeof (short) * CHAR_BIT - 1))); |
| 2447 | + TEST_COMPARE (stdc_bit_ceil ((us) ((us) ~0U >> 1)), |
| 2448 | + (1U << (sizeof (short) * CHAR_BIT - 1))); |
| 2449 | + TEST_COMPARE (stdc_bit_ceil (~0U >> 1), |
| 2450 | + (1U << (sizeof (int) * CHAR_BIT - 1))); |
| 2451 | + TEST_COMPARE (stdc_bit_ceil (1U << (sizeof (int) * CHAR_BIT - 1)), |
| 2452 | + (1U << (sizeof (int) * CHAR_BIT - 1))); |
| 2453 | + TEST_COMPARE (stdc_bit_ceil (~0UL >> 1), |
| 2454 | + (1UL << (sizeof (long int) * CHAR_BIT - 1))); |
| 2455 | + TEST_COMPARE (stdc_bit_ceil (~0UL >> 1), |
| 2456 | + (1UL << (sizeof (long int) * CHAR_BIT - 1))); |
| 2457 | + TEST_COMPARE (stdc_bit_ceil (1ULL |
| 2458 | + << (sizeof (long long int) * CHAR_BIT - 1)), |
| 2459 | + (1ULL << (sizeof (long long int) * CHAR_BIT - 1))); |
| 2460 | + TEST_COMPARE (stdc_bit_ceil (~0ULL >> 1), |
| 2461 | + (1ULL << (sizeof (long long int) * CHAR_BIT - 1))); |
| 2462 | + TEST_COMPARE (stdc_bit_ceil ((uc) 1), 1); |
| 2463 | + TEST_COMPARE (stdc_bit_ceil ((uc) 2), 2); |
| 2464 | + TEST_COMPARE (stdc_bit_ceil ((us) 3U), 4); |
| 2465 | + TEST_COMPARE (stdc_bit_ceil ((us) 4U), 4); |
| 2466 | + TEST_COMPARE (stdc_bit_ceil (5U), 8U); |
| 2467 | + TEST_COMPARE (stdc_bit_ceil (269U), 512U); |
| 2468 | + TEST_COMPARE (stdc_bit_ceil (511UL), 512UL); |
| 2469 | + TEST_COMPARE (stdc_bit_ceil (512UL), 512UL); |
| 2470 | + TEST_COMPARE (stdc_bit_ceil (513ULL), 1024ULL); |
| 2471 | + TEST_COMPARE (stdc_bit_ceil (1025ULL), 2048ULL); |
| 2472 | +# ifdef __SIZEOF_INT128__ |
| 2473 | + TEST_COMPARE (stdc_leading_zeros ((unsigned __int128) 0), |
| 2474 | + sizeof (__int128) * CHAR_BIT); |
| 2475 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros ((unsigned __int128) 0), ui), |
| 2476 | + 1); |
| 2477 | + TEST_COMPARE (stdc_leading_zeros (~(unsigned __int128) 0), 0); |
| 2478 | + TEST_COMPARE (stdc_leading_ones ((unsigned __int128) 0), 0); |
| 2479 | + TEST_COMPARE (expr_has_type (stdc_leading_ones ((unsigned __int128) 0), ui), |
| 2480 | + 1); |
| 2481 | + TEST_COMPARE (stdc_leading_ones (~(unsigned __int128) 0), |
| 2482 | + sizeof (__int128) * CHAR_BIT); |
| 2483 | + TEST_COMPARE (stdc_trailing_zeros ((unsigned __int128) 0), |
| 2484 | + sizeof (__int128) * CHAR_BIT); |
| 2485 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros ((unsigned __int128) 0), |
| 2486 | + ui), 1); |
| 2487 | + TEST_COMPARE (stdc_trailing_zeros (~(unsigned __int128) 0), 0); |
| 2488 | + TEST_COMPARE (stdc_trailing_ones ((unsigned __int128) 0), 0); |
| 2489 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones ((unsigned __int128) 0), ui), |
| 2490 | + 1); |
| 2491 | + TEST_COMPARE (stdc_trailing_ones (~(unsigned __int128) 0), |
| 2492 | + sizeof (__int128) * CHAR_BIT); |
| 2493 | + TEST_COMPARE (stdc_first_leading_zero ((unsigned __int128) 0), 1); |
| 2494 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero ((unsigned __int128) 0), |
| 2495 | + ui), 1); |
| 2496 | + TEST_COMPARE (stdc_first_leading_zero (~(unsigned __int128) 0), 0); |
| 2497 | + TEST_COMPARE (stdc_first_leading_one ((unsigned __int128) 0), 0); |
| 2498 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one ((unsigned __int128) 0), |
| 2499 | + ui), 1); |
| 2500 | + TEST_COMPARE (stdc_first_leading_one (~(unsigned __int128) 0), 1); |
| 2501 | + TEST_COMPARE (stdc_first_trailing_zero ((unsigned __int128) 0), 1); |
| 2502 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero ((unsigned __int128) |
| 2503 | + 0), ui), 1); |
| 2504 | + TEST_COMPARE (stdc_first_trailing_zero (~(unsigned __int128) 0), 0); |
| 2505 | + TEST_COMPARE (stdc_first_trailing_one ((unsigned __int128) 0), 0); |
| 2506 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one ((unsigned __int128) 0), |
| 2507 | + ui), 1); |
| 2508 | + TEST_COMPARE (stdc_first_trailing_one (~(unsigned __int128) 0), 1); |
| 2509 | + TEST_COMPARE (stdc_count_zeros ((unsigned __int128) 0), |
| 2510 | + sizeof (__int128) * CHAR_BIT); |
| 2511 | + TEST_COMPARE (expr_has_type (stdc_count_zeros ((unsigned __int128) 0), ui), |
| 2512 | + 1); |
| 2513 | + TEST_COMPARE (stdc_count_zeros (~(unsigned __int128) 0), 0); |
| 2514 | + TEST_COMPARE (stdc_count_ones ((unsigned __int128) 0), 0); |
| 2515 | + TEST_COMPARE (expr_has_type (stdc_count_ones ((unsigned __int128) 0), ui), |
| 2516 | + 1); |
| 2517 | + TEST_COMPARE (stdc_count_ones (~(unsigned __int128) 0), |
| 2518 | + sizeof (__int128) * CHAR_BIT); |
| 2519 | + TEST_COMPARE (stdc_has_single_bit ((unsigned __int128) 0), 0); |
| 2520 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit ((unsigned __int128) 0), |
| 2521 | + _Bool), 1); |
| 2522 | + TEST_COMPARE (stdc_has_single_bit (~(unsigned __int128) 0), 0); |
| 2523 | + TEST_COMPARE (stdc_bit_width ((unsigned __int128) 0), 0); |
| 2524 | + TEST_COMPARE (expr_has_type (stdc_bit_width ((unsigned __int128) 0), ui), 1); |
| 2525 | + TEST_COMPARE (stdc_bit_width (~(unsigned __int128) 0), |
| 2526 | + sizeof (__int128) * CHAR_BIT); |
| 2527 | + TEST_COMPARE (stdc_bit_floor ((unsigned __int128) 0) != 0, 0); |
| 2528 | + TEST_COMPARE (expr_has_type (stdc_bit_floor ((unsigned __int128) 0), |
| 2529 | + unsigned __int128), 1); |
| 2530 | + TEST_COMPARE (stdc_bit_floor (~(unsigned __int128) 0) |
| 2531 | + != ((unsigned __int128) 1) << (sizeof (__int128) |
| 2532 | + * CHAR_BIT - 1), 0); |
| 2533 | + TEST_COMPARE (stdc_bit_ceil ((unsigned __int128) 0) != 1, 0); |
| 2534 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil ((unsigned __int128) 0), |
| 2535 | + unsigned __int128), 1); |
| 2536 | + TEST_COMPARE (stdc_bit_ceil ((unsigned __int128) 1) != 1, 0); |
| 2537 | + TEST_COMPARE (stdc_bit_ceil ((~(unsigned __int128) 0) >> 1) |
| 2538 | + != ((unsigned __int128) 1) << (sizeof (__int128) |
| 2539 | + * CHAR_BIT - 1), 0); |
| 2540 | + TEST_COMPARE (stdc_bit_ceil (~(unsigned __int128) 0) != 0, 0); |
| 2541 | +# endif |
| 2542 | + uc a = 0; |
| 2543 | + TEST_COMPARE (stdc_bit_width (a++), 0); |
| 2544 | + TEST_COMPARE (a, 1); |
| 2545 | + ull b = 0; |
| 2546 | + TEST_COMPARE (stdc_bit_width (b++), 0); |
| 2547 | + TEST_COMPARE (b, 1); |
| 2548 | + TEST_COMPARE (stdc_bit_floor (a++), 1); |
| 2549 | + TEST_COMPARE (a, 2); |
| 2550 | + TEST_COMPARE (stdc_bit_floor (b++), 1); |
| 2551 | + TEST_COMPARE (b, 2); |
| 2552 | + TEST_COMPARE (stdc_bit_ceil (a++), 2); |
| 2553 | + TEST_COMPARE (a, 3); |
| 2554 | + TEST_COMPARE (stdc_bit_ceil (b++), 2); |
| 2555 | + TEST_COMPARE (b, 3); |
| 2556 | + TEST_COMPARE (stdc_leading_zeros (a++), CHAR_BIT - 2); |
| 2557 | + TEST_COMPARE (a, 4); |
| 2558 | + TEST_COMPARE (stdc_leading_zeros (b++), |
| 2559 | + sizeof (long long int) * CHAR_BIT - 2); |
| 2560 | + TEST_COMPARE (b, 4); |
| 2561 | + TEST_COMPARE (stdc_leading_ones (a++), 0); |
| 2562 | + TEST_COMPARE (a, 5); |
| 2563 | + TEST_COMPARE (stdc_leading_ones (b++), 0); |
| 2564 | + TEST_COMPARE (b, 5); |
| 2565 | + TEST_COMPARE (stdc_trailing_zeros (a++), 0); |
| 2566 | + TEST_COMPARE (a, 6); |
| 2567 | + TEST_COMPARE (stdc_trailing_zeros (b++), 0); |
| 2568 | + TEST_COMPARE (b, 6); |
| 2569 | + TEST_COMPARE (stdc_trailing_ones (a++), 0); |
| 2570 | + TEST_COMPARE (a, 7); |
| 2571 | + TEST_COMPARE (stdc_trailing_ones (b++), 0); |
| 2572 | + TEST_COMPARE (b, 7); |
| 2573 | + TEST_COMPARE (stdc_first_leading_zero (a++), 1); |
| 2574 | + TEST_COMPARE (a, 8); |
| 2575 | + TEST_COMPARE (stdc_first_leading_zero (b++), 1); |
| 2576 | + TEST_COMPARE (b, 8); |
| 2577 | + TEST_COMPARE (stdc_first_leading_one (a++), CHAR_BIT - 3); |
| 2578 | + TEST_COMPARE (a, 9); |
| 2579 | + TEST_COMPARE (stdc_first_leading_one (b++), |
| 2580 | + sizeof (long long int) * CHAR_BIT - 3); |
| 2581 | + TEST_COMPARE (b, 9); |
| 2582 | + TEST_COMPARE (stdc_first_trailing_zero (a++), 2); |
| 2583 | + TEST_COMPARE (a, 10); |
| 2584 | + TEST_COMPARE (stdc_first_trailing_zero (b++), 2); |
| 2585 | + TEST_COMPARE (b, 10); |
| 2586 | + TEST_COMPARE (stdc_first_trailing_one (a++), 2); |
| 2587 | + TEST_COMPARE (a, 11); |
| 2588 | + TEST_COMPARE (stdc_first_trailing_one (b++), 2); |
| 2589 | + TEST_COMPARE (b, 11); |
| 2590 | + TEST_COMPARE (stdc_count_zeros (a++), CHAR_BIT - 3); |
| 2591 | + TEST_COMPARE (a, 12); |
| 2592 | + TEST_COMPARE (stdc_count_zeros (b++), |
| 2593 | + sizeof (long long int) * CHAR_BIT - 3); |
| 2594 | + TEST_COMPARE (b, 12); |
| 2595 | + TEST_COMPARE (stdc_count_ones (a++), 2); |
| 2596 | + TEST_COMPARE (a, 13); |
| 2597 | + TEST_COMPARE (stdc_count_ones (b++), 2); |
| 2598 | + TEST_COMPARE (b, 13); |
| 2599 | + TEST_COMPARE (stdc_has_single_bit (a++), 0); |
| 2600 | + TEST_COMPARE (a, 14); |
| 2601 | + TEST_COMPARE (stdc_has_single_bit (b++), 0); |
| 2602 | + TEST_COMPARE (b, 14); |
| 2603 | +# ifdef BITINT_MAXWIDTH |
| 2604 | +# if BITINT_MAXWIDTH >= 64 |
| 2605 | + TEST_COMPARE (stdc_leading_zeros (0uwb), 1); |
| 2606 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros (0uwb), ui), 1); |
| 2607 | + TEST_COMPARE (stdc_leading_zeros (1uwb), 0); |
| 2608 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros (1uwb), ui), 1); |
| 2609 | + TEST_COMPARE (stdc_leading_ones (0uwb), 0); |
| 2610 | + TEST_COMPARE (expr_has_type (stdc_leading_ones (0uwb), ui), 1); |
| 2611 | + TEST_COMPARE (stdc_leading_ones (1uwb), 1); |
| 2612 | + TEST_COMPARE (expr_has_type (stdc_leading_ones (1uwb), ui), 1); |
| 2613 | + TEST_COMPARE (stdc_trailing_zeros (0uwb), 1); |
| 2614 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros (0uwb), ui), 1); |
| 2615 | + TEST_COMPARE (stdc_trailing_zeros (1uwb), 0); |
| 2616 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros (1uwb), ui), 1); |
| 2617 | + TEST_COMPARE (stdc_trailing_ones (0uwb), 0); |
| 2618 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones (0uwb), ui), 1); |
| 2619 | + TEST_COMPARE (stdc_trailing_ones (1uwb), 1); |
| 2620 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones (1uwb), ui), 1); |
| 2621 | + TEST_COMPARE (stdc_first_leading_zero (0uwb), 1); |
| 2622 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero (0uwb), ui), 1); |
| 2623 | + TEST_COMPARE (stdc_first_leading_zero (1uwb), 0); |
| 2624 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero (1uwb), ui), 1); |
| 2625 | + TEST_COMPARE (stdc_first_leading_one (0uwb), 0); |
| 2626 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one (0uwb), ui), 1); |
| 2627 | + TEST_COMPARE (stdc_first_leading_one (1uwb), 1); |
| 2628 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one (1uwb), ui), 1); |
| 2629 | + TEST_COMPARE (stdc_first_trailing_zero (0uwb), 1); |
| 2630 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero (0uwb), ui), 1); |
| 2631 | + TEST_COMPARE (stdc_first_trailing_zero (1uwb), 0); |
| 2632 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero (1uwb), ui), 1); |
| 2633 | + TEST_COMPARE (stdc_first_trailing_one (0uwb), 0); |
| 2634 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one (0uwb), ui), 1); |
| 2635 | + TEST_COMPARE (stdc_first_trailing_one (1uwb), 1); |
| 2636 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one (1uwb), ui), 1); |
| 2637 | + TEST_COMPARE (stdc_count_zeros (0uwb), 1); |
| 2638 | + TEST_COMPARE (expr_has_type (stdc_count_zeros (0uwb), ui), 1); |
| 2639 | + TEST_COMPARE (stdc_count_zeros (1uwb), 0); |
| 2640 | + TEST_COMPARE (expr_has_type (stdc_count_zeros (1uwb), ui), 1); |
| 2641 | + TEST_COMPARE (stdc_count_ones (0uwb), 0); |
| 2642 | + TEST_COMPARE (expr_has_type (stdc_count_ones (0uwb), ui), 1); |
| 2643 | + TEST_COMPARE (stdc_count_ones (1uwb), 1); |
| 2644 | + TEST_COMPARE (expr_has_type (stdc_count_ones (1uwb), ui), 1); |
| 2645 | + TEST_COMPARE (stdc_has_single_bit (0uwb), 0); |
| 2646 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit (0uwb), _Bool), 1); |
| 2647 | + TEST_COMPARE (stdc_has_single_bit (1uwb), 1); |
| 2648 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit (1uwb), _Bool), 1); |
| 2649 | + TEST_COMPARE (stdc_bit_width (0uwb), 0); |
| 2650 | + TEST_COMPARE (expr_has_type (stdc_bit_width (0uwb), ui), 1); |
| 2651 | + TEST_COMPARE (stdc_bit_width (1uwb), 1); |
| 2652 | + TEST_COMPARE (expr_has_type (stdc_bit_width (1uwb), ui), 1); |
| 2653 | + TEST_COMPARE (stdc_bit_floor (0uwb), 0); |
| 2654 | + TEST_COMPARE (expr_has_type (stdc_bit_floor (0uwb), unsigned _BitInt(1)), 1); |
| 2655 | + TEST_COMPARE (stdc_bit_floor (1uwb), 1); |
| 2656 | + TEST_COMPARE (expr_has_type (stdc_bit_floor (1uwb), unsigned _BitInt(1)), 1); |
| 2657 | + TEST_COMPARE (stdc_bit_ceil (0uwb), 1); |
| 2658 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil (0uwb), unsigned _BitInt(1)), 1); |
| 2659 | + TEST_COMPARE (stdc_bit_ceil (1uwb), 1); |
| 2660 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil (1uwb), unsigned _BitInt(1)), 1); |
| 2661 | + unsigned _BitInt(1) c = 0; |
| 2662 | + TEST_COMPARE (stdc_bit_floor (c++), 0); |
| 2663 | + TEST_COMPARE (c, 1); |
| 2664 | + TEST_COMPARE (stdc_bit_floor (c++), 1); |
| 2665 | + TEST_COMPARE (c, 0); |
| 2666 | + TEST_COMPARE (stdc_bit_ceil (c++), 1); |
| 2667 | + TEST_COMPARE (c, 1); |
| 2668 | + TEST_COMPARE (stdc_bit_ceil (c++), 1); |
| 2669 | + TEST_COMPARE (c, 0); |
| 2670 | +# endif |
| 2671 | +# if BITINT_MAXWIDTH >= 512 |
| 2672 | + TEST_COMPARE (stdc_leading_zeros ((unsigned _BitInt(512)) 0), 512); |
| 2673 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros ((unsigned _BitInt(512)) 0), |
| 2674 | + ui), 1); |
| 2675 | + TEST_COMPARE (stdc_leading_zeros ((unsigned _BitInt(373)) 0), 373); |
| 2676 | + TEST_COMPARE (expr_has_type (stdc_leading_zeros ((unsigned _BitInt(373)) 0), |
| 2677 | + ui), 1); |
| 2678 | + TEST_COMPARE (stdc_leading_zeros (~(unsigned _BitInt(512)) 0), 0); |
| 2679 | + TEST_COMPARE (stdc_leading_zeros (~(unsigned _BitInt(373)) 0), 0); |
| 2680 | + TEST_COMPARE (stdc_leading_zeros ((unsigned _BitInt(512)) 275), 512 - 9); |
| 2681 | + TEST_COMPARE (stdc_leading_zeros ((unsigned _BitInt(373)) 512), 373 - 10); |
| 2682 | + TEST_COMPARE (stdc_leading_ones ((unsigned _BitInt(512)) 0), 0); |
| 2683 | + TEST_COMPARE (expr_has_type (stdc_leading_ones ((unsigned _BitInt(512)) 0), |
| 2684 | + ui), 1); |
| 2685 | + TEST_COMPARE (stdc_leading_ones ((unsigned _BitInt(373)) 0), 0); |
| 2686 | + TEST_COMPARE (expr_has_type (stdc_leading_ones ((unsigned _BitInt(373)) 0), |
| 2687 | + ui), 1); |
| 2688 | + TEST_COMPARE (stdc_leading_ones (~(unsigned _BitInt(512)) 0), 512); |
| 2689 | + TEST_COMPARE (stdc_leading_ones (~(unsigned _BitInt(373)) 0), 373); |
| 2690 | + TEST_COMPARE (stdc_leading_ones (~(unsigned _BitInt(512)) 275), 512 - 9); |
| 2691 | + TEST_COMPARE (stdc_leading_ones (~(unsigned _BitInt(373)) 512), 373 - 10); |
| 2692 | + TEST_COMPARE (stdc_trailing_zeros ((unsigned _BitInt(512)) 0), 512); |
| 2693 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros ((unsigned _BitInt(512)) 0), |
| 2694 | + ui), 1); |
| 2695 | + TEST_COMPARE (stdc_trailing_zeros ((unsigned _BitInt(373)) 0), 373); |
| 2696 | + TEST_COMPARE (expr_has_type (stdc_trailing_zeros ((unsigned _BitInt(373)) 0), |
| 2697 | + ui), 1); |
| 2698 | + TEST_COMPARE (stdc_trailing_zeros (~(unsigned _BitInt(512)) 0), 0); |
| 2699 | + TEST_COMPARE (stdc_trailing_zeros (~(unsigned _BitInt(373)) 0), 0); |
| 2700 | + TEST_COMPARE (stdc_trailing_zeros ((unsigned _BitInt(512)) 256), 8); |
| 2701 | + TEST_COMPARE (stdc_trailing_zeros ((unsigned _BitInt(373)) 512), 9); |
| 2702 | + TEST_COMPARE (stdc_trailing_ones ((unsigned _BitInt(512)) 0), 0); |
| 2703 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones ((unsigned _BitInt(512)) 0), |
| 2704 | + ui), 1); |
| 2705 | + TEST_COMPARE (stdc_trailing_ones ((unsigned _BitInt(373)) 0), 0); |
| 2706 | + TEST_COMPARE (expr_has_type (stdc_trailing_ones ((unsigned _BitInt(373)) 0), |
| 2707 | + ui), 1); |
| 2708 | + TEST_COMPARE (stdc_trailing_ones (~(unsigned _BitInt(512)) 0), 512); |
| 2709 | + TEST_COMPARE (stdc_trailing_ones (~(unsigned _BitInt(373)) 0), 373); |
| 2710 | + TEST_COMPARE (stdc_trailing_ones ((unsigned _BitInt(512)) 255), 8); |
| 2711 | + TEST_COMPARE (stdc_trailing_ones ((~(unsigned _BitInt(373)) 0) >> 2), |
| 2712 | + 373 - 2); |
| 2713 | + TEST_COMPARE (stdc_first_leading_zero ((unsigned _BitInt(512)) 0), 1); |
| 2714 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero ((unsigned _BitInt(512)) |
| 2715 | + 0), ui), 1); |
| 2716 | + TEST_COMPARE (stdc_first_leading_zero ((unsigned _BitInt(373)) 0), 1); |
| 2717 | + TEST_COMPARE (expr_has_type (stdc_first_leading_zero ((unsigned _BitInt(373)) |
| 2718 | + 0), ui), 1); |
| 2719 | + TEST_COMPARE (stdc_first_leading_zero (~(unsigned _BitInt(512)) 0), 0); |
| 2720 | + TEST_COMPARE (stdc_first_leading_zero (~(unsigned _BitInt(373)) 0), 0); |
| 2721 | + TEST_COMPARE (stdc_first_leading_zero (~(unsigned _BitInt(512)) 511), |
| 2722 | + 512 - 8); |
| 2723 | + TEST_COMPARE (stdc_first_leading_zero (~(unsigned _BitInt(373)) 1023), |
| 2724 | + 373 - 9); |
| 2725 | + TEST_COMPARE (stdc_first_leading_one ((unsigned _BitInt(512)) 0), 0); |
| 2726 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one ((unsigned _BitInt(512)) |
| 2727 | + 0), ui), 1); |
| 2728 | + TEST_COMPARE (stdc_first_leading_one ((unsigned _BitInt(373)) 0), 0); |
| 2729 | + TEST_COMPARE (expr_has_type (stdc_first_leading_one ((unsigned _BitInt(373)) |
| 2730 | + 0), ui), 1); |
| 2731 | + TEST_COMPARE (stdc_first_leading_one (~(unsigned _BitInt(512)) 0), 1); |
| 2732 | + TEST_COMPARE (stdc_first_leading_one (~(unsigned _BitInt(373)) 0), 1); |
| 2733 | + TEST_COMPARE (stdc_first_leading_one ((unsigned _BitInt(512)) 275), 512 - 8); |
| 2734 | + TEST_COMPARE (stdc_first_leading_one ((unsigned _BitInt(373)) 512), 373 - 9); |
| 2735 | + TEST_COMPARE (stdc_first_trailing_zero ((unsigned _BitInt(512)) 0), 1); |
| 2736 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero ((unsigned |
| 2737 | + _BitInt(512)) 0), |
| 2738 | + ui), 1); |
| 2739 | + TEST_COMPARE (stdc_first_trailing_zero ((unsigned _BitInt(373)) 0), 1); |
| 2740 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_zero ((unsigned |
| 2741 | + _BitInt(373)) 0), |
| 2742 | + ui), 1); |
| 2743 | + TEST_COMPARE (stdc_first_trailing_zero (~(unsigned _BitInt(512)) 0), 0); |
| 2744 | + TEST_COMPARE (stdc_first_trailing_zero (~(unsigned _BitInt(373)) 0), 0); |
| 2745 | + TEST_COMPARE (stdc_first_trailing_zero ((unsigned _BitInt(512)) 255), 9); |
| 2746 | + TEST_COMPARE (stdc_first_trailing_zero ((unsigned _BitInt(373)) 511), 10); |
| 2747 | + TEST_COMPARE (stdc_first_trailing_one ((unsigned _BitInt(512)) 0), 0); |
| 2748 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one ((unsigned _BitInt(512)) |
| 2749 | + 0), ui), 1); |
| 2750 | + TEST_COMPARE (stdc_first_trailing_one ((unsigned _BitInt(373)) 0), 0); |
| 2751 | + TEST_COMPARE (expr_has_type (stdc_first_trailing_one ((unsigned _BitInt(373)) |
| 2752 | + 0), ui), 1); |
| 2753 | + TEST_COMPARE (stdc_first_trailing_one (~(unsigned _BitInt(512)) 0), 1); |
| 2754 | + TEST_COMPARE (stdc_first_trailing_one (~(unsigned _BitInt(373)) 0), 1); |
| 2755 | + TEST_COMPARE (stdc_first_trailing_one (((unsigned _BitInt(512)) 255) << 175), |
| 2756 | + 176); |
| 2757 | + TEST_COMPARE (stdc_first_trailing_one ((~(unsigned _BitInt(373)) 0) << 311), |
| 2758 | + 312); |
| 2759 | + TEST_COMPARE (stdc_count_zeros ((unsigned _BitInt(512)) 0), 512); |
| 2760 | + TEST_COMPARE (expr_has_type (stdc_count_zeros ((unsigned _BitInt(512)) 0), |
| 2761 | + ui), 1); |
| 2762 | + TEST_COMPARE (stdc_count_zeros ((unsigned _BitInt(373)) 0), 373); |
| 2763 | + TEST_COMPARE (expr_has_type (stdc_count_zeros ((unsigned _BitInt(373)) 0), |
| 2764 | + ui), 1); |
| 2765 | + TEST_COMPARE (stdc_count_zeros (~(unsigned _BitInt(512)) 0), 0); |
| 2766 | + TEST_COMPARE (stdc_count_zeros (~(unsigned _BitInt(373)) 0), 0); |
| 2767 | + TEST_COMPARE (stdc_count_zeros ((unsigned _BitInt(512)) 1315), 512 - 5); |
| 2768 | + TEST_COMPARE (stdc_count_zeros ((unsigned _BitInt(373)) 3363), 373 - 6); |
| 2769 | + TEST_COMPARE (stdc_count_ones ((unsigned _BitInt(512)) 0), 0); |
| 2770 | + TEST_COMPARE (expr_has_type (stdc_count_ones ((unsigned _BitInt(512)) 0), |
| 2771 | + ui), 1); |
| 2772 | + TEST_COMPARE (stdc_count_ones ((unsigned _BitInt(373)) 0), 0); |
| 2773 | + TEST_COMPARE (expr_has_type (stdc_count_ones ((unsigned _BitInt(373)) 0), |
| 2774 | + ui), 1); |
| 2775 | + TEST_COMPARE (stdc_count_ones (~(unsigned _BitInt(512)) 0), 512); |
| 2776 | + TEST_COMPARE (stdc_count_ones (~(unsigned _BitInt(373)) 0), 373); |
| 2777 | + TEST_COMPARE (stdc_count_ones (~(unsigned _BitInt(512)) 1315), 512 - 5); |
| 2778 | + TEST_COMPARE (stdc_count_ones (~(unsigned _BitInt(373)) 3363), 373 - 6); |
| 2779 | + TEST_COMPARE (stdc_has_single_bit ((unsigned _BitInt(512)) 0), 0); |
| 2780 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit ((unsigned _BitInt(512)) 0), |
| 2781 | + _Bool), 1); |
| 2782 | + TEST_COMPARE (stdc_has_single_bit ((unsigned _BitInt(373)) 0), 0); |
| 2783 | + TEST_COMPARE (expr_has_type (stdc_has_single_bit ((unsigned _BitInt(373)) 0), |
| 2784 | + _Bool), 1); |
| 2785 | + TEST_COMPARE (stdc_has_single_bit (~(unsigned _BitInt(512)) 0), 0); |
| 2786 | + TEST_COMPARE (stdc_has_single_bit (~(unsigned _BitInt(373)) 0), 0); |
| 2787 | + TEST_COMPARE (stdc_has_single_bit (((unsigned _BitInt(512)) 1022) << 279), |
| 2788 | + 0); |
| 2789 | + TEST_COMPARE (stdc_has_single_bit (((unsigned _BitInt(373)) 12) << 305), 0); |
| 2790 | + TEST_COMPARE (stdc_bit_width ((unsigned _BitInt(512)) 0), 0); |
| 2791 | + TEST_COMPARE (expr_has_type (stdc_bit_width ((unsigned _BitInt(512)) 0), |
| 2792 | + ui), 1); |
| 2793 | + TEST_COMPARE (stdc_bit_width ((unsigned _BitInt(373)) 0), 0); |
| 2794 | + TEST_COMPARE (expr_has_type (stdc_bit_width ((unsigned _BitInt(373)) 0), |
| 2795 | + ui), 1); |
| 2796 | + TEST_COMPARE (stdc_bit_width (~(unsigned _BitInt(512)) 0), 512); |
| 2797 | + TEST_COMPARE (stdc_bit_width (~(unsigned _BitInt(373)) 0), 373); |
| 2798 | + TEST_COMPARE (stdc_bit_width (((unsigned _BitInt(512)) 1023) << 405), |
| 2799 | + 405 + 10); |
| 2800 | + TEST_COMPARE (stdc_bit_width (((unsigned _BitInt(373)) 1024) << 242), |
| 2801 | + 242 + 11); |
| 2802 | + TEST_COMPARE (stdc_bit_floor ((unsigned _BitInt(512)) 0) != 0, 0); |
| 2803 | + TEST_COMPARE (expr_has_type (stdc_bit_floor ((unsigned _BitInt(512)) 0), |
| 2804 | + unsigned _BitInt(512)), 1); |
| 2805 | + TEST_COMPARE (stdc_bit_floor ((unsigned _BitInt(373)) 0) != 0, 0); |
| 2806 | + TEST_COMPARE (expr_has_type (stdc_bit_floor ((unsigned _BitInt(373)) 0), |
| 2807 | + unsigned _BitInt(373)), 1); |
| 2808 | + TEST_COMPARE (stdc_bit_floor (~(unsigned _BitInt(512)) 0) |
| 2809 | + != ((unsigned _BitInt(512)) 1) << (512 - 1), 0); |
| 2810 | + TEST_COMPARE (stdc_bit_floor (~(unsigned _BitInt(373)) 0) |
| 2811 | + != ((unsigned _BitInt(373)) 1) << (373 - 1), 0); |
| 2812 | + TEST_COMPARE (stdc_bit_floor (((unsigned _BitInt(512)) 511) << 405) |
| 2813 | + != (((unsigned _BitInt(512)) 256) << 405), 0); |
| 2814 | + TEST_COMPARE (stdc_bit_floor (((unsigned _BitInt(373)) 512) << 242) |
| 2815 | + != (((unsigned _BitInt(512)) 512) << 242), 0); |
| 2816 | + TEST_COMPARE (stdc_bit_ceil ((unsigned _BitInt(512)) 0) != 1, 0); |
| 2817 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil ((unsigned _BitInt(512)) 0), |
| 2818 | + unsigned _BitInt(512)), 1); |
| 2819 | + TEST_COMPARE (stdc_bit_ceil ((unsigned _BitInt(373)) 0) != 1, 0); |
| 2820 | + TEST_COMPARE (expr_has_type (stdc_bit_ceil ((unsigned _BitInt(373)) 0), |
| 2821 | + unsigned _BitInt(373)), 1); |
| 2822 | + TEST_COMPARE (stdc_bit_ceil (~(unsigned _BitInt(512)) 0) != 0, 0); |
| 2823 | + TEST_COMPARE (stdc_bit_ceil (~(unsigned _BitInt(373)) 0) != 0, 0); |
| 2824 | + TEST_COMPARE (stdc_bit_ceil (((unsigned _BitInt(512)) 1) << (512 - 1)) |
| 2825 | + != ((unsigned _BitInt(512)) 1) << (512 - 1), 0); |
| 2826 | + TEST_COMPARE (stdc_bit_ceil ((~(unsigned _BitInt(373)) 0) >> 1) |
| 2827 | + != ((unsigned _BitInt(373)) 1) << (373 - 1), 0); |
| 2828 | + TEST_COMPARE (stdc_bit_ceil (((unsigned _BitInt(512)) 512) << 405) |
| 2829 | + != (((unsigned _BitInt(512)) 512) << 405), 0); |
| 2830 | + TEST_COMPARE (stdc_bit_ceil (((unsigned _BitInt(373)) 513) << 242) |
| 2831 | + != (((unsigned _BitInt(512)) 1024) << 242), 0); |
| 2832 | + TEST_COMPARE (stdc_bit_floor ((unsigned _BitInt(BITINT_MAXWIDTH)) 0) != 0, |
| 2833 | + 0); |
| 2834 | + TEST_COMPARE (stdc_bit_floor (~(unsigned _BitInt(BITINT_MAXWIDTH)) 0) |
| 2835 | + != ((unsigned _BitInt(BITINT_MAXWIDTH)) 1) << (BITINT_MAXWIDTH |
| 2836 | + - 1), 0); |
| 2837 | + TEST_COMPARE (stdc_bit_floor (((unsigned _BitInt(BITINT_MAXWIDTH)) 511) |
| 2838 | + << 405) |
| 2839 | + != (((unsigned _BitInt(BITINT_MAXWIDTH)) 256) << 405), 0); |
| 2840 | + TEST_COMPARE (stdc_bit_floor (((unsigned _BitInt(BITINT_MAXWIDTH)) 512) |
| 2841 | + << 405) |
| 2842 | + != (((unsigned _BitInt(BITINT_MAXWIDTH)) 512) << 405), 0); |
| 2843 | + TEST_COMPARE (stdc_bit_ceil ((unsigned _BitInt(BITINT_MAXWIDTH)) 0) != 1, 0); |
| 2844 | + TEST_COMPARE (stdc_bit_ceil (~(unsigned _BitInt(BITINT_MAXWIDTH)) 0) != 0, |
| 2845 | + 0); |
| 2846 | + TEST_COMPARE (stdc_bit_ceil (((unsigned _BitInt(BITINT_MAXWIDTH)) 1) |
| 2847 | + << (BITINT_MAXWIDTH - 1)) |
| 2848 | + != ((unsigned _BitInt(BITINT_MAXWIDTH)) 1) << (BITINT_MAXWIDTH |
| 2849 | + - 1), 0); |
| 2850 | + TEST_COMPARE (stdc_bit_ceil (((unsigned _BitInt(BITINT_MAXWIDTH)) 512) |
| 2851 | + << 405) |
| 2852 | + != (((unsigned _BitInt(BITINT_MAXWIDTH)) 512) << 405), 0); |
| 2853 | + TEST_COMPARE (stdc_bit_ceil (((unsigned _BitInt(BITINT_MAXWIDTH)) 513) |
| 2854 | + << 405) |
| 2855 | + != (((unsigned _BitInt(BITINT_MAXWIDTH)) 1024) << 405), 0); |
| 2856 | +# endif |
| 2857 | +# endif |
| 2858 | + return 0; |
| 2859 | +} |
| 2860 | +#else |
| 2861 | +static int |
| 2862 | +do_test (void) |
| 2863 | +{ |
| 2864 | + return 0; |
| 2865 | +} |
| 2866 | +#endif |
| 2867 | + |
| 2868 | +#include <support/test-driver.c> |
| 2869 | diff --git a/support/Makefile b/support/Makefile |
| 2870 | index 362a51f882..aa57207bdc 100644 |
| 2871 | --- a/support/Makefile |
| 2872 | +++ b/support/Makefile |
| 2873 | @@ -131,6 +131,7 @@ libsupport-routines = \ |
| 2874 | xfreopen \ |
| 2875 | xftruncate \ |
| 2876 | xgetline \ |
| 2877 | + xgetpeername \ |
| 2878 | xgetsockname \ |
| 2879 | xlisten \ |
| 2880 | xlseek \ |
| 2881 | diff --git a/support/xgetpeername.c b/support/xgetpeername.c |
| 2882 | new file mode 100644 |
| 2883 | index 0000000000..6f448e456a |
| 2884 | --- /dev/null |
| 2885 | +++ b/support/xgetpeername.c |
| 2886 | @@ -0,0 +1,30 @@ |
| 2887 | +/* getpeername with error checking. |
| 2888 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 2889 | + This file is part of the GNU C Library. |
| 2890 | + |
| 2891 | + The GNU C Library is free software; you can redistribute it and/or |
| 2892 | + modify it under the terms of the GNU Lesser General Public |
| 2893 | + License as published by the Free Software Foundation; either |
| 2894 | + version 2.1 of the License, or (at your option) any later version. |
| 2895 | + |
| 2896 | + The GNU C Library is distributed in the hope that it will be useful, |
| 2897 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 2898 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 2899 | + Lesser General Public License for more details. |
| 2900 | + |
| 2901 | + You should have received a copy of the GNU Lesser General Public |
| 2902 | + License along with the GNU C Library; if not, see |
| 2903 | + <https://www.gnu.org/licenses/>. */ |
| 2904 | + |
| 2905 | +#include <support/xsocket.h> |
| 2906 | + |
| 2907 | +#include <stdio.h> |
| 2908 | +#include <stdlib.h> |
| 2909 | +#include <support/check.h> |
| 2910 | + |
| 2911 | +void |
| 2912 | +xgetpeername (int fd, struct sockaddr *sa, socklen_t *plen) |
| 2913 | +{ |
| 2914 | + if (getpeername (fd, sa, plen) != 0) |
| 2915 | + FAIL_EXIT1 ("getpeername (%d): %m", fd); |
| 2916 | +} |
| 2917 | diff --git a/support/xsocket.h b/support/xsocket.h |
| 2918 | index 3e44103546..4ac0e1f5ff 100644 |
| 2919 | --- a/support/xsocket.h |
| 2920 | +++ b/support/xsocket.h |
| 2921 | @@ -26,6 +26,7 @@ |
| 2922 | int xsocket (int, int, int); |
| 2923 | void xsetsockopt (int, int, int, const void *, socklen_t); |
| 2924 | void xgetsockname (int, struct sockaddr *, socklen_t *); |
| 2925 | +void xgetpeername (int, struct sockaddr *, socklen_t *); |
| 2926 | void xconnect (int, const struct sockaddr *, socklen_t); |
| 2927 | void xbind (int, const struct sockaddr *, socklen_t); |
| 2928 | void xlisten (int, int); |
| 2929 | diff --git a/sysdeps/aarch64/configure b/sysdeps/aarch64/configure |
| 2930 | old mode 100644 |
| 2931 | new mode 100755 |
| 2932 | index ca57edce47..9606137e8d |
| 2933 | --- a/sysdeps/aarch64/configure |
| 2934 | +++ b/sysdeps/aarch64/configure |
| 2935 | @@ -325,9 +325,10 @@ then : |
| 2936 | printf %s "(cached) " >&6 |
| 2937 | else $as_nop |
| 2938 | cat > conftest.s <<\EOF |
| 2939 | - ptrue p0.b |
| 2940 | + .arch armv8.2-a+sve |
| 2941 | + ptrue p0.b |
| 2942 | EOF |
| 2943 | -if { ac_try='${CC-cc} -c -march=armv8.2-a+sve conftest.s 1>&5' |
| 2944 | +if { ac_try='${CC-cc} -c conftest.s 1>&5' |
| 2945 | { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5 |
| 2946 | (eval $ac_try) 2>&5 |
| 2947 | ac_status=$? |
| 2948 | diff --git a/sysdeps/aarch64/configure.ac b/sysdeps/aarch64/configure.ac |
| 2949 | index 27874eceb4..56d12d661d 100644 |
| 2950 | --- a/sysdeps/aarch64/configure.ac |
| 2951 | +++ b/sysdeps/aarch64/configure.ac |
| 2952 | @@ -90,9 +90,10 @@ LIBC_CONFIG_VAR([aarch64-variant-pcs], [$libc_cv_aarch64_variant_pcs]) |
| 2953 | # Check if asm support armv8.2-a+sve |
| 2954 | AC_CACHE_CHECK([for SVE support in assembler], [libc_cv_aarch64_sve_asm], [dnl |
| 2955 | cat > conftest.s <<\EOF |
| 2956 | - ptrue p0.b |
| 2957 | + .arch armv8.2-a+sve |
| 2958 | + ptrue p0.b |
| 2959 | EOF |
| 2960 | -if AC_TRY_COMMAND(${CC-cc} -c -march=armv8.2-a+sve conftest.s 1>&AS_MESSAGE_LOG_FD); then |
| 2961 | +if AC_TRY_COMMAND(${CC-cc} -c conftest.s 1>&AS_MESSAGE_LOG_FD); then |
| 2962 | libc_cv_aarch64_sve_asm=yes |
| 2963 | else |
| 2964 | libc_cv_aarch64_sve_asm=no |
| 2965 | diff --git a/sysdeps/aarch64/cpu-features.h b/sysdeps/aarch64/cpu-features.h |
| 2966 | index 77a782422a..5f2da91ebb 100644 |
| 2967 | --- a/sysdeps/aarch64/cpu-features.h |
| 2968 | +++ b/sysdeps/aarch64/cpu-features.h |
| 2969 | @@ -71,6 +71,7 @@ struct cpu_features |
| 2970 | /* Currently, the GLIBC memory tagging tunable only defines 8 bits. */ |
| 2971 | uint8_t mte_state; |
| 2972 | bool sve; |
| 2973 | + bool prefer_sve_ifuncs; |
| 2974 | bool mops; |
| 2975 | }; |
| 2976 | |
| 2977 | diff --git a/sysdeps/aarch64/fpu/acos_advsimd.c b/sysdeps/aarch64/fpu/acos_advsimd.c |
| 2978 | index a8eabb5e71..0a86c9823a 100644 |
| 2979 | --- a/sysdeps/aarch64/fpu/acos_advsimd.c |
| 2980 | +++ b/sysdeps/aarch64/fpu/acos_advsimd.c |
| 2981 | @@ -40,8 +40,8 @@ static const struct data |
| 2982 | }; |
| 2983 | |
| 2984 | #define AllMask v_u64 (0xffffffffffffffff) |
| 2985 | -#define Oneu (0x3ff0000000000000) |
| 2986 | -#define Small (0x3e50000000000000) /* 2^-53. */ |
| 2987 | +#define Oneu 0x3ff0000000000000 |
| 2988 | +#define Small 0x3e50000000000000 /* 2^-53. */ |
| 2989 | |
| 2990 | #if WANT_SIMD_EXCEPT |
| 2991 | static float64x2_t VPCS_ATTR NOINLINE |
| 2992 | diff --git a/sysdeps/aarch64/fpu/asin_advsimd.c b/sysdeps/aarch64/fpu/asin_advsimd.c |
| 2993 | index 141646e954..2de6eff407 100644 |
| 2994 | --- a/sysdeps/aarch64/fpu/asin_advsimd.c |
| 2995 | +++ b/sysdeps/aarch64/fpu/asin_advsimd.c |
| 2996 | @@ -39,8 +39,8 @@ static const struct data |
| 2997 | }; |
| 2998 | |
| 2999 | #define AllMask v_u64 (0xffffffffffffffff) |
| 3000 | -#define One (0x3ff0000000000000) |
| 3001 | -#define Small (0x3e50000000000000) /* 2^-12. */ |
| 3002 | +#define One 0x3ff0000000000000 |
| 3003 | +#define Small 0x3e50000000000000 /* 2^-12. */ |
| 3004 | |
| 3005 | #if WANT_SIMD_EXCEPT |
| 3006 | static float64x2_t VPCS_ATTR NOINLINE |
| 3007 | diff --git a/sysdeps/aarch64/fpu/atan2_sve.c b/sysdeps/aarch64/fpu/atan2_sve.c |
| 3008 | index 09a4c559b8..04fa71fa37 100644 |
| 3009 | --- a/sysdeps/aarch64/fpu/atan2_sve.c |
| 3010 | +++ b/sysdeps/aarch64/fpu/atan2_sve.c |
| 3011 | @@ -37,9 +37,6 @@ static const struct data |
| 3012 | .pi_over_2 = 0x1.921fb54442d18p+0, |
| 3013 | }; |
| 3014 | |
| 3015 | -/* Useful constants. */ |
| 3016 | -#define SignMask sv_u64 (0x8000000000000000) |
| 3017 | - |
| 3018 | /* Special cases i.e. 0, infinity, nan (fall back to scalar calls). */ |
| 3019 | static svfloat64_t NOINLINE |
| 3020 | special_case (svfloat64_t y, svfloat64_t x, svfloat64_t ret, |
| 3021 | @@ -72,14 +69,15 @@ svfloat64_t SV_NAME_D2 (atan2) (svfloat64_t y, svfloat64_t x, const svbool_t pg) |
| 3022 | svbool_t cmp_y = zeroinfnan (iy, pg); |
| 3023 | svbool_t cmp_xy = svorr_z (pg, cmp_x, cmp_y); |
| 3024 | |
| 3025 | - svuint64_t sign_x = svand_x (pg, ix, SignMask); |
| 3026 | - svuint64_t sign_y = svand_x (pg, iy, SignMask); |
| 3027 | - svuint64_t sign_xy = sveor_x (pg, sign_x, sign_y); |
| 3028 | - |
| 3029 | svfloat64_t ax = svabs_x (pg, x); |
| 3030 | svfloat64_t ay = svabs_x (pg, y); |
| 3031 | + svuint64_t iax = svreinterpret_u64 (ax); |
| 3032 | + svuint64_t iay = svreinterpret_u64 (ay); |
| 3033 | + |
| 3034 | + svuint64_t sign_x = sveor_x (pg, ix, iax); |
| 3035 | + svuint64_t sign_y = sveor_x (pg, iy, iay); |
| 3036 | + svuint64_t sign_xy = sveor_x (pg, sign_x, sign_y); |
| 3037 | |
| 3038 | - svbool_t pred_xlt0 = svcmplt (pg, x, 0.0); |
| 3039 | svbool_t pred_aygtax = svcmpgt (pg, ay, ax); |
| 3040 | |
| 3041 | /* Set up z for call to atan. */ |
| 3042 | @@ -88,8 +86,9 @@ svfloat64_t SV_NAME_D2 (atan2) (svfloat64_t y, svfloat64_t x, const svbool_t pg) |
| 3043 | svfloat64_t z = svdiv_x (pg, n, d); |
| 3044 | |
| 3045 | /* Work out the correct shift. */ |
| 3046 | - svfloat64_t shift = svsel (pred_xlt0, sv_f64 (-2.0), sv_f64 (0.0)); |
| 3047 | - shift = svsel (pred_aygtax, svadd_x (pg, shift, 1.0), shift); |
| 3048 | + svfloat64_t shift = svreinterpret_f64 (svlsr_x (pg, sign_x, 1)); |
| 3049 | + shift = svsel (pred_aygtax, sv_f64 (1.0), shift); |
| 3050 | + shift = svreinterpret_f64 (svorr_x (pg, sign_x, svreinterpret_u64 (shift))); |
| 3051 | shift = svmul_x (pg, shift, data_ptr->pi_over_2); |
| 3052 | |
| 3053 | /* Use split Estrin scheme for P(z^2) with deg(P)=19. */ |
| 3054 | @@ -109,10 +108,10 @@ svfloat64_t SV_NAME_D2 (atan2) (svfloat64_t y, svfloat64_t x, const svbool_t pg) |
| 3055 | ret = svadd_m (pg, ret, shift); |
| 3056 | |
| 3057 | /* Account for the sign of x and y. */ |
| 3058 | - ret = svreinterpret_f64 (sveor_x (pg, svreinterpret_u64 (ret), sign_xy)); |
| 3059 | - |
| 3060 | if (__glibc_unlikely (svptest_any (pg, cmp_xy))) |
| 3061 | - return special_case (y, x, ret, cmp_xy); |
| 3062 | - |
| 3063 | - return ret; |
| 3064 | + return special_case ( |
| 3065 | + y, x, |
| 3066 | + svreinterpret_f64 (sveor_x (pg, svreinterpret_u64 (ret), sign_xy)), |
| 3067 | + cmp_xy); |
| 3068 | + return svreinterpret_f64 (sveor_x (pg, svreinterpret_u64 (ret), sign_xy)); |
| 3069 | } |
| 3070 | diff --git a/sysdeps/aarch64/fpu/atan2f_sve.c b/sysdeps/aarch64/fpu/atan2f_sve.c |
| 3071 | index b92f83cdea..9ea197147c 100644 |
| 3072 | --- a/sysdeps/aarch64/fpu/atan2f_sve.c |
| 3073 | +++ b/sysdeps/aarch64/fpu/atan2f_sve.c |
| 3074 | @@ -32,10 +32,8 @@ static const struct data |
| 3075 | .pi_over_2 = 0x1.921fb6p+0f, |
| 3076 | }; |
| 3077 | |
| 3078 | -#define SignMask sv_u32 (0x80000000) |
| 3079 | - |
| 3080 | /* Special cases i.e. 0, infinity, nan (fall back to scalar calls). */ |
| 3081 | -static inline svfloat32_t |
| 3082 | +static svfloat32_t NOINLINE |
| 3083 | special_case (svfloat32_t y, svfloat32_t x, svfloat32_t ret, |
| 3084 | const svbool_t cmp) |
| 3085 | { |
| 3086 | @@ -67,14 +65,15 @@ svfloat32_t SV_NAME_F2 (atan2) (svfloat32_t y, svfloat32_t x, const svbool_t pg) |
| 3087 | svbool_t cmp_y = zeroinfnan (iy, pg); |
| 3088 | svbool_t cmp_xy = svorr_z (pg, cmp_x, cmp_y); |
| 3089 | |
| 3090 | - svuint32_t sign_x = svand_x (pg, ix, SignMask); |
| 3091 | - svuint32_t sign_y = svand_x (pg, iy, SignMask); |
| 3092 | - svuint32_t sign_xy = sveor_x (pg, sign_x, sign_y); |
| 3093 | - |
| 3094 | svfloat32_t ax = svabs_x (pg, x); |
| 3095 | svfloat32_t ay = svabs_x (pg, y); |
| 3096 | + svuint32_t iax = svreinterpret_u32 (ax); |
| 3097 | + svuint32_t iay = svreinterpret_u32 (ay); |
| 3098 | + |
| 3099 | + svuint32_t sign_x = sveor_x (pg, ix, iax); |
| 3100 | + svuint32_t sign_y = sveor_x (pg, iy, iay); |
| 3101 | + svuint32_t sign_xy = sveor_x (pg, sign_x, sign_y); |
| 3102 | |
| 3103 | - svbool_t pred_xlt0 = svcmplt (pg, x, 0.0); |
| 3104 | svbool_t pred_aygtax = svcmpgt (pg, ay, ax); |
| 3105 | |
| 3106 | /* Set up z for call to atan. */ |
| 3107 | @@ -83,11 +82,12 @@ svfloat32_t SV_NAME_F2 (atan2) (svfloat32_t y, svfloat32_t x, const svbool_t pg) |
| 3108 | svfloat32_t z = svdiv_x (pg, n, d); |
| 3109 | |
| 3110 | /* Work out the correct shift. */ |
| 3111 | - svfloat32_t shift = svsel (pred_xlt0, sv_f32 (-2.0), sv_f32 (0.0)); |
| 3112 | - shift = svsel (pred_aygtax, svadd_x (pg, shift, 1.0), shift); |
| 3113 | + svfloat32_t shift = svreinterpret_f32 (svlsr_x (pg, sign_x, 1)); |
| 3114 | + shift = svsel (pred_aygtax, sv_f32 (1.0), shift); |
| 3115 | + shift = svreinterpret_f32 (svorr_x (pg, sign_x, svreinterpret_u32 (shift))); |
| 3116 | shift = svmul_x (pg, shift, sv_f32 (data_ptr->pi_over_2)); |
| 3117 | |
| 3118 | - /* Use split Estrin scheme for P(z^2) with deg(P)=7. */ |
| 3119 | + /* Use pure Estrin scheme for P(z^2) with deg(P)=7. */ |
| 3120 | svfloat32_t z2 = svmul_x (pg, z, z); |
| 3121 | svfloat32_t z4 = svmul_x (pg, z2, z2); |
| 3122 | svfloat32_t z8 = svmul_x (pg, z4, z4); |
| 3123 | @@ -101,10 +101,12 @@ svfloat32_t SV_NAME_F2 (atan2) (svfloat32_t y, svfloat32_t x, const svbool_t pg) |
| 3124 | ret = svadd_m (pg, ret, shift); |
| 3125 | |
| 3126 | /* Account for the sign of x and y. */ |
| 3127 | - ret = svreinterpret_f32 (sveor_x (pg, svreinterpret_u32 (ret), sign_xy)); |
| 3128 | |
| 3129 | if (__glibc_unlikely (svptest_any (pg, cmp_xy))) |
| 3130 | - return special_case (y, x, ret, cmp_xy); |
| 3131 | + return special_case ( |
| 3132 | + y, x, |
| 3133 | + svreinterpret_f32 (sveor_x (pg, svreinterpret_u32 (ret), sign_xy)), |
| 3134 | + cmp_xy); |
| 3135 | |
| 3136 | - return ret; |
| 3137 | + return svreinterpret_f32 (sveor_x (pg, svreinterpret_u32 (ret), sign_xy)); |
| 3138 | } |
| 3139 | diff --git a/sysdeps/aarch64/fpu/cos_advsimd.c b/sysdeps/aarch64/fpu/cos_advsimd.c |
| 3140 | index 2897e8b909..3924c9ce44 100644 |
| 3141 | --- a/sysdeps/aarch64/fpu/cos_advsimd.c |
| 3142 | +++ b/sysdeps/aarch64/fpu/cos_advsimd.c |
| 3143 | @@ -63,8 +63,7 @@ float64x2_t VPCS_ATTR V_NAME_D1 (cos) (float64x2_t x) |
| 3144 | special-case handler later. */ |
| 3145 | r = vbslq_f64 (cmp, v_f64 (1.0), r); |
| 3146 | #else |
| 3147 | - cmp = vcageq_f64 (d->range_val, x); |
| 3148 | - cmp = vceqzq_u64 (cmp); /* cmp = ~cmp. */ |
| 3149 | + cmp = vcageq_f64 (x, d->range_val); |
| 3150 | r = x; |
| 3151 | #endif |
| 3152 | |
| 3153 | diff --git a/sysdeps/aarch64/fpu/cosf_advsimd.c b/sysdeps/aarch64/fpu/cosf_advsimd.c |
| 3154 | index 60abc8dfcf..d0c285b03a 100644 |
| 3155 | --- a/sysdeps/aarch64/fpu/cosf_advsimd.c |
| 3156 | +++ b/sysdeps/aarch64/fpu/cosf_advsimd.c |
| 3157 | @@ -64,8 +64,7 @@ float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (cos) (float32x4_t x) |
| 3158 | special-case handler later. */ |
| 3159 | r = vbslq_f32 (cmp, v_f32 (1.0f), r); |
| 3160 | #else |
| 3161 | - cmp = vcageq_f32 (d->range_val, x); |
| 3162 | - cmp = vceqzq_u32 (cmp); /* cmp = ~cmp. */ |
| 3163 | + cmp = vcageq_f32 (x, d->range_val); |
| 3164 | r = x; |
| 3165 | #endif |
| 3166 | |
| 3167 | diff --git a/sysdeps/aarch64/fpu/exp10_advsimd.c b/sysdeps/aarch64/fpu/exp10_advsimd.c |
| 3168 | index fe7149b191..eeb31ca839 100644 |
| 3169 | --- a/sysdeps/aarch64/fpu/exp10_advsimd.c |
| 3170 | +++ b/sysdeps/aarch64/fpu/exp10_advsimd.c |
| 3171 | @@ -57,7 +57,7 @@ const static struct data |
| 3172 | # define BigBound v_u64 (0x4070000000000000) /* asuint64 (0x1p8). */ |
| 3173 | # define Thres v_u64 (0x2070000000000000) /* BigBound - TinyBound. */ |
| 3174 | |
| 3175 | -static inline float64x2_t VPCS_ATTR |
| 3176 | +static float64x2_t VPCS_ATTR NOINLINE |
| 3177 | special_case (float64x2_t x, float64x2_t y, uint64x2_t cmp) |
| 3178 | { |
| 3179 | /* If fenv exceptions are to be triggered correctly, fall back to the scalar |
| 3180 | @@ -72,7 +72,7 @@ special_case (float64x2_t x, float64x2_t y, uint64x2_t cmp) |
| 3181 | # define SpecialBias1 v_u64 (0x7000000000000000) /* 0x1p769. */ |
| 3182 | # define SpecialBias2 v_u64 (0x3010000000000000) /* 0x1p-254. */ |
| 3183 | |
| 3184 | -static float64x2_t VPCS_ATTR NOINLINE |
| 3185 | +static inline float64x2_t VPCS_ATTR |
| 3186 | special_case (float64x2_t s, float64x2_t y, float64x2_t n, |
| 3187 | const struct data *d) |
| 3188 | { |
| 3189 | diff --git a/sysdeps/aarch64/fpu/exp10f_advsimd.c b/sysdeps/aarch64/fpu/exp10f_advsimd.c |
| 3190 | index 7ee0c90948..ab117b69da 100644 |
| 3191 | --- a/sysdeps/aarch64/fpu/exp10f_advsimd.c |
| 3192 | +++ b/sysdeps/aarch64/fpu/exp10f_advsimd.c |
| 3193 | @@ -25,7 +25,8 @@ |
| 3194 | static const struct data |
| 3195 | { |
| 3196 | float32x4_t poly[5]; |
| 3197 | - float32x4_t shift, log10_2, log2_10_hi, log2_10_lo; |
| 3198 | + float32x4_t log10_2_and_inv, shift; |
| 3199 | + |
| 3200 | #if !WANT_SIMD_EXCEPT |
| 3201 | float32x4_t scale_thresh; |
| 3202 | #endif |
| 3203 | @@ -38,9 +39,9 @@ static const struct data |
| 3204 | .poly = { V4 (0x1.26bb16p+1f), V4 (0x1.5350d2p+1f), V4 (0x1.04744ap+1f), |
| 3205 | V4 (0x1.2d8176p+0f), V4 (0x1.12b41ap-1f) }, |
| 3206 | .shift = V4 (0x1.8p23f), |
| 3207 | - .log10_2 = V4 (0x1.a934fp+1), |
| 3208 | - .log2_10_hi = V4 (0x1.344136p-2), |
| 3209 | - .log2_10_lo = V4 (-0x1.ec10cp-27), |
| 3210 | + |
| 3211 | + /* Stores constants 1/log10(2), log10(2)_high, log10(2)_low, 0. */ |
| 3212 | + .log10_2_and_inv = { 0x1.a934fp+1, 0x1.344136p-2, -0x1.ec10cp-27, 0 }, |
| 3213 | #if !WANT_SIMD_EXCEPT |
| 3214 | .scale_thresh = V4 (ScaleBound) |
| 3215 | #endif |
| 3216 | @@ -98,24 +99,22 @@ float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (exp10) (float32x4_t x) |
| 3217 | #if WANT_SIMD_EXCEPT |
| 3218 | /* asuint(x) - TinyBound >= BigBound - TinyBound. */ |
| 3219 | uint32x4_t cmp = vcgeq_u32 ( |
| 3220 | - vsubq_u32 (vandq_u32 (vreinterpretq_u32_f32 (x), v_u32 (0x7fffffff)), |
| 3221 | - TinyBound), |
| 3222 | - Thres); |
| 3223 | + vsubq_u32 (vreinterpretq_u32_f32 (vabsq_f32 (x)), TinyBound), Thres); |
| 3224 | float32x4_t xm = x; |
| 3225 | /* If any lanes are special, mask them with 1 and retain a copy of x to allow |
| 3226 | special case handler to fix special lanes later. This is only necessary if |
| 3227 | fenv exceptions are to be triggered correctly. */ |
| 3228 | if (__glibc_unlikely (v_any_u32 (cmp))) |
| 3229 | - x = vbslq_f32 (cmp, v_f32 (1), x); |
| 3230 | + x = v_zerofy_f32 (x, cmp); |
| 3231 | #endif |
| 3232 | |
| 3233 | /* exp10(x) = 2^n * 10^r = 2^n * (1 + poly (r)), |
| 3234 | with poly(r) in [1/sqrt(2), sqrt(2)] and |
| 3235 | x = r + n * log10 (2), with r in [-log10(2)/2, log10(2)/2]. */ |
| 3236 | - float32x4_t z = vfmaq_f32 (d->shift, x, d->log10_2); |
| 3237 | + float32x4_t z = vfmaq_laneq_f32 (d->shift, x, d->log10_2_and_inv, 0); |
| 3238 | float32x4_t n = vsubq_f32 (z, d->shift); |
| 3239 | - float32x4_t r = vfmsq_f32 (x, n, d->log2_10_hi); |
| 3240 | - r = vfmsq_f32 (r, n, d->log2_10_lo); |
| 3241 | + float32x4_t r = vfmsq_laneq_f32 (x, n, d->log10_2_and_inv, 1); |
| 3242 | + r = vfmsq_laneq_f32 (r, n, d->log10_2_and_inv, 2); |
| 3243 | uint32x4_t e = vshlq_n_u32 (vreinterpretq_u32_f32 (z), 23); |
| 3244 | |
| 3245 | float32x4_t scale = vreinterpretq_f32_u32 (vaddq_u32 (e, ExponentBias)); |
| 3246 | diff --git a/sysdeps/aarch64/fpu/exp2_advsimd.c b/sysdeps/aarch64/fpu/exp2_advsimd.c |
| 3247 | index 391a93180c..ae1e63d503 100644 |
| 3248 | --- a/sysdeps/aarch64/fpu/exp2_advsimd.c |
| 3249 | +++ b/sysdeps/aarch64/fpu/exp2_advsimd.c |
| 3250 | @@ -24,6 +24,7 @@ |
| 3251 | #define IndexMask (N - 1) |
| 3252 | #define BigBound 1022.0 |
| 3253 | #define UOFlowBound 1280.0 |
| 3254 | +#define TinyBound 0x2000000000000000 /* asuint64(0x1p-511). */ |
| 3255 | |
| 3256 | static const struct data |
| 3257 | { |
| 3258 | @@ -48,14 +49,13 @@ lookup_sbits (uint64x2_t i) |
| 3259 | |
| 3260 | #if WANT_SIMD_EXCEPT |
| 3261 | |
| 3262 | -# define TinyBound 0x2000000000000000 /* asuint64(0x1p-511). */ |
| 3263 | # define Thres 0x2080000000000000 /* asuint64(512.0) - TinyBound. */ |
| 3264 | |
| 3265 | /* Call scalar exp2 as a fallback. */ |
| 3266 | static float64x2_t VPCS_ATTR NOINLINE |
| 3267 | -special_case (float64x2_t x) |
| 3268 | +special_case (float64x2_t x, float64x2_t y, uint64x2_t is_special) |
| 3269 | { |
| 3270 | - return v_call_f64 (exp2, x, x, v_u64 (0xffffffffffffffff)); |
| 3271 | + return v_call_f64 (exp2, x, y, is_special); |
| 3272 | } |
| 3273 | |
| 3274 | #else |
| 3275 | @@ -65,7 +65,7 @@ special_case (float64x2_t x) |
| 3276 | # define SpecialBias1 0x7000000000000000 /* 0x1p769. */ |
| 3277 | # define SpecialBias2 0x3010000000000000 /* 0x1p-254. */ |
| 3278 | |
| 3279 | -static float64x2_t VPCS_ATTR |
| 3280 | +static inline float64x2_t VPCS_ATTR |
| 3281 | special_case (float64x2_t s, float64x2_t y, float64x2_t n, |
| 3282 | const struct data *d) |
| 3283 | { |
| 3284 | @@ -94,10 +94,10 @@ float64x2_t V_NAME_D1 (exp2) (float64x2_t x) |
| 3285 | #if WANT_SIMD_EXCEPT |
| 3286 | uint64x2_t ia = vreinterpretq_u64_f64 (vabsq_f64 (x)); |
| 3287 | cmp = vcgeq_u64 (vsubq_u64 (ia, v_u64 (TinyBound)), v_u64 (Thres)); |
| 3288 | - /* If any special case (inf, nan, small and large x) is detected, |
| 3289 | - fall back to scalar for all lanes. */ |
| 3290 | - if (__glibc_unlikely (v_any_u64 (cmp))) |
| 3291 | - return special_case (x); |
| 3292 | + /* Mask special lanes and retain a copy of x for passing to special-case |
| 3293 | + handler. */ |
| 3294 | + float64x2_t xc = x; |
| 3295 | + x = v_zerofy_f64 (x, cmp); |
| 3296 | #else |
| 3297 | cmp = vcagtq_f64 (x, d->scale_big_bound); |
| 3298 | #endif |
| 3299 | @@ -120,9 +120,11 @@ float64x2_t V_NAME_D1 (exp2) (float64x2_t x) |
| 3300 | float64x2_t y = v_pairwise_poly_3_f64 (r, r2, d->poly); |
| 3301 | y = vmulq_f64 (r, y); |
| 3302 | |
| 3303 | -#if !WANT_SIMD_EXCEPT |
| 3304 | if (__glibc_unlikely (v_any_u64 (cmp))) |
| 3305 | +#if !WANT_SIMD_EXCEPT |
| 3306 | return special_case (s, y, n, d); |
| 3307 | +#else |
| 3308 | + return special_case (xc, vfmaq_f64 (s, s, y), cmp); |
| 3309 | #endif |
| 3310 | return vfmaq_f64 (s, s, y); |
| 3311 | } |
| 3312 | diff --git a/sysdeps/aarch64/fpu/exp2f_sve.c b/sysdeps/aarch64/fpu/exp2f_sve.c |
| 3313 | index 9a5a523a10..8a686e3e05 100644 |
| 3314 | --- a/sysdeps/aarch64/fpu/exp2f_sve.c |
| 3315 | +++ b/sysdeps/aarch64/fpu/exp2f_sve.c |
| 3316 | @@ -20,6 +20,8 @@ |
| 3317 | #include "sv_math.h" |
| 3318 | #include "poly_sve_f32.h" |
| 3319 | |
| 3320 | +#define Thres 0x1.5d5e2ap+6f |
| 3321 | + |
| 3322 | static const struct data |
| 3323 | { |
| 3324 | float poly[5]; |
| 3325 | @@ -33,7 +35,7 @@ static const struct data |
| 3326 | .shift = 0x1.903f8p17f, |
| 3327 | /* Roughly 87.3. For x < -Thres, the result is subnormal and not handled |
| 3328 | correctly by FEXPA. */ |
| 3329 | - .thres = 0x1.5d5e2ap+6f, |
| 3330 | + .thres = Thres, |
| 3331 | }; |
| 3332 | |
| 3333 | static svfloat32_t NOINLINE |
| 3334 | diff --git a/sysdeps/aarch64/fpu/exp_advsimd.c b/sysdeps/aarch64/fpu/exp_advsimd.c |
| 3335 | index fd215f1d2c..5e3a9a0d44 100644 |
| 3336 | --- a/sysdeps/aarch64/fpu/exp_advsimd.c |
| 3337 | +++ b/sysdeps/aarch64/fpu/exp_advsimd.c |
| 3338 | @@ -54,7 +54,7 @@ const static volatile struct |
| 3339 | # define BigBound v_u64 (0x4080000000000000) /* asuint64 (0x1p9). */ |
| 3340 | # define SpecialBound v_u64 (0x2080000000000000) /* BigBound - TinyBound. */ |
| 3341 | |
| 3342 | -static inline float64x2_t VPCS_ATTR |
| 3343 | +static float64x2_t VPCS_ATTR NOINLINE |
| 3344 | special_case (float64x2_t x, float64x2_t y, uint64x2_t cmp) |
| 3345 | { |
| 3346 | /* If fenv exceptions are to be triggered correctly, fall back to the scalar |
| 3347 | @@ -69,7 +69,7 @@ special_case (float64x2_t x, float64x2_t y, uint64x2_t cmp) |
| 3348 | # define SpecialBias1 v_u64 (0x7000000000000000) /* 0x1p769. */ |
| 3349 | # define SpecialBias2 v_u64 (0x3010000000000000) /* 0x1p-254. */ |
| 3350 | |
| 3351 | -static float64x2_t VPCS_ATTR NOINLINE |
| 3352 | +static inline float64x2_t VPCS_ATTR |
| 3353 | special_case (float64x2_t s, float64x2_t y, float64x2_t n) |
| 3354 | { |
| 3355 | /* 2^(n/N) may overflow, break it up into s1*s2. */ |
| 3356 | diff --git a/sysdeps/aarch64/fpu/expm1_advsimd.c b/sysdeps/aarch64/fpu/expm1_advsimd.c |
| 3357 | index 0b85bd06f3..3628398674 100644 |
| 3358 | --- a/sysdeps/aarch64/fpu/expm1_advsimd.c |
| 3359 | +++ b/sysdeps/aarch64/fpu/expm1_advsimd.c |
| 3360 | @@ -23,7 +23,7 @@ |
| 3361 | static const struct data |
| 3362 | { |
| 3363 | float64x2_t poly[11]; |
| 3364 | - float64x2_t invln2, ln2_lo, ln2_hi, shift; |
| 3365 | + float64x2_t invln2, ln2, shift; |
| 3366 | int64x2_t exponent_bias; |
| 3367 | #if WANT_SIMD_EXCEPT |
| 3368 | uint64x2_t thresh, tiny_bound; |
| 3369 | @@ -38,8 +38,7 @@ static const struct data |
| 3370 | V2 (0x1.71ddf82db5bb4p-19), V2 (0x1.27e517fc0d54bp-22), |
| 3371 | V2 (0x1.af5eedae67435p-26), V2 (0x1.1f143d060a28ap-29) }, |
| 3372 | .invln2 = V2 (0x1.71547652b82fep0), |
| 3373 | - .ln2_hi = V2 (0x1.62e42fefa39efp-1), |
| 3374 | - .ln2_lo = V2 (0x1.abc9e3b39803fp-56), |
| 3375 | + .ln2 = { 0x1.62e42fefa39efp-1, 0x1.abc9e3b39803fp-56 }, |
| 3376 | .shift = V2 (0x1.8p52), |
| 3377 | .exponent_bias = V2 (0x3ff0000000000000), |
| 3378 | #if WANT_SIMD_EXCEPT |
| 3379 | @@ -83,7 +82,7 @@ float64x2_t VPCS_ATTR V_NAME_D1 (expm1) (float64x2_t x) |
| 3380 | x = v_zerofy_f64 (x, special); |
| 3381 | #else |
| 3382 | /* Large input, NaNs and Infs. */ |
| 3383 | - uint64x2_t special = vceqzq_u64 (vcaltq_f64 (x, d->oflow_bound)); |
| 3384 | + uint64x2_t special = vcageq_f64 (x, d->oflow_bound); |
| 3385 | #endif |
| 3386 | |
| 3387 | /* Reduce argument to smaller range: |
| 3388 | @@ -93,8 +92,8 @@ float64x2_t VPCS_ATTR V_NAME_D1 (expm1) (float64x2_t x) |
| 3389 | where 2^i is exact because i is an integer. */ |
| 3390 | float64x2_t n = vsubq_f64 (vfmaq_f64 (d->shift, d->invln2, x), d->shift); |
| 3391 | int64x2_t i = vcvtq_s64_f64 (n); |
| 3392 | - float64x2_t f = vfmsq_f64 (x, n, d->ln2_hi); |
| 3393 | - f = vfmsq_f64 (f, n, d->ln2_lo); |
| 3394 | + float64x2_t f = vfmsq_laneq_f64 (x, n, d->ln2, 0); |
| 3395 | + f = vfmsq_laneq_f64 (f, n, d->ln2, 1); |
| 3396 | |
| 3397 | /* Approximate expm1(f) using polynomial. |
| 3398 | Taylor expansion for expm1(x) has the form: |
| 3399 | diff --git a/sysdeps/aarch64/fpu/expm1f_advsimd.c b/sysdeps/aarch64/fpu/expm1f_advsimd.c |
| 3400 | index 8d4c9a2193..93db200f61 100644 |
| 3401 | --- a/sysdeps/aarch64/fpu/expm1f_advsimd.c |
| 3402 | +++ b/sysdeps/aarch64/fpu/expm1f_advsimd.c |
| 3403 | @@ -23,7 +23,8 @@ |
| 3404 | static const struct data |
| 3405 | { |
| 3406 | float32x4_t poly[5]; |
| 3407 | - float32x4_t invln2, ln2_lo, ln2_hi, shift; |
| 3408 | + float32x4_t invln2_and_ln2; |
| 3409 | + float32x4_t shift; |
| 3410 | int32x4_t exponent_bias; |
| 3411 | #if WANT_SIMD_EXCEPT |
| 3412 | uint32x4_t thresh; |
| 3413 | @@ -34,9 +35,8 @@ static const struct data |
| 3414 | /* Generated using fpminimax with degree=5 in [-log(2)/2, log(2)/2]. */ |
| 3415 | .poly = { V4 (0x1.fffffep-2), V4 (0x1.5554aep-3), V4 (0x1.555736p-5), |
| 3416 | V4 (0x1.12287cp-7), V4 (0x1.6b55a2p-10) }, |
| 3417 | - .invln2 = V4 (0x1.715476p+0f), |
| 3418 | - .ln2_hi = V4 (0x1.62e4p-1f), |
| 3419 | - .ln2_lo = V4 (0x1.7f7d1cp-20f), |
| 3420 | + /* Stores constants: invln2, ln2_hi, ln2_lo, 0. */ |
| 3421 | + .invln2_and_ln2 = { 0x1.715476p+0f, 0x1.62e4p-1f, 0x1.7f7d1cp-20f, 0 }, |
| 3422 | .shift = V4 (0x1.8p23f), |
| 3423 | .exponent_bias = V4 (0x3f800000), |
| 3424 | #if !WANT_SIMD_EXCEPT |
| 3425 | @@ -80,7 +80,7 @@ float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (expm1) (float32x4_t x) |
| 3426 | x = v_zerofy_f32 (x, special); |
| 3427 | #else |
| 3428 | /* Handles very large values (+ve and -ve), +/-NaN, +/-Inf. */ |
| 3429 | - uint32x4_t special = vceqzq_u32 (vcaltq_f32 (x, d->oflow_bound)); |
| 3430 | + uint32x4_t special = vcagtq_f32 (x, d->oflow_bound); |
| 3431 | #endif |
| 3432 | |
| 3433 | /* Reduce argument to smaller range: |
| 3434 | @@ -88,10 +88,11 @@ float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (expm1) (float32x4_t x) |
| 3435 | and f = x - i * ln2, then f is in [-ln2/2, ln2/2]. |
| 3436 | exp(x) - 1 = 2^i * (expm1(f) + 1) - 1 |
| 3437 | where 2^i is exact because i is an integer. */ |
| 3438 | - float32x4_t j = vsubq_f32 (vfmaq_f32 (d->shift, d->invln2, x), d->shift); |
| 3439 | + float32x4_t j = vsubq_f32 ( |
| 3440 | + vfmaq_laneq_f32 (d->shift, x, d->invln2_and_ln2, 0), d->shift); |
| 3441 | int32x4_t i = vcvtq_s32_f32 (j); |
| 3442 | - float32x4_t f = vfmsq_f32 (x, j, d->ln2_hi); |
| 3443 | - f = vfmsq_f32 (f, j, d->ln2_lo); |
| 3444 | + float32x4_t f = vfmsq_laneq_f32 (x, j, d->invln2_and_ln2, 1); |
| 3445 | + f = vfmsq_laneq_f32 (f, j, d->invln2_and_ln2, 2); |
| 3446 | |
| 3447 | /* Approximate expm1(f) using polynomial. |
| 3448 | Taylor expansion for expm1(x) has the form: |
| 3449 | diff --git a/sysdeps/aarch64/fpu/log_advsimd.c b/sysdeps/aarch64/fpu/log_advsimd.c |
| 3450 | index 067ae79613..21df61728c 100644 |
| 3451 | --- a/sysdeps/aarch64/fpu/log_advsimd.c |
| 3452 | +++ b/sysdeps/aarch64/fpu/log_advsimd.c |
| 3453 | @@ -58,8 +58,13 @@ lookup (uint64x2_t i) |
| 3454 | uint64_t i1 = (i[1] >> (52 - V_LOG_TABLE_BITS)) & IndexMask; |
| 3455 | float64x2_t e0 = vld1q_f64 (&__v_log_data.table[i0].invc); |
| 3456 | float64x2_t e1 = vld1q_f64 (&__v_log_data.table[i1].invc); |
| 3457 | +#if __BYTE_ORDER == __LITTLE_ENDIAN |
| 3458 | e.invc = vuzp1q_f64 (e0, e1); |
| 3459 | e.logc = vuzp2q_f64 (e0, e1); |
| 3460 | +#else |
| 3461 | + e.invc = vuzp1q_f64 (e1, e0); |
| 3462 | + e.logc = vuzp2q_f64 (e1, e0); |
| 3463 | +#endif |
| 3464 | return e; |
| 3465 | } |
| 3466 | |
| 3467 | diff --git a/sysdeps/aarch64/fpu/sin_advsimd.c b/sysdeps/aarch64/fpu/sin_advsimd.c |
| 3468 | index efce183e86..a0d9d3b819 100644 |
| 3469 | --- a/sysdeps/aarch64/fpu/sin_advsimd.c |
| 3470 | +++ b/sysdeps/aarch64/fpu/sin_advsimd.c |
| 3471 | @@ -75,8 +75,7 @@ float64x2_t VPCS_ATTR V_NAME_D1 (sin) (float64x2_t x) |
| 3472 | r = vbslq_f64 (cmp, vreinterpretq_f64_u64 (cmp), x); |
| 3473 | #else |
| 3474 | r = x; |
| 3475 | - cmp = vcageq_f64 (d->range_val, x); |
| 3476 | - cmp = vceqzq_u64 (cmp); /* cmp = ~cmp. */ |
| 3477 | + cmp = vcageq_f64 (x, d->range_val); |
| 3478 | #endif |
| 3479 | |
| 3480 | /* n = rint(|x|/pi). */ |
| 3481 | diff --git a/sysdeps/aarch64/fpu/sinf_advsimd.c b/sysdeps/aarch64/fpu/sinf_advsimd.c |
| 3482 | index 60cf3f2ca1..375dfc3331 100644 |
| 3483 | --- a/sysdeps/aarch64/fpu/sinf_advsimd.c |
| 3484 | +++ b/sysdeps/aarch64/fpu/sinf_advsimd.c |
| 3485 | @@ -67,8 +67,7 @@ float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (sin) (float32x4_t x) |
| 3486 | r = vbslq_f32 (cmp, vreinterpretq_f32_u32 (cmp), x); |
| 3487 | #else |
| 3488 | r = x; |
| 3489 | - cmp = vcageq_f32 (d->range_val, x); |
| 3490 | - cmp = vceqzq_u32 (cmp); /* cmp = ~cmp. */ |
| 3491 | + cmp = vcageq_f32 (x, d->range_val); |
| 3492 | #endif |
| 3493 | |
| 3494 | /* n = rint(|x|/pi) */ |
| 3495 | diff --git a/sysdeps/aarch64/fpu/tan_advsimd.c b/sysdeps/aarch64/fpu/tan_advsimd.c |
| 3496 | index d7e5ba7b1a..0459821ab2 100644 |
| 3497 | --- a/sysdeps/aarch64/fpu/tan_advsimd.c |
| 3498 | +++ b/sysdeps/aarch64/fpu/tan_advsimd.c |
| 3499 | @@ -23,7 +23,7 @@ |
| 3500 | static const struct data |
| 3501 | { |
| 3502 | float64x2_t poly[9]; |
| 3503 | - float64x2_t half_pi_hi, half_pi_lo, two_over_pi, shift; |
| 3504 | + float64x2_t half_pi, two_over_pi, shift; |
| 3505 | #if !WANT_SIMD_EXCEPT |
| 3506 | float64x2_t range_val; |
| 3507 | #endif |
| 3508 | @@ -34,8 +34,7 @@ static const struct data |
| 3509 | V2 (0x1.226e5e5ecdfa3p-7), V2 (0x1.d6c7ddbf87047p-9), |
| 3510 | V2 (0x1.7ea75d05b583ep-10), V2 (0x1.289f22964a03cp-11), |
| 3511 | V2 (0x1.4e4fd14147622p-12) }, |
| 3512 | - .half_pi_hi = V2 (0x1.921fb54442d18p0), |
| 3513 | - .half_pi_lo = V2 (0x1.1a62633145c07p-54), |
| 3514 | + .half_pi = { 0x1.921fb54442d18p0, 0x1.1a62633145c07p-54 }, |
| 3515 | .two_over_pi = V2 (0x1.45f306dc9c883p-1), |
| 3516 | .shift = V2 (0x1.8p52), |
| 3517 | #if !WANT_SIMD_EXCEPT |
| 3518 | @@ -56,15 +55,15 @@ special_case (float64x2_t x) |
| 3519 | |
| 3520 | /* Vector approximation for double-precision tan. |
| 3521 | Maximum measured error is 3.48 ULP: |
| 3522 | - __v_tan(0x1.4457047ef78d8p+20) got -0x1.f6ccd8ecf7dedp+37 |
| 3523 | - want -0x1.f6ccd8ecf7deap+37. */ |
| 3524 | + _ZGVnN2v_tan(0x1.4457047ef78d8p+20) got -0x1.f6ccd8ecf7dedp+37 |
| 3525 | + want -0x1.f6ccd8ecf7deap+37. */ |
| 3526 | float64x2_t VPCS_ATTR V_NAME_D1 (tan) (float64x2_t x) |
| 3527 | { |
| 3528 | const struct data *dat = ptr_barrier (&data); |
| 3529 | - /* Our argument reduction cannot calculate q with sufficient accuracy for very |
| 3530 | - large inputs. Fall back to scalar routine for all lanes if any are too |
| 3531 | - large, or Inf/NaN. If fenv exceptions are expected, also fall back for tiny |
| 3532 | - input to avoid underflow. */ |
| 3533 | + /* Our argument reduction cannot calculate q with sufficient accuracy for |
| 3534 | + very large inputs. Fall back to scalar routine for all lanes if any are |
| 3535 | + too large, or Inf/NaN. If fenv exceptions are expected, also fall back for |
| 3536 | + tiny input to avoid underflow. */ |
| 3537 | #if WANT_SIMD_EXCEPT |
| 3538 | uint64x2_t iax = vreinterpretq_u64_f64 (vabsq_f64 (x)); |
| 3539 | /* iax - tiny_bound > range_val - tiny_bound. */ |
| 3540 | @@ -82,8 +81,8 @@ float64x2_t VPCS_ATTR V_NAME_D1 (tan) (float64x2_t x) |
| 3541 | /* Use q to reduce x to r in [-pi/4, pi/4], by: |
| 3542 | r = x - q * pi/2, in extended precision. */ |
| 3543 | float64x2_t r = x; |
| 3544 | - r = vfmsq_f64 (r, q, dat->half_pi_hi); |
| 3545 | - r = vfmsq_f64 (r, q, dat->half_pi_lo); |
| 3546 | + r = vfmsq_laneq_f64 (r, q, dat->half_pi, 0); |
| 3547 | + r = vfmsq_laneq_f64 (r, q, dat->half_pi, 1); |
| 3548 | /* Further reduce r to [-pi/8, pi/8], to be reconstructed using double angle |
| 3549 | formula. */ |
| 3550 | r = vmulq_n_f64 (r, 0.5); |
| 3551 | @@ -106,14 +105,15 @@ float64x2_t VPCS_ATTR V_NAME_D1 (tan) (float64x2_t x) |
| 3552 | and reciprocity around pi/2: |
| 3553 | tan(x) = 1 / (tan(pi/2 - x)) |
| 3554 | to assemble result using change-of-sign and conditional selection of |
| 3555 | - numerator/denominator, dependent on odd/even-ness of q (hence quadrant). */ |
| 3556 | + numerator/denominator, dependent on odd/even-ness of q (hence quadrant). |
| 3557 | + */ |
| 3558 | float64x2_t n = vfmaq_f64 (v_f64 (-1), p, p); |
| 3559 | float64x2_t d = vaddq_f64 (p, p); |
| 3560 | |
| 3561 | uint64x2_t no_recip = vtstq_u64 (vreinterpretq_u64_s64 (qi), v_u64 (1)); |
| 3562 | |
| 3563 | #if !WANT_SIMD_EXCEPT |
| 3564 | - uint64x2_t special = vceqzq_u64 (vcaleq_f64 (x, dat->range_val)); |
| 3565 | + uint64x2_t special = vcageq_f64 (x, dat->range_val); |
| 3566 | if (__glibc_unlikely (v_any_u64 (special))) |
| 3567 | return special_case (x); |
| 3568 | #endif |
| 3569 | diff --git a/sysdeps/aarch64/fpu/tanf_advsimd.c b/sysdeps/aarch64/fpu/tanf_advsimd.c |
| 3570 | index 1f16103f8a..5a7489390a 100644 |
| 3571 | --- a/sysdeps/aarch64/fpu/tanf_advsimd.c |
| 3572 | +++ b/sysdeps/aarch64/fpu/tanf_advsimd.c |
| 3573 | @@ -23,7 +23,8 @@ |
| 3574 | static const struct data |
| 3575 | { |
| 3576 | float32x4_t poly[6]; |
| 3577 | - float32x4_t neg_half_pi_1, neg_half_pi_2, neg_half_pi_3, two_over_pi, shift; |
| 3578 | + float32x4_t pi_consts; |
| 3579 | + float32x4_t shift; |
| 3580 | #if !WANT_SIMD_EXCEPT |
| 3581 | float32x4_t range_val; |
| 3582 | #endif |
| 3583 | @@ -31,10 +32,9 @@ static const struct data |
| 3584 | /* Coefficients generated using FPMinimax. */ |
| 3585 | .poly = { V4 (0x1.55555p-2f), V4 (0x1.11166p-3f), V4 (0x1.b88a78p-5f), |
| 3586 | V4 (0x1.7b5756p-6f), V4 (0x1.4ef4cep-8f), V4 (0x1.0e1e74p-7f) }, |
| 3587 | - .neg_half_pi_1 = V4 (-0x1.921fb6p+0f), |
| 3588 | - .neg_half_pi_2 = V4 (0x1.777a5cp-25f), |
| 3589 | - .neg_half_pi_3 = V4 (0x1.ee59dap-50f), |
| 3590 | - .two_over_pi = V4 (0x1.45f306p-1f), |
| 3591 | + /* Stores constants: (-pi/2)_high, (-pi/2)_mid, (-pi/2)_low, and 2/pi. */ |
| 3592 | + .pi_consts |
| 3593 | + = { -0x1.921fb6p+0f, 0x1.777a5cp-25f, 0x1.ee59dap-50f, 0x1.45f306p-1f }, |
| 3594 | .shift = V4 (0x1.8p+23f), |
| 3595 | #if !WANT_SIMD_EXCEPT |
| 3596 | .range_val = V4 (0x1p15f), |
| 3597 | @@ -58,10 +58,11 @@ eval_poly (float32x4_t z, const struct data *d) |
| 3598 | { |
| 3599 | float32x4_t z2 = vmulq_f32 (z, z); |
| 3600 | #if WANT_SIMD_EXCEPT |
| 3601 | - /* Tiny z (<= 0x1p-31) will underflow when calculating z^4. If fp exceptions |
| 3602 | - are to be triggered correctly, sidestep this by fixing such lanes to 0. */ |
| 3603 | + /* Tiny z (<= 0x1p-31) will underflow when calculating z^4. |
| 3604 | + If fp exceptions are to be triggered correctly, |
| 3605 | + sidestep this by fixing such lanes to 0. */ |
| 3606 | uint32x4_t will_uflow |
| 3607 | - = vcleq_u32 (vreinterpretq_u32_f32 (vabsq_f32 (z)), TinyBound); |
| 3608 | + = vcleq_u32 (vreinterpretq_u32_f32 (vabsq_f32 (z)), TinyBound); |
| 3609 | if (__glibc_unlikely (v_any_u32 (will_uflow))) |
| 3610 | z2 = vbslq_f32 (will_uflow, v_f32 (0), z2); |
| 3611 | #endif |
| 3612 | @@ -94,16 +95,16 @@ float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (tan) (float32x4_t x) |
| 3613 | #endif |
| 3614 | |
| 3615 | /* n = rint(x/(pi/2)). */ |
| 3616 | - float32x4_t q = vfmaq_f32 (d->shift, d->two_over_pi, x); |
| 3617 | + float32x4_t q = vfmaq_laneq_f32 (d->shift, x, d->pi_consts, 3); |
| 3618 | float32x4_t n = vsubq_f32 (q, d->shift); |
| 3619 | /* Determine if x lives in an interval, where |tan(x)| grows to infinity. */ |
| 3620 | uint32x4_t pred_alt = vtstq_u32 (vreinterpretq_u32_f32 (q), v_u32 (1)); |
| 3621 | |
| 3622 | /* r = x - n * (pi/2) (range reduction into -pi./4 .. pi/4). */ |
| 3623 | float32x4_t r; |
| 3624 | - r = vfmaq_f32 (x, d->neg_half_pi_1, n); |
| 3625 | - r = vfmaq_f32 (r, d->neg_half_pi_2, n); |
| 3626 | - r = vfmaq_f32 (r, d->neg_half_pi_3, n); |
| 3627 | + r = vfmaq_laneq_f32 (x, n, d->pi_consts, 0); |
| 3628 | + r = vfmaq_laneq_f32 (r, n, d->pi_consts, 1); |
| 3629 | + r = vfmaq_laneq_f32 (r, n, d->pi_consts, 2); |
| 3630 | |
| 3631 | /* If x lives in an interval, where |tan(x)| |
| 3632 | - is finite, then use a polynomial approximation of the form |
| 3633 | diff --git a/sysdeps/aarch64/multiarch/init-arch.h b/sysdeps/aarch64/multiarch/init-arch.h |
| 3634 | index c52860efb2..61dc40088f 100644 |
| 3635 | --- a/sysdeps/aarch64/multiarch/init-arch.h |
| 3636 | +++ b/sysdeps/aarch64/multiarch/init-arch.h |
| 3637 | @@ -36,5 +36,7 @@ |
| 3638 | MTE_ENABLED (); \ |
| 3639 | bool __attribute__((unused)) sve = \ |
| 3640 | GLRO(dl_aarch64_cpu_features).sve; \ |
| 3641 | + bool __attribute__((unused)) prefer_sve_ifuncs = \ |
| 3642 | + GLRO(dl_aarch64_cpu_features).prefer_sve_ifuncs; \ |
| 3643 | bool __attribute__((unused)) mops = \ |
| 3644 | GLRO(dl_aarch64_cpu_features).mops; |
| 3645 | diff --git a/sysdeps/aarch64/multiarch/memcpy.c b/sysdeps/aarch64/multiarch/memcpy.c |
| 3646 | index d12eccfca5..ce53567dab 100644 |
| 3647 | --- a/sysdeps/aarch64/multiarch/memcpy.c |
| 3648 | +++ b/sysdeps/aarch64/multiarch/memcpy.c |
| 3649 | @@ -47,7 +47,7 @@ select_memcpy_ifunc (void) |
| 3650 | { |
| 3651 | if (IS_A64FX (midr)) |
| 3652 | return __memcpy_a64fx; |
| 3653 | - return __memcpy_sve; |
| 3654 | + return prefer_sve_ifuncs ? __memcpy_sve : __memcpy_generic; |
| 3655 | } |
| 3656 | |
| 3657 | if (IS_THUNDERX (midr)) |
| 3658 | diff --git a/sysdeps/aarch64/multiarch/memmove.c b/sysdeps/aarch64/multiarch/memmove.c |
| 3659 | index 2081eeb4d4..fe95037be3 100644 |
| 3660 | --- a/sysdeps/aarch64/multiarch/memmove.c |
| 3661 | +++ b/sysdeps/aarch64/multiarch/memmove.c |
| 3662 | @@ -47,7 +47,7 @@ select_memmove_ifunc (void) |
| 3663 | { |
| 3664 | if (IS_A64FX (midr)) |
| 3665 | return __memmove_a64fx; |
| 3666 | - return __memmove_sve; |
| 3667 | + return prefer_sve_ifuncs ? __memmove_sve : __memmove_generic; |
| 3668 | } |
| 3669 | |
| 3670 | if (IS_THUNDERX (midr)) |
| 3671 | diff --git a/sysdeps/aarch64/multiarch/memset_generic.S b/sysdeps/aarch64/multiarch/memset_generic.S |
| 3672 | index 81748bdbce..e125a5ed85 100644 |
| 3673 | --- a/sysdeps/aarch64/multiarch/memset_generic.S |
| 3674 | +++ b/sysdeps/aarch64/multiarch/memset_generic.S |
| 3675 | @@ -33,3 +33,7 @@ |
| 3676 | #endif |
| 3677 | |
| 3678 | #include <../memset.S> |
| 3679 | + |
| 3680 | +#if IS_IN (rtld) |
| 3681 | +strong_alias (memset, __memset_generic) |
| 3682 | +#endif |
| 3683 | diff --git a/sysdeps/aarch64/preconfigure b/sysdeps/aarch64/preconfigure |
| 3684 | index d9bd1f8558..19657b627b 100644 |
| 3685 | --- a/sysdeps/aarch64/preconfigure |
| 3686 | +++ b/sysdeps/aarch64/preconfigure |
| 3687 | @@ -2,5 +2,6 @@ case "$machine" in |
| 3688 | aarch64*) |
| 3689 | base_machine=aarch64 |
| 3690 | machine=aarch64 |
| 3691 | + mtls_descriptor=desc |
| 3692 | ;; |
| 3693 | esac |
| 3694 | diff --git a/sysdeps/arc/utmp-size.h b/sysdeps/arc/utmp-size.h |
| 3695 | new file mode 100644 |
| 3696 | index 0000000000..a247fcd3da |
| 3697 | --- /dev/null |
| 3698 | +++ b/sysdeps/arc/utmp-size.h |
| 3699 | @@ -0,0 +1,3 @@ |
| 3700 | +/* arc has less padding than other architectures with 64-bit time_t. */ |
| 3701 | +#define UTMP_SIZE 392 |
| 3702 | +#define LASTLOG_SIZE 296 |
| 3703 | diff --git a/sysdeps/arm/Makefile b/sysdeps/arm/Makefile |
| 3704 | index d5cea717a9..619474eca9 100644 |
| 3705 | --- a/sysdeps/arm/Makefile |
| 3706 | +++ b/sysdeps/arm/Makefile |
| 3707 | @@ -13,15 +13,15 @@ $(objpfx)libgcc-stubs.a: $(objpfx)aeabi_unwind_cpp_pr1.os |
| 3708 | lib-noranlib: $(objpfx)libgcc-stubs.a |
| 3709 | |
| 3710 | ifeq ($(build-shared),yes) |
| 3711 | -ifeq (yes,$(have-mtls-dialect-gnu2)) |
| 3712 | +ifneq (no,$(have-mtls-descriptor)) |
| 3713 | tests += tst-armtlsdescloc tst-armtlsdescextnow tst-armtlsdescextlazy |
| 3714 | modules-names += tst-armtlsdesclocmod |
| 3715 | modules-names += tst-armtlsdescextlazymod tst-armtlsdescextnowmod |
| 3716 | CPPFLAGS-tst-armtlsdescextnowmod.c += -Dstatic= |
| 3717 | CPPFLAGS-tst-armtlsdescextlazymod.c += -Dstatic= |
| 3718 | -CFLAGS-tst-armtlsdesclocmod.c += -mtls-dialect=gnu2 |
| 3719 | -CFLAGS-tst-armtlsdescextnowmod.c += -mtls-dialect=gnu2 |
| 3720 | -CFLAGS-tst-armtlsdescextlazymod.c += -mtls-dialect=gnu2 |
| 3721 | +CFLAGS-tst-armtlsdesclocmod.c += -mtls-dialect=$(have-mtls-descriptor) |
| 3722 | +CFLAGS-tst-armtlsdescextnowmod.c += -mtls-dialect=$(have-mtls-descriptor) |
| 3723 | +CFLAGS-tst-armtlsdescextlazymod.c += -mtls-dialect=$(have-mtls-descriptor) |
| 3724 | LDFLAGS-tst-armtlsdescextnowmod.so += -Wl,-z,now |
| 3725 | tst-armtlsdescloc-ENV = LD_BIND_NOW=1 |
| 3726 | tst-armtlsdescextnow-ENV = LD_BIND_NOW=1 |
| 3727 | diff --git a/sysdeps/arm/bits/wordsize.h b/sysdeps/arm/bits/wordsize.h |
| 3728 | new file mode 100644 |
| 3729 | index 0000000000..6ecbfe7c86 |
| 3730 | --- /dev/null |
| 3731 | +++ b/sysdeps/arm/bits/wordsize.h |
| 3732 | @@ -0,0 +1,21 @@ |
| 3733 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 3734 | + This file is part of the GNU C Library. |
| 3735 | + |
| 3736 | + The GNU C Library is free software; you can redistribute it and/or |
| 3737 | + modify it under the terms of the GNU Lesser General Public |
| 3738 | + License as published by the Free Software Foundation; either |
| 3739 | + version 2.1 of the License, or (at your option) any later version. |
| 3740 | + |
| 3741 | + The GNU C Library is distributed in the hope that it will be useful, |
| 3742 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 3743 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 3744 | + Lesser General Public License for more details. |
| 3745 | + |
| 3746 | + You should have received a copy of the GNU Lesser General Public |
| 3747 | + License along with the GNU C Library; if not, see |
| 3748 | + <https://www.gnu.org/licenses/>. */ |
| 3749 | + |
| 3750 | +#define __WORDSIZE 32 |
| 3751 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 3752 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 3753 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 3754 | diff --git a/sysdeps/arm/configure b/sysdeps/arm/configure |
| 3755 | index 35e2918922..4ef4d46cbd 100644 |
| 3756 | --- a/sysdeps/arm/configure |
| 3757 | +++ b/sysdeps/arm/configure |
| 3758 | @@ -187,6 +187,38 @@ else |
| 3759 | default-abi = soft" |
| 3760 | fi |
| 3761 | |
| 3762 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking whether VFP supports 32 registers" >&5 |
| 3763 | +printf %s "checking whether VFP supports 32 registers... " >&6; } |
| 3764 | +if test ${libc_cv_arm_pcs_vfp_d32+y} |
| 3765 | +then : |
| 3766 | + printf %s "(cached) " >&6 |
| 3767 | +else $as_nop |
| 3768 | + |
| 3769 | +cat confdefs.h - <<_ACEOF >conftest.$ac_ext |
| 3770 | +/* end confdefs.h. */ |
| 3771 | + |
| 3772 | +void foo (void) |
| 3773 | +{ |
| 3774 | + asm volatile ("vldr d16,=17" : : : "d16"); |
| 3775 | +} |
| 3776 | + |
| 3777 | +_ACEOF |
| 3778 | +if ac_fn_c_try_compile "$LINENO" |
| 3779 | +then : |
| 3780 | + libc_cv_arm_pcs_vfp_d32=yes |
| 3781 | +else $as_nop |
| 3782 | + libc_cv_arm_pcs_vfp_d32=no |
| 3783 | +fi |
| 3784 | +rm -f core conftest.err conftest.$ac_objext conftest.beam conftest.$ac_ext |
| 3785 | +fi |
| 3786 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: result: $libc_cv_arm_pcs_vfp_d32" >&5 |
| 3787 | +printf "%s\n" "$libc_cv_arm_pcs_vfp_d32" >&6; } |
| 3788 | +if test "$libc_cv_arm_pcs_vfp_d32" = yes ; |
| 3789 | +then |
| 3790 | + printf "%s\n" "#define HAVE_ARM_PCS_VFP_D32 1" >>confdefs.h |
| 3791 | + |
| 3792 | +fi |
| 3793 | + |
| 3794 | { printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking whether PC-relative relocs in movw/movt work properly" >&5 |
| 3795 | printf %s "checking whether PC-relative relocs in movw/movt work properly... " >&6; } |
| 3796 | if test ${libc_cv_arm_pcrel_movw+y} |
| 3797 | diff --git a/sysdeps/arm/configure.ac b/sysdeps/arm/configure.ac |
| 3798 | index 5172e30bbe..cd00ddc9d9 100644 |
| 3799 | --- a/sysdeps/arm/configure.ac |
| 3800 | +++ b/sysdeps/arm/configure.ac |
| 3801 | @@ -21,6 +21,21 @@ else |
| 3802 | LIBC_CONFIG_VAR([default-abi], [soft]) |
| 3803 | fi |
| 3804 | |
| 3805 | +AC_CACHE_CHECK([whether VFP supports 32 registers], |
| 3806 | + libc_cv_arm_pcs_vfp_d32, [ |
| 3807 | +AC_COMPILE_IFELSE([AC_LANG_SOURCE([[ |
| 3808 | +void foo (void) |
| 3809 | +{ |
| 3810 | + asm volatile ("vldr d16,=17" : : : "d16"); |
| 3811 | +} |
| 3812 | +]])], |
| 3813 | + [libc_cv_arm_pcs_vfp_d32=yes], |
| 3814 | + [libc_cv_arm_pcs_vfp_d32=no])]) |
| 3815 | +if test "$libc_cv_arm_pcs_vfp_d32" = yes ; |
| 3816 | +then |
| 3817 | + AC_DEFINE(HAVE_ARM_PCS_VFP_D32) |
| 3818 | +fi |
| 3819 | + |
| 3820 | AC_CACHE_CHECK([whether PC-relative relocs in movw/movt work properly], |
| 3821 | libc_cv_arm_pcrel_movw, [ |
| 3822 | cat > conftest.s <<\EOF |
| 3823 | diff --git a/sysdeps/arm/dl-machine.h b/sysdeps/arm/dl-machine.h |
| 3824 | index b857bbc868..dd1a0f6b6e 100644 |
| 3825 | --- a/sysdeps/arm/dl-machine.h |
| 3826 | +++ b/sysdeps/arm/dl-machine.h |
| 3827 | @@ -139,7 +139,6 @@ _start:\n\ |
| 3828 | _dl_start_user:\n\ |
| 3829 | adr r6, .L_GET_GOT\n\ |
| 3830 | add sl, sl, r6\n\ |
| 3831 | - ldr r4, [sl, r4]\n\ |
| 3832 | @ save the entry point in another register\n\ |
| 3833 | mov r6, r0\n\ |
| 3834 | @ get the original arg count\n\ |
| 3835 | diff --git a/sysdeps/arm/dl-tlsdesc.S b/sysdeps/arm/dl-tlsdesc.S |
| 3836 | index 764c56e70f..ada106521d 100644 |
| 3837 | --- a/sysdeps/arm/dl-tlsdesc.S |
| 3838 | +++ b/sysdeps/arm/dl-tlsdesc.S |
| 3839 | @@ -19,6 +19,7 @@ |
| 3840 | #include <sysdep.h> |
| 3841 | #include <arm-features.h> |
| 3842 | #include <tls.h> |
| 3843 | +#include <rtld-global-offsets.h> |
| 3844 | #include "tlsdesc.h" |
| 3845 | |
| 3846 | .text |
| 3847 | @@ -83,14 +84,20 @@ _dl_tlsdesc_dynamic(struct tlsdesc *tdp) |
| 3848 | .align 2 |
| 3849 | _dl_tlsdesc_dynamic: |
| 3850 | /* Our calling convention is to clobber r0, r1 and the processor |
| 3851 | - flags. All others that are modified must be saved */ |
| 3852 | - eabi_save ({r2,r3,r4,lr}) |
| 3853 | - push {r2,r3,r4,lr} |
| 3854 | - cfi_adjust_cfa_offset (16) |
| 3855 | + flags. All others that are modified must be saved. r5 is |
| 3856 | + used as the hwcap value to avoid reload after __tls_get_addr |
| 3857 | + call. If required we will save the vector register on the slow |
| 3858 | + path. */ |
| 3859 | + eabi_save ({r2,r3,r4,r5,ip,lr}) |
| 3860 | + push {r2,r3,r4,r5,ip,lr} |
| 3861 | + cfi_adjust_cfa_offset (24) |
| 3862 | cfi_rel_offset (r2,0) |
| 3863 | cfi_rel_offset (r3,4) |
| 3864 | cfi_rel_offset (r4,8) |
| 3865 | - cfi_rel_offset (lr,12) |
| 3866 | + cfi_rel_offset (r5,12) |
| 3867 | + cfi_rel_offset (ip,16) |
| 3868 | + cfi_rel_offset (lr,20) |
| 3869 | + |
| 3870 | ldr r1, [r0] /* td */ |
| 3871 | GET_TLS (lr) |
| 3872 | mov r4, r0 /* r4 = tp */ |
| 3873 | @@ -113,22 +120,69 @@ _dl_tlsdesc_dynamic: |
| 3874 | rsbne r0, r4, r3 |
| 3875 | bne 2f |
| 3876 | 1: mov r0, r1 |
| 3877 | + |
| 3878 | + /* Load the hwcap to check for vector support. */ |
| 3879 | + ldr r2, 3f |
| 3880 | + ldr r1, .Lrtld_global_ro |
| 3881 | +0: add r2, pc, r2 |
| 3882 | + ldr r2, [r2, r1] |
| 3883 | + ldr r5, [r2, #RTLD_GLOBAL_RO_DL_HWCAP_OFFSET] |
| 3884 | + |
| 3885 | +#ifdef __SOFTFP__ |
| 3886 | + tst r5, #HWCAP_ARM_VFP |
| 3887 | + beq .Lno_vfp |
| 3888 | +#endif |
| 3889 | + |
| 3890 | + /* Store the VFP registers. Don't use VFP instructions directly |
| 3891 | + because this code is used in non-VFP multilibs. */ |
| 3892 | +#define VFP_STACK_REQ (32*8 + 8) |
| 3893 | + sub sp, sp, VFP_STACK_REQ |
| 3894 | + cfi_adjust_cfa_offset (VFP_STACK_REQ) |
| 3895 | + mov r3, sp |
| 3896 | + .inst 0xeca30b20 /* vstmia r3!, {d0-d15} */ |
| 3897 | + tst r5, #HWCAP_ARM_VFPD32 |
| 3898 | + beq 4f |
| 3899 | + .inst 0xece30b20 /* vstmia r3!, {d16-d31} */ |
| 3900 | + /* Store the floating-point status register. */ |
| 3901 | +4: .inst 0xeef12a10 /* vmrs r2, fpscr */ |
| 3902 | + str r2, [r3] |
| 3903 | +.Lno_vfp: |
| 3904 | bl __tls_get_addr |
| 3905 | rsb r0, r4, r0 |
| 3906 | +#ifdef __SOFTFP__ |
| 3907 | + tst r5, #HWCAP_ARM_VFP |
| 3908 | + beq 2f |
| 3909 | +#endif |
| 3910 | + mov r3, sp |
| 3911 | + .inst 0xecb30b20 /* vldmia r3!, {d0-d15} */ |
| 3912 | + tst r5, #HWCAP_ARM_VFPD32 |
| 3913 | + beq 5f |
| 3914 | + .inst 0xecf30b20 /* vldmia r3!, {d16-d31} */ |
| 3915 | + ldr r4, [r3] |
| 3916 | +5: .inst 0xeee14a10 /* vmsr fpscr, r4 */ |
| 3917 | + add sp, sp, VFP_STACK_REQ |
| 3918 | + cfi_adjust_cfa_offset (-VFP_STACK_REQ) |
| 3919 | + |
| 3920 | 2: |
| 3921 | #if ((defined (__ARM_ARCH_4T__) && defined (__THUMB_INTERWORK__)) \ |
| 3922 | || defined (ARM_ALWAYS_BX)) |
| 3923 | - pop {r2,r3,r4, lr} |
| 3924 | - cfi_adjust_cfa_offset (-16) |
| 3925 | + pop {r2,r3,r4,r5,ip, lr} |
| 3926 | + cfi_adjust_cfa_offset (-20) |
| 3927 | cfi_restore (lr) |
| 3928 | + cfi_restore (ip) |
| 3929 | + cfi_restore (r5) |
| 3930 | cfi_restore (r4) |
| 3931 | cfi_restore (r3) |
| 3932 | cfi_restore (r2) |
| 3933 | bx lr |
| 3934 | #else |
| 3935 | - pop {r2,r3,r4, pc} |
| 3936 | + pop {r2,r3,r4,r5,ip, pc} |
| 3937 | #endif |
| 3938 | eabi_fnend |
| 3939 | cfi_endproc |
| 3940 | .size _dl_tlsdesc_dynamic, .-_dl_tlsdesc_dynamic |
| 3941 | + |
| 3942 | +3: .long _GLOBAL_OFFSET_TABLE_ - 0b - PC_OFS |
| 3943 | +.Lrtld_global_ro: |
| 3944 | + .long C_SYMBOL_NAME(_rtld_global_ro)(GOT) |
| 3945 | #endif /* SHARED */ |
| 3946 | diff --git a/sysdeps/arm/tst-gnu2-tls2.h b/sysdeps/arm/tst-gnu2-tls2.h |
| 3947 | new file mode 100644 |
| 3948 | index 0000000000..e413ac21fb |
| 3949 | --- /dev/null |
| 3950 | +++ b/sysdeps/arm/tst-gnu2-tls2.h |
| 3951 | @@ -0,0 +1,128 @@ |
| 3952 | +/* Test TLSDESC relocation. ARM version. |
| 3953 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 3954 | + This file is part of the GNU C Library. |
| 3955 | + |
| 3956 | + The GNU C Library is free software; you can redistribute it and/or |
| 3957 | + modify it under the terms of the GNU Lesser General Public |
| 3958 | + License as published by the Free Software Foundation; either |
| 3959 | + version 2.1 of the License, or (at your option) any later version. |
| 3960 | + |
| 3961 | + The GNU C Library is distributed in the hope that it will be useful, |
| 3962 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 3963 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 3964 | + Lesser General Public License for more details. |
| 3965 | + |
| 3966 | + You should have received a copy of the GNU Lesser General Public |
| 3967 | + License along with the GNU C Library; if not, see |
| 3968 | + <https://www.gnu.org/licenses/>. */ |
| 3969 | + |
| 3970 | +#include <config.h> |
| 3971 | +#include <sys/auxv.h> |
| 3972 | +#include <string.h> |
| 3973 | +#include <stdlib.h> |
| 3974 | +#include <endian.h> |
| 3975 | + |
| 3976 | +#ifndef __SOFTFP__ |
| 3977 | + |
| 3978 | +# ifdef HAVE_ARM_PCS_VFP_D32 |
| 3979 | +# define SAVE_VFP_D32 \ |
| 3980 | + asm volatile ("vldr d16,=17" : : : "d16"); \ |
| 3981 | + asm volatile ("vldr d17,=18" : : : "d17"); \ |
| 3982 | + asm volatile ("vldr d18,=19" : : : "d18"); \ |
| 3983 | + asm volatile ("vldr d19,=20" : : : "d19"); \ |
| 3984 | + asm volatile ("vldr d20,=21" : : : "d20"); \ |
| 3985 | + asm volatile ("vldr d21,=22" : : : "d21"); \ |
| 3986 | + asm volatile ("vldr d22,=23" : : : "d22"); \ |
| 3987 | + asm volatile ("vldr d23,=24" : : : "d23"); \ |
| 3988 | + asm volatile ("vldr d24,=25" : : : "d24"); \ |
| 3989 | + asm volatile ("vldr d25,=26" : : : "d25"); \ |
| 3990 | + asm volatile ("vldr d26,=27" : : : "d26"); \ |
| 3991 | + asm volatile ("vldr d27,=28" : : : "d27"); \ |
| 3992 | + asm volatile ("vldr d28,=29" : : : "d28"); \ |
| 3993 | + asm volatile ("vldr d29,=30" : : : "d29"); \ |
| 3994 | + asm volatile ("vldr d30,=31" : : : "d30"); \ |
| 3995 | + asm volatile ("vldr d31,=32" : : : "d31"); |
| 3996 | +# else |
| 3997 | +# define SAVE_VFP_D32 |
| 3998 | +# endif |
| 3999 | + |
| 4000 | +# define INIT_TLSDESC_CALL() \ |
| 4001 | + unsigned long hwcap = getauxval (AT_HWCAP) |
| 4002 | + |
| 4003 | +/* Set each vector register to a value from 1 to 32 before the TLS access, |
| 4004 | + dump to memory after TLS access, and compare with the expected values. */ |
| 4005 | + |
| 4006 | +# define BEFORE_TLSDESC_CALL() \ |
| 4007 | + if (hwcap & HWCAP_ARM_VFP) \ |
| 4008 | + { \ |
| 4009 | + asm volatile ("vldr d0,=1" : : : "d0"); \ |
| 4010 | + asm volatile ("vldr d1,=2" : : : "d1"); \ |
| 4011 | + asm volatile ("vldr d2,=3" : : : "d1"); \ |
| 4012 | + asm volatile ("vldr d3,=4" : : : "d3"); \ |
| 4013 | + asm volatile ("vldr d4,=5" : : : "d4"); \ |
| 4014 | + asm volatile ("vldr d5,=6" : : : "d5"); \ |
| 4015 | + asm volatile ("vldr d6,=7" : : : "d6"); \ |
| 4016 | + asm volatile ("vldr d7,=8" : : : "d7"); \ |
| 4017 | + asm volatile ("vldr d8,=9" : : : "d8"); \ |
| 4018 | + asm volatile ("vldr d9,=10" : : : "d9"); \ |
| 4019 | + asm volatile ("vldr d10,=11" : : : "d10"); \ |
| 4020 | + asm volatile ("vldr d11,=12" : : : "d11"); \ |
| 4021 | + asm volatile ("vldr d12,=13" : : : "d12"); \ |
| 4022 | + asm volatile ("vldr d13,=14" : : : "d13"); \ |
| 4023 | + asm volatile ("vldr d14,=15" : : : "d14"); \ |
| 4024 | + asm volatile ("vldr d15,=16" : : : "d15"); \ |
| 4025 | + } \ |
| 4026 | + if (hwcap & HWCAP_ARM_VFPD32) \ |
| 4027 | + { \ |
| 4028 | + SAVE_VFP_D32 \ |
| 4029 | + } |
| 4030 | + |
| 4031 | +# define VFP_STACK_REQ (16*8) |
| 4032 | +# if __BYTE_ORDER == __BIG_ENDIAN |
| 4033 | +# define DISP 7 |
| 4034 | +# else |
| 4035 | +# define DISP 0 |
| 4036 | +# endif |
| 4037 | + |
| 4038 | +# ifdef HAVE_ARM_PCS_VFP_D32 |
| 4039 | +# define CHECK_VFP_D32 \ |
| 4040 | + char vfp[VFP_STACK_REQ]; \ |
| 4041 | + asm volatile ("vstmia %0, {d16-d31}\n" \ |
| 4042 | + : \ |
| 4043 | + : "r" (vfp) \ |
| 4044 | + : "memory"); \ |
| 4045 | + \ |
| 4046 | + char expected[VFP_STACK_REQ] = { 0 }; \ |
| 4047 | + for (int i = 0; i < 16; ++i) \ |
| 4048 | + expected[i * 8 + DISP] = i + 17; \ |
| 4049 | + \ |
| 4050 | + if (memcmp (vfp, expected, VFP_STACK_REQ) != 0) \ |
| 4051 | + abort (); |
| 4052 | +# else |
| 4053 | +# define CHECK_VFP_D32 |
| 4054 | +# endif |
| 4055 | + |
| 4056 | +# define AFTER_TLSDESC_CALL() \ |
| 4057 | + if (hwcap & HWCAP_ARM_VFP) \ |
| 4058 | + { \ |
| 4059 | + char vfp[VFP_STACK_REQ]; \ |
| 4060 | + asm volatile ("vstmia %0, {d0-d15}\n" \ |
| 4061 | + : \ |
| 4062 | + : "r" (vfp) \ |
| 4063 | + : "memory"); \ |
| 4064 | + \ |
| 4065 | + char expected[VFP_STACK_REQ] = { 0 }; \ |
| 4066 | + for (int i = 0; i < 16; ++i) \ |
| 4067 | + expected[i * 8 + DISP] = i + 1; \ |
| 4068 | + \ |
| 4069 | + if (memcmp (vfp, expected, VFP_STACK_REQ) != 0) \ |
| 4070 | + abort (); \ |
| 4071 | + } \ |
| 4072 | + if (hwcap & HWCAP_ARM_VFPD32) \ |
| 4073 | + { \ |
| 4074 | + CHECK_VFP_D32 \ |
| 4075 | + } |
| 4076 | + |
| 4077 | +#endif /* __SOFTFP__ */ |
| 4078 | + |
| 4079 | +#include_next <tst-gnu2-tls2.h> |
| 4080 | diff --git a/sysdeps/arm/utmp-size.h b/sysdeps/arm/utmp-size.h |
| 4081 | new file mode 100644 |
| 4082 | index 0000000000..8f21ebe1b6 |
| 4083 | --- /dev/null |
| 4084 | +++ b/sysdeps/arm/utmp-size.h |
| 4085 | @@ -0,0 +1,2 @@ |
| 4086 | +#define UTMP_SIZE 384 |
| 4087 | +#define LASTLOG_SIZE 292 |
| 4088 | diff --git a/sysdeps/csky/bits/wordsize.h b/sysdeps/csky/bits/wordsize.h |
| 4089 | new file mode 100644 |
| 4090 | index 0000000000..6ecbfe7c86 |
| 4091 | --- /dev/null |
| 4092 | +++ b/sysdeps/csky/bits/wordsize.h |
| 4093 | @@ -0,0 +1,21 @@ |
| 4094 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 4095 | + This file is part of the GNU C Library. |
| 4096 | + |
| 4097 | + The GNU C Library is free software; you can redistribute it and/or |
| 4098 | + modify it under the terms of the GNU Lesser General Public |
| 4099 | + License as published by the Free Software Foundation; either |
| 4100 | + version 2.1 of the License, or (at your option) any later version. |
| 4101 | + |
| 4102 | + The GNU C Library is distributed in the hope that it will be useful, |
| 4103 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 4104 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 4105 | + Lesser General Public License for more details. |
| 4106 | + |
| 4107 | + You should have received a copy of the GNU Lesser General Public |
| 4108 | + License along with the GNU C Library; if not, see |
| 4109 | + <https://www.gnu.org/licenses/>. */ |
| 4110 | + |
| 4111 | +#define __WORDSIZE 32 |
| 4112 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 4113 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 4114 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 4115 | diff --git a/sysdeps/csky/utmp-size.h b/sysdeps/csky/utmp-size.h |
| 4116 | new file mode 100644 |
| 4117 | index 0000000000..8f21ebe1b6 |
| 4118 | --- /dev/null |
| 4119 | +++ b/sysdeps/csky/utmp-size.h |
| 4120 | @@ -0,0 +1,2 @@ |
| 4121 | +#define UTMP_SIZE 384 |
| 4122 | +#define LASTLOG_SIZE 292 |
| 4123 | diff --git a/sysdeps/generic/ldsodefs.h b/sysdeps/generic/ldsodefs.h |
| 4124 | index 117c901ccc..50f58a60e3 100644 |
| 4125 | --- a/sysdeps/generic/ldsodefs.h |
| 4126 | +++ b/sysdeps/generic/ldsodefs.h |
| 4127 | @@ -646,6 +646,8 @@ struct rtld_global_ro |
| 4128 | /* Mask for more hardware capabilities that are available on some |
| 4129 | platforms. */ |
| 4130 | EXTERN uint64_t _dl_hwcap2; |
| 4131 | + EXTERN uint64_t _dl_hwcap3; |
| 4132 | + EXTERN uint64_t _dl_hwcap4; |
| 4133 | |
| 4134 | EXTERN enum dso_sort_algorithm _dl_dso_sort_algo; |
| 4135 | |
| 4136 | diff --git a/sysdeps/generic/utmp-size.h b/sysdeps/generic/utmp-size.h |
| 4137 | new file mode 100644 |
| 4138 | index 0000000000..89dbe878b0 |
| 4139 | --- /dev/null |
| 4140 | +++ b/sysdeps/generic/utmp-size.h |
| 4141 | @@ -0,0 +1,23 @@ |
| 4142 | +/* Expected sizes of utmp-related structures stored in files. 64-bit version. |
| 4143 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 4144 | + This file is part of the GNU C Library. |
| 4145 | + |
| 4146 | + The GNU C Library is free software; you can redistribute it and/or |
| 4147 | + modify it under the terms of the GNU Lesser General Public |
| 4148 | + License as published by the Free Software Foundation; either |
| 4149 | + version 2.1 of the License, or (at your option) any later version. |
| 4150 | + |
| 4151 | + The GNU C Library is distributed in the hope that it will be useful, |
| 4152 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 4153 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 4154 | + Lesser General Public License for more details. |
| 4155 | + |
| 4156 | + You should have received a copy of the GNU Lesser General Public |
| 4157 | + License along with the GNU C Library; if not, see |
| 4158 | + <https://www.gnu.org/licenses/>. */ |
| 4159 | + |
| 4160 | +/* Expected size, in bytes, of struct utmp and struct utmpx. */ |
| 4161 | +#define UTMP_SIZE 400 |
| 4162 | + |
| 4163 | +/* Expected size, in bytes, of struct lastlog. */ |
| 4164 | +#define LASTLOG_SIZE 296 |
| 4165 | diff --git a/sysdeps/hppa/utmp-size.h b/sysdeps/hppa/utmp-size.h |
| 4166 | new file mode 100644 |
| 4167 | index 0000000000..8f21ebe1b6 |
| 4168 | --- /dev/null |
| 4169 | +++ b/sysdeps/hppa/utmp-size.h |
| 4170 | @@ -0,0 +1,2 @@ |
| 4171 | +#define UTMP_SIZE 384 |
| 4172 | +#define LASTLOG_SIZE 292 |
| 4173 | diff --git a/sysdeps/i386/dl-machine.h b/sysdeps/i386/dl-machine.h |
| 4174 | index fc1ef96587..50d74fe6e9 100644 |
| 4175 | --- a/sysdeps/i386/dl-machine.h |
| 4176 | +++ b/sysdeps/i386/dl-machine.h |
| 4177 | @@ -347,7 +347,7 @@ and creates an unsatisfiable circular dependency.\n", |
| 4178 | { |
| 4179 | td->arg = _dl_make_tlsdesc_dynamic |
| 4180 | (sym_map, sym->st_value + (ElfW(Word))td->arg); |
| 4181 | - td->entry = _dl_tlsdesc_dynamic; |
| 4182 | + td->entry = GLRO(dl_x86_tlsdesc_dynamic); |
| 4183 | } |
| 4184 | else |
| 4185 | # endif |
| 4186 | diff --git a/sysdeps/i386/dl-tlsdesc-dynamic.h b/sysdeps/i386/dl-tlsdesc-dynamic.h |
| 4187 | new file mode 100644 |
| 4188 | index 0000000000..3627028577 |
| 4189 | --- /dev/null |
| 4190 | +++ b/sysdeps/i386/dl-tlsdesc-dynamic.h |
| 4191 | @@ -0,0 +1,190 @@ |
| 4192 | +/* Thread-local storage handling in the ELF dynamic linker. i386 version. |
| 4193 | + Copyright (C) 2004-2024 Free Software Foundation, Inc. |
| 4194 | + This file is part of the GNU C Library. |
| 4195 | + |
| 4196 | + The GNU C Library is free software; you can redistribute it and/or |
| 4197 | + modify it under the terms of the GNU Lesser General Public |
| 4198 | + License as published by the Free Software Foundation; either |
| 4199 | + version 2.1 of the License, or (at your option) any later version. |
| 4200 | + |
| 4201 | + The GNU C Library is distributed in the hope that it will be useful, |
| 4202 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 4203 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 4204 | + Lesser General Public License for more details. |
| 4205 | + |
| 4206 | + You should have received a copy of the GNU Lesser General Public |
| 4207 | + License along with the GNU C Library; if not, see |
| 4208 | + <https://www.gnu.org/licenses/>. */ |
| 4209 | + |
| 4210 | +#undef REGISTER_SAVE_AREA |
| 4211 | + |
| 4212 | +#if !defined USE_FNSAVE && (STATE_SAVE_ALIGNMENT % 16) != 0 |
| 4213 | +# error STATE_SAVE_ALIGNMENT must be multiple of 16 |
| 4214 | +#endif |
| 4215 | + |
| 4216 | +#if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 4217 | +# ifdef USE_FNSAVE |
| 4218 | +# error USE_FNSAVE shouldn't be defined |
| 4219 | +# endif |
| 4220 | +# ifdef USE_FXSAVE |
| 4221 | +/* Use fxsave to save all registers. */ |
| 4222 | +# define REGISTER_SAVE_AREA 512 |
| 4223 | +# endif |
| 4224 | +#else |
| 4225 | +# ifdef USE_FNSAVE |
| 4226 | +/* Use fnsave to save x87 FPU stack registers. */ |
| 4227 | +# define REGISTER_SAVE_AREA 108 |
| 4228 | +# else |
| 4229 | +# ifndef USE_FXSAVE |
| 4230 | +# error USE_FXSAVE must be defined |
| 4231 | +# endif |
| 4232 | +/* Use fxsave to save all registers. Add 12 bytes to align the stack |
| 4233 | + to 16 bytes. */ |
| 4234 | +# define REGISTER_SAVE_AREA (512 + 12) |
| 4235 | +# endif |
| 4236 | +#endif |
| 4237 | + |
| 4238 | + .hidden _dl_tlsdesc_dynamic |
| 4239 | + .global _dl_tlsdesc_dynamic |
| 4240 | + .type _dl_tlsdesc_dynamic,@function |
| 4241 | + |
| 4242 | + /* This function is used for symbols that need dynamic TLS. |
| 4243 | + |
| 4244 | + %eax points to the TLS descriptor, such that 0(%eax) points to |
| 4245 | + _dl_tlsdesc_dynamic itself, and 4(%eax) points to a struct |
| 4246 | + tlsdesc_dynamic_arg object. It must return in %eax the offset |
| 4247 | + between the thread pointer and the object denoted by the |
| 4248 | + argument, without clobbering any registers. |
| 4249 | + |
| 4250 | + The assembly code that follows is a rendition of the following |
| 4251 | + C code, hand-optimized a little bit. |
| 4252 | + |
| 4253 | +ptrdiff_t |
| 4254 | +__attribute__ ((__regparm__ (1))) |
| 4255 | +_dl_tlsdesc_dynamic (struct tlsdesc *tdp) |
| 4256 | +{ |
| 4257 | + struct tlsdesc_dynamic_arg *td = tdp->arg; |
| 4258 | + dtv_t *dtv = *(dtv_t **)((char *)__thread_pointer + DTV_OFFSET); |
| 4259 | + if (__builtin_expect (td->gen_count <= dtv[0].counter |
| 4260 | + && (dtv[td->tlsinfo.ti_module].pointer.val |
| 4261 | + != TLS_DTV_UNALLOCATED), |
| 4262 | + 1)) |
| 4263 | + return dtv[td->tlsinfo.ti_module].pointer.val + td->tlsinfo.ti_offset |
| 4264 | + - __thread_pointer; |
| 4265 | + |
| 4266 | + return ___tls_get_addr (&td->tlsinfo) - __thread_pointer; |
| 4267 | +} |
| 4268 | +*/ |
| 4269 | + cfi_startproc |
| 4270 | + .align 16 |
| 4271 | +_dl_tlsdesc_dynamic: |
| 4272 | + /* Like all TLS resolvers, preserve call-clobbered registers. |
| 4273 | + We need two scratch regs anyway. */ |
| 4274 | + subl $32, %esp |
| 4275 | + cfi_adjust_cfa_offset (32) |
| 4276 | + movl %ecx, 20(%esp) |
| 4277 | + movl %edx, 24(%esp) |
| 4278 | + movl TLSDESC_ARG(%eax), %eax |
| 4279 | + movl %gs:DTV_OFFSET, %edx |
| 4280 | + movl TLSDESC_GEN_COUNT(%eax), %ecx |
| 4281 | + cmpl (%edx), %ecx |
| 4282 | + ja 2f |
| 4283 | + movl TLSDESC_MODID(%eax), %ecx |
| 4284 | + movl (%edx,%ecx,8), %edx |
| 4285 | + cmpl $-1, %edx |
| 4286 | + je 2f |
| 4287 | + movl TLSDESC_MODOFF(%eax), %eax |
| 4288 | + addl %edx, %eax |
| 4289 | +1: |
| 4290 | + movl 20(%esp), %ecx |
| 4291 | + subl %gs:0, %eax |
| 4292 | + movl 24(%esp), %edx |
| 4293 | + addl $32, %esp |
| 4294 | + cfi_adjust_cfa_offset (-32) |
| 4295 | + ret |
| 4296 | + .p2align 4,,7 |
| 4297 | +2: |
| 4298 | + cfi_adjust_cfa_offset (32) |
| 4299 | +#if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 4300 | + movl %ebx, -28(%esp) |
| 4301 | + movl %esp, %ebx |
| 4302 | + cfi_def_cfa_register(%ebx) |
| 4303 | + and $-STATE_SAVE_ALIGNMENT, %esp |
| 4304 | +#endif |
| 4305 | +#ifdef REGISTER_SAVE_AREA |
| 4306 | + subl $REGISTER_SAVE_AREA, %esp |
| 4307 | +# if !DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 4308 | + cfi_adjust_cfa_offset(REGISTER_SAVE_AREA) |
| 4309 | +# endif |
| 4310 | +#else |
| 4311 | +# if !DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 4312 | +# error DL_RUNTIME_RESOLVE_REALIGN_STACK must be true |
| 4313 | +# endif |
| 4314 | + /* Allocate stack space of the required size to save the state. */ |
| 4315 | + LOAD_PIC_REG (cx) |
| 4316 | + subl RTLD_GLOBAL_RO_DL_X86_CPU_FEATURES_OFFSET+XSAVE_STATE_SIZE_OFFSET+_rtld_local_ro@GOTOFF(%ecx), %esp |
| 4317 | +#endif |
| 4318 | +#ifdef USE_FNSAVE |
| 4319 | + fnsave (%esp) |
| 4320 | +#elif defined USE_FXSAVE |
| 4321 | + fxsave (%esp) |
| 4322 | +#else |
| 4323 | + /* Save the argument for ___tls_get_addr in EAX. */ |
| 4324 | + movl %eax, %ecx |
| 4325 | + movl $TLSDESC_CALL_STATE_SAVE_MASK, %eax |
| 4326 | + xorl %edx, %edx |
| 4327 | + /* Clear the XSAVE Header. */ |
| 4328 | +# ifdef USE_XSAVE |
| 4329 | + movl %edx, (512)(%esp) |
| 4330 | + movl %edx, (512 + 4 * 1)(%esp) |
| 4331 | + movl %edx, (512 + 4 * 2)(%esp) |
| 4332 | + movl %edx, (512 + 4 * 3)(%esp) |
| 4333 | +# endif |
| 4334 | + movl %edx, (512 + 4 * 4)(%esp) |
| 4335 | + movl %edx, (512 + 4 * 5)(%esp) |
| 4336 | + movl %edx, (512 + 4 * 6)(%esp) |
| 4337 | + movl %edx, (512 + 4 * 7)(%esp) |
| 4338 | + movl %edx, (512 + 4 * 8)(%esp) |
| 4339 | + movl %edx, (512 + 4 * 9)(%esp) |
| 4340 | + movl %edx, (512 + 4 * 10)(%esp) |
| 4341 | + movl %edx, (512 + 4 * 11)(%esp) |
| 4342 | + movl %edx, (512 + 4 * 12)(%esp) |
| 4343 | + movl %edx, (512 + 4 * 13)(%esp) |
| 4344 | + movl %edx, (512 + 4 * 14)(%esp) |
| 4345 | + movl %edx, (512 + 4 * 15)(%esp) |
| 4346 | +# ifdef USE_XSAVE |
| 4347 | + xsave (%esp) |
| 4348 | +# else |
| 4349 | + xsavec (%esp) |
| 4350 | +# endif |
| 4351 | + /* Restore the argument for ___tls_get_addr in EAX. */ |
| 4352 | + movl %ecx, %eax |
| 4353 | +#endif |
| 4354 | + call HIDDEN_JUMPTARGET (___tls_get_addr) |
| 4355 | + /* Get register content back. */ |
| 4356 | +#ifdef USE_FNSAVE |
| 4357 | + frstor (%esp) |
| 4358 | +#elif defined USE_FXSAVE |
| 4359 | + fxrstor (%esp) |
| 4360 | +#else |
| 4361 | + /* Save and retore ___tls_get_addr return value stored in EAX. */ |
| 4362 | + movl %eax, %ecx |
| 4363 | + movl $TLSDESC_CALL_STATE_SAVE_MASK, %eax |
| 4364 | + xorl %edx, %edx |
| 4365 | + xrstor (%esp) |
| 4366 | + movl %ecx, %eax |
| 4367 | +#endif |
| 4368 | +#if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 4369 | + mov %ebx, %esp |
| 4370 | + cfi_def_cfa_register(%esp) |
| 4371 | + movl -28(%esp), %ebx |
| 4372 | + cfi_restore(%ebx) |
| 4373 | +#else |
| 4374 | + addl $REGISTER_SAVE_AREA, %esp |
| 4375 | + cfi_adjust_cfa_offset(-REGISTER_SAVE_AREA) |
| 4376 | +#endif |
| 4377 | + jmp 1b |
| 4378 | + cfi_endproc |
| 4379 | + .size _dl_tlsdesc_dynamic, .-_dl_tlsdesc_dynamic |
| 4380 | + |
| 4381 | +#undef STATE_SAVE_ALIGNMENT |
| 4382 | diff --git a/sysdeps/i386/dl-tlsdesc.S b/sysdeps/i386/dl-tlsdesc.S |
| 4383 | index 90d93caa0c..f002feee56 100644 |
| 4384 | --- a/sysdeps/i386/dl-tlsdesc.S |
| 4385 | +++ b/sysdeps/i386/dl-tlsdesc.S |
| 4386 | @@ -18,8 +18,27 @@ |
| 4387 | |
| 4388 | #include <sysdep.h> |
| 4389 | #include <tls.h> |
| 4390 | +#include <cpu-features-offsets.h> |
| 4391 | +#include <features-offsets.h> |
| 4392 | #include "tlsdesc.h" |
| 4393 | |
| 4394 | +#ifndef DL_STACK_ALIGNMENT |
| 4395 | +/* Due to GCC bug: |
| 4396 | + |
| 4397 | + https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58066 |
| 4398 | + |
| 4399 | + __tls_get_addr may be called with 4-byte stack alignment. Although |
| 4400 | + this bug has been fixed in GCC 4.9.4, 5.3 and 6, we can't assume |
| 4401 | + that stack will be always aligned at 16 bytes. */ |
| 4402 | +# define DL_STACK_ALIGNMENT 4 |
| 4403 | +#endif |
| 4404 | + |
| 4405 | +/* True if _dl_tlsdesc_dynamic should align stack for STATE_SAVE or align |
| 4406 | + stack to MINIMUM_ALIGNMENT bytes before calling ___tls_get_addr. */ |
| 4407 | +#define DL_RUNTIME_RESOLVE_REALIGN_STACK \ |
| 4408 | + (STATE_SAVE_ALIGNMENT > DL_STACK_ALIGNMENT \ |
| 4409 | + || MINIMUM_ALIGNMENT > DL_STACK_ALIGNMENT) |
| 4410 | + |
| 4411 | .text |
| 4412 | |
| 4413 | /* This function is used to compute the TP offset for symbols in |
| 4414 | @@ -65,69 +84,35 @@ _dl_tlsdesc_undefweak: |
| 4415 | .size _dl_tlsdesc_undefweak, .-_dl_tlsdesc_undefweak |
| 4416 | |
| 4417 | #ifdef SHARED |
| 4418 | - .hidden _dl_tlsdesc_dynamic |
| 4419 | - .global _dl_tlsdesc_dynamic |
| 4420 | - .type _dl_tlsdesc_dynamic,@function |
| 4421 | - |
| 4422 | - /* This function is used for symbols that need dynamic TLS. |
| 4423 | - |
| 4424 | - %eax points to the TLS descriptor, such that 0(%eax) points to |
| 4425 | - _dl_tlsdesc_dynamic itself, and 4(%eax) points to a struct |
| 4426 | - tlsdesc_dynamic_arg object. It must return in %eax the offset |
| 4427 | - between the thread pointer and the object denoted by the |
| 4428 | - argument, without clobbering any registers. |
| 4429 | - |
| 4430 | - The assembly code that follows is a rendition of the following |
| 4431 | - C code, hand-optimized a little bit. |
| 4432 | - |
| 4433 | -ptrdiff_t |
| 4434 | -__attribute__ ((__regparm__ (1))) |
| 4435 | -_dl_tlsdesc_dynamic (struct tlsdesc *tdp) |
| 4436 | -{ |
| 4437 | - struct tlsdesc_dynamic_arg *td = tdp->arg; |
| 4438 | - dtv_t *dtv = *(dtv_t **)((char *)__thread_pointer + DTV_OFFSET); |
| 4439 | - if (__builtin_expect (td->gen_count <= dtv[0].counter |
| 4440 | - && (dtv[td->tlsinfo.ti_module].pointer.val |
| 4441 | - != TLS_DTV_UNALLOCATED), |
| 4442 | - 1)) |
| 4443 | - return dtv[td->tlsinfo.ti_module].pointer.val + td->tlsinfo.ti_offset |
| 4444 | - - __thread_pointer; |
| 4445 | - |
| 4446 | - return ___tls_get_addr (&td->tlsinfo) - __thread_pointer; |
| 4447 | -} |
| 4448 | -*/ |
| 4449 | - cfi_startproc |
| 4450 | - .align 16 |
| 4451 | -_dl_tlsdesc_dynamic: |
| 4452 | - /* Like all TLS resolvers, preserve call-clobbered registers. |
| 4453 | - We need two scratch regs anyway. */ |
| 4454 | - subl $28, %esp |
| 4455 | - cfi_adjust_cfa_offset (28) |
| 4456 | - movl %ecx, 20(%esp) |
| 4457 | - movl %edx, 24(%esp) |
| 4458 | - movl TLSDESC_ARG(%eax), %eax |
| 4459 | - movl %gs:DTV_OFFSET, %edx |
| 4460 | - movl TLSDESC_GEN_COUNT(%eax), %ecx |
| 4461 | - cmpl (%edx), %ecx |
| 4462 | - ja .Lslow |
| 4463 | - movl TLSDESC_MODID(%eax), %ecx |
| 4464 | - movl (%edx,%ecx,8), %edx |
| 4465 | - cmpl $-1, %edx |
| 4466 | - je .Lslow |
| 4467 | - movl TLSDESC_MODOFF(%eax), %eax |
| 4468 | - addl %edx, %eax |
| 4469 | -.Lret: |
| 4470 | - movl 20(%esp), %ecx |
| 4471 | - subl %gs:0, %eax |
| 4472 | - movl 24(%esp), %edx |
| 4473 | - addl $28, %esp |
| 4474 | - cfi_adjust_cfa_offset (-28) |
| 4475 | - ret |
| 4476 | - .p2align 4,,7 |
| 4477 | -.Lslow: |
| 4478 | - cfi_adjust_cfa_offset (28) |
| 4479 | - call HIDDEN_JUMPTARGET (___tls_get_addr) |
| 4480 | - jmp .Lret |
| 4481 | - cfi_endproc |
| 4482 | - .size _dl_tlsdesc_dynamic, .-_dl_tlsdesc_dynamic |
| 4483 | +# define USE_FNSAVE |
| 4484 | +# define MINIMUM_ALIGNMENT 4 |
| 4485 | +# define STATE_SAVE_ALIGNMENT 4 |
| 4486 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_fnsave |
| 4487 | +# include "dl-tlsdesc-dynamic.h" |
| 4488 | +# undef _dl_tlsdesc_dynamic |
| 4489 | +# undef MINIMUM_ALIGNMENT |
| 4490 | +# undef USE_FNSAVE |
| 4491 | + |
| 4492 | +# define MINIMUM_ALIGNMENT 16 |
| 4493 | + |
| 4494 | +# define USE_FXSAVE |
| 4495 | +# define STATE_SAVE_ALIGNMENT 16 |
| 4496 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_fxsave |
| 4497 | +# include "dl-tlsdesc-dynamic.h" |
| 4498 | +# undef _dl_tlsdesc_dynamic |
| 4499 | +# undef USE_FXSAVE |
| 4500 | + |
| 4501 | +# define USE_XSAVE |
| 4502 | +# define STATE_SAVE_ALIGNMENT 64 |
| 4503 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_xsave |
| 4504 | +# include "dl-tlsdesc-dynamic.h" |
| 4505 | +# undef _dl_tlsdesc_dynamic |
| 4506 | +# undef USE_XSAVE |
| 4507 | + |
| 4508 | +# define USE_XSAVEC |
| 4509 | +# define STATE_SAVE_ALIGNMENT 64 |
| 4510 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_xsavec |
| 4511 | +# include "dl-tlsdesc-dynamic.h" |
| 4512 | +# undef _dl_tlsdesc_dynamic |
| 4513 | +# undef USE_XSAVEC |
| 4514 | #endif /* SHARED */ |
| 4515 | diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps |
| 4516 | index 84e6686eba..f2139fc172 100644 |
| 4517 | --- a/sysdeps/i386/fpu/libm-test-ulps |
| 4518 | +++ b/sysdeps/i386/fpu/libm-test-ulps |
| 4519 | @@ -1232,6 +1232,7 @@ ldouble: 6 |
| 4520 | |
| 4521 | Function: "hypot": |
| 4522 | double: 1 |
| 4523 | +float: 1 |
| 4524 | float128: 1 |
| 4525 | ldouble: 1 |
| 4526 | |
| 4527 | diff --git a/sysdeps/i386/i586/memcpy.S b/sysdeps/i386/i586/memcpy.S |
| 4528 | index 3e26f112d6..79856d498a 100644 |
| 4529 | --- a/sysdeps/i386/i586/memcpy.S |
| 4530 | +++ b/sysdeps/i386/i586/memcpy.S |
| 4531 | @@ -26,7 +26,7 @@ |
| 4532 | #define LEN SRC+4 |
| 4533 | |
| 4534 | .text |
| 4535 | -#if defined PIC && IS_IN (libc) |
| 4536 | +#if defined SHARED && IS_IN (libc) |
| 4537 | ENTRY (__memcpy_chk) |
| 4538 | movl 12(%esp), %eax |
| 4539 | cmpl %eax, 16(%esp) |
| 4540 | diff --git a/sysdeps/i386/i686/memmove.S b/sysdeps/i386/i686/memmove.S |
| 4541 | index f230359ad6..effd958120 100644 |
| 4542 | --- a/sysdeps/i386/i686/memmove.S |
| 4543 | +++ b/sysdeps/i386/i686/memmove.S |
| 4544 | @@ -29,7 +29,7 @@ |
| 4545 | #define SRC DEST+4 |
| 4546 | #define LEN SRC+4 |
| 4547 | |
| 4548 | -#if defined PIC && IS_IN (libc) |
| 4549 | +#if defined SHARED && IS_IN (libc) |
| 4550 | ENTRY_CHK (__memmove_chk) |
| 4551 | movl 12(%esp), %eax |
| 4552 | cmpl %eax, 16(%esp) |
| 4553 | diff --git a/sysdeps/i386/i686/memset.S b/sysdeps/i386/i686/memset.S |
| 4554 | index f02f5a6df7..ab06771ea0 100644 |
| 4555 | --- a/sysdeps/i386/i686/memset.S |
| 4556 | +++ b/sysdeps/i386/i686/memset.S |
| 4557 | @@ -27,7 +27,7 @@ |
| 4558 | #define LEN CHR+4 |
| 4559 | |
| 4560 | .text |
| 4561 | -#if defined PIC && IS_IN (libc) |
| 4562 | +#if defined SHARED && IS_IN (libc) |
| 4563 | ENTRY_CHK (__memset_chk) |
| 4564 | movl 12(%esp), %eax |
| 4565 | cmpl %eax, 16(%esp) |
| 4566 | diff --git a/sysdeps/i386/i686/multiarch/memrchr-c.c b/sysdeps/i386/i686/multiarch/memrchr-c.c |
| 4567 | index ef7bbbe792..20bfdf3af3 100644 |
| 4568 | --- a/sysdeps/i386/i686/multiarch/memrchr-c.c |
| 4569 | +++ b/sysdeps/i386/i686/multiarch/memrchr-c.c |
| 4570 | @@ -5,3 +5,4 @@ extern void *__memrchr_ia32 (const void *, int, size_t); |
| 4571 | #endif |
| 4572 | |
| 4573 | #include "string/memrchr.c" |
| 4574 | +strong_alias (__memrchr_ia32, __GI___memrchr) |
| 4575 | diff --git a/sysdeps/i386/i686/multiarch/memrchr-sse2.S b/sysdeps/i386/i686/multiarch/memrchr-sse2.S |
| 4576 | index d9dae04171..e123f87435 100644 |
| 4577 | --- a/sysdeps/i386/i686/multiarch/memrchr-sse2.S |
| 4578 | +++ b/sysdeps/i386/i686/multiarch/memrchr-sse2.S |
| 4579 | @@ -720,5 +720,4 @@ L(ret_null): |
| 4580 | ret |
| 4581 | |
| 4582 | END (__memrchr_sse2) |
| 4583 | -strong_alias (__memrchr_sse2, __GI___memrchr) |
| 4584 | #endif |
| 4585 | diff --git a/sysdeps/loongarch/fpu/e_scalbf.c b/sysdeps/loongarch/fpu/e_scalbf.c |
| 4586 | index 9f05485236..7c0395fbb5 100644 |
| 4587 | --- a/sysdeps/loongarch/fpu/e_scalbf.c |
| 4588 | +++ b/sysdeps/loongarch/fpu/e_scalbf.c |
| 4589 | @@ -57,4 +57,4 @@ __ieee754_scalbf (float x, float fn) |
| 4590 | |
| 4591 | return x; |
| 4592 | } |
| 4593 | -libm_alias_finite (__ieee754_scalb, __scalb) |
| 4594 | +libm_alias_finite (__ieee754_scalbf, __scalbf) |
| 4595 | diff --git a/sysdeps/loongarch/lp64/multiarch/Makefile b/sysdeps/loongarch/lp64/multiarch/Makefile |
| 4596 | index fe863e1ba4..01762ef526 100644 |
| 4597 | --- a/sysdeps/loongarch/lp64/multiarch/Makefile |
| 4598 | +++ b/sysdeps/loongarch/lp64/multiarch/Makefile |
| 4599 | @@ -1,52 +1,52 @@ |
| 4600 | ifeq ($(subdir),string) |
| 4601 | sysdep_routines += \ |
| 4602 | - strlen-aligned \ |
| 4603 | - strlen-lsx \ |
| 4604 | - strlen-lasx \ |
| 4605 | - strnlen-aligned \ |
| 4606 | - strnlen-lsx \ |
| 4607 | - strnlen-lasx \ |
| 4608 | + memchr-aligned \ |
| 4609 | + memchr-lasx \ |
| 4610 | + memchr-lsx \ |
| 4611 | + memcmp-aligned \ |
| 4612 | + memcmp-lasx \ |
| 4613 | + memcmp-lsx \ |
| 4614 | + memcpy-aligned \ |
| 4615 | + memcpy-unaligned \ |
| 4616 | + memmove-lasx \ |
| 4617 | + memmove-lsx \ |
| 4618 | + memmove-unaligned \ |
| 4619 | + memrchr-generic \ |
| 4620 | + memrchr-lasx \ |
| 4621 | + memrchr-lsx \ |
| 4622 | + memset-aligned \ |
| 4623 | + memset-lasx \ |
| 4624 | + memset-lsx \ |
| 4625 | + memset-unaligned \ |
| 4626 | + rawmemchr-aligned \ |
| 4627 | + rawmemchr-lasx \ |
| 4628 | + rawmemchr-lsx \ |
| 4629 | + stpcpy-aligned \ |
| 4630 | + stpcpy-lasx \ |
| 4631 | + stpcpy-lsx \ |
| 4632 | + stpcpy-unaligned \ |
| 4633 | strchr-aligned \ |
| 4634 | - strchr-lsx \ |
| 4635 | strchr-lasx \ |
| 4636 | - strrchr-aligned \ |
| 4637 | - strrchr-lsx \ |
| 4638 | - strrchr-lasx \ |
| 4639 | + strchr-lsx \ |
| 4640 | strchrnul-aligned \ |
| 4641 | - strchrnul-lsx \ |
| 4642 | strchrnul-lasx \ |
| 4643 | + strchrnul-lsx \ |
| 4644 | strcmp-aligned \ |
| 4645 | strcmp-lsx \ |
| 4646 | - strncmp-aligned \ |
| 4647 | - strncmp-lsx \ |
| 4648 | strcpy-aligned \ |
| 4649 | - strcpy-unaligned \ |
| 4650 | - strcpy-lsx \ |
| 4651 | strcpy-lasx \ |
| 4652 | - stpcpy-aligned \ |
| 4653 | - stpcpy-unaligned \ |
| 4654 | - stpcpy-lsx \ |
| 4655 | - stpcpy-lasx \ |
| 4656 | - memcpy-aligned \ |
| 4657 | - memcpy-unaligned \ |
| 4658 | - memmove-unaligned \ |
| 4659 | - memmove-lsx \ |
| 4660 | - memmove-lasx \ |
| 4661 | - rawmemchr-aligned \ |
| 4662 | - rawmemchr-lsx \ |
| 4663 | - rawmemchr-lasx \ |
| 4664 | - memchr-aligned \ |
| 4665 | - memchr-lsx \ |
| 4666 | - memchr-lasx \ |
| 4667 | - memrchr-generic \ |
| 4668 | - memrchr-lsx \ |
| 4669 | - memrchr-lasx \ |
| 4670 | - memset-aligned \ |
| 4671 | - memset-unaligned \ |
| 4672 | - memset-lsx \ |
| 4673 | - memset-lasx \ |
| 4674 | - memcmp-aligned \ |
| 4675 | - memcmp-lsx \ |
| 4676 | - memcmp-lasx \ |
| 4677 | + strcpy-lsx \ |
| 4678 | + strcpy-unaligned \ |
| 4679 | + strlen-aligned \ |
| 4680 | + strlen-lasx \ |
| 4681 | + strlen-lsx \ |
| 4682 | + strncmp-aligned \ |
| 4683 | + strncmp-lsx \ |
| 4684 | + strnlen-aligned \ |
| 4685 | + strnlen-lasx \ |
| 4686 | + strnlen-lsx \ |
| 4687 | + strrchr-aligned \ |
| 4688 | + strrchr-lasx \ |
| 4689 | + strrchr-lsx \ |
| 4690 | # sysdep_routines |
| 4691 | endif |
| 4692 | diff --git a/sysdeps/loongarch/lp64/multiarch/dl-symbol-redir-ifunc.h b/sysdeps/loongarch/lp64/multiarch/dl-symbol-redir-ifunc.h |
| 4693 | index cb640d77b7..a73390b12f 100644 |
| 4694 | --- a/sysdeps/loongarch/lp64/multiarch/dl-symbol-redir-ifunc.h |
| 4695 | +++ b/sysdeps/loongarch/lp64/multiarch/dl-symbol-redir-ifunc.h |
| 4696 | @@ -19,6 +19,9 @@ |
| 4697 | #ifndef _DL_IFUNC_GENERIC_H |
| 4698 | #define _DL_IFUNC_GENERIC_H |
| 4699 | |
| 4700 | +#ifndef SHARED |
| 4701 | asm ("memset = __memset_aligned"); |
| 4702 | +asm ("memcmp = __memcmp_aligned"); |
| 4703 | +#endif |
| 4704 | |
| 4705 | #endif |
| 4706 | diff --git a/sysdeps/m68k/bits/wordsize.h b/sysdeps/m68k/bits/wordsize.h |
| 4707 | new file mode 100644 |
| 4708 | index 0000000000..6ecbfe7c86 |
| 4709 | --- /dev/null |
| 4710 | +++ b/sysdeps/m68k/bits/wordsize.h |
| 4711 | @@ -0,0 +1,21 @@ |
| 4712 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 4713 | + This file is part of the GNU C Library. |
| 4714 | + |
| 4715 | + The GNU C Library is free software; you can redistribute it and/or |
| 4716 | + modify it under the terms of the GNU Lesser General Public |
| 4717 | + License as published by the Free Software Foundation; either |
| 4718 | + version 2.1 of the License, or (at your option) any later version. |
| 4719 | + |
| 4720 | + The GNU C Library is distributed in the hope that it will be useful, |
| 4721 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 4722 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 4723 | + Lesser General Public License for more details. |
| 4724 | + |
| 4725 | + You should have received a copy of the GNU Lesser General Public |
| 4726 | + License along with the GNU C Library; if not, see |
| 4727 | + <https://www.gnu.org/licenses/>. */ |
| 4728 | + |
| 4729 | +#define __WORDSIZE 32 |
| 4730 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 4731 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 4732 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 4733 | diff --git a/sysdeps/m68k/utmp-size.h b/sysdeps/m68k/utmp-size.h |
| 4734 | new file mode 100644 |
| 4735 | index 0000000000..5946685819 |
| 4736 | --- /dev/null |
| 4737 | +++ b/sysdeps/m68k/utmp-size.h |
| 4738 | @@ -0,0 +1,3 @@ |
| 4739 | +/* m68k has 2-byte alignment. */ |
| 4740 | +#define UTMP_SIZE 382 |
| 4741 | +#define LASTLOG_SIZE 292 |
| 4742 | diff --git a/sysdeps/microblaze/bits/wordsize.h b/sysdeps/microblaze/bits/wordsize.h |
| 4743 | new file mode 100644 |
| 4744 | index 0000000000..6ecbfe7c86 |
| 4745 | --- /dev/null |
| 4746 | +++ b/sysdeps/microblaze/bits/wordsize.h |
| 4747 | @@ -0,0 +1,21 @@ |
| 4748 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 4749 | + This file is part of the GNU C Library. |
| 4750 | + |
| 4751 | + The GNU C Library is free software; you can redistribute it and/or |
| 4752 | + modify it under the terms of the GNU Lesser General Public |
| 4753 | + License as published by the Free Software Foundation; either |
| 4754 | + version 2.1 of the License, or (at your option) any later version. |
| 4755 | + |
| 4756 | + The GNU C Library is distributed in the hope that it will be useful, |
| 4757 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 4758 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 4759 | + Lesser General Public License for more details. |
| 4760 | + |
| 4761 | + You should have received a copy of the GNU Lesser General Public |
| 4762 | + License along with the GNU C Library; if not, see |
| 4763 | + <https://www.gnu.org/licenses/>. */ |
| 4764 | + |
| 4765 | +#define __WORDSIZE 32 |
| 4766 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 4767 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 4768 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 4769 | diff --git a/sysdeps/microblaze/utmp-size.h b/sysdeps/microblaze/utmp-size.h |
| 4770 | new file mode 100644 |
| 4771 | index 0000000000..8f21ebe1b6 |
| 4772 | --- /dev/null |
| 4773 | +++ b/sysdeps/microblaze/utmp-size.h |
| 4774 | @@ -0,0 +1,2 @@ |
| 4775 | +#define UTMP_SIZE 384 |
| 4776 | +#define LASTLOG_SIZE 292 |
| 4777 | diff --git a/sysdeps/mips/bits/wordsize.h b/sysdeps/mips/bits/wordsize.h |
| 4778 | index 57f0f2a22f..30dd3fd85d 100644 |
| 4779 | --- a/sysdeps/mips/bits/wordsize.h |
| 4780 | +++ b/sysdeps/mips/bits/wordsize.h |
| 4781 | @@ -19,11 +19,7 @@ |
| 4782 | |
| 4783 | #define __WORDSIZE _MIPS_SZPTR |
| 4784 | |
| 4785 | -#if _MIPS_SIM == _ABI64 |
| 4786 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 4787 | -#else |
| 4788 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 4789 | -#endif |
| 4790 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 4791 | |
| 4792 | #if __WORDSIZE == 32 |
| 4793 | #define __WORDSIZE32_SIZE_ULONG 0 |
| 4794 | diff --git a/sysdeps/mips/mips64/libm-test-ulps b/sysdeps/mips/mips64/libm-test-ulps |
| 4795 | index 78969745b2..933aba4735 100644 |
| 4796 | --- a/sysdeps/mips/mips64/libm-test-ulps |
| 4797 | +++ b/sysdeps/mips/mips64/libm-test-ulps |
| 4798 | @@ -1066,17 +1066,17 @@ double: 1 |
| 4799 | ldouble: 1 |
| 4800 | |
| 4801 | Function: "j0": |
| 4802 | -double: 2 |
| 4803 | +double: 3 |
| 4804 | float: 9 |
| 4805 | ldouble: 2 |
| 4806 | |
| 4807 | Function: "j0_downward": |
| 4808 | -double: 5 |
| 4809 | +double: 6 |
| 4810 | float: 9 |
| 4811 | ldouble: 9 |
| 4812 | |
| 4813 | Function: "j0_towardzero": |
| 4814 | -double: 6 |
| 4815 | +double: 7 |
| 4816 | float: 9 |
| 4817 | ldouble: 9 |
| 4818 | |
| 4819 | @@ -1146,6 +1146,7 @@ float: 6 |
| 4820 | ldouble: 8 |
| 4821 | |
| 4822 | Function: "log": |
| 4823 | +double: 1 |
| 4824 | float: 1 |
| 4825 | ldouble: 1 |
| 4826 | |
| 4827 | diff --git a/sysdeps/mips/utmp-size.h b/sysdeps/mips/utmp-size.h |
| 4828 | new file mode 100644 |
| 4829 | index 0000000000..8f21ebe1b6 |
| 4830 | --- /dev/null |
| 4831 | +++ b/sysdeps/mips/utmp-size.h |
| 4832 | @@ -0,0 +1,2 @@ |
| 4833 | +#define UTMP_SIZE 384 |
| 4834 | +#define LASTLOG_SIZE 292 |
| 4835 | diff --git a/sysdeps/nios2/bits/wordsize.h b/sysdeps/nios2/bits/wordsize.h |
| 4836 | new file mode 100644 |
| 4837 | index 0000000000..6ecbfe7c86 |
| 4838 | --- /dev/null |
| 4839 | +++ b/sysdeps/nios2/bits/wordsize.h |
| 4840 | @@ -0,0 +1,21 @@ |
| 4841 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 4842 | + This file is part of the GNU C Library. |
| 4843 | + |
| 4844 | + The GNU C Library is free software; you can redistribute it and/or |
| 4845 | + modify it under the terms of the GNU Lesser General Public |
| 4846 | + License as published by the Free Software Foundation; either |
| 4847 | + version 2.1 of the License, or (at your option) any later version. |
| 4848 | + |
| 4849 | + The GNU C Library is distributed in the hope that it will be useful, |
| 4850 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 4851 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 4852 | + Lesser General Public License for more details. |
| 4853 | + |
| 4854 | + You should have received a copy of the GNU Lesser General Public |
| 4855 | + License along with the GNU C Library; if not, see |
| 4856 | + <https://www.gnu.org/licenses/>. */ |
| 4857 | + |
| 4858 | +#define __WORDSIZE 32 |
| 4859 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 4860 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 4861 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 4862 | diff --git a/sysdeps/nios2/utmp-size.h b/sysdeps/nios2/utmp-size.h |
| 4863 | new file mode 100644 |
| 4864 | index 0000000000..8f21ebe1b6 |
| 4865 | --- /dev/null |
| 4866 | +++ b/sysdeps/nios2/utmp-size.h |
| 4867 | @@ -0,0 +1,2 @@ |
| 4868 | +#define UTMP_SIZE 384 |
| 4869 | +#define LASTLOG_SIZE 292 |
| 4870 | diff --git a/sysdeps/or1k/utmp-size.h b/sysdeps/or1k/utmp-size.h |
| 4871 | new file mode 100644 |
| 4872 | index 0000000000..6b3653aa4d |
| 4873 | --- /dev/null |
| 4874 | +++ b/sysdeps/or1k/utmp-size.h |
| 4875 | @@ -0,0 +1,3 @@ |
| 4876 | +/* or1k has less padding than other architectures with 64-bit time_t. */ |
| 4877 | +#define UTMP_SIZE 392 |
| 4878 | +#define LASTLOG_SIZE 296 |
| 4879 | diff --git a/sysdeps/powerpc/dl-procinfo.c b/sysdeps/powerpc/dl-procinfo.c |
| 4880 | index a76bb6e5b0..8cf00aa7e3 100644 |
| 4881 | --- a/sysdeps/powerpc/dl-procinfo.c |
| 4882 | +++ b/sysdeps/powerpc/dl-procinfo.c |
| 4883 | @@ -38,6 +38,10 @@ |
| 4884 | needed. |
| 4885 | */ |
| 4886 | |
| 4887 | +/* The total number of available bits (including those prior to |
| 4888 | + _DL_HWCAP_FIRST). Some of these bits might not be used. */ |
| 4889 | +#define _DL_HWCAP_COUNT 128 |
| 4890 | + |
| 4891 | #ifndef PROCINFO_CLASS |
| 4892 | # define PROCINFO_CLASS |
| 4893 | #endif |
| 4894 | @@ -61,7 +65,7 @@ PROCINFO_CLASS struct cpu_features _dl_powerpc_cpu_features |
| 4895 | #if !defined PROCINFO_DECL && defined SHARED |
| 4896 | ._dl_powerpc_cap_flags |
| 4897 | #else |
| 4898 | -PROCINFO_CLASS const char _dl_powerpc_cap_flags[64][15] |
| 4899 | +PROCINFO_CLASS const char _dl_powerpc_cap_flags[_DL_HWCAP_COUNT][15] |
| 4900 | #endif |
| 4901 | #ifndef PROCINFO_DECL |
| 4902 | = { |
| 4903 | diff --git a/sysdeps/powerpc/dl-procinfo.h b/sysdeps/powerpc/dl-procinfo.h |
| 4904 | index 68f4241095..b36697ba44 100644 |
| 4905 | --- a/sysdeps/powerpc/dl-procinfo.h |
| 4906 | +++ b/sysdeps/powerpc/dl-procinfo.h |
| 4907 | @@ -22,22 +22,23 @@ |
| 4908 | #include <ldsodefs.h> |
| 4909 | #include <sysdep.h> /* This defines the PPC_FEATURE[2]_* macros. */ |
| 4910 | |
| 4911 | -/* The total number of available bits (including those prior to |
| 4912 | - _DL_HWCAP_FIRST). Some of these bits might not be used. */ |
| 4913 | -#define _DL_HWCAP_COUNT 64 |
| 4914 | +/* Feature masks are all 32-bits in size. */ |
| 4915 | +#define _DL_HWCAP_SIZE 32 |
| 4916 | |
| 4917 | -/* Features started at bit 31 and decremented as new features were added. */ |
| 4918 | -#define _DL_HWCAP_LAST 31 |
| 4919 | +/* AT_HWCAP2 feature strings follow the AT_HWCAP feature strings. */ |
| 4920 | +#define _DL_HWCAP2_OFFSET _DL_HWCAP_SIZE |
| 4921 | |
| 4922 | -/* AT_HWCAP2 features started at bit 31 and decremented as new features were |
| 4923 | - added. HWCAP2 feature bits start at bit 0. */ |
| 4924 | -#define _DL_HWCAP2_LAST 31 |
| 4925 | +/* AT_HWCAP3 feature strings follow the AT_HWCAP2 feature strings. */ |
| 4926 | +#define _DL_HWCAP3_OFFSET (_DL_HWCAP2_OFFSET + _DL_HWCAP_SIZE) |
| 4927 | + |
| 4928 | +/* AT_HWCAP4 feature strings follow the AT_HWCAP3 feature strings. */ |
| 4929 | +#define _DL_HWCAP4_OFFSET (_DL_HWCAP3_OFFSET + _DL_HWCAP_SIZE) |
| 4930 | |
| 4931 | /* These bits influence library search. */ |
| 4932 | #define HWCAP_IMPORTANT (PPC_FEATURE_HAS_ALTIVEC \ |
| 4933 | + PPC_FEATURE_HAS_DFP) |
| 4934 | |
| 4935 | -#define _DL_PLATFORMS_COUNT 16 |
| 4936 | +#define _DL_PLATFORMS_COUNT 17 |
| 4937 | |
| 4938 | #define _DL_FIRST_PLATFORM 32 |
| 4939 | /* Mask to filter out platforms. */ |
| 4940 | @@ -61,6 +62,7 @@ |
| 4941 | #define PPC_PLATFORM_POWER8 13 |
| 4942 | #define PPC_PLATFORM_POWER9 14 |
| 4943 | #define PPC_PLATFORM_POWER10 15 |
| 4944 | +#define PPC_PLATFORM_POWER11 16 |
| 4945 | |
| 4946 | static inline const char * |
| 4947 | __attribute__ ((unused)) |
| 4948 | @@ -88,6 +90,11 @@ _dl_string_platform (const char *str) |
| 4949 | ret = _DL_FIRST_PLATFORM + PPC_PLATFORM_POWER10; |
| 4950 | str++; |
| 4951 | } |
| 4952 | + else if (str[1] == '1') |
| 4953 | + { |
| 4954 | + ret = _DL_FIRST_PLATFORM + PPC_PLATFORM_POWER11; |
| 4955 | + str++; |
| 4956 | + } |
| 4957 | else |
| 4958 | return -1; |
| 4959 | break; |
| 4960 | @@ -187,21 +194,42 @@ _dl_procinfo (unsigned int type, unsigned long int word) |
| 4961 | case AT_HWCAP: |
| 4962 | _dl_printf ("AT_HWCAP: "); |
| 4963 | |
| 4964 | - for (int i = 0; i <= _DL_HWCAP_LAST; ++i) |
| 4965 | + for (int i = 0; i < _DL_HWCAP_SIZE; ++i) |
| 4966 | if (word & (1 << i)) |
| 4967 | _dl_printf (" %s", _dl_hwcap_string (i)); |
| 4968 | break; |
| 4969 | case AT_HWCAP2: |
| 4970 | { |
| 4971 | - unsigned int offset = _DL_HWCAP_LAST + 1; |
| 4972 | |
| 4973 | _dl_printf ("AT_HWCAP2: "); |
| 4974 | |
| 4975 | - /* We have to go through them all because the kernel added the |
| 4976 | - AT_HWCAP2 features starting with the high bits. */ |
| 4977 | - for (int i = 0; i <= _DL_HWCAP2_LAST; ++i) |
| 4978 | - if (word & (1 << i)) |
| 4979 | - _dl_printf (" %s", _dl_hwcap_string (offset + i)); |
| 4980 | + /* We have to go through them all because the kernel added the |
| 4981 | + AT_HWCAP2 features starting with the high bits. */ |
| 4982 | + for (int i = 0; i < _DL_HWCAP_SIZE; ++i) |
| 4983 | + if (word & (1 << i)) |
| 4984 | + _dl_printf (" %s", _dl_hwcap_string (_DL_HWCAP2_OFFSET + i)); |
| 4985 | + break; |
| 4986 | + } |
| 4987 | + case AT_HWCAP3: |
| 4988 | + { |
| 4989 | + _dl_printf ("AT_HWCAP3: "); |
| 4990 | + |
| 4991 | + /* We have to go through them all because the kernel added the |
| 4992 | + AT_HWCAP3 features starting with the high bits. */ |
| 4993 | + for (int i = 0; i < _DL_HWCAP_SIZE; ++i) |
| 4994 | + if (word & (1 << i)) |
| 4995 | + _dl_printf (" %s", _dl_hwcap_string (_DL_HWCAP3_OFFSET + i)); |
| 4996 | + break; |
| 4997 | + } |
| 4998 | + case AT_HWCAP4: |
| 4999 | + { |
| 5000 | + _dl_printf ("AT_HWCAP4: "); |
| 5001 | + |
| 5002 | + /* We have to go through them all because the kernel added the |
| 5003 | + AT_HWCAP4 features starting with the high bits. */ |
| 5004 | + for (int i = 0; i <= _DL_HWCAP_SIZE; ++i) |
| 5005 | + if (word & (1 << i)) |
| 5006 | + _dl_printf (" %s", _dl_hwcap_string (_DL_HWCAP4_OFFSET + i)); |
| 5007 | break; |
| 5008 | } |
| 5009 | case AT_L1I_CACHEGEOMETRY: |
| 5010 | diff --git a/sysdeps/powerpc/hwcapinfo.c b/sysdeps/powerpc/hwcapinfo.c |
| 5011 | index 76344f285a..f6fede15a7 100644 |
| 5012 | --- a/sysdeps/powerpc/hwcapinfo.c |
| 5013 | +++ b/sysdeps/powerpc/hwcapinfo.c |
| 5014 | @@ -31,7 +31,7 @@ void |
| 5015 | __tcb_parse_hwcap_and_convert_at_platform (void) |
| 5016 | { |
| 5017 | |
| 5018 | - uint64_t h1, h2; |
| 5019 | + uint64_t h1, h2, h3, h4; |
| 5020 | |
| 5021 | /* Read AT_PLATFORM string from auxv and convert it to a number. */ |
| 5022 | __tcb.at_platform = _dl_string_platform (GLRO (dl_platform)); |
| 5023 | @@ -39,6 +39,8 @@ __tcb_parse_hwcap_and_convert_at_platform (void) |
| 5024 | /* Read HWCAP and HWCAP2 from auxv. */ |
| 5025 | h1 = GLRO (dl_hwcap); |
| 5026 | h2 = GLRO (dl_hwcap2); |
| 5027 | + h3 = GLRO (dl_hwcap3); |
| 5028 | + h4 = GLRO (dl_hwcap4); |
| 5029 | |
| 5030 | /* hwcap contains only the latest supported ISA, the code checks which is |
| 5031 | and fills the previous supported ones. */ |
| 5032 | @@ -64,13 +66,16 @@ __tcb_parse_hwcap_and_convert_at_platform (void) |
| 5033 | else if (h1 & PPC_FEATURE_POWER5) |
| 5034 | h1 |= PPC_FEATURE_POWER4; |
| 5035 | |
| 5036 | - uint64_t array_hwcaps[] = { h1, h2 }; |
| 5037 | + uint64_t array_hwcaps[] = { h1, h2, h3, h4 }; |
| 5038 | init_cpu_features (&GLRO(dl_powerpc_cpu_features), array_hwcaps); |
| 5039 | |
| 5040 | /* Consolidate both HWCAP and HWCAP2 into a single doubleword so that |
| 5041 | we can read both in a single load later. */ |
| 5042 | __tcb.hwcap = (h1 << 32) | (h2 & 0xffffffff); |
| 5043 | - __tcb.hwcap_extn = 0x0; |
| 5044 | + |
| 5045 | + /* Consolidate both HWCAP3 and HWCAP4 into a single doubleword so that |
| 5046 | + we can read both in a single load later. */ |
| 5047 | + __tcb.hwcap_extn = (h3 << 32) | (h4 & 0xffffffff); |
| 5048 | |
| 5049 | } |
| 5050 | #if IS_IN (rtld) |
| 5051 | diff --git a/sysdeps/powerpc/powerpc32/bits/wordsize.h b/sysdeps/powerpc/powerpc32/bits/wordsize.h |
| 5052 | index 04ca9debf0..6993fb6b29 100644 |
| 5053 | --- a/sysdeps/powerpc/powerpc32/bits/wordsize.h |
| 5054 | +++ b/sysdeps/powerpc/powerpc32/bits/wordsize.h |
| 5055 | @@ -2,10 +2,9 @@ |
| 5056 | |
| 5057 | #if defined __powerpc64__ |
| 5058 | # define __WORDSIZE 64 |
| 5059 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 5060 | #else |
| 5061 | # define __WORDSIZE 32 |
| 5062 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 5063 | # define __WORDSIZE32_SIZE_ULONG 0 |
| 5064 | # define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5065 | #endif |
| 5066 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5067 | diff --git a/sysdeps/powerpc/powerpc32/power11/Implies b/sysdeps/powerpc/powerpc32/power11/Implies |
| 5068 | new file mode 100644 |
| 5069 | index 0000000000..051cbe0f79 |
| 5070 | --- /dev/null |
| 5071 | +++ b/sysdeps/powerpc/powerpc32/power11/Implies |
| 5072 | @@ -0,0 +1,2 @@ |
| 5073 | +powerpc/powerpc32/power10/fpu |
| 5074 | +powerpc/powerpc32/power10 |
| 5075 | diff --git a/sysdeps/powerpc/powerpc32/power11/fpu/multiarch/Implies b/sysdeps/powerpc/powerpc32/power11/fpu/multiarch/Implies |
| 5076 | new file mode 100644 |
| 5077 | index 0000000000..58edb2861d |
| 5078 | --- /dev/null |
| 5079 | +++ b/sysdeps/powerpc/powerpc32/power11/fpu/multiarch/Implies |
| 5080 | @@ -0,0 +1 @@ |
| 5081 | +powerpc/powerpc32/power10/fpu/multiarch |
| 5082 | diff --git a/sysdeps/powerpc/powerpc32/power11/multiarch/Implies b/sysdeps/powerpc/powerpc32/power11/multiarch/Implies |
| 5083 | new file mode 100644 |
| 5084 | index 0000000000..c70f0428ba |
| 5085 | --- /dev/null |
| 5086 | +++ b/sysdeps/powerpc/powerpc32/power11/multiarch/Implies |
| 5087 | @@ -0,0 +1 @@ |
| 5088 | +powerpc/powerpc32/power10/multiarch |
| 5089 | diff --git a/sysdeps/powerpc/powerpc64/be/power11/Implies b/sysdeps/powerpc/powerpc64/be/power11/Implies |
| 5090 | new file mode 100644 |
| 5091 | index 0000000000..de481d1c13 |
| 5092 | --- /dev/null |
| 5093 | +++ b/sysdeps/powerpc/powerpc64/be/power11/Implies |
| 5094 | @@ -0,0 +1,2 @@ |
| 5095 | +powerpc/powerpc64/be/power10/fpu |
| 5096 | +powerpc/powerpc64/be/power10 |
| 5097 | diff --git a/sysdeps/powerpc/powerpc64/be/power11/fpu/Implies b/sysdeps/powerpc/powerpc64/be/power11/fpu/Implies |
| 5098 | new file mode 100644 |
| 5099 | index 0000000000..dff0e13064 |
| 5100 | --- /dev/null |
| 5101 | +++ b/sysdeps/powerpc/powerpc64/be/power11/fpu/Implies |
| 5102 | @@ -0,0 +1 @@ |
| 5103 | +powerpc/powerpc64/be/power10/fpu |
| 5104 | diff --git a/sysdeps/powerpc/powerpc64/be/power11/fpu/multiarch/Implies b/sysdeps/powerpc/powerpc64/be/power11/fpu/multiarch/Implies |
| 5105 | new file mode 100644 |
| 5106 | index 0000000000..c3f259e009 |
| 5107 | --- /dev/null |
| 5108 | +++ b/sysdeps/powerpc/powerpc64/be/power11/fpu/multiarch/Implies |
| 5109 | @@ -0,0 +1 @@ |
| 5110 | +powerpc/powerpc64/be/power10/fpu/multiarch |
| 5111 | diff --git a/sysdeps/powerpc/powerpc64/be/power11/multiarch/Implies b/sysdeps/powerpc/powerpc64/be/power11/multiarch/Implies |
| 5112 | new file mode 100644 |
| 5113 | index 0000000000..9491a394c9 |
| 5114 | --- /dev/null |
| 5115 | +++ b/sysdeps/powerpc/powerpc64/be/power11/multiarch/Implies |
| 5116 | @@ -0,0 +1 @@ |
| 5117 | +powerpc/powerpc64/be/power10/multiarch |
| 5118 | diff --git a/sysdeps/powerpc/powerpc64/bits/wordsize.h b/sysdeps/powerpc/powerpc64/bits/wordsize.h |
| 5119 | index 04ca9debf0..6993fb6b29 100644 |
| 5120 | --- a/sysdeps/powerpc/powerpc64/bits/wordsize.h |
| 5121 | +++ b/sysdeps/powerpc/powerpc64/bits/wordsize.h |
| 5122 | @@ -2,10 +2,9 @@ |
| 5123 | |
| 5124 | #if defined __powerpc64__ |
| 5125 | # define __WORDSIZE 64 |
| 5126 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 5127 | #else |
| 5128 | # define __WORDSIZE 32 |
| 5129 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 5130 | # define __WORDSIZE32_SIZE_ULONG 0 |
| 5131 | # define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5132 | #endif |
| 5133 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5134 | diff --git a/sysdeps/powerpc/powerpc64/dl-machine.h b/sysdeps/powerpc/powerpc64/dl-machine.h |
| 5135 | index c6682f3445..2b6f5d2b08 100644 |
| 5136 | --- a/sysdeps/powerpc/powerpc64/dl-machine.h |
| 5137 | +++ b/sysdeps/powerpc/powerpc64/dl-machine.h |
| 5138 | @@ -78,6 +78,7 @@ elf_host_tolerates_class (const Elf64_Ehdr *ehdr) |
| 5139 | static inline Elf64_Addr |
| 5140 | elf_machine_load_address (void) __attribute__ ((const)); |
| 5141 | |
| 5142 | +#ifndef __PCREL__ |
| 5143 | static inline Elf64_Addr |
| 5144 | elf_machine_load_address (void) |
| 5145 | { |
| 5146 | @@ -105,6 +106,24 @@ elf_machine_dynamic (void) |
| 5147 | /* Then subtract off the load address offset. */ |
| 5148 | return runtime_dynamic - elf_machine_load_address() ; |
| 5149 | } |
| 5150 | +#else /* __PCREL__ */ |
| 5151 | +/* In PCREL mode, r2 may have been clobbered. Rely on relative |
| 5152 | + relocations instead. */ |
| 5153 | + |
| 5154 | +static inline ElfW(Addr) |
| 5155 | +elf_machine_load_address (void) |
| 5156 | +{ |
| 5157 | + extern const ElfW(Ehdr) __ehdr_start attribute_hidden; |
| 5158 | + return (ElfW(Addr)) &__ehdr_start; |
| 5159 | +} |
| 5160 | + |
| 5161 | +static inline ElfW(Addr) |
| 5162 | +elf_machine_dynamic (void) |
| 5163 | +{ |
| 5164 | + extern ElfW(Dyn) _DYNAMIC[] attribute_hidden; |
| 5165 | + return (ElfW(Addr)) _DYNAMIC - elf_machine_load_address (); |
| 5166 | +} |
| 5167 | +#endif /* __PCREL__ */ |
| 5168 | |
| 5169 | /* The PLT uses Elf64_Rela relocs. */ |
| 5170 | #define elf_machine_relplt elf_machine_rela |
| 5171 | diff --git a/sysdeps/powerpc/powerpc64/le/power11/Implies b/sysdeps/powerpc/powerpc64/le/power11/Implies |
| 5172 | new file mode 100644 |
| 5173 | index 0000000000..e18182dcc1 |
| 5174 | --- /dev/null |
| 5175 | +++ b/sysdeps/powerpc/powerpc64/le/power11/Implies |
| 5176 | @@ -0,0 +1,2 @@ |
| 5177 | +powerpc/powerpc64/le/power10/fpu |
| 5178 | +powerpc/powerpc64/le/power10 |
| 5179 | diff --git a/sysdeps/powerpc/powerpc64/le/power11/fpu/Implies b/sysdeps/powerpc/powerpc64/le/power11/fpu/Implies |
| 5180 | new file mode 100644 |
| 5181 | index 0000000000..e41bd55684 |
| 5182 | --- /dev/null |
| 5183 | +++ b/sysdeps/powerpc/powerpc64/le/power11/fpu/Implies |
| 5184 | @@ -0,0 +1 @@ |
| 5185 | +powerpc/powerpc64/le/power10/fpu |
| 5186 | diff --git a/sysdeps/powerpc/powerpc64/le/power11/fpu/multiarch/Implies b/sysdeps/powerpc/powerpc64/le/power11/fpu/multiarch/Implies |
| 5187 | new file mode 100644 |
| 5188 | index 0000000000..c838d50931 |
| 5189 | --- /dev/null |
| 5190 | +++ b/sysdeps/powerpc/powerpc64/le/power11/fpu/multiarch/Implies |
| 5191 | @@ -0,0 +1 @@ |
| 5192 | +powerpc/powerpc64/le/power10/fpu/multiarch |
| 5193 | diff --git a/sysdeps/powerpc/powerpc64/le/power11/multiarch/Implies b/sysdeps/powerpc/powerpc64/le/power11/multiarch/Implies |
| 5194 | new file mode 100644 |
| 5195 | index 0000000000..687248c3c2 |
| 5196 | --- /dev/null |
| 5197 | +++ b/sysdeps/powerpc/powerpc64/le/power11/multiarch/Implies |
| 5198 | @@ -0,0 +1 @@ |
| 5199 | +powerpc/powerpc64/le/power10/multiarch |
| 5200 | diff --git a/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c b/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c |
| 5201 | index 77465d9133..65d3e69303 100644 |
| 5202 | --- a/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c |
| 5203 | +++ b/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c |
| 5204 | @@ -36,9 +36,11 @@ compute_level (void) |
| 5205 | return 9; |
| 5206 | if (strcmp (platform, "power10") == 0) |
| 5207 | return 10; |
| 5208 | + if (strcmp (platform, "power11") == 0) |
| 5209 | + return 11; |
| 5210 | printf ("warning: unrecognized AT_PLATFORM value: %s\n", platform); |
| 5211 | - /* Assume that the new platform supports POWER10. */ |
| 5212 | - return 10; |
| 5213 | + /* Assume that the new platform supports POWER11. */ |
| 5214 | + return 11; |
| 5215 | } |
| 5216 | |
| 5217 | static int |
| 5218 | diff --git a/sysdeps/powerpc/preconfigure b/sysdeps/powerpc/preconfigure |
| 5219 | index 4de94089a3..9e5a07ab6d 100644 |
| 5220 | --- a/sysdeps/powerpc/preconfigure |
| 5221 | +++ b/sysdeps/powerpc/preconfigure |
| 5222 | @@ -58,7 +58,7 @@ fi |
| 5223 | |
| 5224 | ;; |
| 5225 | |
| 5226 | - a2|970|power[4-9]|power5x|power6+|power10) |
| 5227 | + a2|970|power[4-9]|power5x|power6+|power10|power11) |
| 5228 | submachine=${archcpu} |
| 5229 | if test ${libc_cv_cc_submachine+y} |
| 5230 | then : |
| 5231 | diff --git a/sysdeps/powerpc/preconfigure.ac b/sysdeps/powerpc/preconfigure.ac |
| 5232 | index 6c63bd8257..14b6dafd4a 100644 |
| 5233 | --- a/sysdeps/powerpc/preconfigure.ac |
| 5234 | +++ b/sysdeps/powerpc/preconfigure.ac |
| 5235 | @@ -46,7 +46,7 @@ case "${machine}:${submachine}" in |
| 5236 | AC_CACHE_VAL(libc_cv_cc_submachine,libc_cv_cc_submachine="") |
| 5237 | ;; |
| 5238 | |
| 5239 | - a2|970|power[[4-9]]|power5x|power6+|power10) |
| 5240 | + a2|970|power[[4-9]]|power5x|power6+|power10|power11) |
| 5241 | submachine=${archcpu} |
| 5242 | AC_CACHE_VAL(libc_cv_cc_submachine,libc_cv_cc_submachine="") |
| 5243 | ;; |
| 5244 | diff --git a/sysdeps/powerpc/utmp-size.h b/sysdeps/powerpc/utmp-size.h |
| 5245 | new file mode 100644 |
| 5246 | index 0000000000..8f21ebe1b6 |
| 5247 | --- /dev/null |
| 5248 | +++ b/sysdeps/powerpc/utmp-size.h |
| 5249 | @@ -0,0 +1,2 @@ |
| 5250 | +#define UTMP_SIZE 384 |
| 5251 | +#define LASTLOG_SIZE 292 |
| 5252 | diff --git a/sysdeps/pthread/tst-cancel30.c b/sysdeps/pthread/tst-cancel30.c |
| 5253 | index 3030660e5f..94ad6281bc 100644 |
| 5254 | --- a/sysdeps/pthread/tst-cancel30.c |
| 5255 | +++ b/sysdeps/pthread/tst-cancel30.c |
| 5256 | @@ -18,6 +18,7 @@ |
| 5257 | License along with the GNU C Library; if not, see |
| 5258 | <https://www.gnu.org/licenses/>. */ |
| 5259 | |
| 5260 | +#include <errno.h> |
| 5261 | #include <support/check.h> |
| 5262 | #include <support/xstdio.h> |
| 5263 | #include <support/xthread.h> |
| 5264 | @@ -46,13 +47,19 @@ tf (void *arg) |
| 5265 | |
| 5266 | /* Wait indefinitely for cancellation, which only works if asynchronous |
| 5267 | cancellation is enabled. */ |
| 5268 | -#if defined SYS_ppoll || defined SYS_ppoll_time64 |
| 5269 | -# ifndef SYS_ppoll_time64 |
| 5270 | -# define SYS_ppoll_time64 SYS_ppoll |
| 5271 | +#ifdef SYS_ppoll_time64 |
| 5272 | + long int ret = syscall (SYS_ppoll_time64, NULL, 0, NULL, NULL); |
| 5273 | + (void) ret; |
| 5274 | +# ifdef SYS_ppoll |
| 5275 | + if (ret == -1 && errno == ENOSYS) |
| 5276 | + syscall (SYS_ppoll, NULL, 0, NULL, NULL); |
| 5277 | # endif |
| 5278 | - syscall (SYS_ppoll_time64, NULL, 0, NULL, NULL); |
| 5279 | #else |
| 5280 | +# ifdef SYS_ppoll |
| 5281 | + syscall (SYS_ppoll, NULL, 0, NULL, NULL); |
| 5282 | +# else |
| 5283 | for (;;); |
| 5284 | +# endif |
| 5285 | #endif |
| 5286 | |
| 5287 | return 0; |
| 5288 | diff --git a/sysdeps/riscv/utmp-size.h b/sysdeps/riscv/utmp-size.h |
| 5289 | new file mode 100644 |
| 5290 | index 0000000000..8f21ebe1b6 |
| 5291 | --- /dev/null |
| 5292 | +++ b/sysdeps/riscv/utmp-size.h |
| 5293 | @@ -0,0 +1,2 @@ |
| 5294 | +#define UTMP_SIZE 384 |
| 5295 | +#define LASTLOG_SIZE 292 |
| 5296 | diff --git a/sysdeps/sh/bits/wordsize.h b/sysdeps/sh/bits/wordsize.h |
| 5297 | new file mode 100644 |
| 5298 | index 0000000000..6ecbfe7c86 |
| 5299 | --- /dev/null |
| 5300 | +++ b/sysdeps/sh/bits/wordsize.h |
| 5301 | @@ -0,0 +1,21 @@ |
| 5302 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 5303 | + This file is part of the GNU C Library. |
| 5304 | + |
| 5305 | + The GNU C Library is free software; you can redistribute it and/or |
| 5306 | + modify it under the terms of the GNU Lesser General Public |
| 5307 | + License as published by the Free Software Foundation; either |
| 5308 | + version 2.1 of the License, or (at your option) any later version. |
| 5309 | + |
| 5310 | + The GNU C Library is distributed in the hope that it will be useful, |
| 5311 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 5312 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 5313 | + Lesser General Public License for more details. |
| 5314 | + |
| 5315 | + You should have received a copy of the GNU Lesser General Public |
| 5316 | + License along with the GNU C Library; if not, see |
| 5317 | + <https://www.gnu.org/licenses/>. */ |
| 5318 | + |
| 5319 | +#define __WORDSIZE 32 |
| 5320 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5321 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 5322 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5323 | diff --git a/sysdeps/sh/utmp-size.h b/sysdeps/sh/utmp-size.h |
| 5324 | new file mode 100644 |
| 5325 | index 0000000000..8f21ebe1b6 |
| 5326 | --- /dev/null |
| 5327 | +++ b/sysdeps/sh/utmp-size.h |
| 5328 | @@ -0,0 +1,2 @@ |
| 5329 | +#define UTMP_SIZE 384 |
| 5330 | +#define LASTLOG_SIZE 292 |
| 5331 | diff --git a/sysdeps/sparc/sparc32/bits/wordsize.h b/sysdeps/sparc/sparc32/bits/wordsize.h |
| 5332 | index 4bbd2e63b4..a2e79e0fa9 100644 |
| 5333 | --- a/sysdeps/sparc/sparc32/bits/wordsize.h |
| 5334 | +++ b/sysdeps/sparc/sparc32/bits/wordsize.h |
| 5335 | @@ -1,6 +1,6 @@ |
| 5336 | /* Determine the wordsize from the preprocessor defines. */ |
| 5337 | |
| 5338 | #define __WORDSIZE 32 |
| 5339 | -#define __WORDSIZE_TIME64_COMPAT32 0 |
| 5340 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5341 | #define __WORDSIZE32_SIZE_ULONG 0 |
| 5342 | #define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5343 | diff --git a/sysdeps/sparc/sparc64/bits/wordsize.h b/sysdeps/sparc/sparc64/bits/wordsize.h |
| 5344 | index 2f66f10d72..ea103e5970 100644 |
| 5345 | --- a/sysdeps/sparc/sparc64/bits/wordsize.h |
| 5346 | +++ b/sysdeps/sparc/sparc64/bits/wordsize.h |
| 5347 | @@ -2,10 +2,9 @@ |
| 5348 | |
| 5349 | #if defined __arch64__ || defined __sparcv9 |
| 5350 | # define __WORDSIZE 64 |
| 5351 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 5352 | #else |
| 5353 | # define __WORDSIZE 32 |
| 5354 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 5355 | # define __WORDSIZE32_SIZE_ULONG 0 |
| 5356 | # define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5357 | #endif |
| 5358 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5359 | diff --git a/sysdeps/sparc/sparc64/rtld-memset.c b/sysdeps/sparc/sparc64/rtld-memset.c |
| 5360 | index 55f3835790..a19202a620 100644 |
| 5361 | --- a/sysdeps/sparc/sparc64/rtld-memset.c |
| 5362 | +++ b/sysdeps/sparc/sparc64/rtld-memset.c |
| 5363 | @@ -1 +1,4 @@ |
| 5364 | #include <string/memset.c> |
| 5365 | +#if IS_IN(rtld) |
| 5366 | +strong_alias (memset, __memset_ultra1) |
| 5367 | +#endif |
| 5368 | diff --git a/sysdeps/sparc/utmp-size.h b/sysdeps/sparc/utmp-size.h |
| 5369 | new file mode 100644 |
| 5370 | index 0000000000..8f21ebe1b6 |
| 5371 | --- /dev/null |
| 5372 | +++ b/sysdeps/sparc/utmp-size.h |
| 5373 | @@ -0,0 +1,2 @@ |
| 5374 | +#define UTMP_SIZE 384 |
| 5375 | +#define LASTLOG_SIZE 292 |
| 5376 | diff --git a/sysdeps/unix/sysv/linux/aarch64/cpu-features.c b/sysdeps/unix/sysv/linux/aarch64/cpu-features.c |
| 5377 | index b1a3f673f0..c0b047bc0d 100644 |
| 5378 | --- a/sysdeps/unix/sysv/linux/aarch64/cpu-features.c |
| 5379 | +++ b/sysdeps/unix/sysv/linux/aarch64/cpu-features.c |
| 5380 | @@ -21,6 +21,7 @@ |
| 5381 | #include <sys/auxv.h> |
| 5382 | #include <elf/dl-hwcaps.h> |
| 5383 | #include <sys/prctl.h> |
| 5384 | +#include <sys/utsname.h> |
| 5385 | #include <dl-tunables-parse.h> |
| 5386 | |
| 5387 | #define DCZID_DZP_MASK (1 << 4) |
| 5388 | @@ -62,6 +63,46 @@ get_midr_from_mcpu (const struct tunable_str_t *mcpu) |
| 5389 | return UINT64_MAX; |
| 5390 | } |
| 5391 | |
| 5392 | +#if __LINUX_KERNEL_VERSION < 0x060200 |
| 5393 | + |
| 5394 | +/* Return true if we prefer using SVE in string ifuncs. Old kernels disable |
| 5395 | + SVE after every system call which results in unnecessary traps if memcpy |
| 5396 | + uses SVE. This is true for kernels between 4.15.0 and before 6.2.0, except |
| 5397 | + for 5.14.0 which was patched. For these versions return false to avoid using |
| 5398 | + SVE ifuncs. |
| 5399 | + Parse the kernel version into a 24-bit kernel.major.minor value without |
| 5400 | + calling any library functions. If uname() is not supported or if the version |
| 5401 | + format is not recognized, assume the kernel is modern and return true. */ |
| 5402 | + |
| 5403 | +static inline bool |
| 5404 | +prefer_sve_ifuncs (void) |
| 5405 | +{ |
| 5406 | + struct utsname buf; |
| 5407 | + const char *p = &buf.release[0]; |
| 5408 | + int kernel = 0; |
| 5409 | + int val; |
| 5410 | + |
| 5411 | + if (__uname (&buf) < 0) |
| 5412 | + return true; |
| 5413 | + |
| 5414 | + for (int shift = 16; shift >= 0; shift -= 8) |
| 5415 | + { |
| 5416 | + for (val = 0; *p >= '0' && *p <= '9'; p++) |
| 5417 | + val = val * 10 + *p - '0'; |
| 5418 | + kernel |= (val & 255) << shift; |
| 5419 | + if (*p++ != '.') |
| 5420 | + break; |
| 5421 | + } |
| 5422 | + |
| 5423 | + if (kernel >= 0x060200 || kernel == 0x050e00) |
| 5424 | + return true; |
| 5425 | + if (kernel >= 0x040f00) |
| 5426 | + return false; |
| 5427 | + return true; |
| 5428 | +} |
| 5429 | + |
| 5430 | +#endif |
| 5431 | + |
| 5432 | static inline void |
| 5433 | init_cpu_features (struct cpu_features *cpu_features) |
| 5434 | { |
| 5435 | @@ -126,6 +167,13 @@ init_cpu_features (struct cpu_features *cpu_features) |
| 5436 | /* Check if SVE is supported. */ |
| 5437 | cpu_features->sve = GLRO (dl_hwcap) & HWCAP_SVE; |
| 5438 | |
| 5439 | + cpu_features->prefer_sve_ifuncs = cpu_features->sve; |
| 5440 | + |
| 5441 | +#if __LINUX_KERNEL_VERSION < 0x060200 |
| 5442 | + if (cpu_features->sve) |
| 5443 | + cpu_features->prefer_sve_ifuncs = prefer_sve_ifuncs (); |
| 5444 | +#endif |
| 5445 | + |
| 5446 | /* Check if MOPS is supported. */ |
| 5447 | cpu_features->mops = GLRO (dl_hwcap2) & HWCAP2_MOPS; |
| 5448 | } |
| 5449 | diff --git a/sysdeps/unix/sysv/linux/dl-parse_auxv.h b/sysdeps/unix/sysv/linux/dl-parse_auxv.h |
| 5450 | index e3d758b163..ea2a58ecb1 100644 |
| 5451 | --- a/sysdeps/unix/sysv/linux/dl-parse_auxv.h |
| 5452 | +++ b/sysdeps/unix/sysv/linux/dl-parse_auxv.h |
| 5453 | @@ -47,6 +47,8 @@ void _dl_parse_auxv (ElfW(auxv_t) *av, dl_parse_auxv_t auxv_values) |
| 5454 | GLRO(dl_platform) = (void *) auxv_values[AT_PLATFORM]; |
| 5455 | GLRO(dl_hwcap) = auxv_values[AT_HWCAP]; |
| 5456 | GLRO(dl_hwcap2) = auxv_values[AT_HWCAP2]; |
| 5457 | + GLRO(dl_hwcap3) = auxv_values[AT_HWCAP3]; |
| 5458 | + GLRO(dl_hwcap4) = auxv_values[AT_HWCAP4]; |
| 5459 | GLRO(dl_clktck) = auxv_values[AT_CLKTCK]; |
| 5460 | GLRO(dl_fpu_control) = auxv_values[AT_FPUCW]; |
| 5461 | _dl_random = (void *) auxv_values[AT_RANDOM]; |
| 5462 | diff --git a/sysdeps/unix/sysv/linux/dl-sysdep.c b/sysdeps/unix/sysv/linux/dl-sysdep.c |
| 5463 | index ad3692d738..e1b14e9eb3 100644 |
| 5464 | --- a/sysdeps/unix/sysv/linux/dl-sysdep.c |
| 5465 | +++ b/sysdeps/unix/sysv/linux/dl-sysdep.c |
| 5466 | @@ -197,6 +197,8 @@ _dl_show_auxv (void) |
| 5467 | [AT_SYSINFO_EHDR - 2] = { "SYSINFO_EHDR: 0x", hex }, |
| 5468 | [AT_RANDOM - 2] = { "RANDOM: 0x", hex }, |
| 5469 | [AT_HWCAP2 - 2] = { "HWCAP2: 0x", hex }, |
| 5470 | + [AT_HWCAP3 - 2] = { "HWCAP3: 0x", hex }, |
| 5471 | + [AT_HWCAP4 - 2] = { "HWCAP4: 0x", hex }, |
| 5472 | [AT_MINSIGSTKSZ - 2] = { "MINSIGSTKSZ: ", dec }, |
| 5473 | [AT_L1I_CACHESIZE - 2] = { "L1I_CACHESIZE: ", dec }, |
| 5474 | [AT_L1I_CACHEGEOMETRY - 2] = { "L1I_CACHEGEOMETRY: 0x", hex }, |
| 5475 | diff --git a/sysdeps/unix/sysv/linux/hppa/bits/wordsize.h b/sysdeps/unix/sysv/linux/hppa/bits/wordsize.h |
| 5476 | new file mode 100644 |
| 5477 | index 0000000000..6ecbfe7c86 |
| 5478 | --- /dev/null |
| 5479 | +++ b/sysdeps/unix/sysv/linux/hppa/bits/wordsize.h |
| 5480 | @@ -0,0 +1,21 @@ |
| 5481 | +/* Copyright (C) 1999-2024 Free Software Foundation, Inc. |
| 5482 | + This file is part of the GNU C Library. |
| 5483 | + |
| 5484 | + The GNU C Library is free software; you can redistribute it and/or |
| 5485 | + modify it under the terms of the GNU Lesser General Public |
| 5486 | + License as published by the Free Software Foundation; either |
| 5487 | + version 2.1 of the License, or (at your option) any later version. |
| 5488 | + |
| 5489 | + The GNU C Library is distributed in the hope that it will be useful, |
| 5490 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 5491 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 5492 | + Lesser General Public License for more details. |
| 5493 | + |
| 5494 | + You should have received a copy of the GNU Lesser General Public |
| 5495 | + License along with the GNU C Library; if not, see |
| 5496 | + <https://www.gnu.org/licenses/>. */ |
| 5497 | + |
| 5498 | +#define __WORDSIZE 32 |
| 5499 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5500 | +#define __WORDSIZE32_SIZE_ULONG 0 |
| 5501 | +#define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5502 | diff --git a/sysdeps/unix/sysv/linux/mips/clone3.S b/sysdeps/unix/sysv/linux/mips/clone3.S |
| 5503 | index e9fec2fa47..481b8ae963 100644 |
| 5504 | --- a/sysdeps/unix/sysv/linux/mips/clone3.S |
| 5505 | +++ b/sysdeps/unix/sysv/linux/mips/clone3.S |
| 5506 | @@ -37,11 +37,6 @@ |
| 5507 | |
| 5508 | .text |
| 5509 | .set nomips16 |
| 5510 | -#if _MIPS_SIM == _ABIO32 |
| 5511 | -# define EXTRA_LOCALS 1 |
| 5512 | -#else |
| 5513 | -# define EXTRA_LOCALS 0 |
| 5514 | -#endif |
| 5515 | #define FRAMESZ ((NARGSAVE*SZREG)+ALSZ)&ALMASK |
| 5516 | GPOFF= FRAMESZ-(1*SZREG) |
| 5517 | NESTED(__clone3, SZREG, sp) |
| 5518 | @@ -68,8 +63,31 @@ NESTED(__clone3, SZREG, sp) |
| 5519 | beqz a0, L(error) /* No NULL cl_args pointer. */ |
| 5520 | beqz a2, L(error) /* No NULL function pointer. */ |
| 5521 | |
| 5522 | +#if _MIPS_SIM == _ABIO32 |
| 5523 | + /* Both stack and stack_size on clone_args are defined as uint64_t, and |
| 5524 | + there is no need to handle values larger than to 32 bits for o32. */ |
| 5525 | +# if __BYTE_ORDER == __BIG_ENDIAN |
| 5526 | +# define CL_STACKPOINTER_OFFSET 44 |
| 5527 | +# define CL_STACKSIZE_OFFSET 52 |
| 5528 | +# else |
| 5529 | +# define CL_STACKPOINTER_OFFSET 40 |
| 5530 | +# define CL_STACKSIZE_OFFSET 48 |
| 5531 | +# endif |
| 5532 | + |
| 5533 | + /* For o32 we need to setup a minimal stack frame to allow cprestore |
| 5534 | + on __thread_start_clone3. Also there is no guarantee by kABI that |
| 5535 | + $8 will be preserved after syscall execution (so we need to save it |
| 5536 | + on the provided stack). */ |
| 5537 | + lw t0, CL_STACKPOINTER_OFFSET(a0) /* Load the stack pointer. */ |
| 5538 | + lw t1, CL_STACKSIZE_OFFSET(a0) /* Load the stack_size. */ |
| 5539 | + addiu t1, -32 /* Update the stack size. */ |
| 5540 | + addu t2, t1, t0 /* Calculate the thread stack. */ |
| 5541 | + sw a3, 0(t2) /* Save argument pointer. */ |
| 5542 | + sw t1, CL_STACKSIZE_OFFSET(a0) /* Save the new stack size. */ |
| 5543 | +#else |
| 5544 | move $8, a3 /* a3 is set to 0/1 for syscall success/error |
| 5545 | while a4/$8 is returned unmodified. */ |
| 5546 | +#endif |
| 5547 | |
| 5548 | /* Do the system call, the kernel expects: |
| 5549 | v0: system call number |
| 5550 | @@ -125,7 +143,11 @@ L(thread_start_clone3): |
| 5551 | |
| 5552 | /* Restore the arg for user's function. */ |
| 5553 | move t9, a2 /* Function pointer. */ |
| 5554 | +#if _MIPS_SIM == _ABIO32 |
| 5555 | + PTR_L a0, 0(sp) |
| 5556 | +#else |
| 5557 | move a0, $8 /* Argument pointer. */ |
| 5558 | +#endif |
| 5559 | |
| 5560 | /* Call the user's function. */ |
| 5561 | jal t9 |
| 5562 | diff --git a/sysdeps/unix/sysv/linux/powerpc/bits/wordsize.h b/sysdeps/unix/sysv/linux/powerpc/bits/wordsize.h |
| 5563 | index 04ca9debf0..6993fb6b29 100644 |
| 5564 | --- a/sysdeps/unix/sysv/linux/powerpc/bits/wordsize.h |
| 5565 | +++ b/sysdeps/unix/sysv/linux/powerpc/bits/wordsize.h |
| 5566 | @@ -2,10 +2,9 @@ |
| 5567 | |
| 5568 | #if defined __powerpc64__ |
| 5569 | # define __WORDSIZE 64 |
| 5570 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 5571 | #else |
| 5572 | # define __WORDSIZE 32 |
| 5573 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 5574 | # define __WORDSIZE32_SIZE_ULONG 0 |
| 5575 | # define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5576 | #endif |
| 5577 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5578 | diff --git a/sysdeps/unix/sysv/linux/powerpc/cpu-features.c b/sysdeps/unix/sysv/linux/powerpc/cpu-features.c |
| 5579 | index 8e8a5ec2ea..a947d62db6 100644 |
| 5580 | --- a/sysdeps/unix/sysv/linux/powerpc/cpu-features.c |
| 5581 | +++ b/sysdeps/unix/sysv/linux/powerpc/cpu-features.c |
| 5582 | @@ -94,6 +94,8 @@ init_cpu_features (struct cpu_features *cpu_features, uint64_t hwcaps[]) |
| 5583 | which are set by __tcb_parse_hwcap_and_convert_at_platform. */ |
| 5584 | cpu_features->hwcap = hwcaps[0]; |
| 5585 | cpu_features->hwcap2 = hwcaps[1]; |
| 5586 | + cpu_features->hwcap3 = hwcaps[2]; |
| 5587 | + cpu_features->hwcap4 = hwcaps[3]; |
| 5588 | /* Default is to use aligned memory access on optimized function unless |
| 5589 | tunables is enable, since for this case user can explicit disable |
| 5590 | unaligned optimizations. */ |
| 5591 | diff --git a/sysdeps/unix/sysv/linux/powerpc/cpu-features.h b/sysdeps/unix/sysv/linux/powerpc/cpu-features.h |
| 5592 | index 1294f0b601..e9eb6a13c8 100644 |
| 5593 | --- a/sysdeps/unix/sysv/linux/powerpc/cpu-features.h |
| 5594 | +++ b/sysdeps/unix/sysv/linux/powerpc/cpu-features.h |
| 5595 | @@ -26,6 +26,8 @@ struct cpu_features |
| 5596 | bool use_cached_memopt; |
| 5597 | unsigned long int hwcap; |
| 5598 | unsigned long int hwcap2; |
| 5599 | + unsigned long int hwcap3; |
| 5600 | + unsigned long int hwcap4; |
| 5601 | }; |
| 5602 | |
| 5603 | static const char hwcap_names[] = { |
| 5604 | diff --git a/sysdeps/unix/sysv/linux/powerpc/libc-start.c b/sysdeps/unix/sysv/linux/powerpc/libc-start.c |
| 5605 | index a4705daf1c..6a00cd88cd 100644 |
| 5606 | --- a/sysdeps/unix/sysv/linux/powerpc/libc-start.c |
| 5607 | +++ b/sysdeps/unix/sysv/linux/powerpc/libc-start.c |
| 5608 | @@ -87,6 +87,12 @@ __libc_start_main_impl (int argc, char **argv, |
| 5609 | case AT_HWCAP2: |
| 5610 | _dl_hwcap2 = (unsigned long int) av->a_un.a_val; |
| 5611 | break; |
| 5612 | + case AT_HWCAP3: |
| 5613 | + _dl_hwcap3 = (unsigned long int) av->a_un.a_val; |
| 5614 | + break; |
| 5615 | + case AT_HWCAP4: |
| 5616 | + _dl_hwcap4 = (unsigned long int) av->a_un.a_val; |
| 5617 | + break; |
| 5618 | case AT_PLATFORM: |
| 5619 | _dl_platform = (void *) av->a_un.a_val; |
| 5620 | break; |
| 5621 | diff --git a/sysdeps/unix/sysv/linux/s390/s390-32/clone.S b/sysdeps/unix/sysv/linux/s390/s390-32/clone.S |
| 5622 | index 4c882ef2ee..a7a863242c 100644 |
| 5623 | --- a/sysdeps/unix/sysv/linux/s390/s390-32/clone.S |
| 5624 | +++ b/sysdeps/unix/sysv/linux/s390/s390-32/clone.S |
| 5625 | @@ -53,6 +53,7 @@ ENTRY(__clone) |
| 5626 | br %r14 |
| 5627 | error: |
| 5628 | lhi %r2,-EINVAL |
| 5629 | + lm %r6,%r7,24(%r15) /* Load registers. */ |
| 5630 | j SYSCALL_ERROR_LABEL |
| 5631 | PSEUDO_END (__clone) |
| 5632 | |
| 5633 | diff --git a/sysdeps/unix/sysv/linux/s390/s390-64/clone.S b/sysdeps/unix/sysv/linux/s390/s390-64/clone.S |
| 5634 | index 4eb104be71..c552a6b8de 100644 |
| 5635 | --- a/sysdeps/unix/sysv/linux/s390/s390-64/clone.S |
| 5636 | +++ b/sysdeps/unix/sysv/linux/s390/s390-64/clone.S |
| 5637 | @@ -54,6 +54,7 @@ ENTRY(__clone) |
| 5638 | br %r14 |
| 5639 | error: |
| 5640 | lghi %r2,-EINVAL |
| 5641 | + lmg %r6,%r7,48(%r15) /* Restore registers. */ |
| 5642 | jg SYSCALL_ERROR_LABEL |
| 5643 | PSEUDO_END (__clone) |
| 5644 | |
| 5645 | diff --git a/sysdeps/unix/sysv/linux/sched_getcpu.c b/sysdeps/unix/sysv/linux/sched_getcpu.c |
| 5646 | index dfb884568d..72a3360550 100644 |
| 5647 | --- a/sysdeps/unix/sysv/linux/sched_getcpu.c |
| 5648 | +++ b/sysdeps/unix/sysv/linux/sched_getcpu.c |
| 5649 | @@ -33,17 +33,9 @@ vsyscall_sched_getcpu (void) |
| 5650 | return r == -1 ? r : cpu; |
| 5651 | } |
| 5652 | |
| 5653 | -#ifdef RSEQ_SIG |
| 5654 | int |
| 5655 | sched_getcpu (void) |
| 5656 | { |
| 5657 | int cpu_id = THREAD_GETMEM_VOLATILE (THREAD_SELF, rseq_area.cpu_id); |
| 5658 | return __glibc_likely (cpu_id >= 0) ? cpu_id : vsyscall_sched_getcpu (); |
| 5659 | } |
| 5660 | -#else /* RSEQ_SIG */ |
| 5661 | -int |
| 5662 | -sched_getcpu (void) |
| 5663 | -{ |
| 5664 | - return vsyscall_sched_getcpu (); |
| 5665 | -} |
| 5666 | -#endif /* RSEQ_SIG */ |
| 5667 | diff --git a/sysdeps/unix/sysv/linux/sparc/bits/wordsize.h b/sysdeps/unix/sysv/linux/sparc/bits/wordsize.h |
| 5668 | index 7562875ee2..ea103e5970 100644 |
| 5669 | --- a/sysdeps/unix/sysv/linux/sparc/bits/wordsize.h |
| 5670 | +++ b/sysdeps/unix/sysv/linux/sparc/bits/wordsize.h |
| 5671 | @@ -2,10 +2,9 @@ |
| 5672 | |
| 5673 | #if defined __arch64__ || defined __sparcv9 |
| 5674 | # define __WORDSIZE 64 |
| 5675 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 5676 | #else |
| 5677 | # define __WORDSIZE 32 |
| 5678 | # define __WORDSIZE32_SIZE_ULONG 0 |
| 5679 | # define __WORDSIZE32_PTRDIFF_LONG 0 |
| 5680 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 5681 | #endif |
| 5682 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 5683 | diff --git a/sysdeps/unix/sysv/linux/timespec_get.c b/sysdeps/unix/sysv/linux/timespec_get.c |
| 5684 | index c6e5e66289..778d1e3354 100644 |
| 5685 | --- a/sysdeps/unix/sysv/linux/timespec_get.c |
| 5686 | +++ b/sysdeps/unix/sysv/linux/timespec_get.c |
| 5687 | @@ -5,7 +5,7 @@ |
| 5688 | The GNU C Library is free software; you can redistribute it and/or |
| 5689 | modify it under the terms of the GNU Lesser General Public |
| 5690 | License as published by the Free Software Foundation; either |
| 5691 | - version 2.1 of the License. |
| 5692 | + version 2.1 of the License, or (at your option) any later version. |
| 5693 | |
| 5694 | The GNU C Library is distributed in the hope that it will be useful, |
| 5695 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 5696 | diff --git a/sysdeps/unix/sysv/linux/timespec_getres.c b/sysdeps/unix/sysv/linux/timespec_getres.c |
| 5697 | index 5acebe2a2c..2eef9e512c 100644 |
| 5698 | --- a/sysdeps/unix/sysv/linux/timespec_getres.c |
| 5699 | +++ b/sysdeps/unix/sysv/linux/timespec_getres.c |
| 5700 | @@ -5,7 +5,7 @@ |
| 5701 | The GNU C Library is free software; you can redistribute it and/or |
| 5702 | modify it under the terms of the GNU Lesser General Public |
| 5703 | License as published by the Free Software Foundation; either |
| 5704 | - version 2.1 of the License. |
| 5705 | + version 2.1 of the License, or (at your option) any later version. |
| 5706 | |
| 5707 | The GNU C Library is distributed in the hope that it will be useful, |
| 5708 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 5709 | diff --git a/sysdeps/unix/sysv/linux/tst-clone.c b/sysdeps/unix/sysv/linux/tst-clone.c |
| 5710 | index 470676ab2b..2bc7124983 100644 |
| 5711 | --- a/sysdeps/unix/sysv/linux/tst-clone.c |
| 5712 | +++ b/sysdeps/unix/sysv/linux/tst-clone.c |
| 5713 | @@ -16,12 +16,16 @@ |
| 5714 | License along with the GNU C Library; if not, see |
| 5715 | <https://www.gnu.org/licenses/>. */ |
| 5716 | |
| 5717 | -/* BZ #2386 */ |
| 5718 | +/* BZ #2386, BZ #31402 */ |
| 5719 | #include <errno.h> |
| 5720 | #include <stdio.h> |
| 5721 | #include <stdlib.h> |
| 5722 | #include <unistd.h> |
| 5723 | #include <sched.h> |
| 5724 | +#include <stackinfo.h> /* For _STACK_GROWS_{UP,DOWN}. */ |
| 5725 | +#include <support/check.h> |
| 5726 | + |
| 5727 | +volatile unsigned v = 0xdeadbeef; |
| 5728 | |
| 5729 | int child_fn(void *arg) |
| 5730 | { |
| 5731 | @@ -30,22 +34,67 @@ int child_fn(void *arg) |
| 5732 | } |
| 5733 | |
| 5734 | static int |
| 5735 | -do_test (void) |
| 5736 | +__attribute__((noinline)) |
| 5737 | +do_clone (int (*fn)(void *), void *stack) |
| 5738 | { |
| 5739 | int result; |
| 5740 | + unsigned int a = v; |
| 5741 | + unsigned int b = v; |
| 5742 | + unsigned int c = v; |
| 5743 | + unsigned int d = v; |
| 5744 | + unsigned int e = v; |
| 5745 | + unsigned int f = v; |
| 5746 | + unsigned int g = v; |
| 5747 | + unsigned int h = v; |
| 5748 | + unsigned int i = v; |
| 5749 | + unsigned int j = v; |
| 5750 | + unsigned int k = v; |
| 5751 | + unsigned int l = v; |
| 5752 | + unsigned int m = v; |
| 5753 | + unsigned int n = v; |
| 5754 | + unsigned int o = v; |
| 5755 | + |
| 5756 | + result = clone (fn, stack, 0, NULL); |
| 5757 | + |
| 5758 | + /* Check that clone does not clobber call-saved registers. */ |
| 5759 | + TEST_VERIFY (a == v && b == v && c == v && d == v && e == v && f == v |
| 5760 | + && g == v && h == v && i == v && j == v && k == v && l == v |
| 5761 | + && m == v && n == v && o == v); |
| 5762 | + |
| 5763 | + return result; |
| 5764 | +} |
| 5765 | + |
| 5766 | +static void |
| 5767 | +__attribute__((noinline)) |
| 5768 | +do_test_single (int (*fn)(void *), void *stack) |
| 5769 | +{ |
| 5770 | + printf ("%s (fn=%p, stack=%p)\n", __FUNCTION__, fn, stack); |
| 5771 | + errno = 0; |
| 5772 | + |
| 5773 | + int result = do_clone (fn, stack); |
| 5774 | + |
| 5775 | + TEST_COMPARE (errno, EINVAL); |
| 5776 | + TEST_COMPARE (result, -1); |
| 5777 | +} |
| 5778 | |
| 5779 | - result = clone (child_fn, NULL, 0, NULL); |
| 5780 | +static int |
| 5781 | +do_test (void) |
| 5782 | +{ |
| 5783 | + char st[128 * 1024] __attribute__ ((aligned)); |
| 5784 | + void *stack = NULL; |
| 5785 | +#if _STACK_GROWS_DOWN |
| 5786 | + stack = st + sizeof (st); |
| 5787 | +#elif _STACK_GROWS_UP |
| 5788 | + stack = st; |
| 5789 | +#else |
| 5790 | +# error "Define either _STACK_GROWS_DOWN or _STACK_GROWS_UP" |
| 5791 | +#endif |
| 5792 | |
| 5793 | - if (errno != EINVAL || result != -1) |
| 5794 | - { |
| 5795 | - printf ("FAIL: clone()=%d (wanted -1) errno=%d (wanted %d)\n", |
| 5796 | - result, errno, EINVAL); |
| 5797 | - return 1; |
| 5798 | - } |
| 5799 | + do_test_single (child_fn, NULL); |
| 5800 | + do_test_single (NULL, stack); |
| 5801 | + do_test_single (NULL, NULL); |
| 5802 | |
| 5803 | - puts ("All OK"); |
| 5804 | return 0; |
| 5805 | } |
| 5806 | |
| 5807 | -#define TEST_FUNCTION do_test () |
| 5808 | -#include "../test-skeleton.c" |
| 5809 | +#include <support/test-driver.c> |
| 5810 | diff --git a/sysdeps/unix/sysv/linux/x86_64/Makefile b/sysdeps/unix/sysv/linux/x86_64/Makefile |
| 5811 | index 4223feb95f..9a1e7aa646 100644 |
| 5812 | --- a/sysdeps/unix/sysv/linux/x86_64/Makefile |
| 5813 | +++ b/sysdeps/unix/sysv/linux/x86_64/Makefile |
| 5814 | @@ -63,6 +63,33 @@ $(objpfx)libx86-64-isa-level%.os: $(..)/sysdeps/unix/sysv/linux/x86_64/x86-64-is |
| 5815 | $(objpfx)libx86-64-isa-level.so: $(objpfx)libx86-64-isa-level-1.so |
| 5816 | cp $< $@ |
| 5817 | endif |
| 5818 | + |
| 5819 | +ifeq (yes,$(have-mamx-tile)) |
| 5820 | +tests += \ |
| 5821 | + tst-gnu2-tls2-amx \ |
| 5822 | +# tests |
| 5823 | + |
| 5824 | +modules-names += \ |
| 5825 | + tst-gnu2-tls2-amx-mod0 \ |
| 5826 | + tst-gnu2-tls2-amx-mod1 \ |
| 5827 | + tst-gnu2-tls2-amx-mod2 \ |
| 5828 | +# modules-names |
| 5829 | + |
| 5830 | +$(objpfx)tst-gnu2-tls2-amx: $(shared-thread-library) |
| 5831 | +$(objpfx)tst-gnu2-tls2-amx.out: \ |
| 5832 | + $(objpfx)tst-gnu2-tls2-amx-mod0.so \ |
| 5833 | + $(objpfx)tst-gnu2-tls2-amx-mod1.so \ |
| 5834 | + $(objpfx)tst-gnu2-tls2-amx-mod2.so |
| 5835 | +$(objpfx)tst-gnu2-tls2-amx-mod0.so: $(libsupport) |
| 5836 | +$(objpfx)tst-gnu2-tls2-amx-mod1.so: $(libsupport) |
| 5837 | +$(objpfx)tst-gnu2-tls2-amx-mod2.so: $(libsupport) |
| 5838 | + |
| 5839 | +CFLAGS-tst-gnu2-tls2-amx.c += -mamx-tile |
| 5840 | +CFLAGS-tst-gnu2-tls2-amx-mod0.c += -mamx-tile -mtls-dialect=gnu2 |
| 5841 | +CFLAGS-tst-gnu2-tls2-amx-mod1.c += -mamx-tile -mtls-dialect=gnu2 |
| 5842 | +CFLAGS-tst-gnu2-tls2-amx-mod2.c += -mamx-tile -mtls-dialect=gnu2 |
| 5843 | +endif |
| 5844 | + |
| 5845 | endif # $(subdir) == elf |
| 5846 | |
| 5847 | ifneq ($(enable-cet),no) |
| 5848 | diff --git a/sysdeps/unix/sysv/linux/x86_64/include/asm/prctl.h b/sysdeps/unix/sysv/linux/x86_64/include/asm/prctl.h |
| 5849 | index 2f511321ad..ef4631bf4b 100644 |
| 5850 | --- a/sysdeps/unix/sysv/linux/x86_64/include/asm/prctl.h |
| 5851 | +++ b/sysdeps/unix/sysv/linux/x86_64/include/asm/prctl.h |
| 5852 | @@ -20,3 +20,8 @@ |
| 5853 | # define ARCH_SHSTK_SHSTK 0x1 |
| 5854 | # define ARCH_SHSTK_WRSS 0x2 |
| 5855 | #endif |
| 5856 | + |
| 5857 | +#ifndef ARCH_GET_XCOMP_PERM |
| 5858 | +# define ARCH_GET_XCOMP_PERM 0x1022 |
| 5859 | +# define ARCH_REQ_XCOMP_PERM 0x1023 |
| 5860 | +#endif |
| 5861 | diff --git a/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod0.c b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod0.c |
| 5862 | new file mode 100644 |
| 5863 | index 0000000000..2e0c7b91b7 |
| 5864 | --- /dev/null |
| 5865 | +++ b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod0.c |
| 5866 | @@ -0,0 +1,2 @@ |
| 5867 | +#include "tst-gnu2-tls2-amx.h" |
| 5868 | +#include <tst-gnu2-tls2mod0.c> |
| 5869 | diff --git a/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod1.c b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod1.c |
| 5870 | new file mode 100644 |
| 5871 | index 0000000000..b8a8ccf1c1 |
| 5872 | --- /dev/null |
| 5873 | +++ b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod1.c |
| 5874 | @@ -0,0 +1,2 @@ |
| 5875 | +#include "tst-gnu2-tls2-amx.h" |
| 5876 | +#include <tst-gnu2-tls2mod1.c> |
| 5877 | diff --git a/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod2.c b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod2.c |
| 5878 | new file mode 100644 |
| 5879 | index 0000000000..cdf4a8f363 |
| 5880 | --- /dev/null |
| 5881 | +++ b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx-mod2.c |
| 5882 | @@ -0,0 +1,2 @@ |
| 5883 | +#include "tst-gnu2-tls2-amx.h" |
| 5884 | +#include <tst-gnu2-tls2mod2.c> |
| 5885 | diff --git a/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx.c b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx.c |
| 5886 | new file mode 100644 |
| 5887 | index 0000000000..ae4dd82556 |
| 5888 | --- /dev/null |
| 5889 | +++ b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx.c |
| 5890 | @@ -0,0 +1,83 @@ |
| 5891 | +/* Test TLSDESC relocation with AMX. |
| 5892 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 5893 | + This file is part of the GNU C Library. |
| 5894 | + |
| 5895 | + The GNU C Library is free software; you can redistribute it and/or |
| 5896 | + modify it under the terms of the GNU Lesser General Public |
| 5897 | + License as published by the Free Software Foundation; either |
| 5898 | + version 2.1 of the License, or (at your option) any later version. |
| 5899 | + |
| 5900 | + The GNU C Library is distributed in the hope that it will be useful, |
| 5901 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 5902 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 5903 | + Lesser General Public License for more details. |
| 5904 | + |
| 5905 | + You should have received a copy of the GNU Lesser General Public |
| 5906 | + License along with the GNU C Library; if not, see |
| 5907 | + <http://www.gnu.org/licenses/>. */ |
| 5908 | + |
| 5909 | +#include <stdbool.h> |
| 5910 | +#include <asm/prctl.h> |
| 5911 | +#include <support/check.h> |
| 5912 | +#include "tst-gnu2-tls2-amx.h" |
| 5913 | + |
| 5914 | +extern int arch_prctl (int, ...); |
| 5915 | + |
| 5916 | +#define X86_XSTATE_TILECFG_ID 17 |
| 5917 | +#define X86_XSTATE_TILEDATA_ID 18 |
| 5918 | + |
| 5919 | +/* Initialize tile config. */ |
| 5920 | +__attribute__ ((noinline, noclone)) |
| 5921 | +static void |
| 5922 | +init_tile_config (__tilecfg *tileinfo) |
| 5923 | +{ |
| 5924 | + int i; |
| 5925 | + tileinfo->palette_id = 1; |
| 5926 | + tileinfo->start_row = 0; |
| 5927 | + |
| 5928 | + tileinfo->colsb[0] = MAX_ROWS; |
| 5929 | + tileinfo->rows[0] = MAX_ROWS; |
| 5930 | + |
| 5931 | + for (i = 1; i < 4; ++i) |
| 5932 | + { |
| 5933 | + tileinfo->colsb[i] = MAX_COLS; |
| 5934 | + tileinfo->rows[i] = MAX_ROWS; |
| 5935 | + } |
| 5936 | + |
| 5937 | + _tile_loadconfig (tileinfo); |
| 5938 | +} |
| 5939 | + |
| 5940 | +static bool |
| 5941 | +enable_amx (void) |
| 5942 | +{ |
| 5943 | + uint64_t bitmask; |
| 5944 | + if (arch_prctl (ARCH_GET_XCOMP_PERM, &bitmask) != 0) |
| 5945 | + return false; |
| 5946 | + |
| 5947 | + if ((bitmask & (1 << X86_XSTATE_TILECFG_ID)) == 0) |
| 5948 | + return false; |
| 5949 | + |
| 5950 | + if (arch_prctl (ARCH_REQ_XCOMP_PERM, X86_XSTATE_TILEDATA_ID) != 0) |
| 5951 | + return false; |
| 5952 | + |
| 5953 | + /* Load tile configuration. */ |
| 5954 | + __tilecfg tile_data = { 0 }; |
| 5955 | + init_tile_config (&tile_data); |
| 5956 | + |
| 5957 | + return true; |
| 5958 | +} |
| 5959 | + |
| 5960 | +/* An architecture can define it to clobber caller-saved registers in |
| 5961 | + malloc below to verify that the implicit TLSDESC call won't change |
| 5962 | + caller-saved registers. */ |
| 5963 | +static void |
| 5964 | +clear_tile_register (void) |
| 5965 | +{ |
| 5966 | + _tile_zero (2); |
| 5967 | +} |
| 5968 | + |
| 5969 | +#define MOD(i) "tst-gnu2-tls2-amx-mod" #i ".so" |
| 5970 | +#define IS_SUPPORTED() enable_amx () |
| 5971 | +#define PREPARE_MALLOC() clear_tile_register () |
| 5972 | + |
| 5973 | +#include <elf/tst-gnu2-tls2.c> |
| 5974 | diff --git a/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx.h b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx.h |
| 5975 | new file mode 100644 |
| 5976 | index 0000000000..1845a3caba |
| 5977 | --- /dev/null |
| 5978 | +++ b/sysdeps/unix/sysv/linux/x86_64/tst-gnu2-tls2-amx.h |
| 5979 | @@ -0,0 +1,63 @@ |
| 5980 | +/* Test TLSDESC relocation with AMX. |
| 5981 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 5982 | + This file is part of the GNU C Library. |
| 5983 | + |
| 5984 | + The GNU C Library is free software; you can redistribute it and/or |
| 5985 | + modify it under the terms of the GNU Lesser General Public |
| 5986 | + License as published by the Free Software Foundation; either |
| 5987 | + version 2.1 of the License, or (at your option) any later version. |
| 5988 | + |
| 5989 | + The GNU C Library is distributed in the hope that it will be useful, |
| 5990 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 5991 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 5992 | + Lesser General Public License for more details. |
| 5993 | + |
| 5994 | + You should have received a copy of the GNU Lesser General Public |
| 5995 | + License along with the GNU C Library; if not, see |
| 5996 | + <http://www.gnu.org/licenses/>. */ |
| 5997 | + |
| 5998 | +#include <stdint.h> |
| 5999 | +#include <string.h> |
| 6000 | +#include <x86intrin.h> |
| 6001 | +#include <support/check.h> |
| 6002 | + |
| 6003 | +#define MAX_ROWS 16 |
| 6004 | +#define MAX_COLS 64 |
| 6005 | +#define MAX 1024 |
| 6006 | +#define STRIDE 64 |
| 6007 | + |
| 6008 | +typedef struct __tile_config |
| 6009 | +{ |
| 6010 | + uint8_t palette_id; |
| 6011 | + uint8_t start_row; |
| 6012 | + uint8_t reserved_0[14]; |
| 6013 | + uint16_t colsb[16]; |
| 6014 | + uint8_t rows[16]; |
| 6015 | +} __tilecfg __attribute__ ((aligned (64))); |
| 6016 | + |
| 6017 | +/* Initialize int8_t buffer */ |
| 6018 | +static inline void |
| 6019 | +init_buffer (int8_t *buf, int8_t value) |
| 6020 | +{ |
| 6021 | + int rows, colsb, i, j; |
| 6022 | + rows = MAX_ROWS; |
| 6023 | + colsb = MAX_COLS; |
| 6024 | + |
| 6025 | + for (i = 0; i < rows; i++) |
| 6026 | + for (j = 0; j < colsb; j++) |
| 6027 | + buf[i * colsb + j] = value; |
| 6028 | +} |
| 6029 | + |
| 6030 | +#define BEFORE_TLSDESC_CALL() \ |
| 6031 | + int8_t src[MAX]; \ |
| 6032 | + int8_t res[MAX]; \ |
| 6033 | + /* Initialize src with data */ \ |
| 6034 | + init_buffer (src, 2); \ |
| 6035 | + /* Load tile rows from memory. */ \ |
| 6036 | + _tile_loadd (2, src, STRIDE); |
| 6037 | + |
| 6038 | +#define AFTER_TLSDESC_CALL() \ |
| 6039 | + /* Store the tile data to memory. */ \ |
| 6040 | + _tile_stored (2, res, STRIDE); \ |
| 6041 | + _tile_release (); \ |
| 6042 | + TEST_VERIFY_EXIT (memcmp (src, res, sizeof (res)) == 0); |
| 6043 | diff --git a/sysdeps/x86/Makefile b/sysdeps/x86/Makefile |
| 6044 | index 4d50b327b5..5311b594af 100644 |
| 6045 | --- a/sysdeps/x86/Makefile |
| 6046 | +++ b/sysdeps/x86/Makefile |
| 6047 | @@ -1,5 +1,5 @@ |
| 6048 | ifeq ($(subdir),csu) |
| 6049 | -gen-as-const-headers += cpu-features-offsets.sym |
| 6050 | +gen-as-const-headers += cpu-features-offsets.sym features-offsets.sym |
| 6051 | endif |
| 6052 | |
| 6053 | ifeq ($(subdir),elf) |
| 6054 | @@ -15,18 +15,18 @@ CFLAGS-dl-get-cpu-features.os += $(rtld-early-cflags) |
| 6055 | CFLAGS-get-cpuid-feature-leaf.o += $(no-stack-protector) |
| 6056 | |
| 6057 | tests += \ |
| 6058 | - tst-get-cpu-features \ |
| 6059 | - tst-get-cpu-features-static \ |
| 6060 | tst-cpu-features-cpuinfo \ |
| 6061 | tst-cpu-features-cpuinfo-static \ |
| 6062 | tst-cpu-features-supports \ |
| 6063 | tst-cpu-features-supports-static \ |
| 6064 | + tst-get-cpu-features \ |
| 6065 | + tst-get-cpu-features-static \ |
| 6066 | tst-hwcap-tunables \ |
| 6067 | # tests |
| 6068 | tests-static += \ |
| 6069 | - tst-get-cpu-features-static \ |
| 6070 | tst-cpu-features-cpuinfo-static \ |
| 6071 | tst-cpu-features-supports-static \ |
| 6072 | + tst-get-cpu-features-static \ |
| 6073 | # tests-static |
| 6074 | ifeq (yes,$(have-ifunc)) |
| 6075 | ifeq (yes,$(have-gcc-ifunc)) |
| 6076 | @@ -86,6 +86,11 @@ endif |
| 6077 | tst-ifunc-isa-2-ENV = GLIBC_TUNABLES=glibc.cpu.hwcaps=-SSE4_2,-AVX,-AVX2,-AVX512F |
| 6078 | tst-ifunc-isa-2-static-ENV = $(tst-ifunc-isa-2-ENV) |
| 6079 | tst-hwcap-tunables-ARGS = -- $(host-test-program-cmd) |
| 6080 | + |
| 6081 | +CFLAGS-tst-gnu2-tls2.c += -msse |
| 6082 | +CFLAGS-tst-gnu2-tls2mod0.c += -msse2 -mtune=haswell |
| 6083 | +CFLAGS-tst-gnu2-tls2mod1.c += -msse2 -mtune=haswell |
| 6084 | +CFLAGS-tst-gnu2-tls2mod2.c += -msse2 -mtune=haswell |
| 6085 | endif |
| 6086 | |
| 6087 | ifeq ($(subdir),math) |
| 6088 | diff --git a/sysdeps/x86/bits/wordsize.h b/sysdeps/x86/bits/wordsize.h |
| 6089 | index 70f652bca1..3f40aa76f9 100644 |
| 6090 | --- a/sysdeps/x86/bits/wordsize.h |
| 6091 | +++ b/sysdeps/x86/bits/wordsize.h |
| 6092 | @@ -8,10 +8,9 @@ |
| 6093 | #define __WORDSIZE32_PTRDIFF_LONG 0 |
| 6094 | #endif |
| 6095 | |
| 6096 | +#define __WORDSIZE_TIME64_COMPAT32 1 |
| 6097 | + |
| 6098 | #ifdef __x86_64__ |
| 6099 | -# define __WORDSIZE_TIME64_COMPAT32 1 |
| 6100 | /* Both x86-64 and x32 use the 64-bit system call interface. */ |
| 6101 | # define __SYSCALL_WORDSIZE 64 |
| 6102 | -#else |
| 6103 | -# define __WORDSIZE_TIME64_COMPAT32 0 |
| 6104 | #endif |
| 6105 | diff --git a/sysdeps/x86/configure b/sysdeps/x86/configure |
| 6106 | index 1f4c2d67fd..d28d9bcb29 100644 |
| 6107 | --- a/sysdeps/x86/configure |
| 6108 | +++ b/sysdeps/x86/configure |
| 6109 | @@ -98,6 +98,7 @@ printf "%s\n" "$libc_cv_have_x86_lahf_sahf" >&6; } |
| 6110 | if test $libc_cv_have_x86_lahf_sahf = yes; then |
| 6111 | printf "%s\n" "#define HAVE_X86_LAHF_SAHF 1" >>confdefs.h |
| 6112 | |
| 6113 | + ISAFLAG="-DHAVE_X86_LAHF_SAHF" |
| 6114 | fi |
| 6115 | { printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking for MOVBE instruction support" >&5 |
| 6116 | printf %s "checking for MOVBE instruction support... " >&6; } |
| 6117 | @@ -120,8 +121,47 @@ printf "%s\n" "$libc_cv_have_x86_movbe" >&6; } |
| 6118 | if test $libc_cv_have_x86_movbe = yes; then |
| 6119 | printf "%s\n" "#define HAVE_X86_MOVBE 1" >>confdefs.h |
| 6120 | |
| 6121 | + ISAFLAG="$ISAFLAG -DHAVE_X86_MOVBE" |
| 6122 | fi |
| 6123 | + |
| 6124 | + # Check for ISA level support. |
| 6125 | + { printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking for ISA level support" >&5 |
| 6126 | +printf %s "checking for ISA level support... " >&6; } |
| 6127 | +if test ${libc_cv_have_x86_isa_level+y} |
| 6128 | +then : |
| 6129 | + printf %s "(cached) " >&6 |
| 6130 | +else $as_nop |
| 6131 | + cat > conftest.c <<EOF |
| 6132 | +#include <sysdeps/x86/isa-level.h> |
| 6133 | +#if MINIMUM_X86_ISA_LEVEL >= 4 |
| 6134 | +libc_cv_have_x86_isa_level=4 |
| 6135 | +#elif MINIMUM_X86_ISA_LEVEL == 3 |
| 6136 | +libc_cv_have_x86_isa_level=3 |
| 6137 | +#elif MINIMUM_X86_ISA_LEVEL == 2 |
| 6138 | +libc_cv_have_x86_isa_level=2 |
| 6139 | +#else |
| 6140 | +libc_cv_have_x86_isa_level=baseline |
| 6141 | +#endif |
| 6142 | +EOF |
| 6143 | + eval `${CC-cc} $CFLAGS $CPPFLAGS $ISAFLAG -I$srcdir -E conftest.c | grep libc_cv_have_x86_isa_level` |
| 6144 | + rm -rf conftest* |
| 6145 | fi |
| 6146 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: result: $libc_cv_have_x86_isa_level" >&5 |
| 6147 | +printf "%s\n" "$libc_cv_have_x86_isa_level" >&6; } |
| 6148 | +else |
| 6149 | + libc_cv_have_x86_isa_level=baseline |
| 6150 | +fi |
| 6151 | +if test $libc_cv_have_x86_isa_level = baseline; then |
| 6152 | + printf "%s\n" "#define MINIMUM_X86_ISA_LEVEL 1" >>confdefs.h |
| 6153 | + |
| 6154 | +else |
| 6155 | + printf "%s\n" "#define MINIMUM_X86_ISA_LEVEL $libc_cv_have_x86_isa_level" >>confdefs.h |
| 6156 | + |
| 6157 | +fi |
| 6158 | +config_vars="$config_vars |
| 6159 | +have-x86-isa-level = $libc_cv_have_x86_isa_level" |
| 6160 | +config_vars="$config_vars |
| 6161 | +x86-isa-level-3-or-above = 3 4" |
| 6162 | config_vars="$config_vars |
| 6163 | enable-x86-isa-level = $libc_cv_include_x86_isa_level" |
| 6164 | |
| 6165 | diff --git a/sysdeps/x86/configure.ac b/sysdeps/x86/configure.ac |
| 6166 | index 437a50623b..5b0acd03d2 100644 |
| 6167 | --- a/sysdeps/x86/configure.ac |
| 6168 | +++ b/sysdeps/x86/configure.ac |
| 6169 | @@ -72,6 +72,7 @@ if test $libc_cv_include_x86_isa_level = yes; then |
| 6170 | fi]) |
| 6171 | if test $libc_cv_have_x86_lahf_sahf = yes; then |
| 6172 | AC_DEFINE(HAVE_X86_LAHF_SAHF) |
| 6173 | + ISAFLAG="-DHAVE_X86_LAHF_SAHF" |
| 6174 | fi |
| 6175 | AC_CACHE_CHECK([for MOVBE instruction support], |
| 6176 | libc_cv_have_x86_movbe, [dnl |
| 6177 | @@ -81,8 +82,36 @@ if test $libc_cv_include_x86_isa_level = yes; then |
| 6178 | fi]) |
| 6179 | if test $libc_cv_have_x86_movbe = yes; then |
| 6180 | AC_DEFINE(HAVE_X86_MOVBE) |
| 6181 | + ISAFLAG="$ISAFLAG -DHAVE_X86_MOVBE" |
| 6182 | fi |
| 6183 | + |
| 6184 | + # Check for ISA level support. |
| 6185 | + AC_CACHE_CHECK([for ISA level support], |
| 6186 | + libc_cv_have_x86_isa_level, [dnl |
| 6187 | +cat > conftest.c <<EOF |
| 6188 | +#include <sysdeps/x86/isa-level.h> |
| 6189 | +#if MINIMUM_X86_ISA_LEVEL >= 4 |
| 6190 | +libc_cv_have_x86_isa_level=4 |
| 6191 | +#elif MINIMUM_X86_ISA_LEVEL == 3 |
| 6192 | +libc_cv_have_x86_isa_level=3 |
| 6193 | +#elif MINIMUM_X86_ISA_LEVEL == 2 |
| 6194 | +libc_cv_have_x86_isa_level=2 |
| 6195 | +#else |
| 6196 | +libc_cv_have_x86_isa_level=baseline |
| 6197 | +#endif |
| 6198 | +EOF |
| 6199 | + eval `${CC-cc} $CFLAGS $CPPFLAGS $ISAFLAG -I$srcdir -E conftest.c | grep libc_cv_have_x86_isa_level` |
| 6200 | + rm -rf conftest*]) |
| 6201 | +else |
| 6202 | + libc_cv_have_x86_isa_level=baseline |
| 6203 | +fi |
| 6204 | +if test $libc_cv_have_x86_isa_level = baseline; then |
| 6205 | + AC_DEFINE_UNQUOTED(MINIMUM_X86_ISA_LEVEL, 1) |
| 6206 | +else |
| 6207 | + AC_DEFINE_UNQUOTED(MINIMUM_X86_ISA_LEVEL, $libc_cv_have_x86_isa_level) |
| 6208 | fi |
| 6209 | +LIBC_CONFIG_VAR([have-x86-isa-level], [$libc_cv_have_x86_isa_level]) |
| 6210 | +LIBC_CONFIG_VAR([x86-isa-level-3-or-above], [3 4]) |
| 6211 | LIBC_CONFIG_VAR([enable-x86-isa-level], [$libc_cv_include_x86_isa_level]) |
| 6212 | |
| 6213 | dnl Static PIE is supported. |
| 6214 | diff --git a/sysdeps/x86/cpu-features-offsets.sym b/sysdeps/x86/cpu-features-offsets.sym |
| 6215 | index 6a8fd29813..21fc88d651 100644 |
| 6216 | --- a/sysdeps/x86/cpu-features-offsets.sym |
| 6217 | +++ b/sysdeps/x86/cpu-features-offsets.sym |
| 6218 | @@ -3,3 +3,4 @@ |
| 6219 | #include <ldsodefs.h> |
| 6220 | |
| 6221 | XSAVE_STATE_SIZE_OFFSET offsetof (struct cpu_features, xsave_state_size) |
| 6222 | +XSAVE_STATE_FULL_SIZE_OFFSET offsetof (struct cpu_features, xsave_state_full_size) |
| 6223 | diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c |
| 6224 | index 25e6622a79..3d7c2819d7 100644 |
| 6225 | --- a/sysdeps/x86/cpu-features.c |
| 6226 | +++ b/sysdeps/x86/cpu-features.c |
| 6227 | @@ -18,6 +18,7 @@ |
| 6228 | |
| 6229 | #include <dl-hwcap.h> |
| 6230 | #include <libc-pointer-arith.h> |
| 6231 | +#include <isa-level.h> |
| 6232 | #include <get-isa-level.h> |
| 6233 | #include <cacheinfo.h> |
| 6234 | #include <dl-cacheinfo.h> |
| 6235 | @@ -27,8 +28,13 @@ |
| 6236 | extern void TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *) |
| 6237 | attribute_hidden; |
| 6238 | |
| 6239 | -#if defined SHARED && defined __x86_64__ |
| 6240 | -# include <dl-plt-rewrite.h> |
| 6241 | +#if defined SHARED |
| 6242 | +extern void _dl_tlsdesc_dynamic_fxsave (void) attribute_hidden; |
| 6243 | +extern void _dl_tlsdesc_dynamic_xsave (void) attribute_hidden; |
| 6244 | +extern void _dl_tlsdesc_dynamic_xsavec (void) attribute_hidden; |
| 6245 | + |
| 6246 | +# ifdef __x86_64__ |
| 6247 | +# include <dl-plt-rewrite.h> |
| 6248 | |
| 6249 | static void |
| 6250 | TUNABLE_CALLBACK (set_plt_rewrite) (tunable_val_t *valp) |
| 6251 | @@ -47,6 +53,15 @@ TUNABLE_CALLBACK (set_plt_rewrite) (tunable_val_t *valp) |
| 6252 | : plt_rewrite_jmp); |
| 6253 | } |
| 6254 | } |
| 6255 | +# else |
| 6256 | +extern void _dl_tlsdesc_dynamic_fnsave (void) attribute_hidden; |
| 6257 | +# endif |
| 6258 | +#endif |
| 6259 | + |
| 6260 | +#ifdef __x86_64__ |
| 6261 | +extern void _dl_runtime_resolve_fxsave (void) attribute_hidden; |
| 6262 | +extern void _dl_runtime_resolve_xsave (void) attribute_hidden; |
| 6263 | +extern void _dl_runtime_resolve_xsavec (void) attribute_hidden; |
| 6264 | #endif |
| 6265 | |
| 6266 | #ifdef __LP64__ |
| 6267 | @@ -293,8 +308,10 @@ update_active (struct cpu_features *cpu_features) |
| 6268 | __cpuid_count (0xd, 0, eax, ebx, ecx, edx); |
| 6269 | if (ebx != 0) |
| 6270 | { |
| 6271 | + /* NB: On AMX capable processors, ebx always includes AMX |
| 6272 | + states. */ |
| 6273 | unsigned int xsave_state_full_size |
| 6274 | - = ALIGN_UP (ebx + STATE_SAVE_OFFSET, 64); |
| 6275 | + = ALIGN_UP (ebx + TLSDESC_CALL_REGISTER_SAVE_AREA, 64); |
| 6276 | |
| 6277 | cpu_features->xsave_state_size |
| 6278 | = xsave_state_full_size; |
| 6279 | @@ -306,6 +323,11 @@ update_active (struct cpu_features *cpu_features) |
| 6280 | { |
| 6281 | unsigned int xstate_comp_offsets[32]; |
| 6282 | unsigned int xstate_comp_sizes[32]; |
| 6283 | +#ifdef __x86_64__ |
| 6284 | + unsigned int xstate_amx_comp_offsets[32]; |
| 6285 | + unsigned int xstate_amx_comp_sizes[32]; |
| 6286 | + unsigned int amx_ecx; |
| 6287 | +#endif |
| 6288 | unsigned int i; |
| 6289 | |
| 6290 | xstate_comp_offsets[0] = 0; |
| 6291 | @@ -313,16 +335,39 @@ update_active (struct cpu_features *cpu_features) |
| 6292 | xstate_comp_offsets[2] = 576; |
| 6293 | xstate_comp_sizes[0] = 160; |
| 6294 | xstate_comp_sizes[1] = 256; |
| 6295 | +#ifdef __x86_64__ |
| 6296 | + xstate_amx_comp_offsets[0] = 0; |
| 6297 | + xstate_amx_comp_offsets[1] = 160; |
| 6298 | + xstate_amx_comp_offsets[2] = 576; |
| 6299 | + xstate_amx_comp_sizes[0] = 160; |
| 6300 | + xstate_amx_comp_sizes[1] = 256; |
| 6301 | +#endif |
| 6302 | |
| 6303 | for (i = 2; i < 32; i++) |
| 6304 | { |
| 6305 | - if ((STATE_SAVE_MASK & (1 << i)) != 0) |
| 6306 | + if ((FULL_STATE_SAVE_MASK & (1 << i)) != 0) |
| 6307 | { |
| 6308 | __cpuid_count (0xd, i, eax, ebx, ecx, edx); |
| 6309 | - xstate_comp_sizes[i] = eax; |
| 6310 | +#ifdef __x86_64__ |
| 6311 | + /* Include this in xsave_state_full_size. */ |
| 6312 | + amx_ecx = ecx; |
| 6313 | + xstate_amx_comp_sizes[i] = eax; |
| 6314 | + if ((AMX_STATE_SAVE_MASK & (1 << i)) != 0) |
| 6315 | + { |
| 6316 | + /* Exclude this from xsave_state_size. */ |
| 6317 | + ecx = 0; |
| 6318 | + xstate_comp_sizes[i] = 0; |
| 6319 | + } |
| 6320 | + else |
| 6321 | +#endif |
| 6322 | + xstate_comp_sizes[i] = eax; |
| 6323 | } |
| 6324 | else |
| 6325 | { |
| 6326 | +#ifdef __x86_64__ |
| 6327 | + amx_ecx = 0; |
| 6328 | + xstate_amx_comp_sizes[i] = 0; |
| 6329 | +#endif |
| 6330 | ecx = 0; |
| 6331 | xstate_comp_sizes[i] = 0; |
| 6332 | } |
| 6333 | @@ -335,6 +380,15 @@ update_active (struct cpu_features *cpu_features) |
| 6334 | if ((ecx & (1 << 1)) != 0) |
| 6335 | xstate_comp_offsets[i] |
| 6336 | = ALIGN_UP (xstate_comp_offsets[i], 64); |
| 6337 | +#ifdef __x86_64__ |
| 6338 | + xstate_amx_comp_offsets[i] |
| 6339 | + = (xstate_amx_comp_offsets[i - 1] |
| 6340 | + + xstate_amx_comp_sizes[i - 1]); |
| 6341 | + if ((amx_ecx & (1 << 1)) != 0) |
| 6342 | + xstate_amx_comp_offsets[i] |
| 6343 | + = ALIGN_UP (xstate_amx_comp_offsets[i], |
| 6344 | + 64); |
| 6345 | +#endif |
| 6346 | } |
| 6347 | } |
| 6348 | |
| 6349 | @@ -343,8 +397,23 @@ update_active (struct cpu_features *cpu_features) |
| 6350 | = xstate_comp_offsets[31] + xstate_comp_sizes[31]; |
| 6351 | if (size) |
| 6352 | { |
| 6353 | +#ifdef __x86_64__ |
| 6354 | + unsigned int amx_size |
| 6355 | + = (xstate_amx_comp_offsets[31] |
| 6356 | + + xstate_amx_comp_sizes[31]); |
| 6357 | + amx_size |
| 6358 | + = ALIGN_UP ((amx_size |
| 6359 | + + TLSDESC_CALL_REGISTER_SAVE_AREA), |
| 6360 | + 64); |
| 6361 | + /* Set xsave_state_full_size to the compact AMX |
| 6362 | + state size for XSAVEC. NB: xsave_state_full_size |
| 6363 | + is only used in _dl_tlsdesc_dynamic_xsave and |
| 6364 | + _dl_tlsdesc_dynamic_xsavec. */ |
| 6365 | + cpu_features->xsave_state_full_size = amx_size; |
| 6366 | +#endif |
| 6367 | cpu_features->xsave_state_size |
| 6368 | - = ALIGN_UP (size + STATE_SAVE_OFFSET, 64); |
| 6369 | + = ALIGN_UP (size + TLSDESC_CALL_REGISTER_SAVE_AREA, |
| 6370 | + 64); |
| 6371 | CPU_FEATURE_SET (cpu_features, XSAVEC); |
| 6372 | } |
| 6373 | } |
| 6374 | @@ -1130,6 +1199,45 @@ no_cpuid: |
| 6375 | TUNABLE_CALLBACK (set_x86_shstk)); |
| 6376 | #endif |
| 6377 | |
| 6378 | + if (MINIMUM_X86_ISA_LEVEL >= AVX_X86_ISA_LEVEL |
| 6379 | + || (GLRO(dl_x86_cpu_features).xsave_state_size != 0)) |
| 6380 | + { |
| 6381 | + if (CPU_FEATURE_USABLE_P (cpu_features, XSAVEC)) |
| 6382 | + { |
| 6383 | +#ifdef __x86_64__ |
| 6384 | + GLRO(dl_x86_64_runtime_resolve) = _dl_runtime_resolve_xsavec; |
| 6385 | +#endif |
| 6386 | +#ifdef SHARED |
| 6387 | + GLRO(dl_x86_tlsdesc_dynamic) = _dl_tlsdesc_dynamic_xsavec; |
| 6388 | +#endif |
| 6389 | + } |
| 6390 | + else |
| 6391 | + { |
| 6392 | +#ifdef __x86_64__ |
| 6393 | + GLRO(dl_x86_64_runtime_resolve) = _dl_runtime_resolve_xsave; |
| 6394 | +#endif |
| 6395 | +#ifdef SHARED |
| 6396 | + GLRO(dl_x86_tlsdesc_dynamic) = _dl_tlsdesc_dynamic_xsave; |
| 6397 | +#endif |
| 6398 | + } |
| 6399 | + } |
| 6400 | + else |
| 6401 | + { |
| 6402 | +#ifdef __x86_64__ |
| 6403 | + GLRO(dl_x86_64_runtime_resolve) = _dl_runtime_resolve_fxsave; |
| 6404 | +# ifdef SHARED |
| 6405 | + GLRO(dl_x86_tlsdesc_dynamic) = _dl_tlsdesc_dynamic_fxsave; |
| 6406 | +# endif |
| 6407 | +#else |
| 6408 | +# ifdef SHARED |
| 6409 | + if (CPU_FEATURE_USABLE_P (cpu_features, FXSR)) |
| 6410 | + GLRO(dl_x86_tlsdesc_dynamic) = _dl_tlsdesc_dynamic_fxsave; |
| 6411 | + else |
| 6412 | + GLRO(dl_x86_tlsdesc_dynamic) = _dl_tlsdesc_dynamic_fnsave; |
| 6413 | +# endif |
| 6414 | +#endif |
| 6415 | + } |
| 6416 | + |
| 6417 | #ifdef SHARED |
| 6418 | # ifdef __x86_64__ |
| 6419 | TUNABLE_GET (plt_rewrite, tunable_val_t *, |
| 6420 | diff --git a/sysdeps/x86/dl-cacheinfo.h b/sysdeps/x86/dl-cacheinfo.h |
| 6421 | index d5101615e3..5a98f70364 100644 |
| 6422 | --- a/sysdeps/x86/dl-cacheinfo.h |
| 6423 | +++ b/sysdeps/x86/dl-cacheinfo.h |
| 6424 | @@ -791,7 +791,6 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6425 | long int data = -1; |
| 6426 | long int shared = -1; |
| 6427 | long int shared_per_thread = -1; |
| 6428 | - long int core = -1; |
| 6429 | unsigned int threads = 0; |
| 6430 | unsigned long int level1_icache_size = -1; |
| 6431 | unsigned long int level1_icache_linesize = -1; |
| 6432 | @@ -809,7 +808,6 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6433 | if (cpu_features->basic.kind == arch_kind_intel) |
| 6434 | { |
| 6435 | data = handle_intel (_SC_LEVEL1_DCACHE_SIZE, cpu_features); |
| 6436 | - core = handle_intel (_SC_LEVEL2_CACHE_SIZE, cpu_features); |
| 6437 | shared = handle_intel (_SC_LEVEL3_CACHE_SIZE, cpu_features); |
| 6438 | shared_per_thread = shared; |
| 6439 | |
| 6440 | @@ -822,7 +820,8 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6441 | = handle_intel (_SC_LEVEL1_DCACHE_ASSOC, cpu_features); |
| 6442 | level1_dcache_linesize |
| 6443 | = handle_intel (_SC_LEVEL1_DCACHE_LINESIZE, cpu_features); |
| 6444 | - level2_cache_size = core; |
| 6445 | + level2_cache_size |
| 6446 | + = handle_intel (_SC_LEVEL2_CACHE_SIZE, cpu_features); |
| 6447 | level2_cache_assoc |
| 6448 | = handle_intel (_SC_LEVEL2_CACHE_ASSOC, cpu_features); |
| 6449 | level2_cache_linesize |
| 6450 | @@ -835,12 +834,12 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6451 | level4_cache_size |
| 6452 | = handle_intel (_SC_LEVEL4_CACHE_SIZE, cpu_features); |
| 6453 | |
| 6454 | - get_common_cache_info (&shared, &shared_per_thread, &threads, core); |
| 6455 | + get_common_cache_info (&shared, &shared_per_thread, &threads, |
| 6456 | + level2_cache_size); |
| 6457 | } |
| 6458 | else if (cpu_features->basic.kind == arch_kind_zhaoxin) |
| 6459 | { |
| 6460 | data = handle_zhaoxin (_SC_LEVEL1_DCACHE_SIZE); |
| 6461 | - core = handle_zhaoxin (_SC_LEVEL2_CACHE_SIZE); |
| 6462 | shared = handle_zhaoxin (_SC_LEVEL3_CACHE_SIZE); |
| 6463 | shared_per_thread = shared; |
| 6464 | |
| 6465 | @@ -849,19 +848,19 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6466 | level1_dcache_size = data; |
| 6467 | level1_dcache_assoc = handle_zhaoxin (_SC_LEVEL1_DCACHE_ASSOC); |
| 6468 | level1_dcache_linesize = handle_zhaoxin (_SC_LEVEL1_DCACHE_LINESIZE); |
| 6469 | - level2_cache_size = core; |
| 6470 | + level2_cache_size = handle_zhaoxin (_SC_LEVEL2_CACHE_SIZE); |
| 6471 | level2_cache_assoc = handle_zhaoxin (_SC_LEVEL2_CACHE_ASSOC); |
| 6472 | level2_cache_linesize = handle_zhaoxin (_SC_LEVEL2_CACHE_LINESIZE); |
| 6473 | level3_cache_size = shared; |
| 6474 | level3_cache_assoc = handle_zhaoxin (_SC_LEVEL3_CACHE_ASSOC); |
| 6475 | level3_cache_linesize = handle_zhaoxin (_SC_LEVEL3_CACHE_LINESIZE); |
| 6476 | |
| 6477 | - get_common_cache_info (&shared, &shared_per_thread, &threads, core); |
| 6478 | + get_common_cache_info (&shared, &shared_per_thread, &threads, |
| 6479 | + level2_cache_size); |
| 6480 | } |
| 6481 | else if (cpu_features->basic.kind == arch_kind_amd) |
| 6482 | { |
| 6483 | data = handle_amd (_SC_LEVEL1_DCACHE_SIZE); |
| 6484 | - core = handle_amd (_SC_LEVEL2_CACHE_SIZE); |
| 6485 | shared = handle_amd (_SC_LEVEL3_CACHE_SIZE); |
| 6486 | |
| 6487 | level1_icache_size = handle_amd (_SC_LEVEL1_ICACHE_SIZE); |
| 6488 | @@ -869,7 +868,7 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6489 | level1_dcache_size = data; |
| 6490 | level1_dcache_assoc = handle_amd (_SC_LEVEL1_DCACHE_ASSOC); |
| 6491 | level1_dcache_linesize = handle_amd (_SC_LEVEL1_DCACHE_LINESIZE); |
| 6492 | - level2_cache_size = core; |
| 6493 | + level2_cache_size = handle_amd (_SC_LEVEL2_CACHE_SIZE);; |
| 6494 | level2_cache_assoc = handle_amd (_SC_LEVEL2_CACHE_ASSOC); |
| 6495 | level2_cache_linesize = handle_amd (_SC_LEVEL2_CACHE_LINESIZE); |
| 6496 | level3_cache_size = shared; |
| 6497 | @@ -880,12 +879,12 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6498 | if (shared <= 0) |
| 6499 | { |
| 6500 | /* No shared L3 cache. All we have is the L2 cache. */ |
| 6501 | - shared = core; |
| 6502 | + shared = level2_cache_size; |
| 6503 | } |
| 6504 | else if (cpu_features->basic.family < 0x17) |
| 6505 | { |
| 6506 | /* Account for exclusive L2 and L3 caches. */ |
| 6507 | - shared += core; |
| 6508 | + shared += level2_cache_size; |
| 6509 | } |
| 6510 | |
| 6511 | shared_per_thread = shared; |
| 6512 | @@ -987,6 +986,12 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6513 | if (CPU_FEATURE_USABLE_P (cpu_features, FSRM)) |
| 6514 | rep_movsb_threshold = 2112; |
| 6515 | |
| 6516 | + /* For AMD CPUs that support ERMS (Zen3+), REP MOVSB is in a lot of |
| 6517 | + cases slower than the vectorized path (and for some alignments, |
| 6518 | + it is really slow, check BZ #30994). */ |
| 6519 | + if (cpu_features->basic.kind == arch_kind_amd) |
| 6520 | + rep_movsb_threshold = non_temporal_threshold; |
| 6521 | + |
| 6522 | /* The default threshold to use Enhanced REP STOSB. */ |
| 6523 | unsigned long int rep_stosb_threshold = 2048; |
| 6524 | |
| 6525 | @@ -1016,6 +1021,11 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6526 | minimum value is fixed. */ |
| 6527 | rep_stosb_threshold = TUNABLE_GET (x86_rep_stosb_threshold, |
| 6528 | long int, NULL); |
| 6529 | + if (cpu_features->basic.kind == arch_kind_amd |
| 6530 | + && !TUNABLE_IS_INITIALIZED (x86_rep_stosb_threshold)) |
| 6531 | + /* For AMD Zen3+ architecture, the performance of the vectorized loop is |
| 6532 | + slightly better than ERMS. */ |
| 6533 | + rep_stosb_threshold = SIZE_MAX; |
| 6534 | |
| 6535 | TUNABLE_SET_WITH_BOUNDS (x86_data_cache_size, data, 0, SIZE_MAX); |
| 6536 | TUNABLE_SET_WITH_BOUNDS (x86_shared_cache_size, shared, 0, SIZE_MAX); |
| 6537 | @@ -1028,16 +1038,9 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) |
| 6538 | SIZE_MAX); |
| 6539 | |
| 6540 | unsigned long int rep_movsb_stop_threshold; |
| 6541 | - /* ERMS feature is implemented from AMD Zen3 architecture and it is |
| 6542 | - performing poorly for data above L2 cache size. Henceforth, adding |
| 6543 | - an upper bound threshold parameter to limit the usage of Enhanced |
| 6544 | - REP MOVSB operations and setting its value to L2 cache size. */ |
| 6545 | - if (cpu_features->basic.kind == arch_kind_amd) |
| 6546 | - rep_movsb_stop_threshold = core; |
| 6547 | /* Setting the upper bound of ERMS to the computed value of |
| 6548 | - non-temporal threshold for architectures other than AMD. */ |
| 6549 | - else |
| 6550 | - rep_movsb_stop_threshold = non_temporal_threshold; |
| 6551 | + non-temporal threshold for all architectures. */ |
| 6552 | + rep_movsb_stop_threshold = non_temporal_threshold; |
| 6553 | |
| 6554 | cpu_features->data_cache_size = data; |
| 6555 | cpu_features->shared_cache_size = shared; |
| 6556 | diff --git a/sysdeps/x86/dl-procinfo.c b/sysdeps/x86/dl-procinfo.c |
| 6557 | index ee957b4d70..5920d4b320 100644 |
| 6558 | --- a/sysdeps/x86/dl-procinfo.c |
| 6559 | +++ b/sysdeps/x86/dl-procinfo.c |
| 6560 | @@ -86,3 +86,19 @@ PROCINFO_CLASS const char _dl_x86_platforms[4][9] |
| 6561 | #else |
| 6562 | , |
| 6563 | #endif |
| 6564 | + |
| 6565 | +#if defined SHARED && !IS_IN (ldconfig) |
| 6566 | +# if !defined PROCINFO_DECL |
| 6567 | + ._dl_x86_tlsdesc_dynamic |
| 6568 | +# else |
| 6569 | +PROCINFO_CLASS void * _dl_x86_tlsdesc_dynamic |
| 6570 | +# endif |
| 6571 | +# ifndef PROCINFO_DECL |
| 6572 | += NULL |
| 6573 | +# endif |
| 6574 | +# ifdef PROCINFO_DECL |
| 6575 | +; |
| 6576 | +# else |
| 6577 | +, |
| 6578 | +# endif |
| 6579 | +#endif |
| 6580 | diff --git a/sysdeps/x86_64/features-offsets.sym b/sysdeps/x86/features-offsets.sym |
| 6581 | similarity index 89% |
| 6582 | rename from sysdeps/x86_64/features-offsets.sym |
| 6583 | rename to sysdeps/x86/features-offsets.sym |
| 6584 | index 9e4be3393a..77e990c705 100644 |
| 6585 | --- a/sysdeps/x86_64/features-offsets.sym |
| 6586 | +++ b/sysdeps/x86/features-offsets.sym |
| 6587 | @@ -3,4 +3,6 @@ |
| 6588 | #include <ldsodefs.h> |
| 6589 | |
| 6590 | RTLD_GLOBAL_RO_DL_X86_CPU_FEATURES_OFFSET offsetof (struct rtld_global_ro, _dl_x86_cpu_features) |
| 6591 | +#ifdef __x86_64__ |
| 6592 | RTLD_GLOBAL_DL_X86_FEATURE_1_OFFSET offsetof (struct rtld_global, _dl_x86_feature_1) |
| 6593 | +#endif |
| 6594 | diff --git a/sysdeps/x86/include/cpu-features.h b/sysdeps/x86/include/cpu-features.h |
| 6595 | index b9bf3115b6..cd7bd27cf3 100644 |
| 6596 | --- a/sysdeps/x86/include/cpu-features.h |
| 6597 | +++ b/sysdeps/x86/include/cpu-features.h |
| 6598 | @@ -934,6 +934,8 @@ struct cpu_features |
| 6599 | /* The full state size for XSAVE when XSAVEC is disabled by |
| 6600 | |
| 6601 | GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVEC |
| 6602 | + |
| 6603 | + and the AMX state size when XSAVEC is available. |
| 6604 | */ |
| 6605 | unsigned int xsave_state_full_size; |
| 6606 | /* Data cache size for use in memory and string routines, typically |
| 6607 | diff --git a/sysdeps/x86/isa-level.h b/sysdeps/x86/isa-level.h |
| 6608 | index 11fe1ca90c..2c7f74212b 100644 |
| 6609 | --- a/sysdeps/x86/isa-level.h |
| 6610 | +++ b/sysdeps/x86/isa-level.h |
| 6611 | @@ -61,8 +61,10 @@ |
| 6612 | # define __X86_ISA_V4 0 |
| 6613 | #endif |
| 6614 | |
| 6615 | -#define MINIMUM_X86_ISA_LEVEL \ |
| 6616 | +#ifndef MINIMUM_X86_ISA_LEVEL |
| 6617 | +# define MINIMUM_X86_ISA_LEVEL \ |
| 6618 | (__X86_ISA_V1 + __X86_ISA_V2 + __X86_ISA_V3 + __X86_ISA_V4) |
| 6619 | +#endif |
| 6620 | |
| 6621 | /* Depending on the minimum ISA level, a feature check result can be a |
| 6622 | compile-time constant.. */ |
| 6623 | diff --git a/sysdeps/x86/sysdep.h b/sysdeps/x86/sysdep.h |
| 6624 | index 85d0a8c943..7359149e17 100644 |
| 6625 | --- a/sysdeps/x86/sysdep.h |
| 6626 | +++ b/sysdeps/x86/sysdep.h |
| 6627 | @@ -21,14 +21,118 @@ |
| 6628 | |
| 6629 | #include <sysdeps/generic/sysdep.h> |
| 6630 | |
| 6631 | +/* The extended state feature IDs in the state component bitmap. */ |
| 6632 | +#define X86_XSTATE_X87_ID 0 |
| 6633 | +#define X86_XSTATE_SSE_ID 1 |
| 6634 | +#define X86_XSTATE_AVX_ID 2 |
| 6635 | +#define X86_XSTATE_BNDREGS_ID 3 |
| 6636 | +#define X86_XSTATE_BNDCFG_ID 4 |
| 6637 | +#define X86_XSTATE_K_ID 5 |
| 6638 | +#define X86_XSTATE_ZMM_H_ID 6 |
| 6639 | +#define X86_XSTATE_ZMM_ID 7 |
| 6640 | +#define X86_XSTATE_PKRU_ID 9 |
| 6641 | +#define X86_XSTATE_TILECFG_ID 17 |
| 6642 | +#define X86_XSTATE_TILEDATA_ID 18 |
| 6643 | +#define X86_XSTATE_APX_F_ID 19 |
| 6644 | + |
| 6645 | +#ifdef __x86_64__ |
| 6646 | /* Offset for fxsave/xsave area used by _dl_runtime_resolve. Also need |
| 6647 | space to preserve RCX, RDX, RSI, RDI, R8, R9 and RAX. It must be |
| 6648 | - aligned to 16 bytes for fxsave and 64 bytes for xsave. */ |
| 6649 | -#define STATE_SAVE_OFFSET (8 * 7 + 8) |
| 6650 | + aligned to 16 bytes for fxsave and 64 bytes for xsave. It is non-zero |
| 6651 | + because MOV, instead of PUSH, is used to save registers onto stack. |
| 6652 | + |
| 6653 | + +==================+<- stack frame start aligned at 8 or 16 bytes |
| 6654 | + | |<- paddings for stack realignment of 64 bytes |
| 6655 | + |------------------|<- xsave buffer end aligned at 64 bytes |
| 6656 | + | |<- |
| 6657 | + | |<- |
| 6658 | + | |<- |
| 6659 | + |------------------|<- xsave buffer start at STATE_SAVE_OFFSET(%rsp) |
| 6660 | + | |<- 8-byte padding for 64-byte alignment |
| 6661 | + | |<- R9 |
| 6662 | + | |<- R8 |
| 6663 | + | |<- RDI |
| 6664 | + | |<- RSI |
| 6665 | + | |<- RDX |
| 6666 | + | |<- RCX |
| 6667 | + | |<- RAX |
| 6668 | + +==================+<- RSP aligned at 64 bytes |
| 6669 | + |
| 6670 | + */ |
| 6671 | +# define STATE_SAVE_OFFSET (8 * 7 + 8) |
| 6672 | + |
| 6673 | +/* _dl_tlsdesc_dynamic preserves RDI, RSI and RBX before realigning |
| 6674 | + stack. After realigning stack, it saves RCX, RDX, R8, R9, R10 and |
| 6675 | + R11. Allocate space for RDI, RSI and RBX to avoid clobbering saved |
| 6676 | + RDI, RSI and RBX values on stack by xsave. |
| 6677 | + |
| 6678 | + +==================+<- stack frame start aligned at 8 or 16 bytes |
| 6679 | + | |<- RDI saved in the red zone |
| 6680 | + | |<- RSI saved in the red zone |
| 6681 | + | |<- RBX saved in the red zone |
| 6682 | + | |<- paddings for stack realignment of 64 bytes |
| 6683 | + |------------------|<- xsave buffer end aligned at 64 bytes |
| 6684 | + | |<- |
| 6685 | + | |<- |
| 6686 | + | |<- |
| 6687 | + |------------------|<- xsave buffer start at STATE_SAVE_OFFSET(%rsp) |
| 6688 | + | |<- 8-byte padding for 64-byte alignment |
| 6689 | + | |<- 8-byte padding for 64-byte alignment |
| 6690 | + | |<- R11 |
| 6691 | + | |<- R10 |
| 6692 | + | |<- R9 |
| 6693 | + | |<- R8 |
| 6694 | + | |<- RDX |
| 6695 | + | |<- RCX |
| 6696 | + +==================+<- RSP aligned at 64 bytes |
| 6697 | + |
| 6698 | + Define the total register save area size for all integer registers by |
| 6699 | + adding 24 to STATE_SAVE_OFFSET since RDI, RSI and RBX are saved onto |
| 6700 | + stack without adjusting stack pointer first, using the red-zone. */ |
| 6701 | +# define TLSDESC_CALL_REGISTER_SAVE_AREA (STATE_SAVE_OFFSET + 24) |
| 6702 | + |
| 6703 | +/* Save SSE, AVX, AVX512, mask, bound and APX registers. Bound and APX |
| 6704 | + registers are mutually exclusive. */ |
| 6705 | +# define STATE_SAVE_MASK \ |
| 6706 | + ((1 << X86_XSTATE_SSE_ID) \ |
| 6707 | + | (1 << X86_XSTATE_AVX_ID) \ |
| 6708 | + | (1 << X86_XSTATE_BNDREGS_ID) \ |
| 6709 | + | (1 << X86_XSTATE_K_ID) \ |
| 6710 | + | (1 << X86_XSTATE_ZMM_H_ID) \ |
| 6711 | + | (1 << X86_XSTATE_ZMM_ID) \ |
| 6712 | + | (1 << X86_XSTATE_APX_F_ID)) |
| 6713 | + |
| 6714 | +/* AMX state mask. */ |
| 6715 | +# define AMX_STATE_SAVE_MASK \ |
| 6716 | + ((1 << X86_XSTATE_TILECFG_ID) | (1 << X86_XSTATE_TILEDATA_ID)) |
| 6717 | + |
| 6718 | +/* States to be included in xsave_state_full_size. */ |
| 6719 | +# define FULL_STATE_SAVE_MASK \ |
| 6720 | + (STATE_SAVE_MASK | AMX_STATE_SAVE_MASK) |
| 6721 | +#else |
| 6722 | +/* Offset for fxsave/xsave area used by _dl_tlsdesc_dynamic. Since i386 |
| 6723 | + uses PUSH to save registers onto stack, use 0 here. */ |
| 6724 | +# define STATE_SAVE_OFFSET 0 |
| 6725 | +# define TLSDESC_CALL_REGISTER_SAVE_AREA 0 |
| 6726 | + |
| 6727 | +/* Save SSE, AVX, AXV512, mask and bound registers. */ |
| 6728 | +# define STATE_SAVE_MASK \ |
| 6729 | + ((1 << X86_XSTATE_SSE_ID) \ |
| 6730 | + | (1 << X86_XSTATE_AVX_ID) \ |
| 6731 | + | (1 << X86_XSTATE_BNDREGS_ID) \ |
| 6732 | + | (1 << X86_XSTATE_K_ID) \ |
| 6733 | + | (1 << X86_XSTATE_ZMM_H_ID)) |
| 6734 | + |
| 6735 | +/* States to be included in xsave_state_size. */ |
| 6736 | +# define FULL_STATE_SAVE_MASK STATE_SAVE_MASK |
| 6737 | +#endif |
| 6738 | |
| 6739 | -/* Save SSE, AVX, AVX512, mask and bound registers. */ |
| 6740 | -#define STATE_SAVE_MASK \ |
| 6741 | - ((1 << 1) | (1 << 2) | (1 << 3) | (1 << 5) | (1 << 6) | (1 << 7)) |
| 6742 | +/* States which should be saved for TLSDESC_CALL and TLS_DESC_CALL. |
| 6743 | + Compiler assumes that all registers, including AMX and x87 FPU |
| 6744 | + stack registers, are unchanged after CALL, except for EFLAGS and |
| 6745 | + RAX/EAX. */ |
| 6746 | +#define TLSDESC_CALL_STATE_SAVE_MASK \ |
| 6747 | + (FULL_STATE_SAVE_MASK | (1 << X86_XSTATE_X87_ID)) |
| 6748 | |
| 6749 | /* Constants for bits in __x86_string_control: */ |
| 6750 | |
| 6751 | diff --git a/sysdeps/x86/tst-gnu2-tls2.c b/sysdeps/x86/tst-gnu2-tls2.c |
| 6752 | new file mode 100644 |
| 6753 | index 0000000000..de900a423b |
| 6754 | --- /dev/null |
| 6755 | +++ b/sysdeps/x86/tst-gnu2-tls2.c |
| 6756 | @@ -0,0 +1,20 @@ |
| 6757 | +#ifndef __x86_64__ |
| 6758 | +#include <sys/platform/x86.h> |
| 6759 | + |
| 6760 | +#define IS_SUPPORTED() CPU_FEATURE_ACTIVE (SSE2) |
| 6761 | +#endif |
| 6762 | + |
| 6763 | +/* Clear XMM0...XMM7 */ |
| 6764 | +#define PREPARE_MALLOC() \ |
| 6765 | +{ \ |
| 6766 | + asm volatile ("xorps %%xmm0, %%xmm0" : : : "xmm0" ); \ |
| 6767 | + asm volatile ("xorps %%xmm1, %%xmm1" : : : "xmm1" ); \ |
| 6768 | + asm volatile ("xorps %%xmm2, %%xmm2" : : : "xmm2" ); \ |
| 6769 | + asm volatile ("xorps %%xmm3, %%xmm3" : : : "xmm3" ); \ |
| 6770 | + asm volatile ("xorps %%xmm4, %%xmm4" : : : "xmm4" ); \ |
| 6771 | + asm volatile ("xorps %%xmm5, %%xmm5" : : : "xmm5" ); \ |
| 6772 | + asm volatile ("xorps %%xmm6, %%xmm6" : : : "xmm6" ); \ |
| 6773 | + asm volatile ("xorps %%xmm7, %%xmm7" : : : "xmm7" ); \ |
| 6774 | +} |
| 6775 | + |
| 6776 | +#include <elf/tst-gnu2-tls2.c> |
| 6777 | diff --git a/sysdeps/x86/utmp-size.h b/sysdeps/x86/utmp-size.h |
| 6778 | new file mode 100644 |
| 6779 | index 0000000000..8f21ebe1b6 |
| 6780 | --- /dev/null |
| 6781 | +++ b/sysdeps/x86/utmp-size.h |
| 6782 | @@ -0,0 +1,2 @@ |
| 6783 | +#define UTMP_SIZE 384 |
| 6784 | +#define LASTLOG_SIZE 292 |
| 6785 | diff --git a/sysdeps/x86_64/Makefile b/sysdeps/x86_64/Makefile |
| 6786 | index 90f4ecfd26..0ede447405 100644 |
| 6787 | --- a/sysdeps/x86_64/Makefile |
| 6788 | +++ b/sysdeps/x86_64/Makefile |
| 6789 | @@ -10,7 +10,7 @@ LDFLAGS-rtld += -Wl,-z,nomark-plt |
| 6790 | endif |
| 6791 | |
| 6792 | ifeq ($(subdir),csu) |
| 6793 | -gen-as-const-headers += features-offsets.sym link-defines.sym |
| 6794 | +gen-as-const-headers += link-defines.sym |
| 6795 | endif |
| 6796 | |
| 6797 | ifeq ($(subdir),gmon) |
| 6798 | @@ -210,6 +210,8 @@ tst-plt-rewrite2-ENV = GLIBC_TUNABLES=glibc.cpu.plt_rewrite=2 |
| 6799 | $(objpfx)tst-plt-rewrite2: $(objpfx)tst-plt-rewritemod2.so |
| 6800 | endif |
| 6801 | |
| 6802 | +test-internal-extras += tst-gnu2-tls2mod1 |
| 6803 | + |
| 6804 | endif # $(subdir) == elf |
| 6805 | |
| 6806 | ifeq ($(subdir),csu) |
| 6807 | @@ -250,6 +252,10 @@ sysdep-dl-routines += dl-cet |
| 6808 | |
| 6809 | tests += \ |
| 6810 | tst-cet-legacy-1 \ |
| 6811 | + tst-cet-legacy-10 \ |
| 6812 | + tst-cet-legacy-10-static \ |
| 6813 | + tst-cet-legacy-10a \ |
| 6814 | + tst-cet-legacy-10a-static \ |
| 6815 | tst-cet-legacy-1a \ |
| 6816 | tst-cet-legacy-2 \ |
| 6817 | tst-cet-legacy-2a \ |
| 6818 | @@ -261,15 +267,11 @@ tests += \ |
| 6819 | tst-cet-legacy-8 \ |
| 6820 | tst-cet-legacy-9 \ |
| 6821 | tst-cet-legacy-9-static \ |
| 6822 | - tst-cet-legacy-10 \ |
| 6823 | - tst-cet-legacy-10-static \ |
| 6824 | - tst-cet-legacy-10a \ |
| 6825 | - tst-cet-legacy-10a-static \ |
| 6826 | # tests |
| 6827 | tests-static += \ |
| 6828 | - tst-cet-legacy-9-static \ |
| 6829 | tst-cet-legacy-10-static \ |
| 6830 | tst-cet-legacy-10a-static \ |
| 6831 | + tst-cet-legacy-9-static \ |
| 6832 | # tests-static |
| 6833 | tst-cet-legacy-1a-ARGS = -- $(host-test-program-cmd) |
| 6834 | |
| 6835 | diff --git a/sysdeps/x86_64/configure b/sysdeps/x86_64/configure |
| 6836 | index 418cc4a9b8..04a534fa12 100755 |
| 6837 | --- a/sysdeps/x86_64/configure |
| 6838 | +++ b/sysdeps/x86_64/configure |
| 6839 | @@ -134,6 +134,34 @@ fi |
| 6840 | config_vars="$config_vars |
| 6841 | enable-cet = $enable_cet" |
| 6842 | |
| 6843 | +# Check if -mamx-tile works properly. |
| 6844 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: checking whether -mamx-tile works properly" >&5 |
| 6845 | +printf %s "checking whether -mamx-tile works properly... " >&6; } |
| 6846 | +if test ${libc_cv_x86_have_amx_tile+y} |
| 6847 | +then : |
| 6848 | + printf %s "(cached) " >&6 |
| 6849 | +else $as_nop |
| 6850 | + cat > conftest.c <<EOF |
| 6851 | +#include <x86intrin.h> |
| 6852 | +EOF |
| 6853 | + libc_cv_x86_have_amx_tile=no |
| 6854 | + if { ac_try='${CC-cc} -E $CFLAGS -mamx-tile conftest.c > conftest.i' |
| 6855 | + { { eval echo "\"\$as_me\":${as_lineno-$LINENO}: \"$ac_try\""; } >&5 |
| 6856 | + (eval $ac_try) 2>&5 |
| 6857 | + ac_status=$? |
| 6858 | + printf "%s\n" "$as_me:${as_lineno-$LINENO}: \$? = $ac_status" >&5 |
| 6859 | + test $ac_status = 0; }; }; then |
| 6860 | + if grep -q __builtin_ia32_ldtilecfg conftest.i; then |
| 6861 | + libc_cv_x86_have_amx_tile=yes |
| 6862 | + fi |
| 6863 | + fi |
| 6864 | + rm -rf conftest* |
| 6865 | +fi |
| 6866 | +{ printf "%s\n" "$as_me:${as_lineno-$LINENO}: result: $libc_cv_x86_have_amx_tile" >&5 |
| 6867 | +printf "%s\n" "$libc_cv_x86_have_amx_tile" >&6; } |
| 6868 | +config_vars="$config_vars |
| 6869 | +have-mamx-tile = $libc_cv_x86_have_amx_tile" |
| 6870 | + |
| 6871 | test -n "$critic_missing" && as_fn_error $? " |
| 6872 | *** $critic_missing" "$LINENO" 5 |
| 6873 | |
| 6874 | diff --git a/sysdeps/x86_64/configure.ac b/sysdeps/x86_64/configure.ac |
| 6875 | index d1f803c02e..c714c47351 100644 |
| 6876 | --- a/sysdeps/x86_64/configure.ac |
| 6877 | +++ b/sysdeps/x86_64/configure.ac |
| 6878 | @@ -61,5 +61,20 @@ elif test $enable_cet = permissive; then |
| 6879 | fi |
| 6880 | LIBC_CONFIG_VAR([enable-cet], [$enable_cet]) |
| 6881 | |
| 6882 | +# Check if -mamx-tile works properly. |
| 6883 | +AC_CACHE_CHECK(whether -mamx-tile works properly, |
| 6884 | + libc_cv_x86_have_amx_tile, [dnl |
| 6885 | +cat > conftest.c <<EOF |
| 6886 | +#include <x86intrin.h> |
| 6887 | +EOF |
| 6888 | + libc_cv_x86_have_amx_tile=no |
| 6889 | + if AC_TRY_COMMAND(${CC-cc} -E $CFLAGS -mamx-tile conftest.c > conftest.i); then |
| 6890 | + if grep -q __builtin_ia32_ldtilecfg conftest.i; then |
| 6891 | + libc_cv_x86_have_amx_tile=yes |
| 6892 | + fi |
| 6893 | + fi |
| 6894 | + rm -rf conftest*]) |
| 6895 | +LIBC_CONFIG_VAR([have-mamx-tile], [$libc_cv_x86_have_amx_tile]) |
| 6896 | + |
| 6897 | test -n "$critic_missing" && AC_MSG_ERROR([ |
| 6898 | *** $critic_missing]) |
| 6899 | diff --git a/sysdeps/x86_64/dl-machine.h b/sysdeps/x86_64/dl-machine.h |
| 6900 | index 6d605d0d32..ff5d45f7cb 100644 |
| 6901 | --- a/sysdeps/x86_64/dl-machine.h |
| 6902 | +++ b/sysdeps/x86_64/dl-machine.h |
| 6903 | @@ -71,9 +71,6 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], |
| 6904 | int lazy, int profile) |
| 6905 | { |
| 6906 | Elf64_Addr *got; |
| 6907 | - extern void _dl_runtime_resolve_fxsave (ElfW(Word)) attribute_hidden; |
| 6908 | - extern void _dl_runtime_resolve_xsave (ElfW(Word)) attribute_hidden; |
| 6909 | - extern void _dl_runtime_resolve_xsavec (ElfW(Word)) attribute_hidden; |
| 6910 | extern void _dl_runtime_profile_sse (ElfW(Word)) attribute_hidden; |
| 6911 | extern void _dl_runtime_profile_avx (ElfW(Word)) attribute_hidden; |
| 6912 | extern void _dl_runtime_profile_avx512 (ElfW(Word)) attribute_hidden; |
| 6913 | @@ -96,8 +93,6 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], |
| 6914 | /* Identify this shared object. */ |
| 6915 | *(ElfW(Addr) *) (got + 1) = (ElfW(Addr)) l; |
| 6916 | |
| 6917 | - const struct cpu_features* cpu_features = __get_cpu_features (); |
| 6918 | - |
| 6919 | #ifdef SHARED |
| 6920 | /* The got[2] entry contains the address of a function which gets |
| 6921 | called to get the address of a so far unresolved function and |
| 6922 | @@ -107,6 +102,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], |
| 6923 | end in this function. */ |
| 6924 | if (__glibc_unlikely (profile)) |
| 6925 | { |
| 6926 | + const struct cpu_features* cpu_features = __get_cpu_features (); |
| 6927 | if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512F)) |
| 6928 | *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile_avx512; |
| 6929 | else if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX)) |
| 6930 | @@ -126,15 +122,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], |
| 6931 | /* This function will get called to fix up the GOT entry |
| 6932 | indicated by the offset on the stack, and then jump to |
| 6933 | the resolved address. */ |
| 6934 | - if (MINIMUM_X86_ISA_LEVEL >= AVX_X86_ISA_LEVEL |
| 6935 | - || GLRO(dl_x86_cpu_features).xsave_state_size != 0) |
| 6936 | - *(ElfW(Addr) *) (got + 2) |
| 6937 | - = (CPU_FEATURE_USABLE_P (cpu_features, XSAVEC) |
| 6938 | - ? (ElfW(Addr)) &_dl_runtime_resolve_xsavec |
| 6939 | - : (ElfW(Addr)) &_dl_runtime_resolve_xsave); |
| 6940 | - else |
| 6941 | - *(ElfW(Addr) *) (got + 2) |
| 6942 | - = (ElfW(Addr)) &_dl_runtime_resolve_fxsave; |
| 6943 | + *(ElfW(Addr) *) (got + 2) |
| 6944 | + = (ElfW(Addr)) GLRO(dl_x86_64_runtime_resolve); |
| 6945 | } |
| 6946 | } |
| 6947 | |
| 6948 | @@ -383,7 +372,7 @@ and creates an unsatisfiable circular dependency.\n", |
| 6949 | { |
| 6950 | td->arg = _dl_make_tlsdesc_dynamic |
| 6951 | (sym_map, sym->st_value + reloc->r_addend); |
| 6952 | - td->entry = _dl_tlsdesc_dynamic; |
| 6953 | + td->entry = GLRO(dl_x86_tlsdesc_dynamic); |
| 6954 | } |
| 6955 | else |
| 6956 | # endif |
| 6957 | diff --git a/sysdeps/x86_64/dl-procinfo.c b/sysdeps/x86_64/dl-procinfo.c |
| 6958 | index 4d1d790fbb..06637a8154 100644 |
| 6959 | --- a/sysdeps/x86_64/dl-procinfo.c |
| 6960 | +++ b/sysdeps/x86_64/dl-procinfo.c |
| 6961 | @@ -41,5 +41,21 @@ |
| 6962 | |
| 6963 | #include <sysdeps/x86/dl-procinfo.c> |
| 6964 | |
| 6965 | +#if !IS_IN (ldconfig) |
| 6966 | +# if !defined PROCINFO_DECL && defined SHARED |
| 6967 | + ._dl_x86_64_runtime_resolve |
| 6968 | +# else |
| 6969 | +PROCINFO_CLASS void * _dl_x86_64_runtime_resolve |
| 6970 | +# endif |
| 6971 | +# ifndef PROCINFO_DECL |
| 6972 | += NULL |
| 6973 | +# endif |
| 6974 | +# if !defined SHARED || defined PROCINFO_DECL |
| 6975 | +; |
| 6976 | +# else |
| 6977 | +, |
| 6978 | +# endif |
| 6979 | +#endif |
| 6980 | + |
| 6981 | #undef PROCINFO_DECL |
| 6982 | #undef PROCINFO_CLASS |
| 6983 | diff --git a/sysdeps/x86_64/dl-tlsdesc-dynamic.h b/sysdeps/x86_64/dl-tlsdesc-dynamic.h |
| 6984 | new file mode 100644 |
| 6985 | index 0000000000..9f02cfc3eb |
| 6986 | --- /dev/null |
| 6987 | +++ b/sysdeps/x86_64/dl-tlsdesc-dynamic.h |
| 6988 | @@ -0,0 +1,166 @@ |
| 6989 | +/* Thread-local storage handling in the ELF dynamic linker. x86_64 version. |
| 6990 | + Copyright (C) 2004-2024 Free Software Foundation, Inc. |
| 6991 | + This file is part of the GNU C Library. |
| 6992 | + |
| 6993 | + The GNU C Library is free software; you can redistribute it and/or |
| 6994 | + modify it under the terms of the GNU Lesser General Public |
| 6995 | + License as published by the Free Software Foundation; either |
| 6996 | + version 2.1 of the License, or (at your option) any later version. |
| 6997 | + |
| 6998 | + The GNU C Library is distributed in the hope that it will be useful, |
| 6999 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 7000 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 7001 | + Lesser General Public License for more details. |
| 7002 | + |
| 7003 | + You should have received a copy of the GNU Lesser General Public |
| 7004 | + License along with the GNU C Library; if not, see |
| 7005 | + <https://www.gnu.org/licenses/>. */ |
| 7006 | + |
| 7007 | +#ifndef SECTION |
| 7008 | +# define SECTION(p) p |
| 7009 | +#endif |
| 7010 | + |
| 7011 | +#undef REGISTER_SAVE_AREA |
| 7012 | +#undef LOCAL_STORAGE_AREA |
| 7013 | +#undef BASE |
| 7014 | + |
| 7015 | +#include "dl-trampoline-state.h" |
| 7016 | + |
| 7017 | + .section SECTION(.text),"ax",@progbits |
| 7018 | + |
| 7019 | + .hidden _dl_tlsdesc_dynamic |
| 7020 | + .global _dl_tlsdesc_dynamic |
| 7021 | + .type _dl_tlsdesc_dynamic,@function |
| 7022 | + |
| 7023 | + /* %rax points to the TLS descriptor, such that 0(%rax) points to |
| 7024 | + _dl_tlsdesc_dynamic itself, and 8(%rax) points to a struct |
| 7025 | + tlsdesc_dynamic_arg object. It must return in %rax the offset |
| 7026 | + between the thread pointer and the object denoted by the |
| 7027 | + argument, without clobbering any registers. |
| 7028 | + |
| 7029 | + The assembly code that follows is a rendition of the following |
| 7030 | + C code, hand-optimized a little bit. |
| 7031 | + |
| 7032 | +ptrdiff_t |
| 7033 | +_dl_tlsdesc_dynamic (register struct tlsdesc *tdp asm ("%rax")) |
| 7034 | +{ |
| 7035 | + struct tlsdesc_dynamic_arg *td = tdp->arg; |
| 7036 | + dtv_t *dtv = *(dtv_t **)((char *)__thread_pointer + DTV_OFFSET); |
| 7037 | + if (__builtin_expect (td->gen_count <= dtv[0].counter |
| 7038 | + && (dtv[td->tlsinfo.ti_module].pointer.val |
| 7039 | + != TLS_DTV_UNALLOCATED), |
| 7040 | + 1)) |
| 7041 | + return dtv[td->tlsinfo.ti_module].pointer.val + td->tlsinfo.ti_offset |
| 7042 | + - __thread_pointer; |
| 7043 | + |
| 7044 | + return __tls_get_addr_internal (&td->tlsinfo) - __thread_pointer; |
| 7045 | +} |
| 7046 | +*/ |
| 7047 | + cfi_startproc |
| 7048 | + .align 16 |
| 7049 | +_dl_tlsdesc_dynamic: |
| 7050 | + _CET_ENDBR |
| 7051 | + /* Preserve call-clobbered registers that we modify. |
| 7052 | + We need two scratch regs anyway. */ |
| 7053 | + movq %rsi, -16(%rsp) |
| 7054 | + mov %fs:DTV_OFFSET, %RSI_LP |
| 7055 | + movq %rdi, -8(%rsp) |
| 7056 | + movq TLSDESC_ARG(%rax), %rdi |
| 7057 | + movq (%rsi), %rax |
| 7058 | + cmpq %rax, TLSDESC_GEN_COUNT(%rdi) |
| 7059 | + ja 2f |
| 7060 | + movq TLSDESC_MODID(%rdi), %rax |
| 7061 | + salq $4, %rax |
| 7062 | + movq (%rax,%rsi), %rax |
| 7063 | + cmpq $-1, %rax |
| 7064 | + je 2f |
| 7065 | + addq TLSDESC_MODOFF(%rdi), %rax |
| 7066 | +1: |
| 7067 | + movq -16(%rsp), %rsi |
| 7068 | + sub %fs:0, %RAX_LP |
| 7069 | + movq -8(%rsp), %rdi |
| 7070 | + ret |
| 7071 | +2: |
| 7072 | +#if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 7073 | + movq %rbx, -24(%rsp) |
| 7074 | + mov %RSP_LP, %RBX_LP |
| 7075 | + cfi_def_cfa_register(%rbx) |
| 7076 | + and $-STATE_SAVE_ALIGNMENT, %RSP_LP |
| 7077 | +#endif |
| 7078 | +#ifdef REGISTER_SAVE_AREA |
| 7079 | +# if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 7080 | + /* STATE_SAVE_OFFSET has space for 8 integer registers. But we |
| 7081 | + need space for RCX, RDX, RSI, RDI, R8, R9, R10 and R11, plus |
| 7082 | + RBX above. */ |
| 7083 | + sub $(REGISTER_SAVE_AREA + STATE_SAVE_ALIGNMENT), %RSP_LP |
| 7084 | +# else |
| 7085 | + sub $REGISTER_SAVE_AREA, %RSP_LP |
| 7086 | + cfi_adjust_cfa_offset(REGISTER_SAVE_AREA) |
| 7087 | +# endif |
| 7088 | +#else |
| 7089 | + /* Allocate stack space of the required size to save the state. */ |
| 7090 | + sub _rtld_local_ro+RTLD_GLOBAL_RO_DL_X86_CPU_FEATURES_OFFSET+XSAVE_STATE_FULL_SIZE_OFFSET(%rip), %RSP_LP |
| 7091 | +#endif |
| 7092 | + /* Besides rdi and rsi, saved above, save rcx, rdx, r8, r9, |
| 7093 | + r10 and r11. */ |
| 7094 | + movq %rcx, REGISTER_SAVE_RCX(%rsp) |
| 7095 | + movq %rdx, REGISTER_SAVE_RDX(%rsp) |
| 7096 | + movq %r8, REGISTER_SAVE_R8(%rsp) |
| 7097 | + movq %r9, REGISTER_SAVE_R9(%rsp) |
| 7098 | + movq %r10, REGISTER_SAVE_R10(%rsp) |
| 7099 | + movq %r11, REGISTER_SAVE_R11(%rsp) |
| 7100 | +#ifdef USE_FXSAVE |
| 7101 | + fxsave STATE_SAVE_OFFSET(%rsp) |
| 7102 | +#else |
| 7103 | + movl $TLSDESC_CALL_STATE_SAVE_MASK, %eax |
| 7104 | + xorl %edx, %edx |
| 7105 | + /* Clear the XSAVE Header. */ |
| 7106 | +# ifdef USE_XSAVE |
| 7107 | + movq %rdx, (STATE_SAVE_OFFSET + 512)(%rsp) |
| 7108 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8)(%rsp) |
| 7109 | +# endif |
| 7110 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8 * 2)(%rsp) |
| 7111 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8 * 3)(%rsp) |
| 7112 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8 * 4)(%rsp) |
| 7113 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8 * 5)(%rsp) |
| 7114 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8 * 6)(%rsp) |
| 7115 | + movq %rdx, (STATE_SAVE_OFFSET + 512 + 8 * 7)(%rsp) |
| 7116 | +# ifdef USE_XSAVE |
| 7117 | + xsave STATE_SAVE_OFFSET(%rsp) |
| 7118 | +# else |
| 7119 | + xsavec STATE_SAVE_OFFSET(%rsp) |
| 7120 | +# endif |
| 7121 | +#endif |
| 7122 | + /* %rdi already points to the tlsinfo data structure. */ |
| 7123 | + call HIDDEN_JUMPTARGET (__tls_get_addr) |
| 7124 | + # Get register content back. |
| 7125 | +#ifdef USE_FXSAVE |
| 7126 | + fxrstor STATE_SAVE_OFFSET(%rsp) |
| 7127 | +#else |
| 7128 | + /* Save and retore __tls_get_addr return value stored in RAX. */ |
| 7129 | + mov %RAX_LP, %RCX_LP |
| 7130 | + movl $TLSDESC_CALL_STATE_SAVE_MASK, %eax |
| 7131 | + xorl %edx, %edx |
| 7132 | + xrstor STATE_SAVE_OFFSET(%rsp) |
| 7133 | + mov %RCX_LP, %RAX_LP |
| 7134 | +#endif |
| 7135 | + movq REGISTER_SAVE_R11(%rsp), %r11 |
| 7136 | + movq REGISTER_SAVE_R10(%rsp), %r10 |
| 7137 | + movq REGISTER_SAVE_R9(%rsp), %r9 |
| 7138 | + movq REGISTER_SAVE_R8(%rsp), %r8 |
| 7139 | + movq REGISTER_SAVE_RDX(%rsp), %rdx |
| 7140 | + movq REGISTER_SAVE_RCX(%rsp), %rcx |
| 7141 | +#if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 7142 | + mov %RBX_LP, %RSP_LP |
| 7143 | + cfi_def_cfa_register(%rsp) |
| 7144 | + movq -24(%rsp), %rbx |
| 7145 | + cfi_restore(%rbx) |
| 7146 | +#else |
| 7147 | + add $REGISTER_SAVE_AREA, %RSP_LP |
| 7148 | + cfi_adjust_cfa_offset(-REGISTER_SAVE_AREA) |
| 7149 | +#endif |
| 7150 | + jmp 1b |
| 7151 | + cfi_endproc |
| 7152 | + .size _dl_tlsdesc_dynamic, .-_dl_tlsdesc_dynamic |
| 7153 | + |
| 7154 | +#undef STATE_SAVE_ALIGNMENT |
| 7155 | diff --git a/sysdeps/x86_64/dl-tlsdesc.S b/sysdeps/x86_64/dl-tlsdesc.S |
| 7156 | index f748af2ece..057a10862a 100644 |
| 7157 | --- a/sysdeps/x86_64/dl-tlsdesc.S |
| 7158 | +++ b/sysdeps/x86_64/dl-tlsdesc.S |
| 7159 | @@ -18,7 +18,20 @@ |
| 7160 | |
| 7161 | #include <sysdep.h> |
| 7162 | #include <tls.h> |
| 7163 | +#include <cpu-features-offsets.h> |
| 7164 | +#include <features-offsets.h> |
| 7165 | +#include <isa-level.h> |
| 7166 | #include "tlsdesc.h" |
| 7167 | +#include "dl-trampoline-save.h" |
| 7168 | + |
| 7169 | +/* Area on stack to save and restore registers used for parameter |
| 7170 | + passing when calling _dl_tlsdesc_dynamic. */ |
| 7171 | +#define REGISTER_SAVE_RCX 0 |
| 7172 | +#define REGISTER_SAVE_RDX (REGISTER_SAVE_RCX + 8) |
| 7173 | +#define REGISTER_SAVE_R8 (REGISTER_SAVE_RDX + 8) |
| 7174 | +#define REGISTER_SAVE_R9 (REGISTER_SAVE_R8 + 8) |
| 7175 | +#define REGISTER_SAVE_R10 (REGISTER_SAVE_R9 + 8) |
| 7176 | +#define REGISTER_SAVE_R11 (REGISTER_SAVE_R10 + 8) |
| 7177 | |
| 7178 | .text |
| 7179 | |
| 7180 | @@ -67,80 +80,26 @@ _dl_tlsdesc_undefweak: |
| 7181 | .size _dl_tlsdesc_undefweak, .-_dl_tlsdesc_undefweak |
| 7182 | |
| 7183 | #ifdef SHARED |
| 7184 | - .hidden _dl_tlsdesc_dynamic |
| 7185 | - .global _dl_tlsdesc_dynamic |
| 7186 | - .type _dl_tlsdesc_dynamic,@function |
| 7187 | - |
| 7188 | - /* %rax points to the TLS descriptor, such that 0(%rax) points to |
| 7189 | - _dl_tlsdesc_dynamic itself, and 8(%rax) points to a struct |
| 7190 | - tlsdesc_dynamic_arg object. It must return in %rax the offset |
| 7191 | - between the thread pointer and the object denoted by the |
| 7192 | - argument, without clobbering any registers. |
| 7193 | - |
| 7194 | - The assembly code that follows is a rendition of the following |
| 7195 | - C code, hand-optimized a little bit. |
| 7196 | - |
| 7197 | -ptrdiff_t |
| 7198 | -_dl_tlsdesc_dynamic (register struct tlsdesc *tdp asm ("%rax")) |
| 7199 | -{ |
| 7200 | - struct tlsdesc_dynamic_arg *td = tdp->arg; |
| 7201 | - dtv_t *dtv = *(dtv_t **)((char *)__thread_pointer + DTV_OFFSET); |
| 7202 | - if (__builtin_expect (td->gen_count <= dtv[0].counter |
| 7203 | - && (dtv[td->tlsinfo.ti_module].pointer.val |
| 7204 | - != TLS_DTV_UNALLOCATED), |
| 7205 | - 1)) |
| 7206 | - return dtv[td->tlsinfo.ti_module].pointer.val + td->tlsinfo.ti_offset |
| 7207 | - - __thread_pointer; |
| 7208 | - |
| 7209 | - return __tls_get_addr_internal (&td->tlsinfo) - __thread_pointer; |
| 7210 | -} |
| 7211 | -*/ |
| 7212 | - cfi_startproc |
| 7213 | - .align 16 |
| 7214 | -_dl_tlsdesc_dynamic: |
| 7215 | - _CET_ENDBR |
| 7216 | - /* Preserve call-clobbered registers that we modify. |
| 7217 | - We need two scratch regs anyway. */ |
| 7218 | - movq %rsi, -16(%rsp) |
| 7219 | - mov %fs:DTV_OFFSET, %RSI_LP |
| 7220 | - movq %rdi, -8(%rsp) |
| 7221 | - movq TLSDESC_ARG(%rax), %rdi |
| 7222 | - movq (%rsi), %rax |
| 7223 | - cmpq %rax, TLSDESC_GEN_COUNT(%rdi) |
| 7224 | - ja .Lslow |
| 7225 | - movq TLSDESC_MODID(%rdi), %rax |
| 7226 | - salq $4, %rax |
| 7227 | - movq (%rax,%rsi), %rax |
| 7228 | - cmpq $-1, %rax |
| 7229 | - je .Lslow |
| 7230 | - addq TLSDESC_MODOFF(%rdi), %rax |
| 7231 | -.Lret: |
| 7232 | - movq -16(%rsp), %rsi |
| 7233 | - sub %fs:0, %RAX_LP |
| 7234 | - movq -8(%rsp), %rdi |
| 7235 | - ret |
| 7236 | -.Lslow: |
| 7237 | - /* Besides rdi and rsi, saved above, save rdx, rcx, r8, r9, |
| 7238 | - r10 and r11. Also, align the stack, that's off by 8 bytes. */ |
| 7239 | - subq $72, %rsp |
| 7240 | - cfi_adjust_cfa_offset (72) |
| 7241 | - movq %rdx, 8(%rsp) |
| 7242 | - movq %rcx, 16(%rsp) |
| 7243 | - movq %r8, 24(%rsp) |
| 7244 | - movq %r9, 32(%rsp) |
| 7245 | - movq %r10, 40(%rsp) |
| 7246 | - movq %r11, 48(%rsp) |
| 7247 | - /* %rdi already points to the tlsinfo data structure. */ |
| 7248 | - call HIDDEN_JUMPTARGET (__tls_get_addr) |
| 7249 | - movq 8(%rsp), %rdx |
| 7250 | - movq 16(%rsp), %rcx |
| 7251 | - movq 24(%rsp), %r8 |
| 7252 | - movq 32(%rsp), %r9 |
| 7253 | - movq 40(%rsp), %r10 |
| 7254 | - movq 48(%rsp), %r11 |
| 7255 | - addq $72, %rsp |
| 7256 | - cfi_adjust_cfa_offset (-72) |
| 7257 | - jmp .Lret |
| 7258 | - cfi_endproc |
| 7259 | - .size _dl_tlsdesc_dynamic, .-_dl_tlsdesc_dynamic |
| 7260 | +# if MINIMUM_X86_ISA_LEVEL < AVX_X86_ISA_LEVEL |
| 7261 | +# define USE_FXSAVE |
| 7262 | +# define STATE_SAVE_ALIGNMENT 16 |
| 7263 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_fxsave |
| 7264 | +# include "dl-tlsdesc-dynamic.h" |
| 7265 | +# undef _dl_tlsdesc_dynamic |
| 7266 | +# undef USE_FXSAVE |
| 7267 | +# endif |
| 7268 | + |
| 7269 | +# define USE_XSAVE |
| 7270 | +# define STATE_SAVE_ALIGNMENT 64 |
| 7271 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_xsave |
| 7272 | +# include "dl-tlsdesc-dynamic.h" |
| 7273 | +# undef _dl_tlsdesc_dynamic |
| 7274 | +# undef USE_XSAVE |
| 7275 | + |
| 7276 | +# define USE_XSAVEC |
| 7277 | +# define STATE_SAVE_ALIGNMENT 64 |
| 7278 | +# define _dl_tlsdesc_dynamic _dl_tlsdesc_dynamic_xsavec |
| 7279 | +# include "dl-tlsdesc-dynamic.h" |
| 7280 | +# undef _dl_tlsdesc_dynamic |
| 7281 | +# undef USE_XSAVEC |
| 7282 | #endif /* SHARED */ |
| 7283 | diff --git a/sysdeps/x86_64/dl-trampoline-save.h b/sysdeps/x86_64/dl-trampoline-save.h |
| 7284 | new file mode 100644 |
| 7285 | index 0000000000..84eac4a8ac |
| 7286 | --- /dev/null |
| 7287 | +++ b/sysdeps/x86_64/dl-trampoline-save.h |
| 7288 | @@ -0,0 +1,34 @@ |
| 7289 | +/* x86-64 PLT trampoline register save macros. |
| 7290 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 7291 | + This file is part of the GNU C Library. |
| 7292 | + |
| 7293 | + The GNU C Library is free software; you can redistribute it and/or |
| 7294 | + modify it under the terms of the GNU Lesser General Public |
| 7295 | + License as published by the Free Software Foundation; either |
| 7296 | + version 2.1 of the License, or (at your option) any later version. |
| 7297 | + |
| 7298 | + The GNU C Library is distributed in the hope that it will be useful, |
| 7299 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 7300 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 7301 | + Lesser General Public License for more details. |
| 7302 | + |
| 7303 | + You should have received a copy of the GNU Lesser General Public |
| 7304 | + License along with the GNU C Library; if not, see |
| 7305 | + <https://www.gnu.org/licenses/>. */ |
| 7306 | + |
| 7307 | +#ifndef DL_STACK_ALIGNMENT |
| 7308 | +/* Due to GCC bug: |
| 7309 | + |
| 7310 | + https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58066 |
| 7311 | + |
| 7312 | + __tls_get_addr may be called with 8-byte stack alignment. Although |
| 7313 | + this bug has been fixed in GCC 4.9.4, 5.3 and 6, we can't assume |
| 7314 | + that stack will be always aligned at 16 bytes. */ |
| 7315 | +# define DL_STACK_ALIGNMENT 8 |
| 7316 | +#endif |
| 7317 | + |
| 7318 | +/* True if _dl_runtime_resolve should align stack for STATE_SAVE or align |
| 7319 | + stack to 16 bytes before calling _dl_fixup. */ |
| 7320 | +#define DL_RUNTIME_RESOLVE_REALIGN_STACK \ |
| 7321 | + (STATE_SAVE_ALIGNMENT > DL_STACK_ALIGNMENT \ |
| 7322 | + || 16 > DL_STACK_ALIGNMENT) |
| 7323 | diff --git a/sysdeps/x86_64/dl-trampoline-state.h b/sysdeps/x86_64/dl-trampoline-state.h |
| 7324 | new file mode 100644 |
| 7325 | index 0000000000..575f120797 |
| 7326 | --- /dev/null |
| 7327 | +++ b/sysdeps/x86_64/dl-trampoline-state.h |
| 7328 | @@ -0,0 +1,51 @@ |
| 7329 | +/* x86-64 PLT dl-trampoline state macros. |
| 7330 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 7331 | + This file is part of the GNU C Library. |
| 7332 | + |
| 7333 | + The GNU C Library is free software; you can redistribute it and/or |
| 7334 | + modify it under the terms of the GNU Lesser General Public |
| 7335 | + License as published by the Free Software Foundation; either |
| 7336 | + version 2.1 of the License, or (at your option) any later version. |
| 7337 | + |
| 7338 | + The GNU C Library is distributed in the hope that it will be useful, |
| 7339 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 7340 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 7341 | + Lesser General Public License for more details. |
| 7342 | + |
| 7343 | + You should have received a copy of the GNU Lesser General Public |
| 7344 | + License along with the GNU C Library; if not, see |
| 7345 | + <https://www.gnu.org/licenses/>. */ |
| 7346 | + |
| 7347 | +#if (STATE_SAVE_ALIGNMENT % 16) != 0 |
| 7348 | +# error STATE_SAVE_ALIGNMENT must be multiple of 16 |
| 7349 | +#endif |
| 7350 | + |
| 7351 | +#if (STATE_SAVE_OFFSET % STATE_SAVE_ALIGNMENT) != 0 |
| 7352 | +# error STATE_SAVE_OFFSET must be multiple of STATE_SAVE_ALIGNMENT |
| 7353 | +#endif |
| 7354 | + |
| 7355 | +#if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 7356 | +/* Local stack area before jumping to function address: RBX. */ |
| 7357 | +# define LOCAL_STORAGE_AREA 8 |
| 7358 | +# define BASE rbx |
| 7359 | +# ifdef USE_FXSAVE |
| 7360 | +/* Use fxsave to save XMM registers. */ |
| 7361 | +# define REGISTER_SAVE_AREA (512 + STATE_SAVE_OFFSET) |
| 7362 | +# if (REGISTER_SAVE_AREA % 16) != 0 |
| 7363 | +# error REGISTER_SAVE_AREA must be multiple of 16 |
| 7364 | +# endif |
| 7365 | +# endif |
| 7366 | +#else |
| 7367 | +# ifndef USE_FXSAVE |
| 7368 | +# error USE_FXSAVE must be defined |
| 7369 | +# endif |
| 7370 | +/* Use fxsave to save XMM registers. */ |
| 7371 | +# define REGISTER_SAVE_AREA (512 + STATE_SAVE_OFFSET + 8) |
| 7372 | +/* Local stack area before jumping to function address: All saved |
| 7373 | + registers. */ |
| 7374 | +# define LOCAL_STORAGE_AREA REGISTER_SAVE_AREA |
| 7375 | +# define BASE rsp |
| 7376 | +# if (REGISTER_SAVE_AREA % 16) != 8 |
| 7377 | +# error REGISTER_SAVE_AREA must be odd multiple of 8 |
| 7378 | +# endif |
| 7379 | +#endif |
| 7380 | diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S |
| 7381 | index b2e7e0f69b..87c5137837 100644 |
| 7382 | --- a/sysdeps/x86_64/dl-trampoline.S |
| 7383 | +++ b/sysdeps/x86_64/dl-trampoline.S |
| 7384 | @@ -22,25 +22,7 @@ |
| 7385 | #include <features-offsets.h> |
| 7386 | #include <link-defines.h> |
| 7387 | #include <isa-level.h> |
| 7388 | - |
| 7389 | -#ifndef DL_STACK_ALIGNMENT |
| 7390 | -/* Due to GCC bug: |
| 7391 | - |
| 7392 | - https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58066 |
| 7393 | - |
| 7394 | - __tls_get_addr may be called with 8-byte stack alignment. Although |
| 7395 | - this bug has been fixed in GCC 4.9.4, 5.3 and 6, we can't assume |
| 7396 | - that stack will be always aligned at 16 bytes. We use unaligned |
| 7397 | - 16-byte move to load and store SSE registers, which has no penalty |
| 7398 | - on modern processors if stack is 16-byte aligned. */ |
| 7399 | -# define DL_STACK_ALIGNMENT 8 |
| 7400 | -#endif |
| 7401 | - |
| 7402 | -/* True if _dl_runtime_resolve should align stack for STATE_SAVE or align |
| 7403 | - stack to 16 bytes before calling _dl_fixup. */ |
| 7404 | -#define DL_RUNTIME_RESOLVE_REALIGN_STACK \ |
| 7405 | - (STATE_SAVE_ALIGNMENT > DL_STACK_ALIGNMENT \ |
| 7406 | - || 16 > DL_STACK_ALIGNMENT) |
| 7407 | +#include "dl-trampoline-save.h" |
| 7408 | |
| 7409 | /* Area on stack to save and restore registers used for parameter |
| 7410 | passing when calling _dl_fixup. */ |
| 7411 | diff --git a/sysdeps/x86_64/dl-trampoline.h b/sysdeps/x86_64/dl-trampoline.h |
| 7412 | index f55c6ea040..d9ccfb40d4 100644 |
| 7413 | --- a/sysdeps/x86_64/dl-trampoline.h |
| 7414 | +++ b/sysdeps/x86_64/dl-trampoline.h |
| 7415 | @@ -27,39 +27,7 @@ |
| 7416 | # undef LOCAL_STORAGE_AREA |
| 7417 | # undef BASE |
| 7418 | |
| 7419 | -# if (STATE_SAVE_ALIGNMENT % 16) != 0 |
| 7420 | -# error STATE_SAVE_ALIGNMENT must be multiple of 16 |
| 7421 | -# endif |
| 7422 | - |
| 7423 | -# if (STATE_SAVE_OFFSET % STATE_SAVE_ALIGNMENT) != 0 |
| 7424 | -# error STATE_SAVE_OFFSET must be multiple of STATE_SAVE_ALIGNMENT |
| 7425 | -# endif |
| 7426 | - |
| 7427 | -# if DL_RUNTIME_RESOLVE_REALIGN_STACK |
| 7428 | -/* Local stack area before jumping to function address: RBX. */ |
| 7429 | -# define LOCAL_STORAGE_AREA 8 |
| 7430 | -# define BASE rbx |
| 7431 | -# ifdef USE_FXSAVE |
| 7432 | -/* Use fxsave to save XMM registers. */ |
| 7433 | -# define REGISTER_SAVE_AREA (512 + STATE_SAVE_OFFSET) |
| 7434 | -# if (REGISTER_SAVE_AREA % 16) != 0 |
| 7435 | -# error REGISTER_SAVE_AREA must be multiple of 16 |
| 7436 | -# endif |
| 7437 | -# endif |
| 7438 | -# else |
| 7439 | -# ifndef USE_FXSAVE |
| 7440 | -# error USE_FXSAVE must be defined |
| 7441 | -# endif |
| 7442 | -/* Use fxsave to save XMM registers. */ |
| 7443 | -# define REGISTER_SAVE_AREA (512 + STATE_SAVE_OFFSET + 8) |
| 7444 | -/* Local stack area before jumping to function address: All saved |
| 7445 | - registers. */ |
| 7446 | -# define LOCAL_STORAGE_AREA REGISTER_SAVE_AREA |
| 7447 | -# define BASE rsp |
| 7448 | -# if (REGISTER_SAVE_AREA % 16) != 8 |
| 7449 | -# error REGISTER_SAVE_AREA must be odd multiple of 8 |
| 7450 | -# endif |
| 7451 | -# endif |
| 7452 | +# include "dl-trampoline-state.h" |
| 7453 | |
| 7454 | .globl _dl_runtime_resolve |
| 7455 | .hidden _dl_runtime_resolve |
| 7456 | diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile |
| 7457 | index ea81753b70..6ddd50240c 100644 |
| 7458 | --- a/sysdeps/x86_64/fpu/multiarch/Makefile |
| 7459 | +++ b/sysdeps/x86_64/fpu/multiarch/Makefile |
| 7460 | @@ -1,49 +1,4 @@ |
| 7461 | ifeq ($(subdir),math) |
| 7462 | -libm-sysdep_routines += \ |
| 7463 | - s_ceil-c \ |
| 7464 | - s_ceilf-c \ |
| 7465 | - s_floor-c \ |
| 7466 | - s_floorf-c \ |
| 7467 | - s_rint-c \ |
| 7468 | - s_rintf-c \ |
| 7469 | - s_nearbyint-c \ |
| 7470 | - s_nearbyintf-c \ |
| 7471 | - s_roundeven-c \ |
| 7472 | - s_roundevenf-c \ |
| 7473 | - s_trunc-c \ |
| 7474 | - s_truncf-c \ |
| 7475 | -# libm-sysdep_routines |
| 7476 | - |
| 7477 | -libm-sysdep_routines += \ |
| 7478 | - s_ceil-sse4_1 \ |
| 7479 | - s_ceilf-sse4_1 \ |
| 7480 | - s_floor-sse4_1 \ |
| 7481 | - s_floorf-sse4_1 \ |
| 7482 | - s_nearbyint-sse4_1 \ |
| 7483 | - s_nearbyintf-sse4_1 \ |
| 7484 | - s_roundeven-sse4_1 \ |
| 7485 | - s_roundevenf-sse4_1 \ |
| 7486 | - s_rint-sse4_1 \ |
| 7487 | - s_rintf-sse4_1 \ |
| 7488 | - s_trunc-sse4_1 \ |
| 7489 | - s_truncf-sse4_1 \ |
| 7490 | -# libm-sysdep_routines |
| 7491 | - |
| 7492 | -libm-sysdep_routines += \ |
| 7493 | - e_asin-fma \ |
| 7494 | - e_atan2-fma \ |
| 7495 | - e_exp-fma \ |
| 7496 | - e_log-fma \ |
| 7497 | - e_log2-fma \ |
| 7498 | - e_pow-fma \ |
| 7499 | - s_atan-fma \ |
| 7500 | - s_expm1-fma \ |
| 7501 | - s_log1p-fma \ |
| 7502 | - s_sin-fma \ |
| 7503 | - s_sincos-fma \ |
| 7504 | - s_tan-fma \ |
| 7505 | -# libm-sysdep_routines |
| 7506 | - |
| 7507 | CFLAGS-e_asin-fma.c = -mfma -mavx2 |
| 7508 | CFLAGS-e_atan2-fma.c = -mfma -mavx2 |
| 7509 | CFLAGS-e_exp-fma.c = -mfma -mavx2 |
| 7510 | @@ -57,23 +12,6 @@ CFLAGS-s_sin-fma.c = -mfma -mavx2 |
| 7511 | CFLAGS-s_tan-fma.c = -mfma -mavx2 |
| 7512 | CFLAGS-s_sincos-fma.c = -mfma -mavx2 |
| 7513 | |
| 7514 | -libm-sysdep_routines += \ |
| 7515 | - s_cosf-sse2 \ |
| 7516 | - s_sincosf-sse2 \ |
| 7517 | - s_sinf-sse2 \ |
| 7518 | -# libm-sysdep_routines |
| 7519 | - |
| 7520 | -libm-sysdep_routines += \ |
| 7521 | - e_exp2f-fma \ |
| 7522 | - e_expf-fma \ |
| 7523 | - e_log2f-fma \ |
| 7524 | - e_logf-fma \ |
| 7525 | - e_powf-fma \ |
| 7526 | - s_cosf-fma \ |
| 7527 | - s_sincosf-fma \ |
| 7528 | - s_sinf-fma \ |
| 7529 | -# libm-sysdep_routines |
| 7530 | - |
| 7531 | CFLAGS-e_exp2f-fma.c = -mfma -mavx2 |
| 7532 | CFLAGS-e_expf-fma.c = -mfma -mavx2 |
| 7533 | CFLAGS-e_log2f-fma.c = -mfma -mavx2 |
| 7534 | @@ -83,17 +21,93 @@ CFLAGS-s_sinf-fma.c = -mfma -mavx2 |
| 7535 | CFLAGS-s_cosf-fma.c = -mfma -mavx2 |
| 7536 | CFLAGS-s_sincosf-fma.c = -mfma -mavx2 |
| 7537 | |
| 7538 | +# Check if ISA level is 3 or above. |
| 7539 | +ifneq (,$(filter $(have-x86-isa-level),$(x86-isa-level-3-or-above))) |
| 7540 | +libm-sysdep_routines += \ |
| 7541 | + s_ceil-avx \ |
| 7542 | + s_ceilf-avx \ |
| 7543 | + s_floor-avx \ |
| 7544 | + s_floorf-avx \ |
| 7545 | + s_nearbyint-avx \ |
| 7546 | + s_nearbyintf-avx \ |
| 7547 | + s_rint-avx \ |
| 7548 | + s_rintf-avx \ |
| 7549 | + s_roundeven-avx \ |
| 7550 | + s_roundevenf-avx \ |
| 7551 | + s_trunc-avx \ |
| 7552 | + s_truncf-avx \ |
| 7553 | +# libm-sysdep_routines |
| 7554 | +else |
| 7555 | libm-sysdep_routines += \ |
| 7556 | + e_asin-fma \ |
| 7557 | + e_asin-fma4 \ |
| 7558 | + e_atan2-avx \ |
| 7559 | + e_atan2-fma \ |
| 7560 | + e_atan2-fma4 \ |
| 7561 | + e_exp-avx \ |
| 7562 | + e_exp-fma \ |
| 7563 | e_exp-fma4 \ |
| 7564 | + e_exp2f-fma \ |
| 7565 | + e_expf-fma \ |
| 7566 | + e_log-avx \ |
| 7567 | + e_log-fma \ |
| 7568 | e_log-fma4 \ |
| 7569 | + e_log2-fma \ |
| 7570 | + e_log2f-fma \ |
| 7571 | + e_logf-fma \ |
| 7572 | + e_pow-fma \ |
| 7573 | e_pow-fma4 \ |
| 7574 | - e_asin-fma4 \ |
| 7575 | + e_powf-fma \ |
| 7576 | + s_atan-avx \ |
| 7577 | + s_atan-fma \ |
| 7578 | s_atan-fma4 \ |
| 7579 | - e_atan2-fma4 \ |
| 7580 | + s_ceil-sse4_1 \ |
| 7581 | + s_ceilf-sse4_1 \ |
| 7582 | + s_cosf-fma \ |
| 7583 | + s_cosf-sse2 \ |
| 7584 | + s_expm1-fma \ |
| 7585 | + s_floor-sse4_1 \ |
| 7586 | + s_floorf-sse4_1 \ |
| 7587 | + s_log1p-fma \ |
| 7588 | + s_nearbyint-sse4_1 \ |
| 7589 | + s_nearbyintf-sse4_1 \ |
| 7590 | + s_rint-sse4_1 \ |
| 7591 | + s_rintf-sse4_1 \ |
| 7592 | + s_roundeven-sse4_1 \ |
| 7593 | + s_roundevenf-sse4_1 \ |
| 7594 | + s_sin-avx \ |
| 7595 | + s_sin-fma \ |
| 7596 | s_sin-fma4 \ |
| 7597 | + s_sincos-avx \ |
| 7598 | + s_sincos-fma \ |
| 7599 | s_sincos-fma4 \ |
| 7600 | + s_sincosf-fma \ |
| 7601 | + s_sincosf-sse2 \ |
| 7602 | + s_sinf-fma \ |
| 7603 | + s_sinf-sse2 \ |
| 7604 | + s_tan-avx \ |
| 7605 | + s_tan-fma \ |
| 7606 | s_tan-fma4 \ |
| 7607 | + s_trunc-sse4_1 \ |
| 7608 | + s_truncf-sse4_1 \ |
| 7609 | +# libm-sysdep_routines |
| 7610 | +ifeq ($(have-x86-isa-level),baseline) |
| 7611 | +libm-sysdep_routines += \ |
| 7612 | + s_ceil-c \ |
| 7613 | + s_ceilf-c \ |
| 7614 | + s_floor-c \ |
| 7615 | + s_floorf-c \ |
| 7616 | + s_nearbyint-c \ |
| 7617 | + s_nearbyintf-c \ |
| 7618 | + s_rint-c \ |
| 7619 | + s_rintf-c \ |
| 7620 | + s_roundeven-c \ |
| 7621 | + s_roundevenf-c \ |
| 7622 | + s_trunc-c \ |
| 7623 | + s_truncf-c \ |
| 7624 | # libm-sysdep_routines |
| 7625 | +endif |
| 7626 | +endif |
| 7627 | |
| 7628 | CFLAGS-e_asin-fma4.c = -mfma4 |
| 7629 | CFLAGS-e_atan2-fma4.c = -mfma4 |
| 7630 | @@ -105,16 +119,6 @@ CFLAGS-s_sin-fma4.c = -mfma4 |
| 7631 | CFLAGS-s_tan-fma4.c = -mfma4 |
| 7632 | CFLAGS-s_sincos-fma4.c = -mfma4 |
| 7633 | |
| 7634 | -libm-sysdep_routines += \ |
| 7635 | - e_exp-avx \ |
| 7636 | - e_log-avx \ |
| 7637 | - s_atan-avx \ |
| 7638 | - e_atan2-avx \ |
| 7639 | - s_sin-avx \ |
| 7640 | - s_sincos-avx \ |
| 7641 | - s_tan-avx \ |
| 7642 | -# libm-sysdep_routines |
| 7643 | - |
| 7644 | CFLAGS-e_atan2-avx.c = -msse2avx -DSSE2AVX |
| 7645 | CFLAGS-e_exp-avx.c = -msse2avx -DSSE2AVX |
| 7646 | CFLAGS-e_log-avx.c = -msse2avx -DSSE2AVX |
| 7647 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_asin.c b/sysdeps/x86_64/fpu/multiarch/e_asin.c |
| 7648 | index 2eaa6c2c04..d64fca2586 100644 |
| 7649 | --- a/sysdeps/x86_64/fpu/multiarch/e_asin.c |
| 7650 | +++ b/sysdeps/x86_64/fpu/multiarch/e_asin.c |
| 7651 | @@ -16,26 +16,29 @@ |
| 7652 | License along with the GNU C Library; if not, see |
| 7653 | <https://www.gnu.org/licenses/>. */ |
| 7654 | |
| 7655 | -#include <libm-alias-finite.h> |
| 7656 | +#include <sysdeps/x86/isa-level.h> |
| 7657 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7658 | +# include <libm-alias-finite.h> |
| 7659 | |
| 7660 | extern double __redirect_ieee754_asin (double); |
| 7661 | extern double __redirect_ieee754_acos (double); |
| 7662 | |
| 7663 | -#define SYMBOL_NAME ieee754_asin |
| 7664 | -#include "ifunc-fma4.h" |
| 7665 | +# define SYMBOL_NAME ieee754_asin |
| 7666 | +# include "ifunc-fma4.h" |
| 7667 | |
| 7668 | libc_ifunc_redirected (__redirect_ieee754_asin, __ieee754_asin, |
| 7669 | IFUNC_SELECTOR ()); |
| 7670 | libm_alias_finite (__ieee754_asin, __asin) |
| 7671 | |
| 7672 | -#undef SYMBOL_NAME |
| 7673 | -#define SYMBOL_NAME ieee754_acos |
| 7674 | -#include "ifunc-fma4.h" |
| 7675 | +# undef SYMBOL_NAME |
| 7676 | +# define SYMBOL_NAME ieee754_acos |
| 7677 | +# include "ifunc-fma4.h" |
| 7678 | |
| 7679 | libc_ifunc_redirected (__redirect_ieee754_acos, __ieee754_acos, |
| 7680 | IFUNC_SELECTOR ()); |
| 7681 | libm_alias_finite (__ieee754_acos, __acos) |
| 7682 | |
| 7683 | -#define __ieee754_acos __ieee754_acos_sse2 |
| 7684 | -#define __ieee754_asin __ieee754_asin_sse2 |
| 7685 | +# define __ieee754_acos __ieee754_acos_sse2 |
| 7686 | +# define __ieee754_asin __ieee754_asin_sse2 |
| 7687 | +#endif |
| 7688 | #include <sysdeps/ieee754/dbl-64/e_asin.c> |
| 7689 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_atan2.c b/sysdeps/x86_64/fpu/multiarch/e_atan2.c |
| 7690 | index 17ee4f3c36..8a86c14ded 100644 |
| 7691 | --- a/sysdeps/x86_64/fpu/multiarch/e_atan2.c |
| 7692 | +++ b/sysdeps/x86_64/fpu/multiarch/e_atan2.c |
| 7693 | @@ -16,16 +16,19 @@ |
| 7694 | License along with the GNU C Library; if not, see |
| 7695 | <https://www.gnu.org/licenses/>. */ |
| 7696 | |
| 7697 | -#include <libm-alias-finite.h> |
| 7698 | +#include <sysdeps/x86/isa-level.h> |
| 7699 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7700 | +# include <libm-alias-finite.h> |
| 7701 | |
| 7702 | extern double __redirect_ieee754_atan2 (double, double); |
| 7703 | |
| 7704 | -#define SYMBOL_NAME ieee754_atan2 |
| 7705 | -#include "ifunc-avx-fma4.h" |
| 7706 | +# define SYMBOL_NAME ieee754_atan2 |
| 7707 | +# include "ifunc-avx-fma4.h" |
| 7708 | |
| 7709 | libc_ifunc_redirected (__redirect_ieee754_atan2, |
| 7710 | __ieee754_atan2, IFUNC_SELECTOR ()); |
| 7711 | libm_alias_finite (__ieee754_atan2, __atan2) |
| 7712 | |
| 7713 | -#define __ieee754_atan2 __ieee754_atan2_sse2 |
| 7714 | +# define __ieee754_atan2 __ieee754_atan2_sse2 |
| 7715 | +#endif |
| 7716 | #include <sysdeps/ieee754/dbl-64/e_atan2.c> |
| 7717 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_exp.c b/sysdeps/x86_64/fpu/multiarch/e_exp.c |
| 7718 | index 406b7ebd44..d56329291a 100644 |
| 7719 | --- a/sysdeps/x86_64/fpu/multiarch/e_exp.c |
| 7720 | +++ b/sysdeps/x86_64/fpu/multiarch/e_exp.c |
| 7721 | @@ -16,17 +16,20 @@ |
| 7722 | License along with the GNU C Library; if not, see |
| 7723 | <https://www.gnu.org/licenses/>. */ |
| 7724 | |
| 7725 | -#include <math.h> |
| 7726 | -#include <libm-alias-finite.h> |
| 7727 | +#include <sysdeps/x86/isa-level.h> |
| 7728 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7729 | +# include <math.h> |
| 7730 | +# include <libm-alias-finite.h> |
| 7731 | |
| 7732 | extern double __redirect_ieee754_exp (double); |
| 7733 | |
| 7734 | -#define SYMBOL_NAME ieee754_exp |
| 7735 | -#include "ifunc-avx-fma4.h" |
| 7736 | +# define SYMBOL_NAME ieee754_exp |
| 7737 | +# include "ifunc-avx-fma4.h" |
| 7738 | |
| 7739 | libc_ifunc_redirected (__redirect_ieee754_exp, __ieee754_exp, |
| 7740 | IFUNC_SELECTOR ()); |
| 7741 | libm_alias_finite (__ieee754_exp, __exp) |
| 7742 | |
| 7743 | -#define __exp __ieee754_exp_sse2 |
| 7744 | +# define __exp __ieee754_exp_sse2 |
| 7745 | +#endif |
| 7746 | #include <sysdeps/ieee754/dbl-64/e_exp.c> |
| 7747 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_exp2f.c b/sysdeps/x86_64/fpu/multiarch/e_exp2f.c |
| 7748 | index 804fd6be85..06fe5028d6 100644 |
| 7749 | --- a/sysdeps/x86_64/fpu/multiarch/e_exp2f.c |
| 7750 | +++ b/sysdeps/x86_64/fpu/multiarch/e_exp2f.c |
| 7751 | @@ -16,25 +16,28 @@ |
| 7752 | License along with the GNU C Library; if not, see |
| 7753 | <https://www.gnu.org/licenses/>. */ |
| 7754 | |
| 7755 | -#include <libm-alias-float.h> |
| 7756 | -#include <libm-alias-finite.h> |
| 7757 | +#include <sysdeps/x86/isa-level.h> |
| 7758 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7759 | +# include <libm-alias-float.h> |
| 7760 | +# include <libm-alias-finite.h> |
| 7761 | |
| 7762 | extern float __redirect_exp2f (float); |
| 7763 | |
| 7764 | -#define SYMBOL_NAME exp2f |
| 7765 | -#include "ifunc-fma.h" |
| 7766 | +# define SYMBOL_NAME exp2f |
| 7767 | +# include "ifunc-fma.h" |
| 7768 | |
| 7769 | libc_ifunc_redirected (__redirect_exp2f, __exp2f, IFUNC_SELECTOR ()); |
| 7770 | |
| 7771 | -#ifdef SHARED |
| 7772 | +# ifdef SHARED |
| 7773 | versioned_symbol (libm, __ieee754_exp2f, exp2f, GLIBC_2_27); |
| 7774 | libm_alias_float_other (__exp2, exp2) |
| 7775 | -#else |
| 7776 | +# else |
| 7777 | libm_alias_float (__exp2, exp2) |
| 7778 | -#endif |
| 7779 | +# endif |
| 7780 | |
| 7781 | strong_alias (__exp2f, __ieee754_exp2f) |
| 7782 | libm_alias_finite (__exp2f, __exp2f) |
| 7783 | |
| 7784 | -#define __exp2f __exp2f_sse2 |
| 7785 | +# define __exp2f __exp2f_sse2 |
| 7786 | +#endif |
| 7787 | #include <sysdeps/ieee754/flt-32/e_exp2f.c> |
| 7788 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_expf.c b/sysdeps/x86_64/fpu/multiarch/e_expf.c |
| 7789 | index 4a7e2a5bce..19d767f636 100644 |
| 7790 | --- a/sysdeps/x86_64/fpu/multiarch/e_expf.c |
| 7791 | +++ b/sysdeps/x86_64/fpu/multiarch/e_expf.c |
| 7792 | @@ -16,28 +16,31 @@ |
| 7793 | License along with the GNU C Library; if not, see |
| 7794 | <https://www.gnu.org/licenses/>. */ |
| 7795 | |
| 7796 | -#include <libm-alias-float.h> |
| 7797 | -#include <libm-alias-finite.h> |
| 7798 | +#include <sysdeps/x86/isa-level.h> |
| 7799 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7800 | +# include <libm-alias-float.h> |
| 7801 | +# include <libm-alias-finite.h> |
| 7802 | |
| 7803 | extern float __redirect_expf (float); |
| 7804 | |
| 7805 | -#define SYMBOL_NAME expf |
| 7806 | -#include "ifunc-fma.h" |
| 7807 | +# define SYMBOL_NAME expf |
| 7808 | +# include "ifunc-fma.h" |
| 7809 | |
| 7810 | libc_ifunc_redirected (__redirect_expf, __expf, IFUNC_SELECTOR ()); |
| 7811 | |
| 7812 | -#ifdef SHARED |
| 7813 | +# ifdef SHARED |
| 7814 | __hidden_ver1 (__expf, __GI___expf, __redirect_expf) |
| 7815 | __attribute__ ((visibility ("hidden"))); |
| 7816 | |
| 7817 | versioned_symbol (libm, __ieee754_expf, expf, GLIBC_2_27); |
| 7818 | libm_alias_float_other (__exp, exp) |
| 7819 | -#else |
| 7820 | +# else |
| 7821 | libm_alias_float (__exp, exp) |
| 7822 | -#endif |
| 7823 | +# endif |
| 7824 | |
| 7825 | strong_alias (__expf, __ieee754_expf) |
| 7826 | libm_alias_finite (__expf, __expf) |
| 7827 | |
| 7828 | -#define __expf __expf_sse2 |
| 7829 | +# define __expf __expf_sse2 |
| 7830 | +#endif |
| 7831 | #include <sysdeps/ieee754/flt-32/e_expf.c> |
| 7832 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_log.c b/sysdeps/x86_64/fpu/multiarch/e_log.c |
| 7833 | index 067fbf58c3..d80c1b1463 100644 |
| 7834 | --- a/sysdeps/x86_64/fpu/multiarch/e_log.c |
| 7835 | +++ b/sysdeps/x86_64/fpu/multiarch/e_log.c |
| 7836 | @@ -16,17 +16,20 @@ |
| 7837 | License along with the GNU C Library; if not, see |
| 7838 | <https://www.gnu.org/licenses/>. */ |
| 7839 | |
| 7840 | -#include <math.h> |
| 7841 | -#include <libm-alias-finite.h> |
| 7842 | +#include <sysdeps/x86/isa-level.h> |
| 7843 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7844 | +# include <math.h> |
| 7845 | +# include <libm-alias-finite.h> |
| 7846 | |
| 7847 | extern double __redirect_ieee754_log (double); |
| 7848 | |
| 7849 | -#define SYMBOL_NAME ieee754_log |
| 7850 | -#include "ifunc-avx-fma4.h" |
| 7851 | +# define SYMBOL_NAME ieee754_log |
| 7852 | +# include "ifunc-avx-fma4.h" |
| 7853 | |
| 7854 | libc_ifunc_redirected (__redirect_ieee754_log, __ieee754_log, |
| 7855 | IFUNC_SELECTOR ()); |
| 7856 | libm_alias_finite (__ieee754_log, __log) |
| 7857 | |
| 7858 | -#define __log __ieee754_log_sse2 |
| 7859 | +# define __log __ieee754_log_sse2 |
| 7860 | +#endif |
| 7861 | #include <sysdeps/ieee754/dbl-64/e_log.c> |
| 7862 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_log2.c b/sysdeps/x86_64/fpu/multiarch/e_log2.c |
| 7863 | index 9c57a2f6cc..9686782c09 100644 |
| 7864 | --- a/sysdeps/x86_64/fpu/multiarch/e_log2.c |
| 7865 | +++ b/sysdeps/x86_64/fpu/multiarch/e_log2.c |
| 7866 | @@ -16,28 +16,31 @@ |
| 7867 | License along with the GNU C Library; if not, see |
| 7868 | <https://www.gnu.org/licenses/>. */ |
| 7869 | |
| 7870 | -#include <libm-alias-double.h> |
| 7871 | -#include <libm-alias-finite.h> |
| 7872 | +#include <sysdeps/x86/isa-level.h> |
| 7873 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7874 | +# include <libm-alias-double.h> |
| 7875 | +# include <libm-alias-finite.h> |
| 7876 | |
| 7877 | extern double __redirect_log2 (double); |
| 7878 | |
| 7879 | -#define SYMBOL_NAME log2 |
| 7880 | -#include "ifunc-fma.h" |
| 7881 | +# define SYMBOL_NAME log2 |
| 7882 | +# include "ifunc-fma.h" |
| 7883 | |
| 7884 | libc_ifunc_redirected (__redirect_log2, __log2, IFUNC_SELECTOR ()); |
| 7885 | |
| 7886 | -#ifdef SHARED |
| 7887 | +# ifdef SHARED |
| 7888 | __hidden_ver1 (__log2, __GI___log2, __redirect_log2) |
| 7889 | __attribute__ ((visibility ("hidden"))); |
| 7890 | |
| 7891 | versioned_symbol (libm, __ieee754_log2, log2, GLIBC_2_29); |
| 7892 | libm_alias_double_other (__log2, log2) |
| 7893 | -#else |
| 7894 | +# else |
| 7895 | libm_alias_double (__log2, log2) |
| 7896 | -#endif |
| 7897 | +# endif |
| 7898 | |
| 7899 | strong_alias (__log2, __ieee754_log2) |
| 7900 | libm_alias_finite (__log2, __log2) |
| 7901 | |
| 7902 | -#define __log2 __log2_sse2 |
| 7903 | +# define __log2 __log2_sse2 |
| 7904 | +#endif |
| 7905 | #include <sysdeps/ieee754/dbl-64/e_log2.c> |
| 7906 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_log2f.c b/sysdeps/x86_64/fpu/multiarch/e_log2f.c |
| 7907 | index 2b45c87f38..8ada46e11e 100644 |
| 7908 | --- a/sysdeps/x86_64/fpu/multiarch/e_log2f.c |
| 7909 | +++ b/sysdeps/x86_64/fpu/multiarch/e_log2f.c |
| 7910 | @@ -16,28 +16,31 @@ |
| 7911 | License along with the GNU C Library; if not, see |
| 7912 | <https://www.gnu.org/licenses/>. */ |
| 7913 | |
| 7914 | -#include <libm-alias-float.h> |
| 7915 | -#include <libm-alias-finite.h> |
| 7916 | +#include <sysdeps/x86/isa-level.h> |
| 7917 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7918 | +# include <libm-alias-float.h> |
| 7919 | +# include <libm-alias-finite.h> |
| 7920 | |
| 7921 | extern float __redirect_log2f (float); |
| 7922 | |
| 7923 | -#define SYMBOL_NAME log2f |
| 7924 | -#include "ifunc-fma.h" |
| 7925 | +# define SYMBOL_NAME log2f |
| 7926 | +# include "ifunc-fma.h" |
| 7927 | |
| 7928 | libc_ifunc_redirected (__redirect_log2f, __log2f, IFUNC_SELECTOR ()); |
| 7929 | |
| 7930 | -#ifdef SHARED |
| 7931 | +# ifdef SHARED |
| 7932 | __hidden_ver1 (__log2f, __GI___log2f, __redirect_log2f) |
| 7933 | __attribute__ ((visibility ("hidden"))); |
| 7934 | |
| 7935 | versioned_symbol (libm, __ieee754_log2f, log2f, GLIBC_2_27); |
| 7936 | libm_alias_float_other (__log2, log2) |
| 7937 | -#else |
| 7938 | +# else |
| 7939 | libm_alias_float (__log2, log2) |
| 7940 | -#endif |
| 7941 | +# endif |
| 7942 | |
| 7943 | strong_alias (__log2f, __ieee754_log2f) |
| 7944 | libm_alias_finite (__log2f, __log2f) |
| 7945 | |
| 7946 | -#define __log2f __log2f_sse2 |
| 7947 | +# define __log2f __log2f_sse2 |
| 7948 | +#endif |
| 7949 | #include <sysdeps/ieee754/flt-32/e_log2f.c> |
| 7950 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_logf.c b/sysdeps/x86_64/fpu/multiarch/e_logf.c |
| 7951 | index 97e23c8fea..a3978d9a8e 100644 |
| 7952 | --- a/sysdeps/x86_64/fpu/multiarch/e_logf.c |
| 7953 | +++ b/sysdeps/x86_64/fpu/multiarch/e_logf.c |
| 7954 | @@ -16,28 +16,31 @@ |
| 7955 | License along with the GNU C Library; if not, see |
| 7956 | <https://www.gnu.org/licenses/>. */ |
| 7957 | |
| 7958 | -#include <libm-alias-float.h> |
| 7959 | -#include <libm-alias-finite.h> |
| 7960 | +#include <sysdeps/x86/isa-level.h> |
| 7961 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 7962 | +# include <libm-alias-float.h> |
| 7963 | +# include <libm-alias-finite.h> |
| 7964 | |
| 7965 | extern float __redirect_logf (float); |
| 7966 | |
| 7967 | -#define SYMBOL_NAME logf |
| 7968 | -#include "ifunc-fma.h" |
| 7969 | +# define SYMBOL_NAME logf |
| 7970 | +# include "ifunc-fma.h" |
| 7971 | |
| 7972 | libc_ifunc_redirected (__redirect_logf, __logf, IFUNC_SELECTOR ()); |
| 7973 | |
| 7974 | -#ifdef SHARED |
| 7975 | +# ifdef SHARED |
| 7976 | __hidden_ver1 (__logf, __GI___logf, __redirect_logf) |
| 7977 | __attribute__ ((visibility ("hidden"))); |
| 7978 | |
| 7979 | versioned_symbol (libm, __ieee754_logf, logf, GLIBC_2_27); |
| 7980 | libm_alias_float_other (__log, log) |
| 7981 | -#else |
| 7982 | +# else |
| 7983 | libm_alias_float (__log, log) |
| 7984 | -#endif |
| 7985 | +# endif |
| 7986 | |
| 7987 | strong_alias (__logf, __ieee754_logf) |
| 7988 | libm_alias_finite (__logf, __logf) |
| 7989 | |
| 7990 | -#define __logf __logf_sse2 |
| 7991 | +# define __logf __logf_sse2 |
| 7992 | +#endif |
| 7993 | #include <sysdeps/ieee754/flt-32/e_logf.c> |
| 7994 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_pow.c b/sysdeps/x86_64/fpu/multiarch/e_pow.c |
| 7995 | index 42618e7112..f8f17aff9f 100644 |
| 7996 | --- a/sysdeps/x86_64/fpu/multiarch/e_pow.c |
| 7997 | +++ b/sysdeps/x86_64/fpu/multiarch/e_pow.c |
| 7998 | @@ -16,17 +16,20 @@ |
| 7999 | License along with the GNU C Library; if not, see |
| 8000 | <https://www.gnu.org/licenses/>. */ |
| 8001 | |
| 8002 | -#include <math.h> |
| 8003 | -#include <libm-alias-finite.h> |
| 8004 | +#include <sysdeps/x86/isa-level.h> |
| 8005 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 8006 | +# include <math.h> |
| 8007 | +# include <libm-alias-finite.h> |
| 8008 | |
| 8009 | extern double __redirect_ieee754_pow (double, double); |
| 8010 | |
| 8011 | -#define SYMBOL_NAME ieee754_pow |
| 8012 | -#include "ifunc-fma4.h" |
| 8013 | +# define SYMBOL_NAME ieee754_pow |
| 8014 | +# include "ifunc-fma4.h" |
| 8015 | |
| 8016 | libc_ifunc_redirected (__redirect_ieee754_pow, |
| 8017 | __ieee754_pow, IFUNC_SELECTOR ()); |
| 8018 | libm_alias_finite (__ieee754_pow, __pow) |
| 8019 | |
| 8020 | -#define __pow __ieee754_pow_sse2 |
| 8021 | +# define __pow __ieee754_pow_sse2 |
| 8022 | +#endif |
| 8023 | #include <sysdeps/ieee754/dbl-64/e_pow.c> |
| 8024 | diff --git a/sysdeps/x86_64/fpu/multiarch/e_powf.c b/sysdeps/x86_64/fpu/multiarch/e_powf.c |
| 8025 | index 8e6ce13cc1..8b1a4c7d04 100644 |
| 8026 | --- a/sysdeps/x86_64/fpu/multiarch/e_powf.c |
| 8027 | +++ b/sysdeps/x86_64/fpu/multiarch/e_powf.c |
| 8028 | @@ -16,31 +16,34 @@ |
| 8029 | License along with the GNU C Library; if not, see |
| 8030 | <https://www.gnu.org/licenses/>. */ |
| 8031 | |
| 8032 | -#include <libm-alias-float.h> |
| 8033 | -#include <libm-alias-finite.h> |
| 8034 | +#include <sysdeps/x86/isa-level.h> |
| 8035 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 8036 | +# include <libm-alias-float.h> |
| 8037 | +# include <libm-alias-finite.h> |
| 8038 | |
| 8039 | -#define powf __redirect_powf |
| 8040 | -#define __DECL_SIMD___redirect_powf |
| 8041 | -#include <math.h> |
| 8042 | -#undef powf |
| 8043 | +# define powf __redirect_powf |
| 8044 | +# define __DECL_SIMD___redirect_powf |
| 8045 | +# include <math.h> |
| 8046 | +# undef powf |
| 8047 | |
| 8048 | -#define SYMBOL_NAME powf |
| 8049 | -#include "ifunc-fma.h" |
| 8050 | +# define SYMBOL_NAME powf |
| 8051 | +# include "ifunc-fma.h" |
| 8052 | |
| 8053 | libc_ifunc_redirected (__redirect_powf, __powf, IFUNC_SELECTOR ()); |
| 8054 | |
| 8055 | -#ifdef SHARED |
| 8056 | +# ifdef SHARED |
| 8057 | __hidden_ver1 (__powf, __GI___powf, __redirect_powf) |
| 8058 | __attribute__ ((visibility ("hidden"))); |
| 8059 | |
| 8060 | versioned_symbol (libm, __ieee754_powf, powf, GLIBC_2_27); |
| 8061 | libm_alias_float_other (__pow, pow) |
| 8062 | -#else |
| 8063 | +# else |
| 8064 | libm_alias_float (__pow, pow) |
| 8065 | -#endif |
| 8066 | +# endif |
| 8067 | |
| 8068 | strong_alias (__powf, __ieee754_powf) |
| 8069 | libm_alias_finite (__powf, __powf) |
| 8070 | |
| 8071 | -#define __powf __powf_sse2 |
| 8072 | +# define __powf __powf_sse2 |
| 8073 | +#endif |
| 8074 | #include <sysdeps/ieee754/flt-32/e_powf.c> |
| 8075 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_atan.c b/sysdeps/x86_64/fpu/multiarch/s_atan.c |
| 8076 | index 71bad096a9..4d2c6ce006 100644 |
| 8077 | --- a/sysdeps/x86_64/fpu/multiarch/s_atan.c |
| 8078 | +++ b/sysdeps/x86_64/fpu/multiarch/s_atan.c |
| 8079 | @@ -16,15 +16,18 @@ |
| 8080 | License along with the GNU C Library; if not, see |
| 8081 | <https://www.gnu.org/licenses/>. */ |
| 8082 | |
| 8083 | -#include <libm-alias-double.h> |
| 8084 | +#include <sysdeps/x86/isa-level.h> |
| 8085 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 8086 | +# include <libm-alias-double.h> |
| 8087 | |
| 8088 | extern double __redirect_atan (double); |
| 8089 | |
| 8090 | -#define SYMBOL_NAME atan |
| 8091 | -#include "ifunc-avx-fma4.h" |
| 8092 | +# define SYMBOL_NAME atan |
| 8093 | +# include "ifunc-avx-fma4.h" |
| 8094 | |
| 8095 | libc_ifunc_redirected (__redirect_atan, __atan, IFUNC_SELECTOR ()); |
| 8096 | libm_alias_double (__atan, atan) |
| 8097 | |
| 8098 | -#define __atan __atan_sse2 |
| 8099 | +# define __atan __atan_sse2 |
| 8100 | +#endif |
| 8101 | #include <sysdeps/ieee754/dbl-64/s_atan.c> |
| 8102 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_ceil-avx.S b/sysdeps/x86_64/fpu/multiarch/s_ceil-avx.S |
| 8103 | new file mode 100644 |
| 8104 | index 0000000000..e6c1106753 |
| 8105 | --- /dev/null |
| 8106 | +++ b/sysdeps/x86_64/fpu/multiarch/s_ceil-avx.S |
| 8107 | @@ -0,0 +1,28 @@ |
| 8108 | +/* AVX implementation of ceil function. |
| 8109 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8110 | + This file is part of the GNU C Library. |
| 8111 | + |
| 8112 | + The GNU C Library is free software; you can redistribute it and/or |
| 8113 | + modify it under the terms of the GNU Lesser General Public |
| 8114 | + License as published by the Free Software Foundation; either |
| 8115 | + version 2.1 of the License, or (at your option) any later version. |
| 8116 | + |
| 8117 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8118 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8119 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8120 | + Lesser General Public License for more details. |
| 8121 | + |
| 8122 | + You should have received a copy of the GNU Lesser General Public |
| 8123 | + License along with the GNU C Library; if not, see |
| 8124 | + <https://www.gnu.org/licenses/>. */ |
| 8125 | + |
| 8126 | +#include <sysdep.h> |
| 8127 | +#include <libm-alias-double.h> |
| 8128 | + |
| 8129 | + .text |
| 8130 | +ENTRY(__ceil) |
| 8131 | + vroundsd $10, %xmm0, %xmm0, %xmm0 |
| 8132 | + ret |
| 8133 | +END(__ceil) |
| 8134 | + |
| 8135 | +libm_alias_double (__ceil, ceil) |
| 8136 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_ceil-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_ceil-sse4_1.S |
| 8137 | index 64119011ad..dba756c38f 100644 |
| 8138 | --- a/sysdeps/x86_64/fpu/multiarch/s_ceil-sse4_1.S |
| 8139 | +++ b/sysdeps/x86_64/fpu/multiarch/s_ceil-sse4_1.S |
| 8140 | @@ -17,8 +17,20 @@ |
| 8141 | |
| 8142 | #include <sysdep.h> |
| 8143 | |
| 8144 | +#include <sysdeps/x86/isa-level.h> |
| 8145 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8146 | +# include <libm-alias-double.h> |
| 8147 | +# define __ceil_sse41 __ceil |
| 8148 | + .text |
| 8149 | +#else |
| 8150 | .section .text.sse4.1,"ax",@progbits |
| 8151 | +#endif |
| 8152 | + |
| 8153 | ENTRY(__ceil_sse41) |
| 8154 | roundsd $10, %xmm0, %xmm0 |
| 8155 | ret |
| 8156 | END(__ceil_sse41) |
| 8157 | + |
| 8158 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8159 | +libm_alias_double (__ceil, ceil) |
| 8160 | +#endif |
| 8161 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_ceil.c b/sysdeps/x86_64/fpu/multiarch/s_ceil.c |
| 8162 | index cc028addee..46c8e91e19 100644 |
| 8163 | --- a/sysdeps/x86_64/fpu/multiarch/s_ceil.c |
| 8164 | +++ b/sysdeps/x86_64/fpu/multiarch/s_ceil.c |
| 8165 | @@ -16,17 +16,20 @@ |
| 8166 | License along with the GNU C Library; if not, see |
| 8167 | <https://www.gnu.org/licenses/>. */ |
| 8168 | |
| 8169 | -#define NO_MATH_REDIRECT |
| 8170 | -#include <libm-alias-double.h> |
| 8171 | +#include <sysdeps/x86/isa-level.h> |
| 8172 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8173 | +# define NO_MATH_REDIRECT |
| 8174 | +# include <libm-alias-double.h> |
| 8175 | |
| 8176 | -#define ceil __redirect_ceil |
| 8177 | -#define __ceil __redirect___ceil |
| 8178 | -#include <math.h> |
| 8179 | -#undef ceil |
| 8180 | -#undef __ceil |
| 8181 | +# define ceil __redirect_ceil |
| 8182 | +# define __ceil __redirect___ceil |
| 8183 | +# include <math.h> |
| 8184 | +# undef ceil |
| 8185 | +# undef __ceil |
| 8186 | |
| 8187 | -#define SYMBOL_NAME ceil |
| 8188 | -#include "ifunc-sse4_1.h" |
| 8189 | +# define SYMBOL_NAME ceil |
| 8190 | +# include "ifunc-sse4_1.h" |
| 8191 | |
| 8192 | libc_ifunc_redirected (__redirect_ceil, __ceil, IFUNC_SELECTOR ()); |
| 8193 | libm_alias_double (__ceil, ceil) |
| 8194 | +#endif |
| 8195 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_ceilf-avx.S b/sysdeps/x86_64/fpu/multiarch/s_ceilf-avx.S |
| 8196 | new file mode 100644 |
| 8197 | index 0000000000..b4d8ac0455 |
| 8198 | --- /dev/null |
| 8199 | +++ b/sysdeps/x86_64/fpu/multiarch/s_ceilf-avx.S |
| 8200 | @@ -0,0 +1,28 @@ |
| 8201 | +/* AVX implementation of ceilf function. |
| 8202 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8203 | + This file is part of the GNU C Library. |
| 8204 | + |
| 8205 | + The GNU C Library is free software; you can redistribute it and/or |
| 8206 | + modify it under the terms of the GNU Lesser General Public |
| 8207 | + License as published by the Free Software Foundation; either |
| 8208 | + version 2.1 of the License, or (at your option) any later version. |
| 8209 | + |
| 8210 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8211 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8212 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8213 | + Lesser General Public License for more details. |
| 8214 | + |
| 8215 | + You should have received a copy of the GNU Lesser General Public |
| 8216 | + License along with the GNU C Library; if not, see |
| 8217 | + <https://www.gnu.org/licenses/>. */ |
| 8218 | + |
| 8219 | +#include <sysdep.h> |
| 8220 | +#include <libm-alias-float.h> |
| 8221 | + |
| 8222 | + .text |
| 8223 | +ENTRY(__ceilf) |
| 8224 | + vroundss $10, %xmm0, %xmm0, %xmm0 |
| 8225 | + ret |
| 8226 | +END(__ceilf) |
| 8227 | + |
| 8228 | +libm_alias_float (__ceil, ceil) |
| 8229 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_ceilf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_ceilf-sse4_1.S |
| 8230 | index dd9a9f6b71..9abc87b91a 100644 |
| 8231 | --- a/sysdeps/x86_64/fpu/multiarch/s_ceilf-sse4_1.S |
| 8232 | +++ b/sysdeps/x86_64/fpu/multiarch/s_ceilf-sse4_1.S |
| 8233 | @@ -17,8 +17,20 @@ |
| 8234 | |
| 8235 | #include <sysdep.h> |
| 8236 | |
| 8237 | +#include <sysdeps/x86/isa-level.h> |
| 8238 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8239 | +# include <libm-alias-float.h> |
| 8240 | +# define __ceilf_sse41 __ceilf |
| 8241 | + .text |
| 8242 | +#else |
| 8243 | .section .text.sse4.1,"ax",@progbits |
| 8244 | +#endif |
| 8245 | + |
| 8246 | ENTRY(__ceilf_sse41) |
| 8247 | roundss $10, %xmm0, %xmm0 |
| 8248 | ret |
| 8249 | END(__ceilf_sse41) |
| 8250 | + |
| 8251 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8252 | +libm_alias_float (__ceil, ceil) |
| 8253 | +#endif |
| 8254 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_ceilf.c b/sysdeps/x86_64/fpu/multiarch/s_ceilf.c |
| 8255 | index 97a0ca7d19..bb53108f73 100644 |
| 8256 | --- a/sysdeps/x86_64/fpu/multiarch/s_ceilf.c |
| 8257 | +++ b/sysdeps/x86_64/fpu/multiarch/s_ceilf.c |
| 8258 | @@ -16,17 +16,20 @@ |
| 8259 | License along with the GNU C Library; if not, see |
| 8260 | <https://www.gnu.org/licenses/>. */ |
| 8261 | |
| 8262 | -#define NO_MATH_REDIRECT |
| 8263 | -#include <libm-alias-float.h> |
| 8264 | +#include <sysdeps/x86/isa-level.h> |
| 8265 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8266 | +# define NO_MATH_REDIRECT |
| 8267 | +# include <libm-alias-float.h> |
| 8268 | |
| 8269 | -#define ceilf __redirect_ceilf |
| 8270 | -#define __ceilf __redirect___ceilf |
| 8271 | -#include <math.h> |
| 8272 | -#undef ceilf |
| 8273 | -#undef __ceilf |
| 8274 | +# define ceilf __redirect_ceilf |
| 8275 | +# define __ceilf __redirect___ceilf |
| 8276 | +# include <math.h> |
| 8277 | +# undef ceilf |
| 8278 | +# undef __ceilf |
| 8279 | |
| 8280 | -#define SYMBOL_NAME ceilf |
| 8281 | -#include "ifunc-sse4_1.h" |
| 8282 | +# define SYMBOL_NAME ceilf |
| 8283 | +# include "ifunc-sse4_1.h" |
| 8284 | |
| 8285 | libc_ifunc_redirected (__redirect_ceilf, __ceilf, IFUNC_SELECTOR ()); |
| 8286 | libm_alias_float (__ceil, ceil) |
| 8287 | +#endif |
| 8288 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_cosf.c b/sysdeps/x86_64/fpu/multiarch/s_cosf.c |
| 8289 | index 2703c576df..8a02e04538 100644 |
| 8290 | --- a/sysdeps/x86_64/fpu/multiarch/s_cosf.c |
| 8291 | +++ b/sysdeps/x86_64/fpu/multiarch/s_cosf.c |
| 8292 | @@ -16,13 +16,18 @@ |
| 8293 | License along with the GNU C Library; if not, see |
| 8294 | <https://www.gnu.org/licenses/>. */ |
| 8295 | |
| 8296 | -#include <libm-alias-float.h> |
| 8297 | +#include <sysdeps/x86/isa-level.h> |
| 8298 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 8299 | +# include <libm-alias-float.h> |
| 8300 | |
| 8301 | extern float __redirect_cosf (float); |
| 8302 | |
| 8303 | -#define SYMBOL_NAME cosf |
| 8304 | -#include "ifunc-fma.h" |
| 8305 | +# define SYMBOL_NAME cosf |
| 8306 | +# include "ifunc-fma.h" |
| 8307 | |
| 8308 | libc_ifunc_redirected (__redirect_cosf, __cosf, IFUNC_SELECTOR ()); |
| 8309 | |
| 8310 | libm_alias_float (__cos, cos) |
| 8311 | +#else |
| 8312 | +# include <sysdeps/ieee754/flt-32/s_cosf.c> |
| 8313 | +#endif |
| 8314 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_expm1.c b/sysdeps/x86_64/fpu/multiarch/s_expm1.c |
| 8315 | index 8a2d69f9b2..d58ef3d8f5 100644 |
| 8316 | --- a/sysdeps/x86_64/fpu/multiarch/s_expm1.c |
| 8317 | +++ b/sysdeps/x86_64/fpu/multiarch/s_expm1.c |
| 8318 | @@ -16,21 +16,24 @@ |
| 8319 | License along with the GNU C Library; if not, see |
| 8320 | <https://www.gnu.org/licenses/>. */ |
| 8321 | |
| 8322 | -#include <libm-alias-double.h> |
| 8323 | +#include <sysdeps/x86/isa-level.h> |
| 8324 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 8325 | +# include <libm-alias-double.h> |
| 8326 | |
| 8327 | extern double __redirect_expm1 (double); |
| 8328 | |
| 8329 | -#define SYMBOL_NAME expm1 |
| 8330 | -#include "ifunc-fma.h" |
| 8331 | +# define SYMBOL_NAME expm1 |
| 8332 | +# include "ifunc-fma.h" |
| 8333 | |
| 8334 | libc_ifunc_redirected (__redirect_expm1, __expm1, IFUNC_SELECTOR ()); |
| 8335 | libm_alias_double (__expm1, expm1) |
| 8336 | |
| 8337 | -#define __expm1 __expm1_sse2 |
| 8338 | +# define __expm1 __expm1_sse2 |
| 8339 | |
| 8340 | /* NB: __expm1 may be expanded to __expm1_sse2 in the following |
| 8341 | prototypes. */ |
| 8342 | extern long double __expm1l (long double); |
| 8343 | extern long double __expm1f128 (long double); |
| 8344 | |
| 8345 | +#endif |
| 8346 | #include <sysdeps/ieee754/dbl-64/s_expm1.c> |
| 8347 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_floor-avx.S b/sysdeps/x86_64/fpu/multiarch/s_floor-avx.S |
| 8348 | new file mode 100644 |
| 8349 | index 0000000000..ff74b5a8bf |
| 8350 | --- /dev/null |
| 8351 | +++ b/sysdeps/x86_64/fpu/multiarch/s_floor-avx.S |
| 8352 | @@ -0,0 +1,28 @@ |
| 8353 | +/* AVX implementation of floor function. |
| 8354 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8355 | + This file is part of the GNU C Library. |
| 8356 | + |
| 8357 | + The GNU C Library is free software; you can redistribute it and/or |
| 8358 | + modify it under the terms of the GNU Lesser General Public |
| 8359 | + License as published by the Free Software Foundation; either |
| 8360 | + version 2.1 of the License, or (at your option) any later version. |
| 8361 | + |
| 8362 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8363 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8364 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8365 | + Lesser General Public License for more details. |
| 8366 | + |
| 8367 | + You should have received a copy of the GNU Lesser General Public |
| 8368 | + License along with the GNU C Library; if not, see |
| 8369 | + <https://www.gnu.org/licenses/>. */ |
| 8370 | + |
| 8371 | +#include <sysdep.h> |
| 8372 | +#include <libm-alias-double.h> |
| 8373 | + |
| 8374 | + .text |
| 8375 | +ENTRY(__floor) |
| 8376 | + vroundsd $9, %xmm0, %xmm0, %xmm0 |
| 8377 | + ret |
| 8378 | +END(__floor) |
| 8379 | + |
| 8380 | +libm_alias_double (__floor, floor) |
| 8381 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_floor-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_floor-sse4_1.S |
| 8382 | index 2f7521f39f..c9b9b0639b 100644 |
| 8383 | --- a/sysdeps/x86_64/fpu/multiarch/s_floor-sse4_1.S |
| 8384 | +++ b/sysdeps/x86_64/fpu/multiarch/s_floor-sse4_1.S |
| 8385 | @@ -17,8 +17,20 @@ |
| 8386 | |
| 8387 | #include <sysdep.h> |
| 8388 | |
| 8389 | +#include <sysdeps/x86/isa-level.h> |
| 8390 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8391 | +# include <libm-alias-double.h> |
| 8392 | +# define __floor_sse41 __floor |
| 8393 | + .text |
| 8394 | +#else |
| 8395 | .section .text.sse4.1,"ax",@progbits |
| 8396 | +#endif |
| 8397 | + |
| 8398 | ENTRY(__floor_sse41) |
| 8399 | roundsd $9, %xmm0, %xmm0 |
| 8400 | ret |
| 8401 | END(__floor_sse41) |
| 8402 | + |
| 8403 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8404 | +libm_alias_double (__floor, floor) |
| 8405 | +#endif |
| 8406 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_floor.c b/sysdeps/x86_64/fpu/multiarch/s_floor.c |
| 8407 | index 8cebd48e10..2c87dd0056 100644 |
| 8408 | --- a/sysdeps/x86_64/fpu/multiarch/s_floor.c |
| 8409 | +++ b/sysdeps/x86_64/fpu/multiarch/s_floor.c |
| 8410 | @@ -16,17 +16,20 @@ |
| 8411 | License along with the GNU C Library; if not, see |
| 8412 | <https://www.gnu.org/licenses/>. */ |
| 8413 | |
| 8414 | -#define NO_MATH_REDIRECT |
| 8415 | -#include <libm-alias-double.h> |
| 8416 | +#include <sysdeps/x86/isa-level.h> |
| 8417 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8418 | +# define NO_MATH_REDIRECT |
| 8419 | +# include <libm-alias-double.h> |
| 8420 | |
| 8421 | -#define floor __redirect_floor |
| 8422 | -#define __floor __redirect___floor |
| 8423 | -#include <math.h> |
| 8424 | -#undef floor |
| 8425 | -#undef __floor |
| 8426 | +# define floor __redirect_floor |
| 8427 | +# define __floor __redirect___floor |
| 8428 | +# include <math.h> |
| 8429 | +# undef floor |
| 8430 | +# undef __floor |
| 8431 | |
| 8432 | -#define SYMBOL_NAME floor |
| 8433 | -#include "ifunc-sse4_1.h" |
| 8434 | +# define SYMBOL_NAME floor |
| 8435 | +# include "ifunc-sse4_1.h" |
| 8436 | |
| 8437 | libc_ifunc_redirected (__redirect_floor, __floor, IFUNC_SELECTOR ()); |
| 8438 | libm_alias_double (__floor, floor) |
| 8439 | +#endif |
| 8440 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_floorf-avx.S b/sysdeps/x86_64/fpu/multiarch/s_floorf-avx.S |
| 8441 | new file mode 100644 |
| 8442 | index 0000000000..c378baae8e |
| 8443 | --- /dev/null |
| 8444 | +++ b/sysdeps/x86_64/fpu/multiarch/s_floorf-avx.S |
| 8445 | @@ -0,0 +1,28 @@ |
| 8446 | +/* AVX implementation of floorf function. |
| 8447 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8448 | + This file is part of the GNU C Library. |
| 8449 | + |
| 8450 | + The GNU C Library is free software; you can redistribute it and/or |
| 8451 | + modify it under the terms of the GNU Lesser General Public |
| 8452 | + License as published by the Free Software Foundation; either |
| 8453 | + version 2.1 of the License, or (at your option) any later version. |
| 8454 | + |
| 8455 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8456 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8457 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8458 | + Lesser General Public License for more details. |
| 8459 | + |
| 8460 | + You should have received a copy of the GNU Lesser General Public |
| 8461 | + License along with the GNU C Library; if not, see |
| 8462 | + <https://www.gnu.org/licenses/>. */ |
| 8463 | + |
| 8464 | +#include <sysdep.h> |
| 8465 | +#include <libm-alias-float.h> |
| 8466 | + |
| 8467 | + .text |
| 8468 | +ENTRY(__floorf) |
| 8469 | + vroundss $9, %xmm0, %xmm0, %xmm0 |
| 8470 | + ret |
| 8471 | +END(__floorf) |
| 8472 | + |
| 8473 | +libm_alias_float (__floor, floor) |
| 8474 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_floorf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_floorf-sse4_1.S |
| 8475 | index 5f6020d27d..c2216899db 100644 |
| 8476 | --- a/sysdeps/x86_64/fpu/multiarch/s_floorf-sse4_1.S |
| 8477 | +++ b/sysdeps/x86_64/fpu/multiarch/s_floorf-sse4_1.S |
| 8478 | @@ -17,8 +17,20 @@ |
| 8479 | |
| 8480 | #include <sysdep.h> |
| 8481 | |
| 8482 | +#include <sysdeps/x86/isa-level.h> |
| 8483 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8484 | +# include <libm-alias-float.h> |
| 8485 | +# define __floorf_sse41 __floorf |
| 8486 | + .text |
| 8487 | +#else |
| 8488 | .section .text.sse4.1,"ax",@progbits |
| 8489 | +#endif |
| 8490 | + |
| 8491 | ENTRY(__floorf_sse41) |
| 8492 | roundss $9, %xmm0, %xmm0 |
| 8493 | ret |
| 8494 | END(__floorf_sse41) |
| 8495 | + |
| 8496 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8497 | +libm_alias_float (__floor, floor) |
| 8498 | +#endif |
| 8499 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_floorf.c b/sysdeps/x86_64/fpu/multiarch/s_floorf.c |
| 8500 | index a14e18b03c..a277802b6d 100644 |
| 8501 | --- a/sysdeps/x86_64/fpu/multiarch/s_floorf.c |
| 8502 | +++ b/sysdeps/x86_64/fpu/multiarch/s_floorf.c |
| 8503 | @@ -16,17 +16,20 @@ |
| 8504 | License along with the GNU C Library; if not, see |
| 8505 | <https://www.gnu.org/licenses/>. */ |
| 8506 | |
| 8507 | -#define NO_MATH_REDIRECT |
| 8508 | -#include <libm-alias-float.h> |
| 8509 | +#include <sysdeps/x86/isa-level.h> |
| 8510 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8511 | +# define NO_MATH_REDIRECT |
| 8512 | +# include <libm-alias-float.h> |
| 8513 | |
| 8514 | -#define floorf __redirect_floorf |
| 8515 | -#define __floorf __redirect___floorf |
| 8516 | -#include <math.h> |
| 8517 | -#undef floorf |
| 8518 | -#undef __floorf |
| 8519 | +# define floorf __redirect_floorf |
| 8520 | +# define __floorf __redirect___floorf |
| 8521 | +# include <math.h> |
| 8522 | +# undef floorf |
| 8523 | +# undef __floorf |
| 8524 | |
| 8525 | -#define SYMBOL_NAME floorf |
| 8526 | -#include "ifunc-sse4_1.h" |
| 8527 | +# define SYMBOL_NAME floorf |
| 8528 | +# include "ifunc-sse4_1.h" |
| 8529 | |
| 8530 | libc_ifunc_redirected (__redirect_floorf, __floorf, IFUNC_SELECTOR ()); |
| 8531 | libm_alias_float (__floor, floor) |
| 8532 | +#endif |
| 8533 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_log1p.c b/sysdeps/x86_64/fpu/multiarch/s_log1p.c |
| 8534 | index a8e1a3f21b..3fa1185d81 100644 |
| 8535 | --- a/sysdeps/x86_64/fpu/multiarch/s_log1p.c |
| 8536 | +++ b/sysdeps/x86_64/fpu/multiarch/s_log1p.c |
| 8537 | @@ -16,14 +16,17 @@ |
| 8538 | License along with the GNU C Library; if not, see |
| 8539 | <https://www.gnu.org/licenses/>. */ |
| 8540 | |
| 8541 | -#include <libm-alias-double.h> |
| 8542 | +#include <sysdeps/x86/isa-level.h> |
| 8543 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 8544 | +# include <libm-alias-double.h> |
| 8545 | |
| 8546 | extern double __redirect_log1p (double); |
| 8547 | |
| 8548 | -#define SYMBOL_NAME log1p |
| 8549 | -#include "ifunc-fma.h" |
| 8550 | +# define SYMBOL_NAME log1p |
| 8551 | +# include "ifunc-fma.h" |
| 8552 | |
| 8553 | libc_ifunc_redirected (__redirect_log1p, __log1p, IFUNC_SELECTOR ()); |
| 8554 | |
| 8555 | -#define __log1p __log1p_sse2 |
| 8556 | +# define __log1p __log1p_sse2 |
| 8557 | +#endif |
| 8558 | #include <sysdeps/ieee754/dbl-64/s_log1p.c> |
| 8559 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_nearbyint-avx.S b/sysdeps/x86_64/fpu/multiarch/s_nearbyint-avx.S |
| 8560 | new file mode 100644 |
| 8561 | index 0000000000..5bfdf73c28 |
| 8562 | --- /dev/null |
| 8563 | +++ b/sysdeps/x86_64/fpu/multiarch/s_nearbyint-avx.S |
| 8564 | @@ -0,0 +1,28 @@ |
| 8565 | +/* AVX implementation of nearbyint function. |
| 8566 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8567 | + This file is part of the GNU C Library. |
| 8568 | + |
| 8569 | + The GNU C Library is free software; you can redistribute it and/or |
| 8570 | + modify it under the terms of the GNU Lesser General Public |
| 8571 | + License as published by the Free Software Foundation; either |
| 8572 | + version 2.1 of the License, or (at your option) any later version. |
| 8573 | + |
| 8574 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8575 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8576 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8577 | + Lesser General Public License for more details. |
| 8578 | + |
| 8579 | + You should have received a copy of the GNU Lesser General Public |
| 8580 | + License along with the GNU C Library; if not, see |
| 8581 | + <https://www.gnu.org/licenses/>. */ |
| 8582 | + |
| 8583 | +#include <sysdep.h> |
| 8584 | +#include <libm-alias-double.h> |
| 8585 | + |
| 8586 | + .text |
| 8587 | +ENTRY(__nearbyint) |
| 8588 | + vroundsd $0xc, %xmm0, %xmm0, %xmm0 |
| 8589 | + ret |
| 8590 | +END(__nearbyint) |
| 8591 | + |
| 8592 | +libm_alias_double (__nearbyint, nearbyint) |
| 8593 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_nearbyint-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_nearbyint-sse4_1.S |
| 8594 | index 674f7eb40a..9d84410a1f 100644 |
| 8595 | --- a/sysdeps/x86_64/fpu/multiarch/s_nearbyint-sse4_1.S |
| 8596 | +++ b/sysdeps/x86_64/fpu/multiarch/s_nearbyint-sse4_1.S |
| 8597 | @@ -17,8 +17,20 @@ |
| 8598 | |
| 8599 | #include <sysdep.h> |
| 8600 | |
| 8601 | +#include <sysdeps/x86/isa-level.h> |
| 8602 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8603 | +# include <libm-alias-double.h> |
| 8604 | +# define __nearbyint_sse41 __nearbyint |
| 8605 | + .text |
| 8606 | +#else |
| 8607 | .section .text.sse4.1,"ax",@progbits |
| 8608 | +#endif |
| 8609 | + |
| 8610 | ENTRY(__nearbyint_sse41) |
| 8611 | roundsd $0xc, %xmm0, %xmm0 |
| 8612 | ret |
| 8613 | END(__nearbyint_sse41) |
| 8614 | + |
| 8615 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8616 | +libm_alias_double (__nearbyint, nearbyint) |
| 8617 | +#endif |
| 8618 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_nearbyint.c b/sysdeps/x86_64/fpu/multiarch/s_nearbyint.c |
| 8619 | index 693e42dd4e..057a7ca60f 100644 |
| 8620 | --- a/sysdeps/x86_64/fpu/multiarch/s_nearbyint.c |
| 8621 | +++ b/sysdeps/x86_64/fpu/multiarch/s_nearbyint.c |
| 8622 | @@ -16,17 +16,20 @@ |
| 8623 | License along with the GNU C Library; if not, see |
| 8624 | <https://www.gnu.org/licenses/>. */ |
| 8625 | |
| 8626 | -#include <libm-alias-double.h> |
| 8627 | +#include <sysdeps/x86/isa-level.h> |
| 8628 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8629 | +# include <libm-alias-double.h> |
| 8630 | |
| 8631 | -#define nearbyint __redirect_nearbyint |
| 8632 | -#define __nearbyint __redirect___nearbyint |
| 8633 | -#include <math.h> |
| 8634 | -#undef nearbyint |
| 8635 | -#undef __nearbyint |
| 8636 | +# define nearbyint __redirect_nearbyint |
| 8637 | +# define __nearbyint __redirect___nearbyint |
| 8638 | +# include <math.h> |
| 8639 | +# undef nearbyint |
| 8640 | +# undef __nearbyint |
| 8641 | |
| 8642 | -#define SYMBOL_NAME nearbyint |
| 8643 | -#include "ifunc-sse4_1.h" |
| 8644 | +# define SYMBOL_NAME nearbyint |
| 8645 | +# include "ifunc-sse4_1.h" |
| 8646 | |
| 8647 | libc_ifunc_redirected (__redirect_nearbyint, __nearbyint, |
| 8648 | IFUNC_SELECTOR ()); |
| 8649 | libm_alias_double (__nearbyint, nearbyint) |
| 8650 | +#endif |
| 8651 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-avx.S b/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-avx.S |
| 8652 | new file mode 100644 |
| 8653 | index 0000000000..1dbaed0324 |
| 8654 | --- /dev/null |
| 8655 | +++ b/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-avx.S |
| 8656 | @@ -0,0 +1,28 @@ |
| 8657 | +/* AVX implmentation of nearbyintf function. |
| 8658 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8659 | + This file is part of the GNU C Library. |
| 8660 | + |
| 8661 | + The GNU C Library is free software; you can redistribute it and/or |
| 8662 | + modify it under the terms of the GNU Lesser General Public |
| 8663 | + License as published by the Free Software Foundation; either |
| 8664 | + version 2.1 of the License, or (at your option) any later version. |
| 8665 | + |
| 8666 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8667 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8668 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8669 | + Lesser General Public License for more details. |
| 8670 | + |
| 8671 | + You should have received a copy of the GNU Lesser General Public |
| 8672 | + License along with the GNU C Library; if not, see |
| 8673 | + <https://www.gnu.org/licenses/>. */ |
| 8674 | + |
| 8675 | +#include <sysdep.h> |
| 8676 | +#include <libm-alias-float.h> |
| 8677 | + |
| 8678 | + .text |
| 8679 | +ENTRY(__nearbyintf) |
| 8680 | + vroundss $0xc, %xmm0, %xmm0, %xmm0 |
| 8681 | + ret |
| 8682 | +END(__nearbyintf) |
| 8683 | + |
| 8684 | +libm_alias_float (__nearbyint, nearbyint) |
| 8685 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-sse4_1.S |
| 8686 | index 5892bd7563..3cf35f92d6 100644 |
| 8687 | --- a/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-sse4_1.S |
| 8688 | +++ b/sysdeps/x86_64/fpu/multiarch/s_nearbyintf-sse4_1.S |
| 8689 | @@ -17,8 +17,20 @@ |
| 8690 | |
| 8691 | #include <sysdep.h> |
| 8692 | |
| 8693 | +#include <sysdeps/x86/isa-level.h> |
| 8694 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8695 | +# include <libm-alias-float.h> |
| 8696 | +# define __nearbyintf_sse41 __nearbyintf |
| 8697 | + .text |
| 8698 | +#else |
| 8699 | .section .text.sse4.1,"ax",@progbits |
| 8700 | +#endif |
| 8701 | + |
| 8702 | ENTRY(__nearbyintf_sse41) |
| 8703 | roundss $0xc, %xmm0, %xmm0 |
| 8704 | ret |
| 8705 | END(__nearbyintf_sse41) |
| 8706 | + |
| 8707 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8708 | +libm_alias_float (__nearbyint, nearbyint) |
| 8709 | +#endif |
| 8710 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_nearbyintf.c b/sysdeps/x86_64/fpu/multiarch/s_nearbyintf.c |
| 8711 | index a0ac009f4b..41f374ba72 100644 |
| 8712 | --- a/sysdeps/x86_64/fpu/multiarch/s_nearbyintf.c |
| 8713 | +++ b/sysdeps/x86_64/fpu/multiarch/s_nearbyintf.c |
| 8714 | @@ -16,17 +16,20 @@ |
| 8715 | License along with the GNU C Library; if not, see |
| 8716 | <https://www.gnu.org/licenses/>. */ |
| 8717 | |
| 8718 | -#include <libm-alias-float.h> |
| 8719 | +#include <sysdeps/x86/isa-level.h> |
| 8720 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8721 | +# include <libm-alias-float.h> |
| 8722 | |
| 8723 | -#define nearbyintf __redirect_nearbyintf |
| 8724 | -#define __nearbyintf __redirect___nearbyintf |
| 8725 | -#include <math.h> |
| 8726 | -#undef nearbyintf |
| 8727 | -#undef __nearbyintf |
| 8728 | +# define nearbyintf __redirect_nearbyintf |
| 8729 | +# define __nearbyintf __redirect___nearbyintf |
| 8730 | +# include <math.h> |
| 8731 | +# undef nearbyintf |
| 8732 | +# undef __nearbyintf |
| 8733 | |
| 8734 | -#define SYMBOL_NAME nearbyintf |
| 8735 | -#include "ifunc-sse4_1.h" |
| 8736 | +# define SYMBOL_NAME nearbyintf |
| 8737 | +# include "ifunc-sse4_1.h" |
| 8738 | |
| 8739 | libc_ifunc_redirected (__redirect_nearbyintf, __nearbyintf, |
| 8740 | IFUNC_SELECTOR ()); |
| 8741 | libm_alias_float (__nearbyint, nearbyint) |
| 8742 | +#endif |
| 8743 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_rint-avx.S b/sysdeps/x86_64/fpu/multiarch/s_rint-avx.S |
| 8744 | new file mode 100644 |
| 8745 | index 0000000000..2b403b331f |
| 8746 | --- /dev/null |
| 8747 | +++ b/sysdeps/x86_64/fpu/multiarch/s_rint-avx.S |
| 8748 | @@ -0,0 +1,28 @@ |
| 8749 | +/* AVX implementation of rint function. |
| 8750 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8751 | + This file is part of the GNU C Library. |
| 8752 | + |
| 8753 | + The GNU C Library is free software; you can redistribute it and/or |
| 8754 | + modify it under the terms of the GNU Lesser General Public |
| 8755 | + License as published by the Free Software Foundation; either |
| 8756 | + version 2.1 of the License, or (at your option) any later version. |
| 8757 | + |
| 8758 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8759 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8760 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8761 | + Lesser General Public License for more details. |
| 8762 | + |
| 8763 | + You should have received a copy of the GNU Lesser General Public |
| 8764 | + License along with the GNU C Library; if not, see |
| 8765 | + <https://www.gnu.org/licenses/>. */ |
| 8766 | + |
| 8767 | +#include <sysdep.h> |
| 8768 | +#include <libm-alias-double.h> |
| 8769 | + |
| 8770 | + .text |
| 8771 | +ENTRY(__rint) |
| 8772 | + vroundsd $4, %xmm0, %xmm0, %xmm0 |
| 8773 | + ret |
| 8774 | +END(__rint) |
| 8775 | + |
| 8776 | +libm_alias_double (__rint, rint) |
| 8777 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_rint-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_rint-sse4_1.S |
| 8778 | index 405372991b..8cd9cf759f 100644 |
| 8779 | --- a/sysdeps/x86_64/fpu/multiarch/s_rint-sse4_1.S |
| 8780 | +++ b/sysdeps/x86_64/fpu/multiarch/s_rint-sse4_1.S |
| 8781 | @@ -17,8 +17,20 @@ |
| 8782 | |
| 8783 | #include <sysdep.h> |
| 8784 | |
| 8785 | +#include <sysdeps/x86/isa-level.h> |
| 8786 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8787 | +# include <libm-alias-double.h> |
| 8788 | +# define __rint_sse41 __rint |
| 8789 | + .text |
| 8790 | +#else |
| 8791 | .section .text.sse4.1,"ax",@progbits |
| 8792 | +#endif |
| 8793 | + |
| 8794 | ENTRY(__rint_sse41) |
| 8795 | roundsd $4, %xmm0, %xmm0 |
| 8796 | ret |
| 8797 | END(__rint_sse41) |
| 8798 | + |
| 8799 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8800 | +libm_alias_double (__rint, rint) |
| 8801 | +#endif |
| 8802 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_rint.c b/sysdeps/x86_64/fpu/multiarch/s_rint.c |
| 8803 | index 754c87e004..18623b7d99 100644 |
| 8804 | --- a/sysdeps/x86_64/fpu/multiarch/s_rint.c |
| 8805 | +++ b/sysdeps/x86_64/fpu/multiarch/s_rint.c |
| 8806 | @@ -16,17 +16,20 @@ |
| 8807 | License along with the GNU C Library; if not, see |
| 8808 | <https://www.gnu.org/licenses/>. */ |
| 8809 | |
| 8810 | -#define NO_MATH_REDIRECT |
| 8811 | -#include <libm-alias-double.h> |
| 8812 | +#include <sysdeps/x86/isa-level.h> |
| 8813 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8814 | +# define NO_MATH_REDIRECT |
| 8815 | +# include <libm-alias-double.h> |
| 8816 | |
| 8817 | -#define rint __redirect_rint |
| 8818 | -#define __rint __redirect___rint |
| 8819 | -#include <math.h> |
| 8820 | -#undef rint |
| 8821 | -#undef __rint |
| 8822 | +# define rint __redirect_rint |
| 8823 | +# define __rint __redirect___rint |
| 8824 | +# include <math.h> |
| 8825 | +# undef rint |
| 8826 | +# undef __rint |
| 8827 | |
| 8828 | -#define SYMBOL_NAME rint |
| 8829 | -#include "ifunc-sse4_1.h" |
| 8830 | +# define SYMBOL_NAME rint |
| 8831 | +# include "ifunc-sse4_1.h" |
| 8832 | |
| 8833 | libc_ifunc_redirected (__redirect_rint, __rint, IFUNC_SELECTOR ()); |
| 8834 | libm_alias_double (__rint, rint) |
| 8835 | +#endif |
| 8836 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_rintf-avx.S b/sysdeps/x86_64/fpu/multiarch/s_rintf-avx.S |
| 8837 | new file mode 100644 |
| 8838 | index 0000000000..171c2867f4 |
| 8839 | --- /dev/null |
| 8840 | +++ b/sysdeps/x86_64/fpu/multiarch/s_rintf-avx.S |
| 8841 | @@ -0,0 +1,28 @@ |
| 8842 | +/* AVX implementation of rintf function. |
| 8843 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8844 | + This file is part of the GNU C Library. |
| 8845 | + |
| 8846 | + The GNU C Library is free software; you can redistribute it and/or |
| 8847 | + modify it under the terms of the GNU Lesser General Public |
| 8848 | + License as published by the Free Software Foundation; either |
| 8849 | + version 2.1 of the License, or (at your option) any later version. |
| 8850 | + |
| 8851 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8852 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8853 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8854 | + Lesser General Public License for more details. |
| 8855 | + |
| 8856 | + You should have received a copy of the GNU Lesser General Public |
| 8857 | + License along with the GNU C Library; if not, see |
| 8858 | + <https://www.gnu.org/licenses/>. */ |
| 8859 | + |
| 8860 | +#include <sysdep.h> |
| 8861 | +#include <libm-alias-float.h> |
| 8862 | + |
| 8863 | + .text |
| 8864 | +ENTRY(__rintf) |
| 8865 | + vroundss $4, %xmm0, %xmm0, %xmm0 |
| 8866 | + ret |
| 8867 | +END(__rintf) |
| 8868 | + |
| 8869 | +libm_alias_float (__rint, rint) |
| 8870 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_rintf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_rintf-sse4_1.S |
| 8871 | index 8ac67ce767..fc1e70f0c9 100644 |
| 8872 | --- a/sysdeps/x86_64/fpu/multiarch/s_rintf-sse4_1.S |
| 8873 | +++ b/sysdeps/x86_64/fpu/multiarch/s_rintf-sse4_1.S |
| 8874 | @@ -17,8 +17,20 @@ |
| 8875 | |
| 8876 | #include <sysdep.h> |
| 8877 | |
| 8878 | +#include <sysdeps/x86/isa-level.h> |
| 8879 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8880 | +# include <libm-alias-float.h> |
| 8881 | +# define __rintf_sse41 __rintf |
| 8882 | + .text |
| 8883 | +#else |
| 8884 | .section .text.sse4.1,"ax",@progbits |
| 8885 | +#endif |
| 8886 | + |
| 8887 | ENTRY(__rintf_sse41) |
| 8888 | roundss $4, %xmm0, %xmm0 |
| 8889 | ret |
| 8890 | END(__rintf_sse41) |
| 8891 | + |
| 8892 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8893 | +libm_alias_float (__rint, rint) |
| 8894 | +#endif |
| 8895 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_rintf.c b/sysdeps/x86_64/fpu/multiarch/s_rintf.c |
| 8896 | index e9d6b7a5f2..e275368dec 100644 |
| 8897 | --- a/sysdeps/x86_64/fpu/multiarch/s_rintf.c |
| 8898 | +++ b/sysdeps/x86_64/fpu/multiarch/s_rintf.c |
| 8899 | @@ -16,17 +16,20 @@ |
| 8900 | License along with the GNU C Library; if not, see |
| 8901 | <https://www.gnu.org/licenses/>. */ |
| 8902 | |
| 8903 | -#define NO_MATH_REDIRECT |
| 8904 | -#include <libm-alias-float.h> |
| 8905 | +#include <sysdeps/x86/isa-level.h> |
| 8906 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8907 | +# define NO_MATH_REDIRECT |
| 8908 | +# include <libm-alias-float.h> |
| 8909 | |
| 8910 | -#define rintf __redirect_rintf |
| 8911 | -#define __rintf __redirect___rintf |
| 8912 | -#include <math.h> |
| 8913 | -#undef rintf |
| 8914 | -#undef __rintf |
| 8915 | +# define rintf __redirect_rintf |
| 8916 | +# define __rintf __redirect___rintf |
| 8917 | +# include <math.h> |
| 8918 | +# undef rintf |
| 8919 | +# undef __rintf |
| 8920 | |
| 8921 | -#define SYMBOL_NAME rintf |
| 8922 | -#include "ifunc-sse4_1.h" |
| 8923 | +# define SYMBOL_NAME rintf |
| 8924 | +# include "ifunc-sse4_1.h" |
| 8925 | |
| 8926 | libc_ifunc_redirected (__redirect_rintf, __rintf, IFUNC_SELECTOR ()); |
| 8927 | libm_alias_float (__rint, rint) |
| 8928 | +#endif |
| 8929 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundeven-avx.S b/sysdeps/x86_64/fpu/multiarch/s_roundeven-avx.S |
| 8930 | new file mode 100644 |
| 8931 | index 0000000000..576790355c |
| 8932 | --- /dev/null |
| 8933 | +++ b/sysdeps/x86_64/fpu/multiarch/s_roundeven-avx.S |
| 8934 | @@ -0,0 +1,28 @@ |
| 8935 | +/* AVX implementation of roundeven function. |
| 8936 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 8937 | + This file is part of the GNU C Library. |
| 8938 | + |
| 8939 | + The GNU C Library is free software; you can redistribute it and/or |
| 8940 | + modify it under the terms of the GNU Lesser General Public |
| 8941 | + License as published by the Free Software Foundation; either |
| 8942 | + version 2.1 of the License, or (at your option) any later version. |
| 8943 | + |
| 8944 | + The GNU C Library is distributed in the hope that it will be useful, |
| 8945 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8946 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 8947 | + Lesser General Public License for more details. |
| 8948 | + |
| 8949 | + You should have received a copy of the GNU Lesser General Public |
| 8950 | + License along with the GNU C Library; if not, see |
| 8951 | + <https://www.gnu.org/licenses/>. */ |
| 8952 | + |
| 8953 | +#include <sysdep.h> |
| 8954 | +#include <libm-alias-double.h> |
| 8955 | + |
| 8956 | + .text |
| 8957 | +ENTRY(__roundeven) |
| 8958 | + vroundsd $8, %xmm0, %xmm0, %xmm0 |
| 8959 | + ret |
| 8960 | +END(__roundeven) |
| 8961 | + |
| 8962 | +libm_alias_double (__roundeven, roundeven) |
| 8963 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S |
| 8964 | index 5ef102336b..f00be56c59 100644 |
| 8965 | --- a/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S |
| 8966 | +++ b/sysdeps/x86_64/fpu/multiarch/s_roundeven-sse4_1.S |
| 8967 | @@ -17,8 +17,20 @@ |
| 8968 | |
| 8969 | #include <sysdep.h> |
| 8970 | |
| 8971 | +#include <sysdeps/x86/isa-level.h> |
| 8972 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8973 | +# include <libm-alias-double.h> |
| 8974 | +# define __roundeven_sse41 __roundeven |
| 8975 | + .text |
| 8976 | +#else |
| 8977 | .section .text.sse4.1,"ax",@progbits |
| 8978 | +#endif |
| 8979 | + |
| 8980 | ENTRY(__roundeven_sse41) |
| 8981 | roundsd $8, %xmm0, %xmm0 |
| 8982 | ret |
| 8983 | END(__roundeven_sse41) |
| 8984 | + |
| 8985 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 8986 | +libm_alias_double (__roundeven, roundeven) |
| 8987 | +#endif |
| 8988 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundeven.c b/sysdeps/x86_64/fpu/multiarch/s_roundeven.c |
| 8989 | index 8737b32e26..139aad088f 100644 |
| 8990 | --- a/sysdeps/x86_64/fpu/multiarch/s_roundeven.c |
| 8991 | +++ b/sysdeps/x86_64/fpu/multiarch/s_roundeven.c |
| 8992 | @@ -16,16 +16,19 @@ |
| 8993 | License along with the GNU C Library; if not, see |
| 8994 | <https://www.gnu.org/licenses/>. */ |
| 8995 | |
| 8996 | -#include <libm-alias-double.h> |
| 8997 | +#include <sysdeps/x86/isa-level.h> |
| 8998 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 8999 | +# include <libm-alias-double.h> |
| 9000 | |
| 9001 | -#define roundeven __redirect_roundeven |
| 9002 | -#define __roundeven __redirect___roundeven |
| 9003 | -#include <math.h> |
| 9004 | -#undef roundeven |
| 9005 | -#undef __roundeven |
| 9006 | +# define roundeven __redirect_roundeven |
| 9007 | +# define __roundeven __redirect___roundeven |
| 9008 | +# include <math.h> |
| 9009 | +# undef roundeven |
| 9010 | +# undef __roundeven |
| 9011 | |
| 9012 | -#define SYMBOL_NAME roundeven |
| 9013 | -#include "ifunc-sse4_1.h" |
| 9014 | +# define SYMBOL_NAME roundeven |
| 9015 | +# include "ifunc-sse4_1.h" |
| 9016 | |
| 9017 | libc_ifunc_redirected (__redirect_roundeven, __roundeven, IFUNC_SELECTOR ()); |
| 9018 | libm_alias_double (__roundeven, roundeven) |
| 9019 | +#endif |
| 9020 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundevenf-avx.S b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-avx.S |
| 9021 | new file mode 100644 |
| 9022 | index 0000000000..42c359f4cd |
| 9023 | --- /dev/null |
| 9024 | +++ b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-avx.S |
| 9025 | @@ -0,0 +1,28 @@ |
| 9026 | +/* AVX implementation of roundevenf function. |
| 9027 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 9028 | + This file is part of the GNU C Library. |
| 9029 | + |
| 9030 | + The GNU C Library is free software; you can redistribute it and/or |
| 9031 | + modify it under the terms of the GNU Lesser General Public |
| 9032 | + License as published by the Free Software Foundation; either |
| 9033 | + version 2.1 of the License, or (at your option) any later version. |
| 9034 | + |
| 9035 | + The GNU C Library is distributed in the hope that it will be useful, |
| 9036 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 9037 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 9038 | + Lesser General Public License for more details. |
| 9039 | + |
| 9040 | + You should have received a copy of the GNU Lesser General Public |
| 9041 | + License along with the GNU C Library; if not, see |
| 9042 | + <https://www.gnu.org/licenses/>. */ |
| 9043 | + |
| 9044 | +#include <sysdep.h> |
| 9045 | +#include <libm-alias-float.h> |
| 9046 | + |
| 9047 | + .text |
| 9048 | +ENTRY(__roundevenf) |
| 9049 | + vroundss $8, %xmm0, %xmm0, %xmm0 |
| 9050 | + ret |
| 9051 | +END(__roundevenf) |
| 9052 | + |
| 9053 | +libm_alias_float (__roundeven, roundeven) |
| 9054 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S |
| 9055 | index 792c90ba07..6b148e4353 100644 |
| 9056 | --- a/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S |
| 9057 | +++ b/sysdeps/x86_64/fpu/multiarch/s_roundevenf-sse4_1.S |
| 9058 | @@ -17,8 +17,20 @@ |
| 9059 | |
| 9060 | #include <sysdep.h> |
| 9061 | |
| 9062 | +#include <sysdeps/x86/isa-level.h> |
| 9063 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 9064 | +# include <libm-alias-float.h> |
| 9065 | +# define __roundevenf_sse41 __roundevenf |
| 9066 | + .text |
| 9067 | +#else |
| 9068 | .section .text.sse4.1,"ax",@progbits |
| 9069 | +#endif |
| 9070 | + |
| 9071 | ENTRY(__roundevenf_sse41) |
| 9072 | roundss $8, %xmm0, %xmm0 |
| 9073 | ret |
| 9074 | END(__roundevenf_sse41) |
| 9075 | + |
| 9076 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 9077 | +libm_alias_float (__roundeven, roundeven) |
| 9078 | +#endif |
| 9079 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c b/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c |
| 9080 | index e96016a4d5..2fb090075d 100644 |
| 9081 | --- a/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c |
| 9082 | +++ b/sysdeps/x86_64/fpu/multiarch/s_roundevenf.c |
| 9083 | @@ -16,16 +16,19 @@ |
| 9084 | License along with the GNU C Library; if not, see |
| 9085 | <https://www.gnu.org/licenses/>. */ |
| 9086 | |
| 9087 | -#include <libm-alias-float.h> |
| 9088 | +#include <sysdeps/x86/isa-level.h> |
| 9089 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 9090 | +# include <libm-alias-float.h> |
| 9091 | |
| 9092 | -#define roundevenf __redirect_roundevenf |
| 9093 | -#define __roundevenf __redirect___roundevenf |
| 9094 | -#include <math.h> |
| 9095 | -#undef roundevenf |
| 9096 | -#undef __roundevenf |
| 9097 | +# define roundevenf __redirect_roundevenf |
| 9098 | +# define __roundevenf __redirect___roundevenf |
| 9099 | +# include <math.h> |
| 9100 | +# undef roundevenf |
| 9101 | +# undef __roundevenf |
| 9102 | |
| 9103 | -#define SYMBOL_NAME roundevenf |
| 9104 | -#include "ifunc-sse4_1.h" |
| 9105 | +# define SYMBOL_NAME roundevenf |
| 9106 | +# include "ifunc-sse4_1.h" |
| 9107 | |
| 9108 | libc_ifunc_redirected (__redirect_roundevenf, __roundevenf, IFUNC_SELECTOR ()); |
| 9109 | libm_alias_float (__roundeven, roundeven) |
| 9110 | +#endif |
| 9111 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_sin.c b/sysdeps/x86_64/fpu/multiarch/s_sin.c |
| 9112 | index 355cc0092e..21e77943a3 100644 |
| 9113 | --- a/sysdeps/x86_64/fpu/multiarch/s_sin.c |
| 9114 | +++ b/sysdeps/x86_64/fpu/multiarch/s_sin.c |
| 9115 | @@ -16,24 +16,27 @@ |
| 9116 | License along with the GNU C Library; if not, see |
| 9117 | <https://www.gnu.org/licenses/>. */ |
| 9118 | |
| 9119 | -#include <libm-alias-double.h> |
| 9120 | +#include <sysdeps/x86/isa-level.h> |
| 9121 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 9122 | +# include <libm-alias-double.h> |
| 9123 | |
| 9124 | extern double __redirect_sin (double); |
| 9125 | extern double __redirect_cos (double); |
| 9126 | |
| 9127 | -#define SYMBOL_NAME sin |
| 9128 | -#include "ifunc-avx-fma4.h" |
| 9129 | +# define SYMBOL_NAME sin |
| 9130 | +# include "ifunc-avx-fma4.h" |
| 9131 | |
| 9132 | libc_ifunc_redirected (__redirect_sin, __sin, IFUNC_SELECTOR ()); |
| 9133 | libm_alias_double (__sin, sin) |
| 9134 | |
| 9135 | -#undef SYMBOL_NAME |
| 9136 | -#define SYMBOL_NAME cos |
| 9137 | -#include "ifunc-avx-fma4.h" |
| 9138 | +# undef SYMBOL_NAME |
| 9139 | +# define SYMBOL_NAME cos |
| 9140 | +# include "ifunc-avx-fma4.h" |
| 9141 | |
| 9142 | libc_ifunc_redirected (__redirect_cos, __cos, IFUNC_SELECTOR ()); |
| 9143 | libm_alias_double (__cos, cos) |
| 9144 | |
| 9145 | -#define __cos __cos_sse2 |
| 9146 | -#define __sin __sin_sse2 |
| 9147 | +# define __cos __cos_sse2 |
| 9148 | +# define __sin __sin_sse2 |
| 9149 | +#endif |
| 9150 | #include <sysdeps/ieee754/dbl-64/s_sin.c> |
| 9151 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_sincos.c b/sysdeps/x86_64/fpu/multiarch/s_sincos.c |
| 9152 | index 70107e999c..b35757f8de 100644 |
| 9153 | --- a/sysdeps/x86_64/fpu/multiarch/s_sincos.c |
| 9154 | +++ b/sysdeps/x86_64/fpu/multiarch/s_sincos.c |
| 9155 | @@ -16,15 +16,18 @@ |
| 9156 | License along with the GNU C Library; if not, see |
| 9157 | <https://www.gnu.org/licenses/>. */ |
| 9158 | |
| 9159 | -#include <libm-alias-double.h> |
| 9160 | +#include <sysdeps/x86/isa-level.h> |
| 9161 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 9162 | +# include <libm-alias-double.h> |
| 9163 | |
| 9164 | extern void __redirect_sincos (double, double *, double *); |
| 9165 | |
| 9166 | -#define SYMBOL_NAME sincos |
| 9167 | -#include "ifunc-fma4.h" |
| 9168 | +# define SYMBOL_NAME sincos |
| 9169 | +# include "ifunc-fma4.h" |
| 9170 | |
| 9171 | libc_ifunc_redirected (__redirect_sincos, __sincos, IFUNC_SELECTOR ()); |
| 9172 | libm_alias_double (__sincos, sincos) |
| 9173 | |
| 9174 | -#define __sincos __sincos_sse2 |
| 9175 | +# define __sincos __sincos_sse2 |
| 9176 | +#endif |
| 9177 | #include <sysdeps/ieee754/dbl-64/s_sincos.c> |
| 9178 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_sincosf.c b/sysdeps/x86_64/fpu/multiarch/s_sincosf.c |
| 9179 | index 80bc028451..0ea9b40e84 100644 |
| 9180 | --- a/sysdeps/x86_64/fpu/multiarch/s_sincosf.c |
| 9181 | +++ b/sysdeps/x86_64/fpu/multiarch/s_sincosf.c |
| 9182 | @@ -16,13 +16,18 @@ |
| 9183 | License along with the GNU C Library; if not, see |
| 9184 | <https://www.gnu.org/licenses/>. */ |
| 9185 | |
| 9186 | -#include <libm-alias-float.h> |
| 9187 | +#include <sysdeps/x86/isa-level.h> |
| 9188 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 9189 | +# include <libm-alias-float.h> |
| 9190 | |
| 9191 | extern void __redirect_sincosf (float, float *, float *); |
| 9192 | |
| 9193 | -#define SYMBOL_NAME sincosf |
| 9194 | -#include "ifunc-fma.h" |
| 9195 | +# define SYMBOL_NAME sincosf |
| 9196 | +# include "ifunc-fma.h" |
| 9197 | |
| 9198 | libc_ifunc_redirected (__redirect_sincosf, __sincosf, IFUNC_SELECTOR ()); |
| 9199 | |
| 9200 | libm_alias_float (__sincos, sincos) |
| 9201 | +#else |
| 9202 | +# include <sysdeps/ieee754/flt-32/s_sincosf.c> |
| 9203 | +#endif |
| 9204 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_sinf.c b/sysdeps/x86_64/fpu/multiarch/s_sinf.c |
| 9205 | index a32b9e9550..c61624e3ee 100644 |
| 9206 | --- a/sysdeps/x86_64/fpu/multiarch/s_sinf.c |
| 9207 | +++ b/sysdeps/x86_64/fpu/multiarch/s_sinf.c |
| 9208 | @@ -16,13 +16,18 @@ |
| 9209 | License along with the GNU C Library; if not, see |
| 9210 | <https://www.gnu.org/licenses/>. */ |
| 9211 | |
| 9212 | -#include <libm-alias-float.h> |
| 9213 | +#include <sysdeps/x86/isa-level.h> |
| 9214 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 9215 | +# include <libm-alias-float.h> |
| 9216 | |
| 9217 | extern float __redirect_sinf (float); |
| 9218 | |
| 9219 | -#define SYMBOL_NAME sinf |
| 9220 | -#include "ifunc-fma.h" |
| 9221 | +# define SYMBOL_NAME sinf |
| 9222 | +# include "ifunc-fma.h" |
| 9223 | |
| 9224 | libc_ifunc_redirected (__redirect_sinf, __sinf, IFUNC_SELECTOR ()); |
| 9225 | |
| 9226 | libm_alias_float (__sin, sin) |
| 9227 | +#else |
| 9228 | +# include <sysdeps/ieee754/flt-32/s_sinf.c> |
| 9229 | +#endif |
| 9230 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_tan.c b/sysdeps/x86_64/fpu/multiarch/s_tan.c |
| 9231 | index f9a2474a13..125d992ba1 100644 |
| 9232 | --- a/sysdeps/x86_64/fpu/multiarch/s_tan.c |
| 9233 | +++ b/sysdeps/x86_64/fpu/multiarch/s_tan.c |
| 9234 | @@ -16,15 +16,18 @@ |
| 9235 | License along with the GNU C Library; if not, see |
| 9236 | <https://www.gnu.org/licenses/>. */ |
| 9237 | |
| 9238 | -#include <libm-alias-double.h> |
| 9239 | +#include <sysdeps/x86/isa-level.h> |
| 9240 | +#if MINIMUM_X86_ISA_LEVEL < AVX2_X86_ISA_LEVEL |
| 9241 | +# include <libm-alias-double.h> |
| 9242 | |
| 9243 | extern double __redirect_tan (double); |
| 9244 | |
| 9245 | -#define SYMBOL_NAME tan |
| 9246 | -#include "ifunc-avx-fma4.h" |
| 9247 | +# define SYMBOL_NAME tan |
| 9248 | +# include "ifunc-avx-fma4.h" |
| 9249 | |
| 9250 | libc_ifunc_redirected (__redirect_tan, __tan, IFUNC_SELECTOR ()); |
| 9251 | libm_alias_double (__tan, tan) |
| 9252 | |
| 9253 | -#define __tan __tan_sse2 |
| 9254 | +# define __tan __tan_sse2 |
| 9255 | +#endif |
| 9256 | #include <sysdeps/ieee754/dbl-64/s_tan.c> |
| 9257 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_trunc-avx.S b/sysdeps/x86_64/fpu/multiarch/s_trunc-avx.S |
| 9258 | new file mode 100644 |
| 9259 | index 0000000000..b3e87e9606 |
| 9260 | --- /dev/null |
| 9261 | +++ b/sysdeps/x86_64/fpu/multiarch/s_trunc-avx.S |
| 9262 | @@ -0,0 +1,28 @@ |
| 9263 | +/* AVX implementation of trunc function. |
| 9264 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 9265 | + This file is part of the GNU C Library. |
| 9266 | + |
| 9267 | + The GNU C Library is free software; you can redistribute it and/or |
| 9268 | + modify it under the terms of the GNU Lesser General Public |
| 9269 | + License as published by the Free Software Foundation; either |
| 9270 | + version 2.1 of the License, or (at your option) any later version. |
| 9271 | + |
| 9272 | + The GNU C Library is distributed in the hope that it will be useful, |
| 9273 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 9274 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 9275 | + Lesser General Public License for more details. |
| 9276 | + |
| 9277 | + You should have received a copy of the GNU Lesser General Public |
| 9278 | + License along with the GNU C Library; if not, see |
| 9279 | + <https://www.gnu.org/licenses/>. */ |
| 9280 | + |
| 9281 | +#include <sysdep.h> |
| 9282 | +#include <libm-alias-double.h> |
| 9283 | + |
| 9284 | + .text |
| 9285 | +ENTRY(__trunc) |
| 9286 | + vroundsd $11, %xmm0, %xmm0, %xmm0 |
| 9287 | + ret |
| 9288 | +END(__trunc) |
| 9289 | + |
| 9290 | +libm_alias_double (__trunc, trunc) |
| 9291 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_trunc-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_trunc-sse4_1.S |
| 9292 | index b496a6ef49..2b79174eed 100644 |
| 9293 | --- a/sysdeps/x86_64/fpu/multiarch/s_trunc-sse4_1.S |
| 9294 | +++ b/sysdeps/x86_64/fpu/multiarch/s_trunc-sse4_1.S |
| 9295 | @@ -18,8 +18,20 @@ |
| 9296 | |
| 9297 | #include <sysdep.h> |
| 9298 | |
| 9299 | +#include <sysdeps/x86/isa-level.h> |
| 9300 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 9301 | +# include <libm-alias-double.h> |
| 9302 | +# define __trunc_sse41 __trunc |
| 9303 | + .text |
| 9304 | +#else |
| 9305 | .section .text.sse4.1,"ax",@progbits |
| 9306 | +#endif |
| 9307 | + |
| 9308 | ENTRY(__trunc_sse41) |
| 9309 | roundsd $11, %xmm0, %xmm0 |
| 9310 | ret |
| 9311 | END(__trunc_sse41) |
| 9312 | + |
| 9313 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 9314 | +libm_alias_double (__trunc, trunc) |
| 9315 | +#endif |
| 9316 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_trunc.c b/sysdeps/x86_64/fpu/multiarch/s_trunc.c |
| 9317 | index 9bc9df8744..ea89c4f85d 100644 |
| 9318 | --- a/sysdeps/x86_64/fpu/multiarch/s_trunc.c |
| 9319 | +++ b/sysdeps/x86_64/fpu/multiarch/s_trunc.c |
| 9320 | @@ -16,17 +16,20 @@ |
| 9321 | License along with the GNU C Library; if not, see |
| 9322 | <https://www.gnu.org/licenses/>. */ |
| 9323 | |
| 9324 | -#define NO_MATH_REDIRECT |
| 9325 | -#include <libm-alias-double.h> |
| 9326 | +#include <sysdeps/x86/isa-level.h> |
| 9327 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 9328 | +# define NO_MATH_REDIRECT |
| 9329 | +# include <libm-alias-double.h> |
| 9330 | |
| 9331 | -#define trunc __redirect_trunc |
| 9332 | -#define __trunc __redirect___trunc |
| 9333 | -#include <math.h> |
| 9334 | -#undef trunc |
| 9335 | -#undef __trunc |
| 9336 | +# define trunc __redirect_trunc |
| 9337 | +# define __trunc __redirect___trunc |
| 9338 | +# include <math.h> |
| 9339 | +# undef trunc |
| 9340 | +# undef __trunc |
| 9341 | |
| 9342 | -#define SYMBOL_NAME trunc |
| 9343 | -#include "ifunc-sse4_1.h" |
| 9344 | +# define SYMBOL_NAME trunc |
| 9345 | +# include "ifunc-sse4_1.h" |
| 9346 | |
| 9347 | libc_ifunc_redirected (__redirect_trunc, __trunc, IFUNC_SELECTOR ()); |
| 9348 | libm_alias_double (__trunc, trunc) |
| 9349 | +#endif |
| 9350 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_truncf-avx.S b/sysdeps/x86_64/fpu/multiarch/s_truncf-avx.S |
| 9351 | new file mode 100644 |
| 9352 | index 0000000000..f31ac7d7f7 |
| 9353 | --- /dev/null |
| 9354 | +++ b/sysdeps/x86_64/fpu/multiarch/s_truncf-avx.S |
| 9355 | @@ -0,0 +1,28 @@ |
| 9356 | +/* AVX implementation of truncf function. |
| 9357 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 9358 | + This file is part of the GNU C Library. |
| 9359 | + |
| 9360 | + The GNU C Library is free software; you can redistribute it and/or |
| 9361 | + modify it under the terms of the GNU Lesser General Public |
| 9362 | + License as published by the Free Software Foundation; either |
| 9363 | + version 2.1 of the License, or (at your option) any later version. |
| 9364 | + |
| 9365 | + The GNU C Library is distributed in the hope that it will be useful, |
| 9366 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 9367 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 9368 | + Lesser General Public License for more details. |
| 9369 | + |
| 9370 | + You should have received a copy of the GNU Lesser General Public |
| 9371 | + License along with the GNU C Library; if not, see |
| 9372 | + <https://www.gnu.org/licenses/>. */ |
| 9373 | + |
| 9374 | +#include <sysdep.h> |
| 9375 | +#include <libm-alias-float.h> |
| 9376 | + |
| 9377 | + .text |
| 9378 | +ENTRY(__truncf) |
| 9379 | + vroundss $11, %xmm0, %xmm0, %xmm0 |
| 9380 | + ret |
| 9381 | +END(__truncf) |
| 9382 | + |
| 9383 | +libm_alias_float (__trunc, trunc) |
| 9384 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_truncf-sse4_1.S b/sysdeps/x86_64/fpu/multiarch/s_truncf-sse4_1.S |
| 9385 | index 22e9a83307..60498b2cb2 100644 |
| 9386 | --- a/sysdeps/x86_64/fpu/multiarch/s_truncf-sse4_1.S |
| 9387 | +++ b/sysdeps/x86_64/fpu/multiarch/s_truncf-sse4_1.S |
| 9388 | @@ -18,8 +18,20 @@ |
| 9389 | |
| 9390 | #include <sysdep.h> |
| 9391 | |
| 9392 | +#include <sysdeps/x86/isa-level.h> |
| 9393 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 9394 | +# include <libm-alias-float.h> |
| 9395 | +# define __truncf_sse41 __truncf |
| 9396 | + .text |
| 9397 | +#else |
| 9398 | .section .text.sse4.1,"ax",@progbits |
| 9399 | +#endif |
| 9400 | + |
| 9401 | ENTRY(__truncf_sse41) |
| 9402 | roundss $11, %xmm0, %xmm0 |
| 9403 | ret |
| 9404 | END(__truncf_sse41) |
| 9405 | + |
| 9406 | +#if MINIMUM_X86_ISA_LEVEL == SSE4_1_X86_ISA_LEVEL |
| 9407 | +libm_alias_float (__trunc, trunc) |
| 9408 | +#endif |
| 9409 | diff --git a/sysdeps/x86_64/fpu/multiarch/s_truncf.c b/sysdeps/x86_64/fpu/multiarch/s_truncf.c |
| 9410 | index dae01d166a..92435ce39d 100644 |
| 9411 | --- a/sysdeps/x86_64/fpu/multiarch/s_truncf.c |
| 9412 | +++ b/sysdeps/x86_64/fpu/multiarch/s_truncf.c |
| 9413 | @@ -16,17 +16,20 @@ |
| 9414 | License along with the GNU C Library; if not, see |
| 9415 | <https://www.gnu.org/licenses/>. */ |
| 9416 | |
| 9417 | -#define NO_MATH_REDIRECT |
| 9418 | -#include <libm-alias-float.h> |
| 9419 | +#include <sysdeps/x86/isa-level.h> |
| 9420 | +#if MINIMUM_X86_ISA_LEVEL < SSE4_1_X86_ISA_LEVEL |
| 9421 | +# define NO_MATH_REDIRECT |
| 9422 | +# include <libm-alias-float.h> |
| 9423 | |
| 9424 | -#define truncf __redirect_truncf |
| 9425 | -#define __truncf __redirect___truncf |
| 9426 | -#include <math.h> |
| 9427 | -#undef truncf |
| 9428 | -#undef __truncf |
| 9429 | +# define truncf __redirect_truncf |
| 9430 | +# define __truncf __redirect___truncf |
| 9431 | +# include <math.h> |
| 9432 | +# undef truncf |
| 9433 | +# undef __truncf |
| 9434 | |
| 9435 | -#define SYMBOL_NAME truncf |
| 9436 | -#include "ifunc-sse4_1.h" |
| 9437 | +# define SYMBOL_NAME truncf |
| 9438 | +# include "ifunc-sse4_1.h" |
| 9439 | |
| 9440 | libc_ifunc_redirected (__redirect_truncf, __truncf, IFUNC_SELECTOR ()); |
| 9441 | libm_alias_float (__trunc, trunc) |
| 9442 | +#endif |
| 9443 | diff --git a/sysdeps/x86_64/fpu/multiarch/w_exp.c b/sysdeps/x86_64/fpu/multiarch/w_exp.c |
| 9444 | index 27eee98a0a..3584187e0e 100644 |
| 9445 | --- a/sysdeps/x86_64/fpu/multiarch/w_exp.c |
| 9446 | +++ b/sysdeps/x86_64/fpu/multiarch/w_exp.c |
| 9447 | @@ -1 +1,6 @@ |
| 9448 | -#include <sysdeps/../math/w_exp.c> |
| 9449 | +#include <sysdeps/x86/isa-level.h> |
| 9450 | +#if MINIMUM_X86_ISA_LEVEL >= AVX2_X86_ISA_LEVEL |
| 9451 | +# include <sysdeps/ieee754/dbl-64/w_exp.c> |
| 9452 | +#else |
| 9453 | +# include <sysdeps/../math/w_exp.c> |
| 9454 | +#endif |
| 9455 | diff --git a/sysdeps/x86_64/fpu/multiarch/w_log.c b/sysdeps/x86_64/fpu/multiarch/w_log.c |
| 9456 | index 9b2b018711..414ca3ca3d 100644 |
| 9457 | --- a/sysdeps/x86_64/fpu/multiarch/w_log.c |
| 9458 | +++ b/sysdeps/x86_64/fpu/multiarch/w_log.c |
| 9459 | @@ -1 +1,6 @@ |
| 9460 | -#include <sysdeps/../math/w_log.c> |
| 9461 | +#include <sysdeps/x86/isa-level.h> |
| 9462 | +#if MINIMUM_X86_ISA_LEVEL >= AVX2_X86_ISA_LEVEL |
| 9463 | +# include <sysdeps/ieee754/dbl-64/w_log.c> |
| 9464 | +#else |
| 9465 | +# include <sysdeps/../math/w_log.c> |
| 9466 | +#endif |
| 9467 | diff --git a/sysdeps/x86_64/fpu/multiarch/w_pow.c b/sysdeps/x86_64/fpu/multiarch/w_pow.c |
| 9468 | index b50c1988de..d5fcc4f871 100644 |
| 9469 | --- a/sysdeps/x86_64/fpu/multiarch/w_pow.c |
| 9470 | +++ b/sysdeps/x86_64/fpu/multiarch/w_pow.c |
| 9471 | @@ -1 +1,6 @@ |
| 9472 | -#include <sysdeps/../math/w_pow.c> |
| 9473 | +#include <sysdeps/x86/isa-level.h> |
| 9474 | +#if MINIMUM_X86_ISA_LEVEL >= AVX2_X86_ISA_LEVEL |
| 9475 | +# include <sysdeps/ieee754/dbl-64/w_pow.c> |
| 9476 | +#else |
| 9477 | +# include <sysdeps/../math/w_pow.c> |
| 9478 | +#endif |
| 9479 | diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile |
| 9480 | index e1e894c963..d3d2270394 100644 |
| 9481 | --- a/sysdeps/x86_64/multiarch/Makefile |
| 9482 | +++ b/sysdeps/x86_64/multiarch/Makefile |
| 9483 | @@ -4,8 +4,8 @@ sysdep_routines += \ |
| 9484 | memchr-avx2 \ |
| 9485 | memchr-avx2-rtm \ |
| 9486 | memchr-evex \ |
| 9487 | - memchr-evex512 \ |
| 9488 | memchr-evex-rtm \ |
| 9489 | + memchr-evex512 \ |
| 9490 | memchr-sse2 \ |
| 9491 | memcmp-avx2-movbe \ |
| 9492 | memcmp-avx2-movbe-rtm \ |
| 9493 | @@ -37,8 +37,8 @@ sysdep_routines += \ |
| 9494 | rawmemchr-avx2 \ |
| 9495 | rawmemchr-avx2-rtm \ |
| 9496 | rawmemchr-evex \ |
| 9497 | - rawmemchr-evex512 \ |
| 9498 | rawmemchr-evex-rtm \ |
| 9499 | + rawmemchr-evex512 \ |
| 9500 | rawmemchr-sse2 \ |
| 9501 | stpcpy-avx2 \ |
| 9502 | stpcpy-avx2-rtm \ |
| 9503 | diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S |
| 9504 | index 9984c3ca0f..97839a2248 100644 |
| 9505 | --- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S |
| 9506 | +++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S |
| 9507 | @@ -21,7 +21,9 @@ |
| 9508 | 2. If size is less than VEC, use integer register stores. |
| 9509 | 3. If size is from VEC_SIZE to 2 * VEC_SIZE, use 2 VEC stores. |
| 9510 | 4. If size is from 2 * VEC_SIZE to 4 * VEC_SIZE, use 4 VEC stores. |
| 9511 | - 5. If size is more to 4 * VEC_SIZE, align to 4 * VEC_SIZE with |
| 9512 | + 5. On machines ERMS feature, if size is greater or equal than |
| 9513 | + __x86_rep_stosb_threshold then REP STOSB will be used. |
| 9514 | + 6. If size is more to 4 * VEC_SIZE, align to 4 * VEC_SIZE with |
| 9515 | 4 VEC stores and store 4 * VEC at a time until done. */ |
| 9516 | |
| 9517 | #include <sysdep.h> |
| 9518 | diff --git a/sysdeps/x86_64/multiarch/wcsncat-evex.S b/sysdeps/x86_64/multiarch/wcsncat-evex.S |
| 9519 | index 392215950a..10bfb0a531 100644 |
| 9520 | --- a/sysdeps/x86_64/multiarch/wcsncat-evex.S |
| 9521 | +++ b/sysdeps/x86_64/multiarch/wcsncat-evex.S |
| 9522 | @@ -1,9 +1,9 @@ |
| 9523 | -#ifndef WCSCAT |
| 9524 | -# define WCSCAT __wcsncat_evex |
| 9525 | +#ifndef WCSNCAT |
| 9526 | +# define WCSNCAT __wcsncat_evex |
| 9527 | #endif |
| 9528 | |
| 9529 | #define USE_AS_WCSCPY |
| 9530 | #define USE_AS_STRCAT |
| 9531 | |
| 9532 | -#define STRNCAT WCSCAT |
| 9533 | +#define STRNCAT WCSNCAT |
| 9534 | #include "strncat-evex.S" |
| 9535 | diff --git a/sysdeps/x86_64/tst-gnu2-tls2mod1.S b/sysdeps/x86_64/tst-gnu2-tls2mod1.S |
| 9536 | new file mode 100644 |
| 9537 | index 0000000000..1d636669ba |
| 9538 | --- /dev/null |
| 9539 | +++ b/sysdeps/x86_64/tst-gnu2-tls2mod1.S |
| 9540 | @@ -0,0 +1,87 @@ |
| 9541 | +/* Check if TLSDESC relocation preserves %rdi, %rsi and %rbx. |
| 9542 | + Copyright (C) 2024 Free Software Foundation, Inc. |
| 9543 | + This file is part of the GNU C Library. |
| 9544 | + |
| 9545 | + The GNU C Library is free software; you can redistribute it and/or |
| 9546 | + modify it under the terms of the GNU Lesser General Public |
| 9547 | + License as published by the Free Software Foundation; either |
| 9548 | + version 2.1 of the License, or (at your option) any later version. |
| 9549 | + |
| 9550 | + The GNU C Library is distributed in the hope that it will be useful, |
| 9551 | + but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 9552 | + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 9553 | + Lesser General Public License for more details. |
| 9554 | + |
| 9555 | + You should have received a copy of the GNU Lesser General Public |
| 9556 | + License along with the GNU C Library; if not, see |
| 9557 | + <http://www.gnu.org/licenses/>. */ |
| 9558 | + |
| 9559 | +#include <sysdep.h> |
| 9560 | + |
| 9561 | +/* On AVX512 machines, OFFSET == 40 caused _dl_tlsdesc_dynamic_xsavec |
| 9562 | + to clobber %rdi, %rsi and %rbx. On Intel AVX CPUs, the state size |
| 9563 | + is 960 bytes and this test didn't fail. It may be due to the unused |
| 9564 | + last 128 bytes. On AMD AVX CPUs, the state size is 832 bytes and |
| 9565 | + this test might fail without the fix. */ |
| 9566 | +#ifndef OFFSET |
| 9567 | +# define OFFSET 40 |
| 9568 | +#endif |
| 9569 | + |
| 9570 | + .text |
| 9571 | + .p2align 4 |
| 9572 | + .globl apply_tls |
| 9573 | + .type apply_tls, @function |
| 9574 | +apply_tls: |
| 9575 | + cfi_startproc |
| 9576 | + _CET_ENDBR |
| 9577 | + pushq %rbp |
| 9578 | + cfi_def_cfa_offset (16) |
| 9579 | + cfi_offset (6, -16) |
| 9580 | + movdqu (%RDI_LP), %xmm0 |
| 9581 | + lea tls_var1@TLSDESC(%rip), %RAX_LP |
| 9582 | + mov %RSP_LP, %RBP_LP |
| 9583 | + cfi_def_cfa_register (6) |
| 9584 | + /* Align stack to 64 bytes. */ |
| 9585 | + and $-64, %RSP_LP |
| 9586 | + sub $OFFSET, %RSP_LP |
| 9587 | + pushq %rbx |
| 9588 | + /* Set %ebx to 0xbadbeef. */ |
| 9589 | + movl $0xbadbeef, %ebx |
| 9590 | + movl $0xbadbeef, %esi |
| 9591 | + movq %rdi, saved_rdi(%rip) |
| 9592 | + movq %rsi, saved_rsi(%rip) |
| 9593 | + call *tls_var1@TLSCALL(%RAX_LP) |
| 9594 | + /* Check if _dl_tlsdesc_dynamic preserves %rdi, %rsi and %rbx. */ |
| 9595 | + cmpq saved_rdi(%rip), %rdi |
| 9596 | + jne L(hlt) |
| 9597 | + cmpq saved_rsi(%rip), %rsi |
| 9598 | + jne L(hlt) |
| 9599 | + cmpl $0xbadbeef, %ebx |
| 9600 | + jne L(hlt) |
| 9601 | + add %fs:0, %RAX_LP |
| 9602 | + movups %xmm0, 32(%RAX_LP) |
| 9603 | + movdqu 16(%RDI_LP), %xmm1 |
| 9604 | + mov %RAX_LP, %RBX_LP |
| 9605 | + movups %xmm1, 48(%RAX_LP) |
| 9606 | + lea 32(%RBX_LP), %RAX_LP |
| 9607 | + pop %rbx |
| 9608 | + leave |
| 9609 | + cfi_def_cfa (7, 8) |
| 9610 | + ret |
| 9611 | +L(hlt): |
| 9612 | + hlt |
| 9613 | + cfi_endproc |
| 9614 | + .size apply_tls, .-apply_tls |
| 9615 | + .hidden tls_var1 |
| 9616 | + .globl tls_var1 |
| 9617 | + .section .tbss,"awT",@nobits |
| 9618 | + .align 16 |
| 9619 | + .type tls_var1, @object |
| 9620 | + .size tls_var1, 3200 |
| 9621 | +tls_var1: |
| 9622 | + .zero 3200 |
| 9623 | + .local saved_rdi |
| 9624 | + .comm saved_rdi,8,8 |
| 9625 | + .local saved_rsi |
| 9626 | + .comm saved_rsi,8,8 |
| 9627 | + .section .note.GNU-stack,"",@progbits |
| 9628 | diff --git a/time/timespec_get.c b/time/timespec_get.c |
| 9629 | index b031e42ca2..26a044bca6 100644 |
| 9630 | --- a/time/timespec_get.c |
| 9631 | +++ b/time/timespec_get.c |
| 9632 | @@ -4,7 +4,7 @@ |
| 9633 | The GNU C Library is free software; you can redistribute it and/or |
| 9634 | modify it under the terms of the GNU Lesser General Public |
| 9635 | License as published by the Free Software Foundation; either |
| 9636 | - version 2.1 of the License. |
| 9637 | + version 2.1 of the License, or (at your option) any later version. |
| 9638 | |
| 9639 | The GNU C Library is distributed in the hope that it will be useful, |
| 9640 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 9641 | diff --git a/time/timespec_getres.c b/time/timespec_getres.c |
| 9642 | index edb397507c..2e18b8bcac 100644 |
| 9643 | --- a/time/timespec_getres.c |
| 9644 | +++ b/time/timespec_getres.c |
| 9645 | @@ -5,7 +5,7 @@ |
| 9646 | The GNU C Library is free software; you can redistribute it and/or |
| 9647 | modify it under the terms of the GNU Lesser General Public |
| 9648 | License as published by the Free Software Foundation; either |
| 9649 | - version 2.1 of the License. |
| 9650 | + version 2.1 of the License, or (at your option) any later version. |
| 9651 | |
| 9652 | The GNU C Library is distributed in the hope that it will be useful, |
| 9653 | but WITHOUT ANY WARRANTY; without even the implied warranty of |