| 1 | ACPI ACPI support is enabled. |
| 2 | AGP AGP (Accelerated Graphics Port) is enabled. |
| 3 | ALSA ALSA sound support is enabled. |
| 4 | APIC APIC support is enabled. |
| 5 | APM Advanced Power Management support is enabled. |
| 6 | APPARMOR AppArmor support is enabled. |
| 7 | ARM ARM architecture is enabled. |
| 8 | ARM64 ARM64 architecture is enabled. |
| 9 | CLK Common clock infrastructure is enabled. |
| 10 | CMA Contiguous Memory Area support is enabled. |
| 11 | CPU_FREQ CPU frequency scaling is enabled. |
| 12 | CPU_IDLE CPU idle time management is enabled. |
| 13 | DRM Direct Rendering Management support is enabled. |
| 14 | DYNAMIC_DEBUG Build in debug messages and enable them at runtime |
| 15 | EARLY Parameter processed too early to be embedded in initrd. |
| 16 | EDD BIOS Enhanced Disk Drive Services (EDD) is enabled |
| 17 | EFI EFI Partitioning (GPT) is enabled |
| 18 | EVM Extended Verification Module |
| 19 | FB The frame buffer device is enabled. |
| 20 | FTRACE Function tracing enabled. |
| 21 | GCOV GCOV profiling is enabled. |
| 22 | HIBERNATION HIBERNATION is enabled. |
| 23 | HW Appropriate hardware is enabled. |
| 24 | HYPER_V HYPERV support is enabled. |
| 25 | IMA Integrity measurement architecture is enabled. |
| 26 | IP_PNP IP DHCP, BOOTP, or RARP is enabled. |
| 27 | IPV6 IPv6 support is enabled. |
| 28 | ISAPNP ISA PnP code is enabled. |
| 29 | ISOL CPU Isolation is enabled. |
| 30 | JOY Appropriate joystick support is enabled. |
| 31 | KGDB Kernel debugger support is enabled. |
| 32 | KVM Kernel Virtual Machine support is enabled. |
| 33 | LIBATA Libata driver is enabled |
| 34 | LOONGARCH LoongArch architecture is enabled. |
| 35 | LOOP Loopback device support is enabled. |
| 36 | LP Printer support is enabled. |
| 37 | M68k M68k architecture is enabled. |
| 38 | These options have more detailed description inside of |
| 39 | Documentation/arch/m68k/kernel-options.rst. |
| 40 | MIPS MIPS architecture is enabled. |
| 41 | MOUSE Appropriate mouse support is enabled. |
| 42 | MSI Message Signaled Interrupts (PCI). |
| 43 | MTD MTD (Memory Technology Device) support is enabled. |
| 44 | NET Appropriate network support is enabled. |
| 45 | NFS Appropriate NFS support is enabled. |
| 46 | NUMA NUMA support is enabled. |
| 47 | OF Devicetree is enabled. |
| 48 | PARISC The PA-RISC architecture is enabled. |
| 49 | PCI PCI bus support is enabled. |
| 50 | PCIE PCI Express support is enabled. |
| 51 | PCMCIA The PCMCIA subsystem is enabled. |
| 52 | PM Power Management support is enabled. |
| 53 | PNP Plug & Play support is enabled. |
| 54 | PPC PowerPC architecture is enabled. |
| 55 | PPT Parallel port support is enabled. |
| 56 | PS2 Appropriate PS/2 support is enabled. |
| 57 | PV_OPS A paravirtualized kernel is enabled. |
| 58 | RAM RAM disk support is enabled. |
| 59 | RDT Intel Resource Director Technology. |
| 60 | RISCV RISCV architecture is enabled. |
| 61 | S390 S390 architecture is enabled. |
| 62 | SCSI Appropriate SCSI support is enabled. |
| 63 | A lot of drivers have their options described inside |
| 64 | the Documentation/scsi/ sub-directory. |
| 65 | SDW SoundWire support is enabled. |
| 66 | SECURITY Different security models are enabled. |
| 67 | SELINUX SELinux support is enabled. |
| 68 | SERIAL Serial support is enabled. |
| 69 | SH SuperH architecture is enabled. |
| 70 | SMP The kernel is an SMP kernel. |
| 71 | SPARC Sparc architecture is enabled. |
| 72 | SUSPEND System suspend states are enabled. |
| 73 | SWSUSP Software suspend (hibernation) is enabled. |
| 74 | TPM TPM drivers are enabled. |
| 75 | UMS USB Mass Storage support is enabled. |
| 76 | USB USB support is enabled. |
| 77 | NVME NVMe support is enabled |
| 78 | USBHID USB Human Interface Device support is enabled. |
| 79 | V4L Video For Linux support is enabled. |
| 80 | VGA The VGA console has been enabled. |
| 81 | VMMIO Driver for memory mapped virtio devices is enabled. |
| 82 | VT Virtual terminal support is enabled. |
| 83 | WDT Watchdog support is enabled. |
| 84 | X86-32 X86-32, aka i386 architecture is enabled. |
| 85 | X86-64 X86-64 architecture is enabled. |
| 86 | X86 Either 32-bit or 64-bit x86 (same as X86-32+X86-64) |
| 87 | X86_UV SGI UV support is enabled. |
| 88 | XEN Xen support is enabled |
| 89 | XTENSA xtensa architecture is enabled. |
| 90 | |
| 91 | In addition, the following text indicates that the option |
| 92 | |
| 93 | BOOT Is a boot loader parameter. |
| 94 | BUGS= Relates to possible processor bugs on the said processor. |
| 95 | KNL Is a kernel start-up parameter. |
| 96 | |
| 97 | |
| 98 | Kernel parameters |
| 99 | |
| 100 | accept_memory= [MM] |
| 101 | Format: { eager | lazy } |
| 102 | default: lazy |
| 103 | By default, unaccepted memory is accepted lazily to |
| 104 | avoid prolonged boot times. The lazy option will add |
| 105 | some runtime overhead until all memory is eventually |
| 106 | accepted. In most cases the overhead is negligible. |
| 107 | For some workloads or for debugging purposes |
| 108 | accept_memory=eager can be used to accept all memory |
| 109 | at once during boot. |
| 110 | |
| 111 | acpi= [HW,ACPI,X86,ARM64,RISCV64,EARLY] |
| 112 | Advanced Configuration and Power Interface |
| 113 | Format: { force | on | off | strict | noirq | rsdt | |
| 114 | copy_dsdt | nospcr } |
| 115 | force -- enable ACPI if default was off |
| 116 | on -- enable ACPI but allow fallback to DT [arm64,riscv64] |
| 117 | off -- disable ACPI if default was on |
| 118 | noirq -- do not use ACPI for IRQ routing |
| 119 | strict -- Be less tolerant of platforms that are not |
| 120 | strictly ACPI specification compliant. |
| 121 | rsdt -- prefer RSDT over (default) XSDT |
| 122 | copy_dsdt -- copy DSDT to memory |
| 123 | nocmcff -- Disable firmware first mode for corrected |
| 124 | errors. This disables parsing the HEST CMC error |
| 125 | source to check if firmware has set the FF flag. This |
| 126 | may result in duplicate corrected error reports. |
| 127 | nospcr -- disable console in ACPI SPCR table as |
| 128 | default _serial_ console on ARM64 |
| 129 | spcr -- enable console in ACPI SPCR table as |
| 130 | default _serial_ console on x86 |
| 131 | For ARM64, ONLY "acpi=off", "acpi=on", "acpi=force" or |
| 132 | "acpi=nospcr" are available |
| 133 | For RISCV64, ONLY "acpi=off", "acpi=on" or "acpi=force" |
| 134 | are available |
| 135 | |
| 136 | See also Documentation/power/runtime_pm.rst, pci=noacpi |
| 137 | |
| 138 | acpi_apic_instance= [ACPI,IOAPIC,EARLY] |
| 139 | Format: <int> |
| 140 | 2: use 2nd APIC table, if available |
| 141 | 1,0: use 1st APIC table |
| 142 | default: 0 |
| 143 | |
| 144 | acpi_backlight= [HW,ACPI] |
| 145 | { vendor | video | native | none } |
| 146 | If set to vendor, prefer vendor-specific driver |
| 147 | (e.g. thinkpad_acpi, sony_acpi, etc.) instead |
| 148 | of the ACPI video.ko driver. |
| 149 | If set to video, use the ACPI video.ko driver. |
| 150 | If set to native, use the device's native backlight mode. |
| 151 | If set to none, disable the ACPI backlight interface. |
| 152 | |
| 153 | acpi_force_32bit_fadt_addr [ACPI,EARLY] |
| 154 | force FADT to use 32 bit addresses rather than the |
| 155 | 64 bit X_* addresses. Some firmware have broken 64 |
| 156 | bit addresses for force ACPI ignore these and use |
| 157 | the older legacy 32 bit addresses. |
| 158 | |
| 159 | acpica_no_return_repair [HW, ACPI] |
| 160 | Disable AML predefined validation mechanism |
| 161 | This mechanism can repair the evaluation result to make |
| 162 | the return objects more ACPI specification compliant. |
| 163 | This option is useful for developers to identify the |
| 164 | root cause of an AML interpreter issue when the issue |
| 165 | has something to do with the repair mechanism. |
| 166 | |
| 167 | acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] |
| 168 | acpi.debug_level= [HW,ACPI,ACPI_DEBUG] |
| 169 | Format: <int> |
| 170 | CONFIG_ACPI_DEBUG must be enabled to produce any ACPI |
| 171 | debug output. Bits in debug_layer correspond to a |
| 172 | _COMPONENT in an ACPI source file, e.g., |
| 173 | #define _COMPONENT ACPI_EVENTS |
| 174 | Bits in debug_level correspond to a level in |
| 175 | ACPI_DEBUG_PRINT statements, e.g., |
| 176 | ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... |
| 177 | The debug_level mask defaults to "info". See |
| 178 | Documentation/firmware-guide/acpi/debug.rst for more information about |
| 179 | debug layers and levels. |
| 180 | |
| 181 | Enable processor driver info messages: |
| 182 | acpi.debug_layer=0x20000000 |
| 183 | Enable AML "Debug" output, i.e., stores to the Debug |
| 184 | object while interpreting AML: |
| 185 | acpi.debug_layer=0xffffffff acpi.debug_level=0x2 |
| 186 | Enable all messages related to ACPI hardware: |
| 187 | acpi.debug_layer=0x2 acpi.debug_level=0xffffffff |
| 188 | |
| 189 | Some values produce so much output that the system is |
| 190 | unusable. The "log_buf_len" parameter may be useful |
| 191 | if you need to capture more output. |
| 192 | |
| 193 | acpi.poweroff_on_fatal= [ACPI] |
| 194 | {0 | 1} |
| 195 | Causes the system to poweroff when the ACPI bytecode signals |
| 196 | a fatal error. The default value of this setting is 1. |
| 197 | Overriding this value should only be done for diagnosing |
| 198 | ACPI firmware problems, as the system might behave erratically |
| 199 | after having encountered a fatal ACPI error. |
| 200 | |
| 201 | acpi_enforce_resources= [ACPI] |
| 202 | { strict | lax | no } |
| 203 | Check for resource conflicts between native drivers |
| 204 | and ACPI OperationRegions (SystemIO and SystemMemory |
| 205 | only). IO ports and memory declared in ACPI might be |
| 206 | used by the ACPI subsystem in arbitrary AML code and |
| 207 | can interfere with legacy drivers. |
| 208 | strict (default): access to resources claimed by ACPI |
| 209 | is denied; legacy drivers trying to access reserved |
| 210 | resources will fail to bind to device using them. |
| 211 | lax: access to resources claimed by ACPI is allowed; |
| 212 | legacy drivers trying to access reserved resources |
| 213 | will bind successfully but a warning message is logged. |
| 214 | no: ACPI OperationRegions are not marked as reserved, |
| 215 | no further checks are performed. |
| 216 | |
| 217 | acpi_force_table_verification [HW,ACPI,EARLY] |
| 218 | Enable table checksum verification during early stage. |
| 219 | By default, this is disabled due to x86 early mapping |
| 220 | size limitation. |
| 221 | |
| 222 | acpi_irq_balance [HW,ACPI] |
| 223 | ACPI will balance active IRQs |
| 224 | default in APIC mode |
| 225 | |
| 226 | acpi_irq_nobalance [HW,ACPI] |
| 227 | ACPI will not move active IRQs (default) |
| 228 | default in PIC mode |
| 229 | |
| 230 | acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA |
| 231 | Format: <irq>,<irq>... |
| 232 | |
| 233 | acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for |
| 234 | use by PCI |
| 235 | Format: <irq>,<irq>... |
| 236 | |
| 237 | acpi_mask_gpe= [HW,ACPI] |
| 238 | Due to the existence of _Lxx/_Exx, some GPEs triggered |
| 239 | by unsupported hardware/firmware features can result in |
| 240 | GPE floodings that cannot be automatically disabled by |
| 241 | the GPE dispatcher. |
| 242 | This facility can be used to prevent such uncontrolled |
| 243 | GPE floodings. |
| 244 | Format: <byte> or <bitmap-list> |
| 245 | |
| 246 | acpi_no_auto_serialize [HW,ACPI] |
| 247 | Disable auto-serialization of AML methods |
| 248 | AML control methods that contain the opcodes to create |
| 249 | named objects will be marked as "Serialized" by the |
| 250 | auto-serialization feature. |
| 251 | This feature is enabled by default. |
| 252 | This option allows to turn off the feature. |
| 253 | |
| 254 | acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump |
| 255 | kernels. |
| 256 | |
| 257 | acpi_no_static_ssdt [HW,ACPI,EARLY] |
| 258 | Disable installation of static SSDTs at early boot time |
| 259 | By default, SSDTs contained in the RSDT/XSDT will be |
| 260 | installed automatically and they will appear under |
| 261 | /sys/firmware/acpi/tables. |
| 262 | This option turns off this feature. |
| 263 | Note that specifying this option does not affect |
| 264 | dynamic table installation which will install SSDT |
| 265 | tables to /sys/firmware/acpi/tables/dynamic. |
| 266 | |
| 267 | acpi_no_watchdog [HW,ACPI,WDT] |
| 268 | Ignore the ACPI-based watchdog interface (WDAT) and let |
| 269 | a native driver control the watchdog device instead. |
| 270 | |
| 271 | acpi_rsdp= [ACPI,EFI,KEXEC,EARLY] |
| 272 | Pass the RSDP address to the kernel, mostly used |
| 273 | on machines running EFI runtime service to boot the |
| 274 | second kernel for kdump. |
| 275 | |
| 276 | acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS |
| 277 | Format: To spoof as Windows 98: ="Microsoft Windows" |
| 278 | |
| 279 | acpi_rev_override [ACPI] Override the _REV object to return 5 (instead |
| 280 | of 2 which is mandated by ACPI 6) as the supported ACPI |
| 281 | specification revision (when using this switch, it may |
| 282 | be necessary to carry out a cold reboot _twice_ in a |
| 283 | row to make it take effect on the platform firmware). |
| 284 | |
| 285 | acpi_osi= [HW,ACPI] Modify list of supported OS interface strings |
| 286 | acpi_osi="string1" # add string1 |
| 287 | acpi_osi="!string2" # remove string2 |
| 288 | acpi_osi=!* # remove all strings |
| 289 | acpi_osi=! # disable all built-in OS vendor |
| 290 | strings |
| 291 | acpi_osi=!! # enable all built-in OS vendor |
| 292 | strings |
| 293 | acpi_osi= # disable all strings |
| 294 | |
| 295 | 'acpi_osi=!' can be used in combination with single or |
| 296 | multiple 'acpi_osi="string1"' to support specific OS |
| 297 | vendor string(s). Note that such command can only |
| 298 | affect the default state of the OS vendor strings, thus |
| 299 | it cannot affect the default state of the feature group |
| 300 | strings and the current state of the OS vendor strings, |
| 301 | specifying it multiple times through kernel command line |
| 302 | is meaningless. This command is useful when one do not |
| 303 | care about the state of the feature group strings which |
| 304 | should be controlled by the OSPM. |
| 305 | Examples: |
| 306 | 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent |
| 307 | to 'acpi_osi="Windows 2000" acpi_osi=!', they all |
| 308 | can make '_OSI("Windows 2000")' TRUE. |
| 309 | |
| 310 | 'acpi_osi=' cannot be used in combination with other |
| 311 | 'acpi_osi=' command lines, the _OSI method will not |
| 312 | exist in the ACPI namespace. NOTE that such command can |
| 313 | only affect the _OSI support state, thus specifying it |
| 314 | multiple times through kernel command line is also |
| 315 | meaningless. |
| 316 | Examples: |
| 317 | 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' |
| 318 | FALSE. |
| 319 | |
| 320 | 'acpi_osi=!*' can be used in combination with single or |
| 321 | multiple 'acpi_osi="string1"' to support specific |
| 322 | string(s). Note that such command can affect the |
| 323 | current state of both the OS vendor strings and the |
| 324 | feature group strings, thus specifying it multiple times |
| 325 | through kernel command line is meaningful. But it may |
| 326 | still not able to affect the final state of a string if |
| 327 | there are quirks related to this string. This command |
| 328 | is useful when one want to control the state of the |
| 329 | feature group strings to debug BIOS issues related to |
| 330 | the OSPM features. |
| 331 | Examples: |
| 332 | 1. 'acpi_osi="Module Device" acpi_osi=!*' can make |
| 333 | '_OSI("Module Device")' FALSE. |
| 334 | 2. 'acpi_osi=!* acpi_osi="Module Device"' can make |
| 335 | '_OSI("Module Device")' TRUE. |
| 336 | 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is |
| 337 | equivalent to |
| 338 | 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' |
| 339 | and |
| 340 | 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', |
| 341 | they all will make '_OSI("Windows 2000")' TRUE. |
| 342 | |
| 343 | acpi_pm_good [X86] |
| 344 | Override the pmtimer bug detection: force the kernel |
| 345 | to assume that this machine's pmtimer latches its value |
| 346 | and always returns good values. |
| 347 | |
| 348 | acpi_sci= [HW,ACPI,EARLY] ACPI System Control Interrupt trigger mode |
| 349 | Format: { level | edge | high | low } |
| 350 | |
| 351 | acpi_skip_timer_override [HW,ACPI,EARLY] |
| 352 | Recognize and ignore IRQ0/pin2 Interrupt Override. |
| 353 | For broken nForce2 BIOS resulting in XT-PIC timer. |
| 354 | |
| 355 | acpi_sleep= [HW,ACPI] Sleep options |
| 356 | Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, |
| 357 | s4_nohwsig, old_ordering, nonvs, |
| 358 | sci_force_enable, nobl } |
| 359 | See Documentation/power/video.rst for information on |
| 360 | s3_bios and s3_mode. |
| 361 | s3_beep is for debugging; it makes the PC's speaker beep |
| 362 | as soon as the kernel's real-mode entry point is called. |
| 363 | s4_hwsig causes the kernel to check the ACPI hardware |
| 364 | signature during resume from hibernation, and gracefully |
| 365 | refuse to resume if it has changed. This complies with |
| 366 | the ACPI specification but not with reality, since |
| 367 | Windows does not do this and many laptops do change it |
| 368 | on docking. So the default behaviour is to allow resume |
| 369 | and simply warn when the signature changes, unless the |
| 370 | s4_hwsig option is enabled. |
| 371 | s4_nohwsig prevents ACPI hardware signature from being |
| 372 | used (or even warned about) during resume. |
| 373 | old_ordering causes the ACPI 1.0 ordering of the _PTS |
| 374 | control method, with respect to putting devices into |
| 375 | low power states, to be enforced (the ACPI 2.0 ordering |
| 376 | of _PTS is used by default). |
| 377 | nonvs prevents the kernel from saving/restoring the |
| 378 | ACPI NVS memory during suspend/hibernation and resume. |
| 379 | sci_force_enable causes the kernel to set SCI_EN directly |
| 380 | on resume from S1/S3 (which is against the ACPI spec, |
| 381 | but some broken systems don't work without it). |
| 382 | nobl causes the internal blacklist of systems known to |
| 383 | behave incorrectly in some ways with respect to system |
| 384 | suspend and resume to be ignored (use wisely). |
| 385 | |
| 386 | acpi_use_timer_override [HW,ACPI,EARLY] |
| 387 | Use timer override. For some broken Nvidia NF5 boards |
| 388 | that require a timer override, but don't have HPET |
| 389 | |
| 390 | add_efi_memmap [EFI,X86,EARLY] Include EFI memory map in |
| 391 | kernel's map of available physical RAM. |
| 392 | |
| 393 | agp= [AGP] |
| 394 | { off | try_unsupported } |
| 395 | off: disable AGP support |
| 396 | try_unsupported: try to drive unsupported chipsets |
| 397 | (may crash computer or cause data corruption) |
| 398 | |
| 399 | ALSA [HW,ALSA] |
| 400 | See Documentation/sound/alsa-configuration.rst |
| 401 | |
| 402 | alignment= [KNL,ARM] |
| 403 | Allow the default userspace alignment fault handler |
| 404 | behaviour to be specified. Bit 0 enables warnings, |
| 405 | bit 1 enables fixups, and bit 2 sends a segfault. |
| 406 | |
| 407 | align_va_addr= [X86-64] |
| 408 | Align virtual addresses by clearing slice [14:12] when |
| 409 | allocating a VMA at process creation time. This option |
| 410 | gives you up to 3% performance improvement on AMD F15h |
| 411 | machines (where it is enabled by default) for a |
| 412 | CPU-intensive style benchmark, and it can vary highly in |
| 413 | a microbenchmark depending on workload and compiler. |
| 414 | |
| 415 | 32: only for 32-bit processes |
| 416 | 64: only for 64-bit processes |
| 417 | on: enable for both 32- and 64-bit processes |
| 418 | off: disable for both 32- and 64-bit processes |
| 419 | |
| 420 | alloc_snapshot [FTRACE] |
| 421 | Allocate the ftrace snapshot buffer on boot up when the |
| 422 | main buffer is allocated. This is handy if debugging |
| 423 | and you need to use tracing_snapshot() on boot up, and |
| 424 | do not want to use tracing_snapshot_alloc() as it needs |
| 425 | to be done where GFP_KERNEL allocations are allowed. |
| 426 | |
| 427 | allow_mismatched_32bit_el0 [ARM64,EARLY] |
| 428 | Allow execve() of 32-bit applications and setting of the |
| 429 | PER_LINUX32 personality on systems where only a strict |
| 430 | subset of the CPUs support 32-bit EL0. When this |
| 431 | parameter is present, the set of CPUs supporting 32-bit |
| 432 | EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 |
| 433 | and hot-unplug operations may be restricted. |
| 434 | |
| 435 | See Documentation/arch/arm64/asymmetric-32bit.rst for more |
| 436 | information. |
| 437 | |
| 438 | amd_iommu= [HW,X86-64] |
| 439 | Pass parameters to the AMD IOMMU driver in the system. |
| 440 | Possible values are: |
| 441 | fullflush - Deprecated, equivalent to iommu.strict=1 |
| 442 | off - do not initialize any AMD IOMMU found in |
| 443 | the system |
| 444 | force_isolation - Force device isolation for all |
| 445 | devices. The IOMMU driver is not |
| 446 | allowed anymore to lift isolation |
| 447 | requirements as needed. This option |
| 448 | does not override iommu=pt |
| 449 | force_enable - Force enable the IOMMU on platforms known |
| 450 | to be buggy with IOMMU enabled. Use this |
| 451 | option with care. |
| 452 | pgtbl_v1 - Use v1 page table for DMA-API (Default). |
| 453 | pgtbl_v2 - Use v2 page table for DMA-API. |
| 454 | irtcachedis - Disable Interrupt Remapping Table (IRT) caching. |
| 455 | nohugepages - Limit page-sizes used for v1 page-tables |
| 456 | to 4 KiB. |
| 457 | v2_pgsizes_only - Limit page-sizes used for v1 page-tables |
| 458 | to 4KiB/2Mib/1GiB. |
| 459 | |
| 460 | |
| 461 | amd_iommu_dump= [HW,X86-64] |
| 462 | Enable AMD IOMMU driver option to dump the ACPI table |
| 463 | for AMD IOMMU. With this option enabled, AMD IOMMU |
| 464 | driver will print ACPI tables for AMD IOMMU during |
| 465 | IOMMU initialization. |
| 466 | |
| 467 | amd_iommu_intr= [HW,X86-64] |
| 468 | Specifies one of the following AMD IOMMU interrupt |
| 469 | remapping modes: |
| 470 | legacy - Use legacy interrupt remapping mode. |
| 471 | vapic - Use virtual APIC mode, which allows IOMMU |
| 472 | to inject interrupts directly into guest. |
| 473 | This mode requires kvm-amd.avic=1. |
| 474 | (Default when IOMMU HW support is present.) |
| 475 | |
| 476 | amd_pstate= [X86,EARLY] |
| 477 | disable |
| 478 | Do not enable amd_pstate as the default |
| 479 | scaling driver for the supported processors |
| 480 | passive |
| 481 | Use amd_pstate with passive mode as a scaling driver. |
| 482 | In this mode autonomous selection is disabled. |
| 483 | Driver requests a desired performance level and platform |
| 484 | tries to match the same performance level if it is |
| 485 | satisfied by guaranteed performance level. |
| 486 | active |
| 487 | Use amd_pstate_epp driver instance as the scaling driver, |
| 488 | driver provides a hint to the hardware if software wants |
| 489 | to bias toward performance (0x0) or energy efficiency (0xff) |
| 490 | to the CPPC firmware. then CPPC power algorithm will |
| 491 | calculate the runtime workload and adjust the realtime cores |
| 492 | frequency. |
| 493 | guided |
| 494 | Activate guided autonomous mode. Driver requests minimum and |
| 495 | maximum performance level and the platform autonomously |
| 496 | selects a performance level in this range and appropriate |
| 497 | to the current workload. |
| 498 | |
| 499 | amd_prefcore= |
| 500 | [X86] |
| 501 | disable |
| 502 | Disable amd-pstate preferred core. |
| 503 | |
| 504 | amd_dynamic_epp= |
| 505 | [X86] |
| 506 | disable |
| 507 | Disable amd-pstate dynamic EPP. |
| 508 | enable |
| 509 | Enable amd-pstate dynamic EPP. |
| 510 | |
| 511 | amijoy.map= [HW,JOY] Amiga joystick support |
| 512 | Map of devices attached to JOY0DAT and JOY1DAT |
| 513 | Format: <a>,<b> |
| 514 | See also Documentation/input/joydev/joystick.rst |
| 515 | |
| 516 | analog.map= [HW,JOY] Analog joystick and gamepad support |
| 517 | Specifies type or capabilities of an analog joystick |
| 518 | connected to one of 16 gameports |
| 519 | Format: <type1>,<type2>,..<type16> |
| 520 | |
| 521 | apc= [HW,SPARC] |
| 522 | Power management functions (SPARCstation-4/5 + deriv.) |
| 523 | Format: noidle |
| 524 | Disable APC CPU standby support. SPARCstation-Fox does |
| 525 | not play well with APC CPU idle - disable it if you have |
| 526 | APC and your system crashes randomly. |
| 527 | |
| 528 | apic [APIC,X86-64] Use IO-APIC. Default. |
| 529 | |
| 530 | apic= [APIC,X86,EARLY] Advanced Programmable Interrupt Controller |
| 531 | Change the output verbosity while booting |
| 532 | Format: { quiet (default) | verbose | debug } |
| 533 | Change the amount of debugging information output |
| 534 | when initialising the APIC and IO-APIC components. |
| 535 | |
| 536 | apic_extnmi= [APIC,X86,EARLY] External NMI delivery setting |
| 537 | Format: { bsp (default) | all | none } |
| 538 | bsp: External NMI is delivered only to CPU 0 |
| 539 | all: External NMIs are broadcast to all CPUs as a |
| 540 | backup of CPU 0 |
| 541 | none: External NMI is masked for all CPUs. This is |
| 542 | useful so that a dump capture kernel won't be |
| 543 | shot down by NMI |
| 544 | |
| 545 | apicpmtimer Do APIC timer calibration using the pmtimer. Implies |
| 546 | apicmaintimer. Useful when your PIT timer is totally |
| 547 | broken. |
| 548 | |
| 549 | autoconf= [IPV6] |
| 550 | See Documentation/networking/ipv6.rst. |
| 551 | |
| 552 | apm= [APM] Advanced Power Management |
| 553 | See header of arch/x86/kernel/apm_32.c. |
| 554 | |
| 555 | apparmor= [APPARMOR] Disable or enable AppArmor at boot time |
| 556 | Format: { "0" | "1" } |
| 557 | See security/apparmor/Kconfig help text |
| 558 | 0 -- disable. |
| 559 | 1 -- enable. |
| 560 | Default value is set via kernel config option. |
| 561 | |
| 562 | arm64.no32bit_el0 [ARM64] Unconditionally disable the execution of |
| 563 | 32 bit applications. |
| 564 | |
| 565 | arm64.nobti [ARM64] Unconditionally disable Branch Target |
| 566 | Identification support |
| 567 | |
| 568 | arm64.nogcs [ARM64] Unconditionally disable Guarded Control Stack |
| 569 | support |
| 570 | |
| 571 | arm64.nohaft [ARM64] Unconditionally disable Hardware managed Access |
| 572 | Flag for Table descriptors support |
| 573 | |
| 574 | arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory |
| 575 | Set instructions support |
| 576 | |
| 577 | arm64.nompam [ARM64] Unconditionally disable Memory Partitioning And |
| 578 | Monitoring support. Only for a machine that does not |
| 579 | boot without it. See Documentation/arch/arm64/mpam.rst |
| 580 | |
| 581 | arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension |
| 582 | support |
| 583 | |
| 584 | arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication |
| 585 | support |
| 586 | |
| 587 | arm64.nosme [ARM64] Unconditionally disable Scalable Matrix |
| 588 | Extension support |
| 589 | |
| 590 | arm64.nosve [ARM64] Unconditionally disable Scalable Vector |
| 591 | Extension support |
| 592 | |
| 593 | arm64.nowfxt [ARM64] Unconditionally disable Wait For Event with |
| 594 | Timeout and Wait For Interrupt with Timeout support |
| 595 | |
| 596 | ataflop= [HW,M68k] |
| 597 | |
| 598 | atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, |
| 599 | EzKey and similar keyboards |
| 600 | |
| 601 | atkbd.reset= [HW] Reset keyboard during initialization |
| 602 | |
| 603 | atkbd.set= [HW] Select keyboard code set |
| 604 | Format: <int> (2 = AT (default), 3 = PS/2) |
| 605 | |
| 606 | atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar |
| 607 | keyboards |
| 608 | |
| 609 | atkbd.softraw= [HW] Choose between synthetic and real raw mode |
| 610 | Format: <bool> (0 = real, 1 = synthetic (default)) |
| 611 | |
| 612 | atkbd.softrepeat= [HW] |
| 613 | Use software keyboard repeat |
| 614 | |
| 615 | audit= [KNL] Enable the audit sub-system |
| 616 | Format: { "0" | "1" | "off" | "on" } |
| 617 | 0 | off - kernel audit is disabled and can not be |
| 618 | enabled until the next reboot |
| 619 | unset - kernel audit is initialized but disabled and |
| 620 | will be fully enabled by the userspace auditd. |
| 621 | 1 | on - kernel audit is initialized and partially |
| 622 | enabled, storing at most audit_backlog_limit |
| 623 | messages in RAM until it is fully enabled by the |
| 624 | userspace auditd. |
| 625 | Default: unset |
| 626 | |
| 627 | audit_backlog_limit= [KNL] Set the audit queue size limit. |
| 628 | Format: <int> (must be >=0) |
| 629 | Default: 64 |
| 630 | |
| 631 | bdev_allow_write_mounted= |
| 632 | Format: <bool> |
| 633 | Control the ability to open a mounted block device |
| 634 | for writing, i.e., allow / disallow writes that bypass |
| 635 | the FS. This was implemented as a means to prevent |
| 636 | fuzzers from crashing the kernel by overwriting the |
| 637 | metadata underneath a mounted FS without its awareness. |
| 638 | This also prevents destructive formatting of mounted |
| 639 | filesystems by naive storage tooling that don't use |
| 640 | O_EXCL. Default is Y and can be changed through the |
| 641 | Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED. |
| 642 | |
| 643 | bert_disable [ACPI] |
| 644 | Disable BERT OS support on buggy BIOSes. |
| 645 | |
| 646 | bgrt_disable [ACPI,X86,EARLY] |
| 647 | Disable BGRT to avoid flickering OEM logo. |
| 648 | |
| 649 | blkdevparts= Manual partition parsing of block device(s) for |
| 650 | embedded devices based on command line input. |
| 651 | See Documentation/block/cmdline-partition.rst |
| 652 | |
| 653 | boot_delay= [KNL,EARLY] |
| 654 | Milliseconds to delay each printk during boot. |
| 655 | Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, |
| 656 | and you may also have to specify "lpj=". Boot_delay |
| 657 | values larger than 10 seconds (10000) are assumed |
| 658 | erroneous and ignored. |
| 659 | Format: integer |
| 660 | |
| 661 | bootconfig [KNL,EARLY] |
| 662 | Extended command line options can be added to an initrd |
| 663 | and this will cause the kernel to look for it. |
| 664 | |
| 665 | See Documentation/admin-guide/bootconfig.rst |
| 666 | |
| 667 | bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) |
| 668 | bttv.radio= Most important insmod options are available as |
| 669 | kernel args too. |
| 670 | bttv.pll= See Documentation/admin-guide/media/bttv.rst |
| 671 | bttv.tuner= |
| 672 | |
| 673 | bulk_remove=off [PPC] This parameter disables the use of the pSeries |
| 674 | firmware feature for flushing multiple hpte entries |
| 675 | at a time. |
| 676 | |
| 677 | c101= [NET] Moxa C101 synchronous serial card |
| 678 | |
| 679 | cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. |
| 680 | Sometimes CPU hardware bugs make them report the cache |
| 681 | size incorrectly. The kernel will attempt work arounds |
| 682 | to fix known problems, but for some CPUs it is not |
| 683 | possible to determine what the correct size should be. |
| 684 | This option provides an override for these situations. |
| 685 | |
| 686 | carrier_timeout= |
| 687 | [NET] Specifies amount of time (in seconds) that |
| 688 | the kernel should wait for a network carrier. By default |
| 689 | it waits 120 seconds. |
| 690 | |
| 691 | ca_keys= [KEYS] This parameter identifies a specific key(s) on |
| 692 | the system trusted keyring to be used for certificate |
| 693 | trust validation. |
| 694 | format: { id:<keyid> | builtin } |
| 695 | |
| 696 | cca= [MIPS,EARLY] Override the kernel pages' cache coherency |
| 697 | algorithm. Accepted values range from 0 to 7 |
| 698 | inclusive. See arch/mips/include/asm/pgtable-bits.h |
| 699 | for platform specific values (SB1, Loongson3 and |
| 700 | others). |
| 701 | |
| 702 | ccw_timeout_log [S390] |
| 703 | See Documentation/arch/s390/common_io.rst for details. |
| 704 | |
| 705 | cfi= [X86-64] Set Control Flow Integrity checking features |
| 706 | when CONFIG_FINEIBT is enabled. |
| 707 | Format: feature[,feature...] |
| 708 | Default: auto |
| 709 | |
| 710 | auto: Use FineIBT if IBT available, otherwise kCFI. |
| 711 | Under FineIBT, enable "paranoid" mode when |
| 712 | FRED is not available. |
| 713 | off: Turn off CFI checking. |
| 714 | kcfi: Use kCFI (disable FineIBT). |
| 715 | fineibt: Use FineIBT (even if IBT not available). |
| 716 | norand: Do not re-randomize CFI hashes. |
| 717 | paranoid: Add caller hash checking under FineIBT. |
| 718 | bhi: Enable register poisoning to stop speculation |
| 719 | across FineIBT. (Disabled by default.) |
| 720 | warn: Do not enforce CFI checking: warn only. |
| 721 | debug: Report CFI initialization details. |
| 722 | |
| 723 | cgroup_disable= [KNL] Disable a particular controller or optional feature |
| 724 | Format: {name of the controller(s) or feature(s) to disable} |
| 725 | The effects of cgroup_disable=foo are: |
| 726 | - foo isn't auto-mounted if you mount all cgroups in |
| 727 | a single hierarchy |
| 728 | - foo isn't visible as an individually mountable |
| 729 | subsystem |
| 730 | - if foo is an optional feature then the feature is |
| 731 | disabled and corresponding cgroup files are not |
| 732 | created |
| 733 | {Currently only "memory" controller deal with this and |
| 734 | cut the overhead, others just disable the usage. So |
| 735 | only cgroup_disable=memory is actually worthy} |
| 736 | Specifying "pressure" disables per-cgroup pressure |
| 737 | stall information accounting feature |
| 738 | |
| 739 | cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 |
| 740 | Format: { { controller | "all" | "named" } |
| 741 | [,{ controller | "all" | "named" }...] } |
| 742 | Like cgroup_disable, but only applies to cgroup v1; |
| 743 | the blacklisted controllers remain available in cgroup2. |
| 744 | "all" blacklists all controllers and "named" disables |
| 745 | named mounts. Specifying both "all" and "named" disables |
| 746 | all v1 hierarchies. |
| 747 | |
| 748 | cgroup_v1_proc= [KNL] Show also missing controllers in /proc/cgroups |
| 749 | Format: { "true" | "false" } |
| 750 | /proc/cgroups lists only v1 controllers by default. |
| 751 | This compatibility option enables listing also v2 |
| 752 | controllers (whose v1 code is not compiled!), so that |
| 753 | semi-legacy software can check this file to decide |
| 754 | about usage of v2 (sic) controllers. |
| 755 | |
| 756 | cgroup_favordynmods= [KNL] Enable or Disable favordynmods. |
| 757 | Format: { "true" | "false" } |
| 758 | Defaults to the value of CONFIG_CGROUP_FAVOR_DYNMODS. |
| 759 | |
| 760 | cgroup.memory= [KNL] Pass options to the cgroup memory controller. |
| 761 | Format: <string> |
| 762 | nosocket -- Disable socket memory accounting. |
| 763 | nokmem -- Disable kernel memory accounting. |
| 764 | nobpf -- Disable BPF memory accounting. |
| 765 | |
| 766 | check_pages= [MM,EARLY] Enable sanity checking of pages after |
| 767 | allocations / before freeing. This adds checks to catch |
| 768 | double-frees, use-after-frees, and other sources of |
| 769 | page corruption by inspecting page internals (flags, |
| 770 | mapcount/refcount, memcg_data, etc.). |
| 771 | Format: { "0" | "1" } |
| 772 | Default: 0 (1 if CONFIG_DEBUG_VM is set) |
| 773 | |
| 774 | checkreqprot= [SELINUX] Set initial checkreqprot flag value. |
| 775 | Format: { "0" | "1" } |
| 776 | See security/selinux/Kconfig help text. |
| 777 | 0 -- check protection applied by kernel (includes |
| 778 | any implied execute protection). |
| 779 | 1 -- check protection requested by application. |
| 780 | Default value is set via a kernel config option. |
| 781 | Value can be changed at runtime via |
| 782 | /sys/fs/selinux/checkreqprot. |
| 783 | Setting checkreqprot to 1 is deprecated. |
| 784 | |
| 785 | cio_ignore= [S390] |
| 786 | See Documentation/arch/s390/common_io.rst for details. |
| 787 | |
| 788 | clk_ignore_unused |
| 789 | [CLK] |
| 790 | Prevents the clock framework from automatically gating |
| 791 | clocks that have not been explicitly enabled by a Linux |
| 792 | device driver but are enabled in hardware at reset or |
| 793 | by the bootloader/firmware. Note that this does not |
| 794 | force such clocks to be always-on nor does it reserve |
| 795 | those clocks in any way. This parameter is useful for |
| 796 | debug and development, but should not be needed on a |
| 797 | platform with proper driver support. For more |
| 798 | information, see Documentation/driver-api/clk.rst. |
| 799 | |
| 800 | clock= [BUGS=X86-32, HW] gettimeofday clocksource override. |
| 801 | [Deprecated] |
| 802 | Forces specified clocksource (if available) to be used |
| 803 | when calculating gettimeofday(). If specified |
| 804 | clocksource is not available, it defaults to PIT. |
| 805 | Format: { pit | tsc | cyclone | pmtmr } |
| 806 | |
| 807 | clocksource= Override the default clocksource |
| 808 | Format: <string> |
| 809 | Override the default clocksource and use the clocksource |
| 810 | with the name specified. |
| 811 | Some clocksource names to choose from, depending on |
| 812 | the platform: |
| 813 | [all] jiffies (this is the base, fallback clocksource) |
| 814 | [ACPI] acpi_pm |
| 815 | [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, |
| 816 | pxa_timer,timer3,32k_counter,timer0_1 |
| 817 | [X86-32] pit,hpet,tsc; |
| 818 | scx200_hrt on Geode; cyclone on IBM x440 |
| 819 | [MIPS] MIPS |
| 820 | [PARISC] cr16 |
| 821 | [S390] tod |
| 822 | [SH] SuperH |
| 823 | [SPARC64] tick |
| 824 | [X86-64] hpet,tsc |
| 825 | |
| 826 | clocksource.arm_arch_timer.evtstrm= |
| 827 | [ARM,ARM64,EARLY] |
| 828 | Format: <bool> |
| 829 | Enable/disable the eventstream feature of the ARM |
| 830 | architected timer so that code using WFE-based polling |
| 831 | loops can be debugged more effectively on production |
| 832 | systems. |
| 833 | |
| 834 | clocksource.verify_n_cpus= [KNL] |
| 835 | Limit the number of CPUs checked for clocksources |
| 836 | marked with CLOCK_SOURCE_VERIFY_PERCPU that |
| 837 | are marked unstable due to excessive skew. |
| 838 | A negative value says to check all CPUs, while |
| 839 | zero says not to check any. Values larger than |
| 840 | nr_cpu_ids are silently truncated to nr_cpu_ids. |
| 841 | The actual CPUs are chosen randomly, with |
| 842 | no replacement if the same CPU is chosen twice. |
| 843 | |
| 844 | clocksource-wdtest.holdoff= [KNL] |
| 845 | Set the time in seconds that the clocksource |
| 846 | watchdog test waits before commencing its tests. |
| 847 | Defaults to zero when built as a module and to |
| 848 | 10 seconds when built into the kernel. |
| 849 | |
| 850 | cma=nn[MG]@[start[MG][-end[MG]]] |
| 851 | [KNL,CMA,EARLY] |
| 852 | Sets the size of kernel global memory area for |
| 853 | contiguous memory allocations and optionally the |
| 854 | placement constraint by the physical address range of |
| 855 | memory allocations. A value of 0 disables CMA |
| 856 | altogether. For more information, see |
| 857 | kernel/dma/contiguous.c |
| 858 | |
| 859 | cma_pernuma=nn[MG] |
| 860 | [KNL,CMA,EARLY] |
| 861 | Sets the size of kernel per-numa memory area for |
| 862 | contiguous memory allocations. A value of 0 disables |
| 863 | per-numa CMA altogether. And If this option is not |
| 864 | specified, the default value is 0. |
| 865 | With per-numa CMA enabled, DMA users on node nid will |
| 866 | first try to allocate buffer from the pernuma area |
| 867 | which is located in node nid, if the allocation fails, |
| 868 | they will fallback to the global default memory area. |
| 869 | |
| 870 | numa_cma=<node>:nn[MG][,<node>:nn[MG]] |
| 871 | [KNL,CMA,EARLY] |
| 872 | Sets the size of kernel numa memory area for |
| 873 | contiguous memory allocations. It will reserve CMA |
| 874 | area for the specified node. |
| 875 | |
| 876 | If it is setup together with upper 'cmd_pernuma=' |
| 877 | (unlikely), its size setting takes priority for the |
| 878 | specified numa nodes. |
| 879 | |
| 880 | With numa CMA enabled, DMA users on node nid will |
| 881 | first try to allocate buffer from the numa area |
| 882 | which is located in node nid, if the allocation fails, |
| 883 | they will fallback to the global default memory area. |
| 884 | |
| 885 | cmo_free_hint= [PPC] Format: { yes | no } |
| 886 | Specify whether pages are marked as being inactive |
| 887 | when they are freed. This is used in CMO environments |
| 888 | to determine OS memory pressure for page stealing by |
| 889 | a hypervisor. |
| 890 | Default: yes |
| 891 | |
| 892 | coherent_pool=nn[KMG] [ARM,KNL,EARLY] |
| 893 | Sets the size of memory pool for coherent, atomic dma |
| 894 | allocations, by default set to 256K. |
| 895 | |
| 896 | condev= [HW,S390] console device |
| 897 | conmode= |
| 898 | |
| 899 | con3215_drop= [S390,EARLY] 3215 console drop mode. |
| 900 | Format: y|n|Y|N|1|0 |
| 901 | When set to true, drop data on the 3215 console when |
| 902 | the console buffer is full. In this case the |
| 903 | operator using a 3270 terminal emulator (for example |
| 904 | x3270) does not have to enter the clear key for the |
| 905 | console output to advance and the kernel to continue. |
| 906 | This leads to a much faster boot time when a 3270 |
| 907 | terminal emulator is active. If no 3270 terminal |
| 908 | emulator is used, this parameter has no effect. |
| 909 | |
| 910 | console= [KNL] Output console device and options. |
| 911 | |
| 912 | tty<n> Use the virtual console device <n>. |
| 913 | |
| 914 | ttyS<n>[,options] |
| 915 | ttyUSB0[,options] |
| 916 | Use the specified serial port. The options are of |
| 917 | the form "bbbbpnf", where "bbbb" is the baud rate, |
| 918 | "p" is parity ("n", "o", or "e"), "n" is number of |
| 919 | bits, and "f" is flow control ("r" for RTS or |
| 920 | omit it). Default is "9600n8". |
| 921 | |
| 922 | See Documentation/admin-guide/serial-console.rst for more |
| 923 | information. See |
| 924 | Documentation/networking/netconsole.rst for an |
| 925 | alternative. |
| 926 | |
| 927 | <DEVNAME>:<n>.<n>[,options] |
| 928 | Use the specified serial port on the serial core bus. |
| 929 | The addressing uses DEVNAME of the physical serial port |
| 930 | device, followed by the serial core controller instance, |
| 931 | and the serial port instance. The options are the same |
| 932 | as documented for the ttyS addressing above. |
| 933 | |
| 934 | The mapping of the serial ports to the tty instances |
| 935 | can be viewed with: |
| 936 | |
| 937 | $ ls -d /sys/bus/serial-base/devices/*:*.*/tty/* |
| 938 | /sys/bus/serial-base/devices/00:04:0.0/tty/ttyS0 |
| 939 | |
| 940 | In the above example, the console can be addressed with |
| 941 | console=00:04:0.0. Note that a console addressed this |
| 942 | way will only get added when the related device driver |
| 943 | is ready. The use of an earlycon parameter in addition to |
| 944 | the console may be desired for console output early on. |
| 945 | |
| 946 | uart[8250],io,<addr>[,options] |
| 947 | uart[8250],mmio,<addr>[,options] |
| 948 | uart[8250],mmio16,<addr>[,options] |
| 949 | uart[8250],mmio32,<addr>[,options] |
| 950 | uart[8250],0x<addr>[,options] |
| 951 | Start an early, polled-mode console on the 8250/16550 |
| 952 | UART at the specified I/O port or MMIO address, |
| 953 | switching to the matching ttyS device later. |
| 954 | MMIO inter-register address stride is either 8-bit |
| 955 | (mmio), 16-bit (mmio16), or 32-bit (mmio32). |
| 956 | If none of [io|mmio|mmio16|mmio32], <addr> is assumed |
| 957 | to be equivalent to 'mmio'. 'options' are specified in |
| 958 | the same format described for ttyS above; if unspecified, |
| 959 | the h/w is not re-initialized. |
| 960 | |
| 961 | hvc<n> Use the hypervisor console device <n>. This is for |
| 962 | both Xen and PowerPC hypervisors. |
| 963 | |
| 964 | { null | "" } |
| 965 | Use to disable console output, i.e., to have kernel |
| 966 | console messages discarded. |
| 967 | This must be the only console= parameter used on the |
| 968 | kernel command line. |
| 969 | |
| 970 | If the device connected to the port is not a TTY but a braille |
| 971 | device, prepend "brl," before the device type, for instance |
| 972 | console=brl,ttyS0 |
| 973 | For now, only VisioBraille is supported. |
| 974 | |
| 975 | console_msg_format= |
| 976 | [KNL] Change console messages format |
| 977 | default |
| 978 | By default we print messages on consoles in |
| 979 | "[time stamp] text\n" format (time stamp may not be |
| 980 | printed, depending on CONFIG_PRINTK_TIME or |
| 981 | `printk_time' param). |
| 982 | syslog |
| 983 | Switch to syslog format: "<%u>[time stamp] text\n" |
| 984 | IOW, each message will have a facility and loglevel |
| 985 | prefix. The format is similar to one used by syslog() |
| 986 | syscall, or to executing "dmesg -S --raw" or to reading |
| 987 | from /proc/kmsg. |
| 988 | |
| 989 | consoleblank= [KNL] The console blank (screen saver) timeout in |
| 990 | seconds. A value of 0 disables the blank timer. |
| 991 | Defaults to 0. |
| 992 | |
| 993 | coredump_filter= |
| 994 | [KNL] Change the default value for |
| 995 | /proc/<pid>/coredump_filter. |
| 996 | See also Documentation/filesystems/proc.rst. |
| 997 | |
| 998 | coresight_cpu_debug.enable |
| 999 | [ARM,ARM64] |
| 1000 | Format: <bool> |
| 1001 | Enable/disable the CPU sampling based debugging. |
| 1002 | 0: default value, disable debugging |
| 1003 | 1: enable debugging at boot time |
| 1004 | |
| 1005 | cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver |
| 1006 | Format: |
| 1007 | <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] |
| 1008 | |
| 1009 | cpuidle.off=1 [CPU_IDLE] |
| 1010 | disable the cpuidle sub-system |
| 1011 | |
| 1012 | cpuidle.governor= |
| 1013 | [CPU_IDLE] Name of the cpuidle governor to use. |
| 1014 | |
| 1015 | cpufreq.off=1 [CPU_FREQ] |
| 1016 | disable the cpufreq sub-system |
| 1017 | |
| 1018 | cpufreq.default_governor= |
| 1019 | [CPU_FREQ] Name of the default cpufreq governor or |
| 1020 | policy to use. This governor must be registered in the |
| 1021 | kernel before the cpufreq driver probes. |
| 1022 | |
| 1023 | cpu_init_udelay=N |
| 1024 | [X86,EARLY] Delay for N microsec between assert and de-assert |
| 1025 | of APIC INIT to start processors. This delay occurs |
| 1026 | on every CPU online, such as boot, and resume from suspend. |
| 1027 | Default: 10000 |
| 1028 | |
| 1029 | cpuhp.parallel= |
| 1030 | [SMP] Enable/disable parallel bringup of secondary CPUs |
| 1031 | Format: <bool> |
| 1032 | Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise |
| 1033 | the parameter has no effect. |
| 1034 | |
| 1035 | crash_kexec_post_notifiers |
| 1036 | Only jump to kdump kernel after running the panic |
| 1037 | notifiers and dumping kmsg. This option increases |
| 1038 | the risks of a kdump failure, since some panic |
| 1039 | notifiers can make the crashed kernel more unstable. |
| 1040 | In configurations where kdump may not be reliable, |
| 1041 | running the panic notifiers could allow collecting |
| 1042 | more data on dmesg, like stack traces from other CPUS |
| 1043 | or extra data dumped by panic_print. Note that some |
| 1044 | configurations enable this option unconditionally, |
| 1045 | like Hyper-V, PowerPC (fadump) and AMD SEV-SNP. |
| 1046 | |
| 1047 | crashkernel=size[KMG][@offset[KMG]] |
| 1048 | [KNL,EARLY] Using kexec, Linux can switch to a 'crash kernel' |
| 1049 | upon panic. This parameter reserves the physical |
| 1050 | memory region [offset, offset + size] for that kernel |
| 1051 | image. If '@offset' is omitted, then a suitable offset |
| 1052 | is selected automatically. |
| 1053 | [KNL, X86-64, ARM64, RISCV, LoongArch] Select a region |
| 1054 | under 4G first, and fall back to reserve region above |
| 1055 | 4G when '@offset' hasn't been specified. |
| 1056 | See Documentation/admin-guide/kdump/kdump.rst for further details. |
| 1057 | |
| 1058 | crashkernel=range1:size1[,range2:size2,...][@offset] |
| 1059 | [KNL] Same as above, but depends on the memory |
| 1060 | in the running system. The syntax of range is |
| 1061 | start-[end] where start and end are both |
| 1062 | a memory unit (amount[KMG]). See also |
| 1063 | Documentation/admin-guide/kdump/kdump.rst for an example. |
| 1064 | |
| 1065 | crashkernel=size[KMG],high |
| 1066 | [KNL, X86-64, ARM64, RISCV, LoongArch] range could be |
| 1067 | above 4G. |
| 1068 | Allow kernel to allocate physical memory region from top, |
| 1069 | so could be above 4G if system have more than 4G ram |
| 1070 | installed. Otherwise memory region will be allocated |
| 1071 | below 4G, if available. |
| 1072 | It will be ignored if crashkernel=X is specified. |
| 1073 | crashkernel=size[KMG],low |
| 1074 | [KNL, X86-64, ARM64, RISCV, LoongArch] range under 4G. |
| 1075 | When crashkernel=X,high is passed, kernel could allocate |
| 1076 | physical memory region above 4G, that cause second kernel |
| 1077 | crash on system that require some amount of low memory, |
| 1078 | e.g. swiotlb requires at least 64M+32K low memory, also |
| 1079 | enough extra low memory is needed to make sure DMA buffers |
| 1080 | for 32-bit devices won't run out. Kernel would try to allocate |
| 1081 | default size of memory below 4G automatically. The default |
| 1082 | size is platform dependent. |
| 1083 | --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) |
| 1084 | --> arm64: 128MiB |
| 1085 | --> riscv: 128MiB |
| 1086 | --> loongarch: 128MiB |
| 1087 | This one lets the user specify own low range under 4G |
| 1088 | for second kernel instead. |
| 1089 | 0: to disable low allocation. |
| 1090 | It will be ignored when crashkernel=X,high is not used |
| 1091 | or memory reserved is below 4G. |
| 1092 | crashkernel=size[KMG],cma |
| 1093 | [KNL, X86, ARM64, RISCV, PPC] Reserve additional crash |
| 1094 | kernel memory from CMA. This reservation is usable by |
| 1095 | the first system's userspace memory and kernel movable |
| 1096 | allocations (memory balloon, zswap). Pages allocated |
| 1097 | from this memory range will not be included in the vmcore |
| 1098 | so this should not be used if dumping of userspace memory |
| 1099 | is intended and it has to be expected that some movable |
| 1100 | kernel pages may be missing from the dump. |
| 1101 | |
| 1102 | A standard crashkernel reservation, as described above, |
| 1103 | is still needed to hold the crash kernel and initrd. |
| 1104 | |
| 1105 | This option increases the risk of a kdump failure: DMA |
| 1106 | transfers configured by the first kernel may end up |
| 1107 | corrupting the second kernel's memory. |
| 1108 | |
| 1109 | This reservation method is intended for systems that |
| 1110 | can't afford to sacrifice enough memory for standard |
| 1111 | crashkernel reservation and where less reliable and |
| 1112 | possibly incomplete kdump is preferable to no kdump at |
| 1113 | all. |
| 1114 | |
| 1115 | cryptomgr.notests |
| 1116 | [KNL] Disable crypto self-tests |
| 1117 | |
| 1118 | cs89x0_dma= [HW,NET] |
| 1119 | Format: <dma> |
| 1120 | |
| 1121 | cs89x0_media= [HW,NET] |
| 1122 | Format: { rj45 | aui | bnc } |
| 1123 | |
| 1124 | csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU |
| 1125 | function call handling. When switched on, |
| 1126 | additional debug data is printed to the console |
| 1127 | in case a hanging CPU is detected, and that |
| 1128 | CPU is pinged again in order to try to resolve |
| 1129 | the hang situation. The default value of this |
| 1130 | option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT |
| 1131 | Kconfig option. |
| 1132 | |
| 1133 | dasd= [HW,NET] |
| 1134 | See header of drivers/s390/block/dasd_devmap.c. |
| 1135 | |
| 1136 | db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port |
| 1137 | (one device per port) |
| 1138 | Format: <port#>,<type> |
| 1139 | See also Documentation/input/devices/joystick-parport.rst |
| 1140 | |
| 1141 | debug [KNL,EARLY] Enable kernel debugging (events log level). |
| 1142 | |
| 1143 | debug_boot_weak_hash |
| 1144 | [KNL,EARLY] Enable printing [hashed] pointers early in the |
| 1145 | boot sequence. If enabled, we use a weak hash instead |
| 1146 | of siphash to hash pointers. Use this option if you are |
| 1147 | seeing instances of '(___ptrval___)') and need to see a |
| 1148 | value (hashed pointer) instead. Cryptographically |
| 1149 | insecure, please do not use on production kernels. |
| 1150 | |
| 1151 | debug_locks_verbose= |
| 1152 | [KNL] verbose locking self-tests |
| 1153 | Format: <int> |
| 1154 | Print debugging info while doing the locking API |
| 1155 | self-tests. |
| 1156 | Bitmask for the various LOCKTYPE_ tests. Defaults to 0 |
| 1157 | (no extra messages), setting it to -1 (all bits set) |
| 1158 | will print _a_lot_ more information - normally only |
| 1159 | useful to lockdep developers. |
| 1160 | |
| 1161 | debug_objects [KNL,EARLY] Enable object debugging |
| 1162 | |
| 1163 | debug_guardpage_minorder= |
| 1164 | [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this |
| 1165 | parameter allows control of the order of pages that will |
| 1166 | be intentionally kept free (and hence protected) by the |
| 1167 | buddy allocator. Bigger value increase the probability |
| 1168 | of catching random memory corruption, but reduce the |
| 1169 | amount of memory for normal system use. The maximum |
| 1170 | possible value is MAX_PAGE_ORDER/2. Setting this |
| 1171 | parameter to 1 or 2 should be enough to identify most |
| 1172 | random memory corruption problems caused by bugs in |
| 1173 | kernel or driver code when a CPU writes to (or reads |
| 1174 | from) a random memory location. Note that there exists |
| 1175 | a class of memory corruptions problems caused by buggy |
| 1176 | H/W or F/W or by drivers badly programming DMA |
| 1177 | (basically when memory is written at bus level and the |
| 1178 | CPU MMU is bypassed) which are not detectable by |
| 1179 | CONFIG_DEBUG_PAGEALLOC, hence this option will not |
| 1180 | help tracking down these problems. |
| 1181 | |
| 1182 | debug_pagealloc= |
| 1183 | [KNL,EARLY] When CONFIG_DEBUG_PAGEALLOC is set, this parameter |
| 1184 | enables the feature at boot time. By default, it is |
| 1185 | disabled and the system will work mostly the same as a |
| 1186 | kernel built without CONFIG_DEBUG_PAGEALLOC. |
| 1187 | Note: to get most of debug_pagealloc error reports, it's |
| 1188 | useful to also enable the page_owner functionality. |
| 1189 | on: enable the feature |
| 1190 | |
| 1191 | debugfs= [KNL,EARLY] This parameter enables what is exposed to |
| 1192 | userspace and debugfs internal clients. |
| 1193 | Format: { on, off } |
| 1194 | on: All functions are enabled. |
| 1195 | off: Filesystem is not registered and clients |
| 1196 | get a -EPERM as result when trying to register files |
| 1197 | or directories within debugfs. |
| 1198 | This is equivalent of the runtime functionality if |
| 1199 | debugfs was not enabled in the kernel at all. |
| 1200 | Default value is set in build-time with a kernel configuration. |
| 1201 | |
| 1202 | debugpat [X86] Enable PAT debugging |
| 1203 | |
| 1204 | default_hugepagesz= |
| 1205 | [HW] The size of the default HugeTLB page. This is |
| 1206 | the size represented by the legacy /proc/ hugepages |
| 1207 | APIs. In addition, this is the default hugetlb size |
| 1208 | used for shmget(), mmap() and mounting hugetlbfs |
| 1209 | filesystems. If not specified, defaults to the |
| 1210 | architecture's default huge page size. Huge page |
| 1211 | sizes are architecture dependent. See also |
| 1212 | Documentation/admin-guide/mm/hugetlbpage.rst. |
| 1213 | Format: size[KMG] |
| 1214 | |
| 1215 | deferred_probe_timeout= |
| 1216 | [KNL] Debugging option to set a timeout in seconds for |
| 1217 | deferred probe to give up waiting on dependencies to |
| 1218 | probe. Only specific dependencies (subsystems or |
| 1219 | drivers) that have opted in will be ignored. A timeout |
| 1220 | of 0 will timeout at the end of initcalls; a negative |
| 1221 | value is treated as an infinite timeout value. If the |
| 1222 | timeout hasn't expired, it'll be restarted by each |
| 1223 | successful driver registration. This option will also |
| 1224 | dump out devices still on the deferred probe list after |
| 1225 | retrying. |
| 1226 | |
| 1227 | delayacct [KNL] Enable per-task delay accounting |
| 1228 | |
| 1229 | dell_smm_hwmon.ignore_dmi= |
| 1230 | [HW] Continue probing hardware even if DMI data |
| 1231 | indicates that the driver is running on unsupported |
| 1232 | hardware. |
| 1233 | |
| 1234 | dell_smm_hwmon.force= |
| 1235 | [HW] Activate driver even if SMM BIOS signature does |
| 1236 | not match list of supported models and enable otherwise |
| 1237 | blacklisted features. |
| 1238 | |
| 1239 | dell_smm_hwmon.power_status= |
| 1240 | [HW] Report power status in /proc/i8k |
| 1241 | (disabled by default). |
| 1242 | |
| 1243 | dell_smm_hwmon.restricted= |
| 1244 | [HW] Allow controlling fans only if SYS_ADMIN |
| 1245 | capability is set. |
| 1246 | |
| 1247 | dell_smm_hwmon.fan_mult= |
| 1248 | [HW] Factor to multiply fan speed with. |
| 1249 | |
| 1250 | dell_smm_hwmon.fan_max= |
| 1251 | [HW] Maximum configurable fan speed. |
| 1252 | |
| 1253 | dfltcc= [HW,S390] |
| 1254 | Format: { on | off | def_only | inf_only | always } |
| 1255 | on: s390 zlib hardware support for compression on |
| 1256 | level 1 and decompression (default) |
| 1257 | off: No s390 zlib hardware support |
| 1258 | def_only: s390 zlib hardware support for deflate |
| 1259 | only (compression on level 1) |
| 1260 | inf_only: s390 zlib hardware support for inflate |
| 1261 | only (decompression) |
| 1262 | always: Same as 'on' but ignores the selected compression |
| 1263 | level always using hardware support (used for debugging) |
| 1264 | |
| 1265 | dhash_entries= [KNL] |
| 1266 | Set number of hash buckets for dentry cache. |
| 1267 | |
| 1268 | disable_1tb_segments [PPC,EARLY] |
| 1269 | Disables the use of 1TB hash page table segments. This |
| 1270 | causes the kernel to fall back to 256MB segments which |
| 1271 | can be useful when debugging issues that require an SLB |
| 1272 | miss to occur. |
| 1273 | |
| 1274 | disable= [IPV6] |
| 1275 | See Documentation/networking/ipv6.rst. |
| 1276 | |
| 1277 | disable_radix [PPC,EARLY] |
| 1278 | Disable RADIX MMU mode on POWER9 |
| 1279 | |
| 1280 | disable_tlbie [PPC] |
| 1281 | Disable TLBIE instruction. Currently does not work |
| 1282 | with KVM, with HASH MMU, or with coherent accelerators. |
| 1283 | |
| 1284 | disable_ddw [PPC/PSERIES,EARLY] |
| 1285 | Disable Dynamic DMA Window support. Use this |
| 1286 | to workaround buggy firmware. |
| 1287 | |
| 1288 | disable_ipv6= [IPV6] |
| 1289 | See Documentation/networking/ipv6.rst. |
| 1290 | |
| 1291 | disable_mtrr_cleanup [X86,EARLY] |
| 1292 | The kernel tries to adjust MTRR layout from continuous |
| 1293 | to discrete, to make X server driver able to add WB |
| 1294 | entry later. This parameter disables that. |
| 1295 | |
| 1296 | disable_mtrr_trim [X86, Intel and AMD only,EARLY] |
| 1297 | By default the kernel will trim any uncacheable |
| 1298 | memory out of your available memory pool based on |
| 1299 | MTRR settings. This parameter disables that behavior, |
| 1300 | possibly causing your machine to run very slowly. |
| 1301 | |
| 1302 | disable_timer_pin_1 [X86,EARLY] |
| 1303 | Disable PIN 1 of APIC timer |
| 1304 | Can be useful to work around chipset bugs. |
| 1305 | |
| 1306 | dis_ucode_ldr [X86] Disable the microcode loader. |
| 1307 | |
| 1308 | dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, |
| 1309 | this option disables the debugging code at boot. |
| 1310 | |
| 1311 | dma_debug_entries=<number> |
| 1312 | This option allows to tune the number of preallocated |
| 1313 | entries for DMA-API debugging code. One entry is |
| 1314 | required per DMA-API allocation. Use this if the |
| 1315 | DMA-API debugging code disables itself because the |
| 1316 | architectural default is too low. |
| 1317 | |
| 1318 | dma_debug_driver=<driver_name> |
| 1319 | With this option the DMA-API debugging driver |
| 1320 | filter feature can be enabled at boot time. Just |
| 1321 | pass the driver to filter for as the parameter. |
| 1322 | The filter can be disabled or changed to another |
| 1323 | driver later using sysfs. |
| 1324 | |
| 1325 | reg_file_data_sampling= |
| 1326 | [X86] Controls mitigation for Register File Data |
| 1327 | Sampling (RFDS) vulnerability. RFDS is a CPU |
| 1328 | vulnerability which may allow userspace to infer |
| 1329 | kernel data values previously stored in floating point |
| 1330 | registers, vector registers, or integer registers. |
| 1331 | RFDS only affects Intel Atom processors. |
| 1332 | |
| 1333 | on: Turns ON the mitigation. |
| 1334 | off: Turns OFF the mitigation. |
| 1335 | |
| 1336 | This parameter overrides the compile time default set |
| 1337 | by CONFIG_MITIGATION_RFDS. Mitigation cannot be |
| 1338 | disabled when other VERW based mitigations (like MDS) |
| 1339 | are enabled. In order to disable RFDS mitigation all |
| 1340 | VERW based mitigations need to be disabled. |
| 1341 | |
| 1342 | For details see: |
| 1343 | Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst |
| 1344 | |
| 1345 | dm_verity.keyring_unsealed= |
| 1346 | [KNL] When set to 1, leave the dm-verity keyring |
| 1347 | unsealed after initialization so userspace can |
| 1348 | provision keys. Once the keyring is restricted |
| 1349 | it becomes active and is searched during signature |
| 1350 | verification. |
| 1351 | |
| 1352 | driver_async_probe= [KNL] |
| 1353 | List of driver names to be probed asynchronously. * |
| 1354 | matches with all driver names. If * is specified, the |
| 1355 | rest of the listed driver names are those that will NOT |
| 1356 | match the *. |
| 1357 | Format: <driver_name1>,<driver_name2>... |
| 1358 | |
| 1359 | drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] |
| 1360 | Broken monitors, graphic adapters, KVMs and EDIDless |
| 1361 | panels may send no or incorrect EDID data sets. |
| 1362 | This parameter allows to specify an EDID data sets |
| 1363 | in the /lib/firmware directory that are used instead. |
| 1364 | An EDID data set will only be used for a particular |
| 1365 | connector, if its name and a colon are prepended to |
| 1366 | the EDID name. Each connector may use a unique EDID |
| 1367 | data set by separating the files with a comma. An EDID |
| 1368 | data set with no connector name will be used for |
| 1369 | any connectors not explicitly specified. |
| 1370 | |
| 1371 | dscc4.setup= [NET] |
| 1372 | |
| 1373 | dt_cpu_ftrs= [PPC,EARLY] |
| 1374 | Format: {"off" | "known"} |
| 1375 | Control how the dt_cpu_ftrs device-tree binding is |
| 1376 | used for CPU feature discovery and setup (if it |
| 1377 | exists). |
| 1378 | off: Do not use it, fall back to legacy cpu table. |
| 1379 | known: Do not pass through unknown features to guests |
| 1380 | or userspace, only those that the kernel is aware of. |
| 1381 | |
| 1382 | dump_apple_properties [X86] |
| 1383 | Dump name and content of EFI device properties on |
| 1384 | x86 Macs. Useful for driver authors to determine |
| 1385 | what data is available or for reverse-engineering. |
| 1386 | |
| 1387 | dyndbg[="val"] [KNL,DYNAMIC_DEBUG] |
| 1388 | <module>.dyndbg[="val"] |
| 1389 | Enable debug messages at boot time. See |
| 1390 | Documentation/admin-guide/dynamic-debug-howto.rst |
| 1391 | for details. |
| 1392 | |
| 1393 | early_ioremap_debug [KNL,EARLY] |
| 1394 | Enable debug messages in early_ioremap support. This |
| 1395 | is useful for tracking down temporary early mappings |
| 1396 | which are not unmapped. |
| 1397 | |
| 1398 | earlycon= [KNL,EARLY] Output early console device and options. |
| 1399 | |
| 1400 | When used with no options, the early console is |
| 1401 | determined by stdout-path property in device tree's |
| 1402 | chosen node or the ACPI SPCR table if supported by |
| 1403 | the platform. |
| 1404 | |
| 1405 | cdns,<addr>[,options] |
| 1406 | Start an early, polled-mode console on a Cadence |
| 1407 | (xuartps) serial port at the specified address. Only |
| 1408 | supported option is baud rate. If baud rate is not |
| 1409 | specified, the serial port must already be setup and |
| 1410 | configured. |
| 1411 | |
| 1412 | uart[8250],io,<addr>[,options[,uartclk]] |
| 1413 | uart[8250],mmio,<addr>[,options[,uartclk]] |
| 1414 | uart[8250],mmio32,<addr>[,options[,uartclk]] |
| 1415 | uart[8250],mmio32be,<addr>[,options[,uartclk]] |
| 1416 | uart[8250],0x<addr>[,options] |
| 1417 | Start an early, polled-mode console on the 8250/16550 |
| 1418 | UART at the specified I/O port or MMIO address. |
| 1419 | MMIO inter-register address stride is either 8-bit |
| 1420 | (mmio) or 32-bit (mmio32 or mmio32be). |
| 1421 | If none of [io|mmio|mmio32|mmio32be], <addr> is assumed |
| 1422 | to be equivalent to 'mmio'. 'options' are specified |
| 1423 | in the same format described for "console=ttyS<n>"; if |
| 1424 | unspecified, the h/w is not initialized. 'uartclk' is |
| 1425 | the uart clock frequency; if unspecified, it is set |
| 1426 | to 'BASE_BAUD' * 16. |
| 1427 | |
| 1428 | pl011,<addr> |
| 1429 | pl011,mmio32,<addr> |
| 1430 | Start an early, polled-mode console on a pl011 serial |
| 1431 | port at the specified address. The pl011 serial port |
| 1432 | must already be setup and configured. Options are not |
| 1433 | yet supported. If 'mmio32' is specified, then only |
| 1434 | the driver will use only 32-bit accessors to read/write |
| 1435 | the device registers. |
| 1436 | |
| 1437 | liteuart,<addr> |
| 1438 | Start an early console on a litex serial port at the |
| 1439 | specified address. The serial port must already be |
| 1440 | setup and configured. Options are not yet supported. |
| 1441 | |
| 1442 | meson,<addr> |
| 1443 | Start an early, polled-mode console on a meson serial |
| 1444 | port at the specified address. The serial port must |
| 1445 | already be setup and configured. Options are not yet |
| 1446 | supported. |
| 1447 | |
| 1448 | msm_serial,<addr> |
| 1449 | Start an early, polled-mode console on an msm serial |
| 1450 | port at the specified address. The serial port |
| 1451 | must already be setup and configured. Options are not |
| 1452 | yet supported. |
| 1453 | |
| 1454 | msm_serial_dm,<addr> |
| 1455 | Start an early, polled-mode console on an msm serial |
| 1456 | dm port at the specified address. The serial port |
| 1457 | must already be setup and configured. Options are not |
| 1458 | yet supported. |
| 1459 | |
| 1460 | owl,<addr> |
| 1461 | Start an early, polled-mode console on a serial port |
| 1462 | of an Actions Semi SoC, such as S500 or S900, at the |
| 1463 | specified address. The serial port must already be |
| 1464 | setup and configured. Options are not yet supported. |
| 1465 | |
| 1466 | rda,<addr> |
| 1467 | Start an early, polled-mode console on a serial port |
| 1468 | of an RDA Micro SoC, such as RDA8810PL, at the |
| 1469 | specified address. The serial port must already be |
| 1470 | setup and configured. Options are not yet supported. |
| 1471 | |
| 1472 | sbi |
| 1473 | Use RISC-V SBI (Supervisor Binary Interface) for early |
| 1474 | console. |
| 1475 | |
| 1476 | smh Use ARM semihosting calls for early console. |
| 1477 | |
| 1478 | s3c2410,<addr> |
| 1479 | s3c2412,<addr> |
| 1480 | s3c2440,<addr> |
| 1481 | s3c6400,<addr> |
| 1482 | s5pv210,<addr> |
| 1483 | exynos4210,<addr> |
| 1484 | Use early console provided by serial driver available |
| 1485 | on Samsung SoCs, requires selecting proper type and |
| 1486 | a correct base address of the selected UART port. The |
| 1487 | serial port must already be setup and configured. |
| 1488 | Options are not yet supported. |
| 1489 | |
| 1490 | lantiq,<addr> |
| 1491 | Start an early, polled-mode console on a lantiq serial |
| 1492 | (lqasc) port at the specified address. The serial port |
| 1493 | must already be setup and configured. Options are not |
| 1494 | yet supported. |
| 1495 | |
| 1496 | lpuart,<addr> |
| 1497 | lpuart32,<addr> |
| 1498 | Use early console provided by Freescale LP UART driver |
| 1499 | found on Freescale Vybrid and QorIQ LS1021A processors. |
| 1500 | A valid base address must be provided, and the serial |
| 1501 | port must already be setup and configured. |
| 1502 | |
| 1503 | ec_imx21,<addr> |
| 1504 | ec_imx6q,<addr> |
| 1505 | Start an early, polled-mode, output-only console on the |
| 1506 | Freescale i.MX UART at the specified address. The UART |
| 1507 | must already be setup and configured. |
| 1508 | |
| 1509 | ar3700_uart,<addr> |
| 1510 | Start an early, polled-mode console on the |
| 1511 | Armada 3700 serial port at the specified |
| 1512 | address. The serial port must already be setup |
| 1513 | and configured. Options are not yet supported. |
| 1514 | |
| 1515 | qcom_geni,<addr> |
| 1516 | Start an early, polled-mode console on a Qualcomm |
| 1517 | Generic Interface (GENI) based serial port at the |
| 1518 | specified address. The serial port must already be |
| 1519 | setup and configured. Options are not yet supported. |
| 1520 | |
| 1521 | efifb,[options] |
| 1522 | Start an early, unaccelerated console on the EFI |
| 1523 | memory mapped framebuffer (if available). On cache |
| 1524 | coherent non-x86 systems that use system memory for |
| 1525 | the framebuffer, pass the 'ram' option so that it is |
| 1526 | mapped with the correct attributes. |
| 1527 | |
| 1528 | linflex,<addr> |
| 1529 | Use early console provided by Freescale LINFlexD UART |
| 1530 | serial driver for NXP S32V234 SoCs. A valid base |
| 1531 | address must be provided, and the serial port must |
| 1532 | already be setup and configured. |
| 1533 | |
| 1534 | earlyprintk= [X86,SH,ARM,M68k,S390,UM,EARLY] |
| 1535 | earlyprintk=vga |
| 1536 | earlyprintk=sclp |
| 1537 | earlyprintk=xen |
| 1538 | earlyprintk=serial[,ttySn[,baudrate]] |
| 1539 | earlyprintk=serial[,0x...[,baudrate]] |
| 1540 | earlyprintk=ttySn[,baudrate] |
| 1541 | earlyprintk=dbgp[debugController#] |
| 1542 | earlyprintk=mmio32,membase[,{nocfg|baudrate}] |
| 1543 | earlyprintk=pciserial[,force],bus:device.function[,{nocfg|baudrate}] |
| 1544 | earlyprintk=xdbc[xhciController#] |
| 1545 | earlyprintk=bios |
| 1546 | |
| 1547 | earlyprintk is useful when the kernel crashes before |
| 1548 | the normal console is initialized. It is not enabled by |
| 1549 | default because it has some cosmetic problems. |
| 1550 | |
| 1551 | Use "nocfg" to skip UART configuration, assume |
| 1552 | BIOS/firmware has configured UART correctly. |
| 1553 | |
| 1554 | Append ",keep" to not disable it when the real console |
| 1555 | takes over. |
| 1556 | |
| 1557 | Only one of vga, serial, or usb debug port can |
| 1558 | be used at a time. |
| 1559 | |
| 1560 | Currently only ttyS0 and ttyS1 may be specified by |
| 1561 | name. Other I/O ports may be explicitly specified |
| 1562 | on some architectures (x86 and arm at least) by |
| 1563 | replacing ttySn with an I/O port address, like this: |
| 1564 | earlyprintk=serial,0x1008,115200 |
| 1565 | You can find the port for a given device in |
| 1566 | /proc/tty/driver/serial: |
| 1567 | 2: uart:ST16650V2 port:00001008 irq:18 ... |
| 1568 | |
| 1569 | Interaction with the standard serial driver is not |
| 1570 | very good. |
| 1571 | |
| 1572 | The VGA output is eventually overwritten by |
| 1573 | the real console. |
| 1574 | |
| 1575 | The xen option can only be used in Xen domains. |
| 1576 | |
| 1577 | The sclp output can only be used on s390. |
| 1578 | |
| 1579 | The bios output can only be used on SuperH. |
| 1580 | |
| 1581 | The optional "force" to "pciserial" enables use of a |
| 1582 | PCI device even when its classcode is not of the |
| 1583 | UART class. |
| 1584 | |
| 1585 | edd= [EDD] |
| 1586 | Format: {"off" | "on" | "skip[mbr]"} |
| 1587 | |
| 1588 | efi= [EFI,EARLY] |
| 1589 | Format: { "debug", "disable_early_pci_dma", |
| 1590 | "nochunk", "noruntime", "nosoftreserve", |
| 1591 | "novamap", "no_disable_early_pci_dma" } |
| 1592 | debug: enable misc debug output. |
| 1593 | disable_early_pci_dma: disable the busmaster bit on all |
| 1594 | PCI bridges while in the EFI boot stub. |
| 1595 | nochunk: disable reading files in "chunks" in the EFI |
| 1596 | boot stub, as chunking can cause problems with some |
| 1597 | firmware implementations. |
| 1598 | noruntime : disable EFI runtime services support |
| 1599 | nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) |
| 1600 | attribute may cause the kernel to reserve the |
| 1601 | memory range for a memory mapping driver to |
| 1602 | claim. Specify efi=nosoftreserve to disable this |
| 1603 | reservation and treat the memory by its base type |
| 1604 | (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). |
| 1605 | novamap: do not call SetVirtualAddressMap(). |
| 1606 | no_disable_early_pci_dma: Leave the busmaster bit set |
| 1607 | on all PCI bridges while in the EFI boot stub |
| 1608 | |
| 1609 | efi_no_storage_paranoia [EFI,X86,EARLY] |
| 1610 | The kernel reserves 5KB of EFI variable storage for |
| 1611 | safety, because some UEFI implementation may fail to |
| 1612 | boot if there's insufficient space in the EFI variable |
| 1613 | storage. |
| 1614 | |
| 1615 | Using this parameter, you can use the 5KB reservation |
| 1616 | in the EFI variable storage. |
| 1617 | |
| 1618 | However, Use this parameter only if you are really |
| 1619 | sure that your UEFI does sane gc and fulfills the spec |
| 1620 | otherwise your board may brick. |
| 1621 | |
| 1622 | efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT |
| 1623 | that is to be dynamically loaded by Linux. If there are |
| 1624 | multiple variables with the same name but with different |
| 1625 | vendor GUIDs, all of them will be loaded. See |
| 1626 | Documentation/admin-guide/acpi/ssdt-overlays.rst for details. |
| 1627 | |
| 1628 | |
| 1629 | eisa_irq_edge= [PARISC,HW] |
| 1630 | See header of drivers/parisc/eisa.c. |
| 1631 | |
| 1632 | ekgdboc= [X86,KGDB,EARLY] Allow early kernel console debugging |
| 1633 | Format: ekgdboc=kbd |
| 1634 | |
| 1635 | This is designed to be used in conjunction with |
| 1636 | the boot argument: earlyprintk=vga |
| 1637 | |
| 1638 | This parameter works in place of the kgdboc parameter |
| 1639 | but can only be used if the backing tty is available |
| 1640 | very early in the boot process. For early debugging |
| 1641 | via a serial port see kgdboc_earlycon instead. |
| 1642 | |
| 1643 | elfcorehdr=[size[KMG]@]offset[KMG] [PPC,SH,X86,S390,EARLY] |
| 1644 | Specifies physical address of start of kernel core |
| 1645 | image elf header and optionally the size. Generally |
| 1646 | kexec loader will pass this option to capture kernel. |
| 1647 | See Documentation/admin-guide/kdump/kdump.rst for details. |
| 1648 | |
| 1649 | enable_mtrr_cleanup [X86,EARLY] |
| 1650 | The kernel tries to adjust MTRR layout from continuous |
| 1651 | to discrete, to make X server driver able to add WB |
| 1652 | entry later. This parameter enables that. |
| 1653 | |
| 1654 | enforcing= [SELINUX] Set initial enforcing status. |
| 1655 | Format: {"0" | "1"} |
| 1656 | See security/selinux/Kconfig help text. |
| 1657 | 0 -- permissive (log only, no denials). |
| 1658 | 1 -- enforcing (deny and log). |
| 1659 | Default value is 0. |
| 1660 | Value can be changed at runtime via |
| 1661 | /sys/fs/selinux/enforce. |
| 1662 | |
| 1663 | erst_disable [ACPI] |
| 1664 | Disable Error Record Serialization Table (ERST) |
| 1665 | support. |
| 1666 | |
| 1667 | ether= [HW,NET] Ethernet cards parameters |
| 1668 | This option is obsoleted by the "netdev=" option, which |
| 1669 | has equivalent usage. See its documentation for details. |
| 1670 | |
| 1671 | evm= [EVM] |
| 1672 | Format: { "fix" } |
| 1673 | Permit 'security.evm' to be updated regardless of |
| 1674 | current integrity status. |
| 1675 | |
| 1676 | early_page_ext [KNL,EARLY] Enforces page_ext initialization to earlier |
| 1677 | stages so cover more early boot allocations. |
| 1678 | Please note that as side effect some optimizations |
| 1679 | might be disabled to achieve that (e.g. parallelized |
| 1680 | memory initialization is disabled) so the boot process |
| 1681 | might take longer, especially on systems with a lot of |
| 1682 | memory. Available with CONFIG_PAGE_EXTENSION=y. |
| 1683 | |
| 1684 | failslab= |
| 1685 | fail_usercopy= |
| 1686 | fail_page_alloc= |
| 1687 | fail_skb_realloc= |
| 1688 | fail_make_request=[KNL] |
| 1689 | General fault injection mechanism. |
| 1690 | Format: <interval>,<probability>,<space>,<times> |
| 1691 | See also Documentation/fault-injection/. |
| 1692 | |
| 1693 | fb_tunnels= [NET] |
| 1694 | Format: { initns | none } |
| 1695 | See Documentation/admin-guide/sysctl/net.rst for |
| 1696 | fb_tunnels_only_for_init_ns |
| 1697 | |
| 1698 | floppy= [HW] |
| 1699 | See Documentation/admin-guide/blockdev/floppy.rst. |
| 1700 | |
| 1701 | forcepae [X86-32] |
| 1702 | Forcefully enable Physical Address Extension (PAE). |
| 1703 | Many Pentium M systems disable PAE but may have a |
| 1704 | functionally usable PAE implementation. |
| 1705 | Warning: use of this parameter will taint the kernel |
| 1706 | and may cause unknown problems. |
| 1707 | |
| 1708 | fred= [X86-64] |
| 1709 | Enable/disable Flexible Return and Event Delivery. |
| 1710 | Format: { on | off } |
| 1711 | on: enable FRED when it's present, the default setting. |
| 1712 | off: disable FRED. |
| 1713 | |
| 1714 | ftrace=[tracer] |
| 1715 | [FTRACE] will set and start the specified tracer |
| 1716 | as early as possible in order to facilitate early |
| 1717 | boot debugging. |
| 1718 | |
| 1719 | ftrace_boot_snapshot |
| 1720 | [FTRACE] On boot up, a snapshot will be taken of the |
| 1721 | ftrace ring buffer that can be read at: |
| 1722 | /sys/kernel/tracing/snapshot. |
| 1723 | This is useful if you need tracing information from kernel |
| 1724 | boot up that is likely to be overridden by user space |
| 1725 | start up functionality. |
| 1726 | |
| 1727 | Optionally, the snapshot can also be defined for a tracing |
| 1728 | instance that was created by the trace_instance= command |
| 1729 | line parameter. |
| 1730 | |
| 1731 | trace_instance=foo,sched_switch ftrace_boot_snapshot=foo |
| 1732 | |
| 1733 | The above will cause the "foo" tracing instance to trigger |
| 1734 | a snapshot at the end of boot up. |
| 1735 | |
| 1736 | ftrace_dump_on_oops[=2(orig_cpu) | =<instance>][,<instance> | |
| 1737 | ,<instance>=2(orig_cpu)] |
| 1738 | [FTRACE] will dump the trace buffers on oops. |
| 1739 | If no parameter is passed, ftrace will dump global |
| 1740 | buffers of all CPUs, if you pass 2 or orig_cpu, it |
| 1741 | will dump only the buffer of the CPU that triggered |
| 1742 | the oops, or the specific instance will be dumped if |
| 1743 | its name is passed. Multiple instance dump is also |
| 1744 | supported, and instances are separated by commas. Each |
| 1745 | instance supports only dump on CPU that triggered the |
| 1746 | oops by passing 2 or orig_cpu to it. |
| 1747 | |
| 1748 | ftrace_dump_on_oops=foo=orig_cpu |
| 1749 | |
| 1750 | The above will dump only the buffer of "foo" instance |
| 1751 | on CPU that triggered the oops. |
| 1752 | |
| 1753 | ftrace_dump_on_oops,foo,bar=orig_cpu |
| 1754 | |
| 1755 | The above will dump global buffer on all CPUs, the |
| 1756 | buffer of "foo" instance on all CPUs and the buffer |
| 1757 | of "bar" instance on CPU that triggered the oops. |
| 1758 | |
| 1759 | ftrace_filter=[function-list] |
| 1760 | [FTRACE] Limit the functions traced by the function |
| 1761 | tracer at boot up. function-list is a comma-separated |
| 1762 | list of functions. This list can be changed at run |
| 1763 | time by the set_ftrace_filter file in the debugfs |
| 1764 | tracing directory. |
| 1765 | |
| 1766 | ftrace_notrace=[function-list] |
| 1767 | [FTRACE] Do not trace the functions specified in |
| 1768 | function-list. This list can be changed at run time |
| 1769 | by the set_ftrace_notrace file in the debugfs |
| 1770 | tracing directory. |
| 1771 | |
| 1772 | ftrace_graph_filter=[function-list] |
| 1773 | [FTRACE] Limit the top level callers functions traced |
| 1774 | by the function graph tracer at boot up. |
| 1775 | function-list is a comma-separated list of functions |
| 1776 | that can be changed at run time by the |
| 1777 | set_graph_function file in the debugfs tracing directory. |
| 1778 | |
| 1779 | ftrace_graph_notrace=[function-list] |
| 1780 | [FTRACE] Do not trace from the functions specified in |
| 1781 | function-list. This list is a comma-separated list of |
| 1782 | functions that can be changed at run time by the |
| 1783 | set_graph_notrace file in the debugfs tracing directory. |
| 1784 | |
| 1785 | ftrace_graph_max_depth=<uint> |
| 1786 | [FTRACE] Used with the function graph tracer. This is |
| 1787 | the max depth it will trace into a function. This value |
| 1788 | can be changed at run time by the max_graph_depth file |
| 1789 | in the tracefs tracing directory. default: 0 (no limit) |
| 1790 | |
| 1791 | fw_devlink= [KNL,EARLY] Create device links between consumer and supplier |
| 1792 | devices by scanning the firmware to infer the |
| 1793 | consumer/supplier relationships. This feature is |
| 1794 | especially useful when drivers are loaded as modules as |
| 1795 | it ensures proper ordering of tasks like device probing |
| 1796 | (suppliers first, then consumers), supplier boot state |
| 1797 | clean up (only after all consumers have probed), |
| 1798 | suspend/resume & runtime PM (consumers first, then |
| 1799 | suppliers). |
| 1800 | Format: { off | permissive | on | rpm } |
| 1801 | off -- Don't create device links from firmware info. |
| 1802 | permissive -- Create device links from firmware info |
| 1803 | but use it only for ordering boot state clean |
| 1804 | up (sync_state() calls). |
| 1805 | on -- Create device links from firmware info and use it |
| 1806 | to enforce probe and suspend/resume ordering. |
| 1807 | rpm -- Like "on", but also use to order runtime PM. |
| 1808 | |
| 1809 | fw_devlink.strict=<bool> |
| 1810 | [KNL,EARLY] Treat all inferred dependencies as mandatory |
| 1811 | dependencies. This only applies for fw_devlink=on|rpm. |
| 1812 | Format: <bool> |
| 1813 | |
| 1814 | fw_devlink.sync_state = |
| 1815 | [KNL,EARLY] When all devices that could probe have finished |
| 1816 | probing, this parameter controls what to do with |
| 1817 | devices that haven't yet received their sync_state() |
| 1818 | calls. |
| 1819 | Format: { strict | timeout } |
| 1820 | strict -- Default. Continue waiting on consumers to |
| 1821 | probe successfully. |
| 1822 | timeout -- Give up waiting on consumers and call |
| 1823 | sync_state() on any devices that haven't yet |
| 1824 | received their sync_state() calls after |
| 1825 | deferred_probe_timeout has expired or by |
| 1826 | late_initcall() if !CONFIG_MODULES. |
| 1827 | |
| 1828 | gamecon.map[2|3]= |
| 1829 | [HW,JOY] Multisystem joystick and NES/SNES/PSX pad |
| 1830 | support via parallel port (up to 5 devices per port) |
| 1831 | Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> |
| 1832 | See also Documentation/input/devices/joystick-parport.rst |
| 1833 | |
| 1834 | gamma= [HW,DRM] |
| 1835 | |
| 1836 | gart_fix_e820= [X86-64,EARLY] disable the fix e820 for K8 GART |
| 1837 | Format: off | on |
| 1838 | default: on |
| 1839 | |
| 1840 | gather_data_sampling= |
| 1841 | [X86,INTEL,EARLY] Control the Gather Data Sampling (GDS) |
| 1842 | mitigation. |
| 1843 | |
| 1844 | Gather Data Sampling is a hardware vulnerability which |
| 1845 | allows unprivileged speculative access to data which was |
| 1846 | previously stored in vector registers. |
| 1847 | |
| 1848 | This issue is mitigated by default in updated microcode. |
| 1849 | The mitigation may have a performance impact but can be |
| 1850 | disabled. On systems without the microcode mitigation |
| 1851 | disabling AVX serves as a mitigation. |
| 1852 | |
| 1853 | force: Disable AVX to mitigate systems without |
| 1854 | microcode mitigation. No effect if the microcode |
| 1855 | mitigation is present. Known to cause crashes in |
| 1856 | userspace with buggy AVX enumeration. |
| 1857 | |
| 1858 | off: Disable GDS mitigation. |
| 1859 | |
| 1860 | gbpages [X86] Use GB pages for kernel direct mappings. |
| 1861 | |
| 1862 | gcov_persist= [GCOV] When non-zero (default), profiling data for |
| 1863 | kernel modules is saved and remains accessible via |
| 1864 | debugfs, even when the module is unloaded/reloaded. |
| 1865 | When zero, profiling data is discarded and associated |
| 1866 | debugfs files are removed at module unload time. |
| 1867 | |
| 1868 | goldfish [X86] Enable the goldfish android emulator platform. |
| 1869 | Don't use this when you are not running on the |
| 1870 | android emulator |
| 1871 | |
| 1872 | gpio-mockup.gpio_mockup_ranges |
| 1873 | [HW] Sets the ranges of gpiochip of for this device. |
| 1874 | Format: <start1>,<end1>,<start2>,<end2>... |
| 1875 | gpio-mockup.gpio_mockup_named_lines |
| 1876 | [HW] Let the driver know GPIO lines should be named. |
| 1877 | |
| 1878 | gpt [EFI] Forces disk with valid GPT signature but |
| 1879 | invalid Protective MBR to be treated as GPT. If the |
| 1880 | primary GPT is corrupted, it enables the backup/alternate |
| 1881 | GPT to be used instead. |
| 1882 | |
| 1883 | grcan.enable0= [HW] Configuration of physical interface 0. Determines |
| 1884 | the "Enable 0" bit of the configuration register. |
| 1885 | Format: 0 | 1 |
| 1886 | Default: 0 |
| 1887 | grcan.enable1= [HW] Configuration of physical interface 1. Determines |
| 1888 | the "Enable 0" bit of the configuration register. |
| 1889 | Format: 0 | 1 |
| 1890 | Default: 0 |
| 1891 | grcan.select= [HW] Select which physical interface to use. |
| 1892 | Format: 0 | 1 |
| 1893 | Default: 0 |
| 1894 | grcan.txsize= [HW] Sets the size of the tx buffer. |
| 1895 | Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. |
| 1896 | Default: 1024 |
| 1897 | grcan.rxsize= [HW] Sets the size of the rx buffer. |
| 1898 | Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. |
| 1899 | Default: 1024 |
| 1900 | |
| 1901 | hardened_usercopy= |
| 1902 | [KNL] Under CONFIG_HARDENED_USERCOPY, whether |
| 1903 | hardening is enabled for this boot. Hardened |
| 1904 | usercopy checking is used to protect the kernel |
| 1905 | from reading or writing beyond known memory |
| 1906 | allocation boundaries as a proactive defense |
| 1907 | against bounds-checking flaws in the kernel's |
| 1908 | copy_to_user()/copy_from_user() interface. |
| 1909 | The default is determined by |
| 1910 | CONFIG_HARDENED_USERCOPY_DEFAULT_ON. |
| 1911 | on Perform hardened usercopy checks. |
| 1912 | off Disable hardened usercopy checks. |
| 1913 | |
| 1914 | hardlockup_all_cpu_backtrace= |
| 1915 | [KNL] Should the hard-lockup detector generate |
| 1916 | backtraces on all cpus. |
| 1917 | Format: 0 | 1 |
| 1918 | |
| 1919 | hash_pointers= |
| 1920 | [KNL,EARLY] |
| 1921 | By default, when pointers are printed to the console |
| 1922 | or buffers via the %p format string, that pointer is |
| 1923 | "hashed", i.e. obscured by hashing the pointer value. |
| 1924 | This is a security feature that hides actual kernel |
| 1925 | addresses from unprivileged users, but it also makes |
| 1926 | debugging the kernel more difficult since unequal |
| 1927 | pointers can no longer be compared. The choices are: |
| 1928 | Format: { auto | always | never } |
| 1929 | Default: auto |
| 1930 | |
| 1931 | auto - Hash pointers unless slab_debug is enabled. |
| 1932 | always - Always hash pointers (even if slab_debug is |
| 1933 | enabled). |
| 1934 | never - Never hash pointers. This option should only |
| 1935 | be specified when debugging the kernel. Do |
| 1936 | not use on production kernels. The boot |
| 1937 | param "no_hash_pointers" is an alias for |
| 1938 | this mode. |
| 1939 | |
| 1940 | For controlling hashing dynamically at runtime, |
| 1941 | use the "kernel.kptr_restrict" sysctl instead. |
| 1942 | |
| 1943 | hashdist= [KNL,NUMA] Large hashes allocated during boot |
| 1944 | are distributed across NUMA nodes. Defaults on |
| 1945 | for 64-bit NUMA, off otherwise. |
| 1946 | Format: 0 | 1 (for off | on) |
| 1947 | |
| 1948 | hd= [EIDE] (E)IDE hard drive subsystem geometry |
| 1949 | Format: <cyl>,<head>,<sect> |
| 1950 | |
| 1951 | hest_disable [ACPI] |
| 1952 | Disable Hardware Error Source Table (HEST) support; |
| 1953 | corresponding firmware-first mode error processing |
| 1954 | logic will be disabled. |
| 1955 | |
| 1956 | hibernate= [HIBERNATION] |
| 1957 | noresume Don't check if there's a hibernation image |
| 1958 | present during boot. |
| 1959 | nocompress Don't compress/decompress hibernation images. |
| 1960 | no Disable hibernation and resume. |
| 1961 | protect_image Turn on image protection during restoration |
| 1962 | (that will set all pages holding image data |
| 1963 | during restoration read-only). |
| 1964 | |
| 1965 | hibernate.compressor= [HIBERNATION] Compression algorithm to be |
| 1966 | used with hibernation. |
| 1967 | Format: { lzo | lz4 } |
| 1968 | Default: lzo |
| 1969 | |
| 1970 | lzo: Select LZO compression algorithm to |
| 1971 | compress/decompress hibernation image. |
| 1972 | |
| 1973 | lz4: Select LZ4 compression algorithm to |
| 1974 | compress/decompress hibernation image. |
| 1975 | |
| 1976 | hibernate.pm_test_delay= |
| 1977 | [HIBERNATION] |
| 1978 | Sets the number of seconds to remain in a hibernation test |
| 1979 | mode before resuming the system (see |
| 1980 | /sys/power/pm_test). Only available when CONFIG_PM_DEBUG |
| 1981 | is set. Default value is 5. |
| 1982 | |
| 1983 | hibernate_compression_threads= |
| 1984 | [HIBERNATION] |
| 1985 | Set the number of threads used for compressing or decompressing |
| 1986 | hibernation images. |
| 1987 | |
| 1988 | Format: <integer> |
| 1989 | Default: 3 |
| 1990 | Minimum: 1 |
| 1991 | Example: hibernate_compression_threads=4 |
| 1992 | |
| 1993 | highmem=nn[KMG] [KNL,BOOT,EARLY] forces the highmem zone to have an exact |
| 1994 | size of <nn>. This works even on boxes that have no |
| 1995 | highmem otherwise. This also works to reduce highmem |
| 1996 | size on bigger boxes. |
| 1997 | |
| 1998 | highres= [KNL] Enable/disable high resolution timer mode. |
| 1999 | Valid parameters: "on", "off" |
| 2000 | Default: "on" |
| 2001 | |
| 2002 | hlt [BUGS=ARM,SH] |
| 2003 | |
| 2004 | hostname= [KNL,EARLY] Set the hostname (aka UTS nodename). |
| 2005 | Format: <string> |
| 2006 | This allows setting the system's hostname during early |
| 2007 | startup. This sets the name returned by gethostname. |
| 2008 | Using this parameter to set the hostname makes it |
| 2009 | possible to ensure the hostname is correctly set before |
| 2010 | any userspace processes run, avoiding the possibility |
| 2011 | that a process may call gethostname before the hostname |
| 2012 | has been explicitly set, resulting in the calling |
| 2013 | process getting an incorrect result. The string must |
| 2014 | not exceed the maximum allowed hostname length (usually |
| 2015 | 64 characters) and will be truncated otherwise. |
| 2016 | |
| 2017 | hpet= [X86-32,HPET] option to control HPET usage |
| 2018 | Format: { enable (default) | disable | force | |
| 2019 | verbose } |
| 2020 | disable: disable HPET and use PIT instead |
| 2021 | force: allow force enabled of undocumented chips (ICH4, |
| 2022 | VIA, nVidia) |
| 2023 | verbose: show contents of HPET registers during setup |
| 2024 | |
| 2025 | hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET |
| 2026 | registers. Default set by CONFIG_HPET_MMAP_DEFAULT. |
| 2027 | |
| 2028 | hugepages= [HW,EARLY] Number of HugeTLB pages to allocate at boot. |
| 2029 | If this follows hugepagesz (below), it specifies |
| 2030 | the number of pages of hugepagesz to be allocated. |
| 2031 | If this is the first HugeTLB parameter on the command |
| 2032 | line, it specifies the number of pages to allocate for |
| 2033 | the default huge page size. If using node format, the |
| 2034 | number of pages to allocate per-node can be specified. |
| 2035 | See also Documentation/admin-guide/mm/hugetlbpage.rst. |
| 2036 | Format: <integer> or (node format) |
| 2037 | <node>:<integer>[,<node>:<integer>] |
| 2038 | |
| 2039 | hugepagesz= |
| 2040 | [HW,EARLY] The size of the HugeTLB pages. This is |
| 2041 | used in conjunction with hugepages (above) to |
| 2042 | allocate huge pages of a specific size at boot. The |
| 2043 | pair hugepagesz=X hugepages=Y can be specified once |
| 2044 | for each supported huge page size. Huge page sizes |
| 2045 | are architecture dependent. See also |
| 2046 | Documentation/admin-guide/mm/hugetlbpage.rst. |
| 2047 | Format: size[KMG] |
| 2048 | |
| 2049 | hugepage_alloc_threads= |
| 2050 | [HW] The number of threads that should be used to |
| 2051 | allocate hugepages during boot. This option can be |
| 2052 | used to improve system bootup time when allocating |
| 2053 | a large amount of huge pages. |
| 2054 | The default value is 25% of the available hardware threads. |
| 2055 | |
| 2056 | Note that this parameter only applies to non-gigantic huge pages. |
| 2057 | |
| 2058 | hugetlb_cma= [HW,CMA,EARLY] The size of a CMA area used for allocation |
| 2059 | of gigantic hugepages. Or using node format, the size |
| 2060 | of a CMA area per node can be specified. |
| 2061 | The size can be an absolute value (e.g., 2G) or a |
| 2062 | percentage of the total memory or node memory (e.g., 20%). |
| 2063 | Percentage-derived sizes are rounded down to a multiple of |
| 2064 | the architecture's gigantic hugepage size and may become |
| 2065 | zero. |
| 2066 | Format: nn[KMGTPE] or nn% or (node format) |
| 2067 | <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] or |
| 2068 | <node>:nn%[,<node>:nn%] |
| 2069 | |
| 2070 | The size must be a multiple of the gigantic page size. |
| 2071 | When using node format, this applies to each per-node size. |
| 2072 | Missaligned values are dropped with a warning. |
| 2073 | |
| 2074 | Reserve a CMA area of given size and allocate gigantic |
| 2075 | hugepages using the CMA allocator. If enabled, the |
| 2076 | boot-time allocation of gigantic hugepages is skipped. |
| 2077 | |
| 2078 | hugetlb_cma_only= |
| 2079 | [HW,CMA,EARLY] When allocating new HugeTLB pages, only |
| 2080 | try to allocate from the CMA areas. |
| 2081 | |
| 2082 | This option does nothing if hugetlb_cma= is not also |
| 2083 | specified. |
| 2084 | |
| 2085 | hugetlb_free_vmemmap= |
| 2086 | [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP |
| 2087 | enabled. |
| 2088 | Control if HugeTLB Vmemmap Optimization (HVO) is enabled. |
| 2089 | Allows heavy hugetlb users to free up some more |
| 2090 | memory (7 * PAGE_SIZE for each 2MB hugetlb page). |
| 2091 | Format: { on | off (default) } |
| 2092 | |
| 2093 | on: enable HVO |
| 2094 | off: disable HVO |
| 2095 | |
| 2096 | Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, |
| 2097 | the default is on. |
| 2098 | |
| 2099 | Note that the vmemmap pages may be allocated from the added |
| 2100 | memory block itself when memory_hotplug.memmap_on_memory is |
| 2101 | enabled, those vmemmap pages cannot be optimized even if this |
| 2102 | feature is enabled. Other vmemmap pages not allocated from |
| 2103 | the added memory block itself do not be affected. |
| 2104 | |
| 2105 | hung_task_panic= |
| 2106 | [KNL] Number of hung tasks to trigger kernel panic. |
| 2107 | Format: <int> |
| 2108 | |
| 2109 | When set to a non-zero value, a kernel panic will be triggered if |
| 2110 | the number of detected hung tasks reaches this value. |
| 2111 | |
| 2112 | 0: don't panic |
| 2113 | 1: panic immediately on first hung task |
| 2114 | N: panic after N hung tasks are detected in a single scan |
| 2115 | |
| 2116 | The default value is controlled by the |
| 2117 | CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time option. The value |
| 2118 | selected by this boot parameter can be changed later by the |
| 2119 | kernel.hung_task_panic sysctl. |
| 2120 | |
| 2121 | hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) |
| 2122 | terminal devices. Valid values: 0..8 |
| 2123 | hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. |
| 2124 | If specified, z/VM IUCV HVC accepts connections |
| 2125 | from listed z/VM user IDs only. |
| 2126 | |
| 2127 | hv_nopvspin [X86,HYPER_V,EARLY] |
| 2128 | Disables the paravirt spinlock optimizations |
| 2129 | which allow the hypervisor to 'idle' the guest |
| 2130 | on lock contention. |
| 2131 | |
| 2132 | hw_protection= [HW] |
| 2133 | Format: reboot | shutdown |
| 2134 | |
| 2135 | Hardware protection action taken on critical events like |
| 2136 | overtemperature or imminent voltage loss. |
| 2137 | |
| 2138 | i2c_bus= [HW] Override the default board specific I2C bus speed |
| 2139 | or register an additional I2C bus that is not |
| 2140 | registered from board initialization code. |
| 2141 | Format: |
| 2142 | <bus_id>,<clkrate> |
| 2143 | |
| 2144 | i2c_touchscreen_props= [HW,ACPI,X86] |
| 2145 | Set device-properties for ACPI-enumerated I2C-attached |
| 2146 | touchscreen, to e.g. fix coordinates of upside-down |
| 2147 | mounted touchscreens. If you need this option please |
| 2148 | submit a drivers/platform/x86/touchscreen_dmi.c patch |
| 2149 | adding a DMI quirk for this. |
| 2150 | |
| 2151 | Format: |
| 2152 | <ACPI_HW_ID>:<prop_name>=<val>[:prop_name=val][:...] |
| 2153 | Where <val> is one of: |
| 2154 | Omit "=<val>" entirely Set a boolean device-property |
| 2155 | Unsigned number Set a u32 device-property |
| 2156 | Anything else Set a string device-property |
| 2157 | |
| 2158 | Examples (split over multiple lines): |
| 2159 | i2c_touchscreen_props=GDIX1001:touchscreen-inverted-x: |
| 2160 | touchscreen-inverted-y |
| 2161 | |
| 2162 | i2c_touchscreen_props=MSSL1680:touchscreen-size-x=1920: |
| 2163 | touchscreen-size-y=1080:touchscreen-inverted-y: |
| 2164 | firmware-name=gsl1680-vendor-model.fw:silead,home-button |
| 2165 | |
| 2166 | i8042.debug [HW] Toggle i8042 debug mode |
| 2167 | i8042.unmask_kbd_data |
| 2168 | [HW] Enable printing of interrupt data from the KBD port |
| 2169 | (disabled by default, and as a pre-condition |
| 2170 | requires that i8042.debug=1 be enabled) |
| 2171 | i8042.direct [HW] Put keyboard port into non-translated mode |
| 2172 | i8042.dumbkbd [HW] Pretend that controller can only read data from |
| 2173 | keyboard and cannot control its state |
| 2174 | (Don't attempt to blink the leds) |
| 2175 | i8042.noaux [HW] Don't check for auxiliary (== mouse) port |
| 2176 | i8042.nokbd [HW] Don't check/create keyboard port |
| 2177 | i8042.noloop [HW] Disable the AUX Loopback command while probing |
| 2178 | for the AUX port |
| 2179 | i8042.nomux [HW] Don't check presence of an active multiplexing |
| 2180 | controller |
| 2181 | i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX |
| 2182 | controllers |
| 2183 | i8042.notimeout [HW] Ignore timeout condition signalled by controller |
| 2184 | i8042.reset [HW] Reset the controller during init, cleanup and |
| 2185 | suspend-to-ram transitions, only during s2r |
| 2186 | transitions, or never reset |
| 2187 | Format: { 1 | Y | y | 0 | N | n } |
| 2188 | 1, Y, y: always reset controller |
| 2189 | 0, N, n: don't ever reset controller |
| 2190 | Default: only on s2r transitions on x86; most other |
| 2191 | architectures force reset to be always executed |
| 2192 | i8042.unlock [HW] Unlock (ignore) the keylock |
| 2193 | i8042.kbdreset [HW] Reset device connected to KBD port |
| 2194 | i8042.probe_defer |
| 2195 | [HW] Allow deferred probing upon i8042 probe errors |
| 2196 | |
| 2197 | i810= [HW,DRM] |
| 2198 | |
| 2199 | i915.invert_brightness= |
| 2200 | [DRM] Invert the sense of the variable that is used to |
| 2201 | set the brightness of the panel backlight. Normally a |
| 2202 | brightness value of 0 indicates backlight switched off, |
| 2203 | and the maximum of the brightness value sets the backlight |
| 2204 | to maximum brightness. If this parameter is set to 0 |
| 2205 | (default) and the machine requires it, or this parameter |
| 2206 | is set to 1, a brightness value of 0 sets the backlight |
| 2207 | to maximum brightness, and the maximum of the brightness |
| 2208 | value switches the backlight off. |
| 2209 | -1 -- never invert brightness |
| 2210 | 0 -- machine default |
| 2211 | 1 -- force brightness inversion |
| 2212 | |
| 2213 | ia32_emulation= [X86-64] |
| 2214 | Format: <bool> |
| 2215 | When true, allows loading 32-bit programs and executing 32-bit |
| 2216 | syscalls, essentially overriding IA32_EMULATION_DEFAULT_DISABLED at |
| 2217 | boot time. When false, unconditionally disables IA32 emulation. |
| 2218 | |
| 2219 | |
| 2220 | idle= [X86,EARLY] |
| 2221 | Format: idle=poll, idle=halt, idle=nomwait |
| 2222 | |
| 2223 | idle=poll: Don't do power saving in the idle loop |
| 2224 | using HLT, but poll for rescheduling event. This will |
| 2225 | make the CPUs eat a lot more power, but may be useful |
| 2226 | to get slightly better performance in multiprocessor |
| 2227 | benchmarks. It also makes some profiling using |
| 2228 | performance counters more accurate. Please note that |
| 2229 | on systems with MONITOR/MWAIT support (like Intel |
| 2230 | EM64T CPUs) this option has no performance advantage |
| 2231 | over the normal idle loop. It may also interact badly |
| 2232 | with hyperthreading. |
| 2233 | |
| 2234 | idle=halt: Halt is forced to be used for CPU idle. |
| 2235 | In such case C2/C3 won't be used again. |
| 2236 | |
| 2237 | idle=nomwait: Disable mwait for CPU C-states |
| 2238 | |
| 2239 | [ARM64,EARLY] |
| 2240 | Format: idle=wfi, idle=yield, idle=nop |
| 2241 | |
| 2242 | idle=wfi: Use the WFI (Wait For Interrupt) hint |
| 2243 | instruction in the idle loop. This is the default and |
| 2244 | allows the CPU to enter a low-power state until an |
| 2245 | interrupt arrives. |
| 2246 | |
| 2247 | idle=yield: Use the YIELD hint instruction instead of |
| 2248 | WFI. CPUs supporting simultaneous multi-threading (SMT), |
| 2249 | can continue executing another thread when the current |
| 2250 | thread reaches the idle loop. This will make the CPUs |
| 2251 | eat more power, but may be useful to get slightly better |
| 2252 | performance in some applications, since the CPUs will |
| 2253 | not enter a low-power state. |
| 2254 | |
| 2255 | idle=nop: Do not execute any idle instruction in the |
| 2256 | idle loop. This is useful on platforms where WFI |
| 2257 | misbehaves, leading to system instability or loss of CPU |
| 2258 | state. This will make the CPUs eat more power, but may |
| 2259 | give slightly better performance in some applications, |
| 2260 | since the CPUs will not enter a low-power state. |
| 2261 | |
| 2262 | idxd.sva= [HW] |
| 2263 | Format: <bool> |
| 2264 | Allow force disabling of Shared Virtual Memory (SVA) |
| 2265 | support for the idxd driver. By default it is set to |
| 2266 | true (1). |
| 2267 | |
| 2268 | idxd.tc_override= [HW] |
| 2269 | Format: <bool> |
| 2270 | Allow override of default traffic class configuration |
| 2271 | for the device. By default it is set to false (0). |
| 2272 | |
| 2273 | ieee754= [MIPS] Select IEEE Std 754 conformance mode |
| 2274 | Format: { strict | legacy | 2008 | relaxed | emulated } |
| 2275 | Default: strict |
| 2276 | |
| 2277 | Choose which programs will be accepted for execution |
| 2278 | based on the IEEE 754 NaN encoding(s) supported by |
| 2279 | the FPU and the NaN encoding requested with the value |
| 2280 | of an ELF file header flag individually set by each |
| 2281 | binary. Hardware implementations are permitted to |
| 2282 | support either or both of the legacy and the 2008 NaN |
| 2283 | encoding mode. |
| 2284 | |
| 2285 | Available settings are as follows: |
| 2286 | strict accept binaries that request a NaN encoding |
| 2287 | supported by the FPU |
| 2288 | legacy only accept legacy-NaN binaries, if supported |
| 2289 | by the FPU |
| 2290 | 2008 only accept 2008-NaN binaries, if supported |
| 2291 | by the FPU |
| 2292 | relaxed accept any binaries regardless of whether |
| 2293 | supported by the FPU |
| 2294 | emulated accept any binaries but enable FPU emulator |
| 2295 | if binary mode is unsupported by the FPU. |
| 2296 | |
| 2297 | The FPU emulator is always able to support both NaN |
| 2298 | encodings, so if no FPU hardware is present or it has |
| 2299 | been disabled with 'nofpu', then the settings of |
| 2300 | 'legacy' and '2008' strap the emulator accordingly, |
| 2301 | 'relaxed' straps the emulator for both legacy-NaN and |
| 2302 | 2008-NaN, whereas 'strict' enables legacy-NaN only on |
| 2303 | legacy processors and both NaN encodings on MIPS32 or |
| 2304 | MIPS64 CPUs. |
| 2305 | |
| 2306 | The setting for ABS.fmt/NEG.fmt instruction execution |
| 2307 | mode generally follows that for the NaN encoding, |
| 2308 | except where unsupported by hardware. |
| 2309 | |
| 2310 | ignore_loglevel [KNL,EARLY] |
| 2311 | Ignore loglevel setting - this will print /all/ |
| 2312 | kernel messages to the console. Useful for debugging. |
| 2313 | We also add it as printk module parameter, so users |
| 2314 | could change it dynamically, usually by |
| 2315 | /sys/module/printk/parameters/ignore_loglevel. |
| 2316 | |
| 2317 | ignore_rlimit_data |
| 2318 | Ignore RLIMIT_DATA setting for data mappings, |
| 2319 | print warning at first misuse. Can be changed via |
| 2320 | /sys/module/kernel/parameters/ignore_rlimit_data. |
| 2321 | |
| 2322 | ihash_entries= [KNL] |
| 2323 | Set number of hash buckets for inode cache. |
| 2324 | |
| 2325 | ima_appraise= [IMA] appraise integrity measurements |
| 2326 | Format: { "off" | "enforce" | "fix" | "log" } |
| 2327 | default: "enforce" |
| 2328 | |
| 2329 | ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. |
| 2330 | The builtin appraise policy appraises all files |
| 2331 | owned by uid=0. |
| 2332 | |
| 2333 | ima_canonical_fmt [IMA] |
| 2334 | Use the canonical format for the binary runtime |
| 2335 | measurements, instead of host native format. |
| 2336 | |
| 2337 | ima_flush_htable [IMA] |
| 2338 | Flush the IMA hash table when deleting all the |
| 2339 | staged measurement records, to achieve maximum |
| 2340 | memory saving at the cost of having duplicate |
| 2341 | records across the staged measurement lists. |
| 2342 | |
| 2343 | ima_hash= [IMA] |
| 2344 | Format: { md5 | sha1 | rmd160 | sha256 | sha384 |
| 2345 | | sha512 | ... } |
| 2346 | default: "sha1" |
| 2347 | |
| 2348 | The list of supported hash algorithms is defined |
| 2349 | in crypto/hash_info.h. |
| 2350 | |
| 2351 | ima_policy= [IMA] |
| 2352 | The builtin policies to load during IMA setup. |
| 2353 | Format: "tcb | appraise_tcb | secure_boot | |
| 2354 | fail_securely | critical_data" |
| 2355 | |
| 2356 | The "tcb" policy measures all programs exec'd, files |
| 2357 | mmap'd for exec, and all files opened with the read |
| 2358 | mode bit set by either the effective uid (euid=0) or |
| 2359 | uid=0. |
| 2360 | |
| 2361 | The "appraise_tcb" policy appraises the integrity of |
| 2362 | all files owned by root. |
| 2363 | |
| 2364 | The "secure_boot" policy appraises the integrity |
| 2365 | of files (eg. kexec kernel image, kernel modules, |
| 2366 | firmware, policy, etc) based on file signatures. |
| 2367 | |
| 2368 | The "fail_securely" policy forces file signature |
| 2369 | verification failure also on privileged mounted |
| 2370 | filesystems with the SB_I_UNVERIFIABLE_SIGNATURE |
| 2371 | flag. |
| 2372 | |
| 2373 | The "critical_data" policy measures kernel integrity |
| 2374 | critical data. |
| 2375 | |
| 2376 | ima_tcb [IMA] Deprecated. Use ima_policy= instead. |
| 2377 | Load a policy which meets the needs of the Trusted |
| 2378 | Computing Base. This means IMA will measure all |
| 2379 | programs exec'd, files mmap'd for exec, and all files |
| 2380 | opened for read by uid=0. |
| 2381 | |
| 2382 | ima_template= [IMA] |
| 2383 | Select one of defined IMA measurements template formats. |
| 2384 | Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | |
| 2385 | "ima-sigv2" } |
| 2386 | Default: "ima-ng" |
| 2387 | |
| 2388 | ima_template_fmt= |
| 2389 | [IMA] Define a custom template format. |
| 2390 | Format: { "field1|...|fieldN" } |
| 2391 | |
| 2392 | ima= [IMA] Enable or disable IMA |
| 2393 | Format: { "off" | "on" } |
| 2394 | Default: "on" |
| 2395 | Note that disabling IMA is limited to kdump kernel. |
| 2396 | |
| 2397 | indirect_target_selection= [X86,Intel] Mitigation control for Indirect |
| 2398 | Target Selection(ITS) bug in Intel CPUs. Updated |
| 2399 | microcode is also required for a fix in IBPB. |
| 2400 | |
| 2401 | on: Enable mitigation (default). |
| 2402 | off: Disable mitigation. |
| 2403 | force: Force the ITS bug and deploy default |
| 2404 | mitigation. |
| 2405 | vmexit: Only deploy mitigation if CPU is affected by |
| 2406 | guest/host isolation part of ITS. |
| 2407 | stuff: Deploy RSB-fill mitigation when retpoline is |
| 2408 | also deployed. Otherwise, deploy the default |
| 2409 | mitigation. |
| 2410 | |
| 2411 | For details see: |
| 2412 | Documentation/admin-guide/hw-vuln/indirect-target-selection.rst |
| 2413 | |
| 2414 | init= [KNL] |
| 2415 | Format: <full_path> |
| 2416 | Run specified binary instead of /sbin/init as init |
| 2417 | process. |
| 2418 | |
| 2419 | initcall_debug [KNL] Trace initcalls as they are executed. Useful |
| 2420 | for working out where the kernel is dying during |
| 2421 | startup. |
| 2422 | |
| 2423 | initcall_blacklist= [KNL] Do not execute a comma-separated list of |
| 2424 | initcall functions. Useful for debugging built-in |
| 2425 | modules and initcalls. |
| 2426 | |
| 2427 | initramfs_async= [KNL] |
| 2428 | Format: <bool> |
| 2429 | Default: 1 |
| 2430 | This parameter controls whether the initramfs |
| 2431 | image is unpacked asynchronously, concurrently |
| 2432 | with devices being probed and |
| 2433 | initialized. This should normally just work, |
| 2434 | but as a debugging aid, one can get the |
| 2435 | historical behaviour of the initramfs |
| 2436 | unpacking being completed before device_ and |
| 2437 | late_ initcalls. |
| 2438 | |
| 2439 | initrd= [BOOT,EARLY] Specify the location of the initial ramdisk |
| 2440 | |
| 2441 | initrdmem= [KNL,EARLY] Specify a physical address and size from which to |
| 2442 | load the initrd. If an initrd is compiled in or |
| 2443 | specified in the bootparams, it takes priority over this |
| 2444 | setting. |
| 2445 | Format: ss[KMG],nn[KMG] |
| 2446 | Default is 0, 0 |
| 2447 | |
| 2448 | init_on_alloc= [MM,EARLY] Fill newly allocated pages and heap objects with |
| 2449 | zeroes. |
| 2450 | Format: 0 | 1 |
| 2451 | Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. |
| 2452 | |
| 2453 | init_on_free= [MM,EARLY] Fill freed pages and heap objects with zeroes. |
| 2454 | Format: 0 | 1 |
| 2455 | Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. |
| 2456 | |
| 2457 | init_pkru= [X86] Specify the default memory protection keys rights |
| 2458 | register contents for all processes. 0x55555554 by |
| 2459 | default (disallow access to all but pkey 0). Can |
| 2460 | override in debugfs after boot. |
| 2461 | |
| 2462 | inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver |
| 2463 | Format: <irq> |
| 2464 | |
| 2465 | int_pln_enable [X86] Enable power limit notification interrupt |
| 2466 | |
| 2467 | integrity_audit=[IMA] |
| 2468 | Format: { "0" | "1" } |
| 2469 | 0 -- basic integrity auditing messages. (Default) |
| 2470 | 1 -- additional integrity auditing messages. |
| 2471 | |
| 2472 | intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option |
| 2473 | on |
| 2474 | Enable intel iommu driver. |
| 2475 | off |
| 2476 | Disable intel iommu driver. |
| 2477 | igfx_off [Default Off] |
| 2478 | By default, gfx is mapped as normal device. If a gfx |
| 2479 | device has a dedicated DMAR unit, the DMAR unit is |
| 2480 | bypassed by not enabling DMAR with this option. In |
| 2481 | this case, gfx device will use physical address for |
| 2482 | DMA. |
| 2483 | strict [Default Off] |
| 2484 | Deprecated, equivalent to iommu.strict=1. |
| 2485 | sp_off [Default Off] |
| 2486 | By default, super page will be supported if Intel IOMMU |
| 2487 | has the capability. With this option, super page will |
| 2488 | not be supported. |
| 2489 | sm_on |
| 2490 | Enable the Intel IOMMU scalable mode if the hardware |
| 2491 | advertises that it has support for the scalable mode |
| 2492 | translation. |
| 2493 | sm_off |
| 2494 | Disallow use of the Intel IOMMU scalable mode. |
| 2495 | tboot_noforce [Default Off] |
| 2496 | Do not force the Intel IOMMU enabled under tboot. |
| 2497 | By default, tboot will force Intel IOMMU on, which |
| 2498 | could harm performance of some high-throughput |
| 2499 | devices like 40GBit network cards, even if identity |
| 2500 | mapping is enabled. |
| 2501 | Note that using this option lowers the security |
| 2502 | provided by tboot because it makes the system |
| 2503 | vulnerable to DMA attacks. |
| 2504 | |
| 2505 | intel_idle.max_cstate= [KNL,HW,ACPI,X86] |
| 2506 | 0 disables intel_idle and fall back on acpi_idle. |
| 2507 | 1 to 9 specify maximum depth of C-state. |
| 2508 | |
| 2509 | intel_pstate= [X86,EARLY] |
| 2510 | disable |
| 2511 | Do not enable intel_pstate as the default |
| 2512 | scaling driver for the supported processors |
| 2513 | active |
| 2514 | Use intel_pstate driver to bypass the scaling |
| 2515 | governors layer of cpufreq and provides it own |
| 2516 | algorithms for p-state selection. There are two |
| 2517 | P-state selection algorithms provided by |
| 2518 | intel_pstate in the active mode: powersave and |
| 2519 | performance. The way they both operate depends |
| 2520 | on whether or not the hardware managed P-states |
| 2521 | (HWP) feature has been enabled in the processor |
| 2522 | and possibly on the processor model. |
| 2523 | passive |
| 2524 | Use intel_pstate as a scaling driver, but configure it |
| 2525 | to work with generic cpufreq governors (instead of |
| 2526 | enabling its internal governor). This mode cannot be |
| 2527 | used along with the hardware-managed P-states (HWP) |
| 2528 | feature. |
| 2529 | force |
| 2530 | Enable intel_pstate on systems that prohibit it by default |
| 2531 | in favor of acpi-cpufreq. Forcing the intel_pstate driver |
| 2532 | instead of acpi-cpufreq may disable platform features, such |
| 2533 | as thermal controls and power capping, that rely on ACPI |
| 2534 | P-States information being indicated to OSPM and therefore |
| 2535 | should be used with caution. This option does not work with |
| 2536 | processors that aren't supported by the intel_pstate driver |
| 2537 | or on platforms that use pcc-cpufreq instead of acpi-cpufreq. |
| 2538 | no_hwp |
| 2539 | Do not enable hardware P state control (HWP) |
| 2540 | if available. |
| 2541 | hwp_only |
| 2542 | Only load intel_pstate on systems which support |
| 2543 | hardware P state control (HWP) if available. |
| 2544 | support_acpi_ppc |
| 2545 | Enforce ACPI _PPC performance limits. If the Fixed ACPI |
| 2546 | Description Table, specifies preferred power management |
| 2547 | profile as "Enterprise Server" or "Performance Server", |
| 2548 | then this feature is turned on by default. |
| 2549 | per_cpu_perf_limits |
| 2550 | Allow per-logical-CPU P-State performance control limits using |
| 2551 | cpufreq sysfs interface |
| 2552 | no_cas |
| 2553 | Do not enable capacity-aware scheduling (CAS) on |
| 2554 | hybrid systems |
| 2555 | |
| 2556 | intremap= [X86-64,Intel-IOMMU,EARLY] |
| 2557 | on enable Interrupt Remapping (default) |
| 2558 | off disable Interrupt Remapping |
| 2559 | nosid disable Source ID checking |
| 2560 | no_x2apic_optout |
| 2561 | BIOS x2APIC opt-out request will be ignored |
| 2562 | nopost disable Interrupt Posting |
| 2563 | posted_msi |
| 2564 | enable MSIs delivered as posted interrupts |
| 2565 | |
| 2566 | iomem= Disable strict checking of access to MMIO memory |
| 2567 | strict regions from userspace. |
| 2568 | relaxed |
| 2569 | |
| 2570 | iommu= [X86,EARLY] |
| 2571 | |
| 2572 | off |
| 2573 | Don't initialize and use any kind of IOMMU. |
| 2574 | |
| 2575 | force |
| 2576 | Force the use of the hardware IOMMU even when |
| 2577 | it is not actually needed (e.g. because < 3 GB |
| 2578 | memory). |
| 2579 | |
| 2580 | noforce |
| 2581 | Don't force hardware IOMMU usage when it is not |
| 2582 | needed. (default). |
| 2583 | |
| 2584 | merge |
| 2585 | Do scatter-gather (SG) merging. Implies "force" |
| 2586 | (experimental). |
| 2587 | |
| 2588 | nomerge |
| 2589 | Don't do scatter-gather (SG) merging. |
| 2590 | |
| 2591 | biomerge |
| 2592 | Do scatter-gather (SG) merging. Implies "force" |
| 2593 | (experimental). [same as "merge"] |
| 2594 | |
| 2595 | panic |
| 2596 | Always panic when IOMMU overflows. |
| 2597 | |
| 2598 | nopanic |
| 2599 | Don't panic on IOMMU overflows. |
| 2600 | |
| 2601 | pt |
| 2602 | Use passththrough mode by default |
| 2603 | (Equivalent to iommu.passthrough=1) |
| 2604 | |
| 2605 | nopt |
| 2606 | Use translated mode for DMA by default |
| 2607 | (Equivalent to iommu.passthrough=0) |
| 2608 | |
| 2609 | soft |
| 2610 | Use software bounce buffering (SWIOTLB) (default for |
| 2611 | Intel machines). This can be used to prevent the usage |
| 2612 | of an available hardware IOMMU. |
| 2613 | |
| 2614 | usedac |
| 2615 | Use the DAC on VIA PCI bridge |
| 2616 | (default: disable the VIA PCI bridge DAC) |
| 2617 | |
| 2618 | AMD Gart HW IOMMU-specific options: (CONFIG_GART_IOMMU) |
| 2619 | |
| 2620 | <size> |
| 2621 | Set the size of the remapping area in bytes. |
| 2622 | |
| 2623 | allowed |
| 2624 | Overwrite iommu off workarounds for specific chipsets |
| 2625 | |
| 2626 | force |
| 2627 | Overwrite iommu off workarounds for specific chipsets |
| 2628 | |
| 2629 | fullflush |
| 2630 | Flush IOMMU on each allocation (default). |
| 2631 | |
| 2632 | nofullflush |
| 2633 | Don't use IOMMU fullflush. |
| 2634 | |
| 2635 | memaper[=<order>] |
| 2636 | Allocate an own aperture over RAM with size |
| 2637 | 32MB<<order. (default: order=1, i.e. 64MB) |
| 2638 | |
| 2639 | noaperture |
| 2640 | Ask the IOMMU not to touch the aperture for AGP. |
| 2641 | |
| 2642 | noagp |
| 2643 | Don't initialize the AGP driver and use full aperture. |
| 2644 | |
| 2645 | iommu= [PPC/POWERNV] |
| 2646 | |
| 2647 | nobypass |
| 2648 | Disable IOMMU bypass, using IOMMU for PCI devices. |
| 2649 | |
| 2650 | iommu.forcedac= [ARM64,X86,EARLY] Control IOVA allocation for PCI devices. |
| 2651 | Format: { "0" | "1" } |
| 2652 | 0 - Try to allocate a 32-bit DMA address first, before |
| 2653 | falling back to the full range if needed. |
| 2654 | 1 - Allocate directly from the full usable range, |
| 2655 | forcing Dual Address Cycle for PCI cards supporting |
| 2656 | greater than 32-bit addressing. |
| 2657 | |
| 2658 | iommu.strict= [ARM64,X86,S390,EARLY] Configure TLB invalidation behaviour |
| 2659 | Format: { "0" | "1" } |
| 2660 | 0 - Lazy mode. |
| 2661 | Request that DMA unmap operations use deferred |
| 2662 | invalidation of hardware TLBs, for increased |
| 2663 | throughput at the cost of reduced device isolation. |
| 2664 | Will fall back to strict mode if not supported by |
| 2665 | the relevant IOMMU driver. |
| 2666 | 1 - Strict mode. |
| 2667 | DMA unmap operations invalidate IOMMU hardware TLBs |
| 2668 | synchronously. |
| 2669 | unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. |
| 2670 | Note: on x86, strict mode specified via one of the |
| 2671 | legacy driver-specific options takes precedence. |
| 2672 | |
| 2673 | iommu.passthrough= |
| 2674 | [ARM64,X86,EARLY] Configure DMA to bypass the IOMMU by default. |
| 2675 | Format: { "0" | "1" } |
| 2676 | 0 - Use IOMMU translation for DMA. |
| 2677 | 1 - Bypass the IOMMU for DMA. |
| 2678 | unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. |
| 2679 | |
| 2680 | iommu.debug_pagealloc= |
| 2681 | [KNL,EARLY] When CONFIG_IOMMU_DEBUG_PAGEALLOC is set, this |
| 2682 | parameter enables the feature at boot time. By default, it |
| 2683 | is disabled and the system behaves the same way as a kernel |
| 2684 | built without CONFIG_IOMMU_DEBUG_PAGEALLOC. |
| 2685 | Format: { "0" | "1" } |
| 2686 | 0 - Sanitizer disabled. |
| 2687 | 1 - Sanitizer enabled, expect runtime overhead. |
| 2688 | |
| 2689 | io7= [HW] IO7 for Marvel-based Alpha systems |
| 2690 | See comment before marvel_specify_io7 in |
| 2691 | arch/alpha/kernel/core_marvel.c. |
| 2692 | |
| 2693 | io_delay= [X86,EARLY] I/O delay method |
| 2694 | 0x80 |
| 2695 | Standard port 0x80 based delay |
| 2696 | 0xed |
| 2697 | Alternate port 0xed based delay (needed on some systems) |
| 2698 | udelay |
| 2699 | Simple two microseconds delay |
| 2700 | none |
| 2701 | No delay |
| 2702 | |
| 2703 | ip= [IP_PNP] |
| 2704 | See Documentation/admin-guide/nfs/nfsroot.rst. |
| 2705 | |
| 2706 | ipcmni_extend [KNL,EARLY] Extend the maximum number of unique System V |
| 2707 | IPC identifiers from 32,768 to 16,777,216. |
| 2708 | |
| 2709 | ipe.enforce= [IPE] |
| 2710 | Format: <bool> |
| 2711 | Determine whether IPE starts in permissive (0) or |
| 2712 | enforce (1) mode. The default is enforce. |
| 2713 | |
| 2714 | ipe.success_audit= |
| 2715 | [IPE] |
| 2716 | Format: <bool> |
| 2717 | Start IPE with success auditing enabled, emitting |
| 2718 | an audit event when a binary is allowed. The default |
| 2719 | is 0. |
| 2720 | |
| 2721 | irqaffinity= [SMP] Set the default irq affinity mask |
| 2722 | The argument is a cpu list, as described above. |
| 2723 | |
| 2724 | irqchip.gicv2_force_probe= |
| 2725 | [ARM,ARM64,EARLY] |
| 2726 | Format: <bool> |
| 2727 | Force the kernel to look for the second 4kB page |
| 2728 | of a GICv2 controller even if the memory range |
| 2729 | exposed by the device tree is too small. |
| 2730 | |
| 2731 | irqchip.gicv3_nolpi= |
| 2732 | [ARM,ARM64,EARLY] |
| 2733 | Force the kernel to ignore the availability of |
| 2734 | LPIs (and by consequence ITSs). Intended for system |
| 2735 | that use the kernel as a bootloader, and thus want |
| 2736 | to let secondary kernels in charge of setting up |
| 2737 | LPIs. |
| 2738 | |
| 2739 | irqchip.gicv3_pseudo_nmi= [ARM64,EARLY] |
| 2740 | Enables support for pseudo-NMIs in the kernel. This |
| 2741 | requires the kernel to be built with |
| 2742 | CONFIG_ARM64_PSEUDO_NMI. |
| 2743 | |
| 2744 | irqchip.riscv_imsic_noipi |
| 2745 | [RISC-V,EARLY] |
| 2746 | Force the kernel to not use IMSIC software injected MSIs |
| 2747 | as IPIs. Intended for system where IMSIC is trap-n-emulated, |
| 2748 | and thus want to reduce MMIO traps when triggering IPIs |
| 2749 | to multiple harts. |
| 2750 | |
| 2751 | irqfixup [HW] |
| 2752 | When an interrupt is not handled search all handlers |
| 2753 | for it. Intended to get systems with badly broken |
| 2754 | firmware running. |
| 2755 | |
| 2756 | irqhandler.duration_warn_us= [KNL] |
| 2757 | Warn if an IRQ handler exceeds the specified duration |
| 2758 | threshold in microseconds. Useful for identifying |
| 2759 | long-running IRQs in the system. |
| 2760 | |
| 2761 | irqpoll [HW] |
| 2762 | When an interrupt is not handled search all handlers |
| 2763 | for it. Also check all handlers each timer |
| 2764 | interrupt. Intended to get systems with badly broken |
| 2765 | firmware running. |
| 2766 | |
| 2767 | isapnp= [ISAPNP] |
| 2768 | Format: <RDP>,<reset>,<pci_scan>,<verbosity> |
| 2769 | |
| 2770 | isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. |
| 2771 | Format: [flag-list,]<cpu-list> |
| 2772 | |
| 2773 | Specify one or more CPUs to isolate from disturbances |
| 2774 | specified in the flag list (default: domain): |
| 2775 | |
| 2776 | nohz |
| 2777 | Disable the tick when a single task runs as well as |
| 2778 | disabling other kernel noises like having RCU callbacks |
| 2779 | offloaded. This is equivalent to the nohz_full parameter. |
| 2780 | |
| 2781 | A residual 1Hz tick is offloaded to workqueues, which you |
| 2782 | need to affine to housekeeping through the global |
| 2783 | workqueue's affinity configured via the |
| 2784 | /sys/devices/virtual/workqueue/cpumask sysfs file, or |
| 2785 | by using the 'domain' flag described below. |
| 2786 | |
| 2787 | NOTE: by default the global workqueue runs on all CPUs, |
| 2788 | so to protect individual CPUs the 'cpumask' file has to |
| 2789 | be configured manually after bootup. |
| 2790 | |
| 2791 | domain |
| 2792 | Isolate from the general SMP balancing and scheduling |
| 2793 | algorithms. Note that performing domain isolation this way |
| 2794 | is irreversible: it's not possible to bring back a CPU to |
| 2795 | the domains once isolated through this boot time |
| 2796 | configuration. Use cpusets for a dynamic configuration |
| 2797 | which can be altered at runtime. For details see |
| 2798 | Documentation/admin-guide/cpu-isolation.rst. |
| 2799 | |
| 2800 | You can move a process onto or off an "isolated" CPU via |
| 2801 | the CPU affinity syscalls or cpuset. |
| 2802 | <cpu number> begins at 0 and the maximum value is |
| 2803 | "number of CPUs in system - 1". |
| 2804 | |
| 2805 | managed_irq |
| 2806 | |
| 2807 | Isolate from being targeted by managed interrupts |
| 2808 | which have an interrupt mask containing isolated |
| 2809 | CPUs. The affinity of managed interrupts is |
| 2810 | handled by the kernel and cannot be changed via |
| 2811 | the /proc/irq/* interfaces. |
| 2812 | |
| 2813 | This isolation is best effort and only effective |
| 2814 | if the automatically assigned interrupt mask of a |
| 2815 | device queue contains isolated and housekeeping |
| 2816 | CPUs. If housekeeping CPUs are online then such |
| 2817 | interrupts are directed to the housekeeping CPU |
| 2818 | so that IO submitted on the housekeeping CPU |
| 2819 | cannot disturb the isolated CPU. |
| 2820 | |
| 2821 | If a queue's affinity mask contains only isolated |
| 2822 | CPUs then this parameter has no effect on the |
| 2823 | interrupt routing decision, though interrupts are |
| 2824 | only delivered when tasks running on those |
| 2825 | isolated CPUs submit IO. IO submitted on |
| 2826 | housekeeping CPUs has no influence on those |
| 2827 | queues. |
| 2828 | |
| 2829 | The format of <cpu-list> is described above. |
| 2830 | |
| 2831 | iucv= [HW,NET] |
| 2832 | |
| 2833 | ivrs_ioapic [HW,X86-64] |
| 2834 | Provide an override to the IOAPIC-ID<->DEVICE-ID |
| 2835 | mapping provided in the IVRS ACPI table. |
| 2836 | By default, PCI segment is 0, and can be omitted. |
| 2837 | |
| 2838 | For example, to map IOAPIC-ID decimal 10 to |
| 2839 | PCI segment 0x1 and PCI device 00:14.0, |
| 2840 | write the parameter as: |
| 2841 | ivrs_ioapic=10@0001:00:14.0 |
| 2842 | |
| 2843 | Deprecated formats: |
| 2844 | * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 |
| 2845 | write the parameter as: |
| 2846 | ivrs_ioapic[10]=00:14.0 |
| 2847 | * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and |
| 2848 | PCI device 00:14.0 write the parameter as: |
| 2849 | ivrs_ioapic[10]=0001:00:14.0 |
| 2850 | |
| 2851 | ivrs_hpet [HW,X86-64] |
| 2852 | Provide an override to the HPET-ID<->DEVICE-ID |
| 2853 | mapping provided in the IVRS ACPI table. |
| 2854 | By default, PCI segment is 0, and can be omitted. |
| 2855 | |
| 2856 | For example, to map HPET-ID decimal 10 to |
| 2857 | PCI segment 0x1 and PCI device 00:14.0, |
| 2858 | write the parameter as: |
| 2859 | ivrs_hpet=10@0001:00:14.0 |
| 2860 | |
| 2861 | Deprecated formats: |
| 2862 | * To map HPET-ID decimal 0 to PCI device 00:14.0 |
| 2863 | write the parameter as: |
| 2864 | ivrs_hpet[0]=00:14.0 |
| 2865 | * To map HPET-ID decimal 10 to PCI segment 0x1 and |
| 2866 | PCI device 00:14.0 write the parameter as: |
| 2867 | ivrs_ioapic[10]=0001:00:14.0 |
| 2868 | |
| 2869 | ivrs_acpihid [HW,X86-64] |
| 2870 | Provide an override to the ACPI-HID:UID<->DEVICE-ID |
| 2871 | mapping provided in the IVRS ACPI table. |
| 2872 | By default, PCI segment is 0, and can be omitted. |
| 2873 | |
| 2874 | For example, to map UART-HID:UID AMD0020:0 to |
| 2875 | PCI segment 0x1 and PCI device ID 00:14.5, |
| 2876 | write the parameter as: |
| 2877 | ivrs_acpihid=AMD0020:0@0001:00:14.5 |
| 2878 | |
| 2879 | Deprecated formats: |
| 2880 | * To map UART-HID:UID AMD0020:0 to PCI segment is 0, |
| 2881 | PCI device ID 00:14.5, write the parameter as: |
| 2882 | ivrs_acpihid[00:14.5]=AMD0020:0 |
| 2883 | * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and |
| 2884 | PCI device ID 00:14.5, write the parameter as: |
| 2885 | ivrs_acpihid[0001:00:14.5]=AMD0020:0 |
| 2886 | |
| 2887 | js= [HW,JOY] Analog joystick |
| 2888 | See Documentation/input/joydev/joystick.rst. |
| 2889 | |
| 2890 | kasan_multi_shot |
| 2891 | [KNL] Enforce KASAN (Kernel Address Sanitizer) to print |
| 2892 | report on every invalid memory access. Without this |
| 2893 | parameter KASAN will print report only for the first |
| 2894 | invalid access. |
| 2895 | |
| 2896 | keep_bootcon [KNL,EARLY] |
| 2897 | Do not unregister boot console at start. This is only |
| 2898 | useful for debugging when something happens in the window |
| 2899 | between unregistering the boot console and initializing |
| 2900 | the real console. |
| 2901 | |
| 2902 | keepinitrd [HW,ARM] See retain_initrd. |
| 2903 | |
| 2904 | kernelcore= [KNL,X86,PPC,EARLY] |
| 2905 | Format: nn[KMGTPE] | nn% | "mirror" |
| 2906 | This parameter specifies the amount of memory usable by |
| 2907 | the kernel for non-movable allocations. The requested |
| 2908 | amount is spread evenly throughout all nodes in the |
| 2909 | system as ZONE_NORMAL. The remaining memory is used for |
| 2910 | movable memory in its own zone, ZONE_MOVABLE. In the |
| 2911 | event, a node is too small to have both ZONE_NORMAL and |
| 2912 | ZONE_MOVABLE, kernelcore memory will take priority and |
| 2913 | other nodes will have a larger ZONE_MOVABLE. |
| 2914 | |
| 2915 | ZONE_MOVABLE is used for the allocation of pages that |
| 2916 | may be reclaimed or moved by the page migration |
| 2917 | subsystem. Note that allocations like PTEs-from-HighMem |
| 2918 | still use the HighMem zone if it exists, and the Normal |
| 2919 | zone if it does not. |
| 2920 | |
| 2921 | It is possible to specify the exact amount of memory in |
| 2922 | the form of "nn[KMGTPE]", a percentage of total system |
| 2923 | memory in the form of "nn%", or "mirror". If "mirror" |
| 2924 | option is specified, mirrored (reliable) memory is used |
| 2925 | for non-movable allocations and remaining memory is used |
| 2926 | for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" |
| 2927 | are exclusive, so you cannot specify multiple forms. |
| 2928 | |
| 2929 | kfence.burst= [MM,KFENCE] The number of additional successive |
| 2930 | allocations to be attempted through KFENCE for each |
| 2931 | sample interval. |
| 2932 | Format: <unsigned integer> |
| 2933 | Default: 0 |
| 2934 | |
| 2935 | kfence.check_on_panic= |
| 2936 | [MM,KFENCE] Whether to check all KFENCE-managed objects' |
| 2937 | canaries on panic. |
| 2938 | Format: <bool> |
| 2939 | Default: false |
| 2940 | |
| 2941 | kfence.deferrable= |
| 2942 | [MM,KFENCE] Whether to use a deferrable timer to trigger |
| 2943 | allocations. This avoids forcing CPU wake-ups if the |
| 2944 | system is idle, at the risk of a less predictable |
| 2945 | sample interval. |
| 2946 | Format: <bool> |
| 2947 | Default: CONFIG_KFENCE_DEFERRABLE |
| 2948 | |
| 2949 | kfence.fault= [MM,KFENCE] Controls the behavior when a KFENCE |
| 2950 | error is detected. |
| 2951 | report - print the error report and continue (default). |
| 2952 | oops - print the error report and oops. |
| 2953 | panic - print the error report and panic. |
| 2954 | |
| 2955 | kfence.sample_interval= |
| 2956 | [MM,KFENCE] KFENCE's sample interval in milliseconds. |
| 2957 | Format: <unsigned integer> |
| 2958 | 0 - Disable KFENCE. |
| 2959 | >0 - Enabled KFENCE with given sample interval. |
| 2960 | Default: CONFIG_KFENCE_SAMPLE_INTERVAL |
| 2961 | |
| 2962 | kfence.skip_covered_thresh= |
| 2963 | [MM,KFENCE] If pool utilization reaches this threshold |
| 2964 | (pool usage%), KFENCE limits currently covered |
| 2965 | allocations of the same source from further filling |
| 2966 | up the pool. |
| 2967 | Format: <unsigned integer> |
| 2968 | Default: 75 |
| 2969 | |
| 2970 | kgdbdbgp= [KGDB,HW,EARLY] kgdb over EHCI usb debug port. |
| 2971 | Format: <Controller#>[,poll interval] |
| 2972 | The controller # is the number of the ehci usb debug |
| 2973 | port as it is probed via PCI. The poll interval is |
| 2974 | optional and is the number seconds in between |
| 2975 | each poll cycle to the debug port in case you need |
| 2976 | the functionality for interrupting the kernel with |
| 2977 | gdb or control-c on the dbgp connection. When |
| 2978 | not using this parameter you use sysrq-g to break into |
| 2979 | the kernel debugger. |
| 2980 | |
| 2981 | kgdboc= [KGDB,HW] kgdb over consoles. |
| 2982 | Requires a tty driver that supports console polling, |
| 2983 | or a supported polling keyboard driver (non-usb). |
| 2984 | Serial only format: <serial_device>[,baud] |
| 2985 | keyboard only format: kbd |
| 2986 | keyboard and serial format: kbd,<serial_device>[,baud] |
| 2987 | Optional Kernel mode setting: |
| 2988 | kms, kbd format: kms,kbd |
| 2989 | kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] |
| 2990 | |
| 2991 | kgdboc_earlycon= [KGDB,HW,EARLY] |
| 2992 | If the boot console provides the ability to read |
| 2993 | characters and can work in polling mode, you can use |
| 2994 | this parameter to tell kgdb to use it as a backend |
| 2995 | until the normal console is registered. Intended to |
| 2996 | be used together with the kgdboc parameter which |
| 2997 | specifies the normal console to transition to. |
| 2998 | |
| 2999 | The name of the early console should be specified |
| 3000 | as the value of this parameter. Note that the name of |
| 3001 | the early console might be different than the tty |
| 3002 | name passed to kgdboc. It's OK to leave the value |
| 3003 | blank and the first boot console that implements |
| 3004 | read() will be picked. |
| 3005 | |
| 3006 | kgdbwait [KGDB,EARLY] Stop kernel execution and enter the |
| 3007 | kernel debugger at the earliest opportunity. |
| 3008 | |
| 3009 | kho= [KEXEC,EARLY] |
| 3010 | Format: { "0" | "1" | "off" | "on" | "y" | "n" } |
| 3011 | Enables or disables Kexec HandOver. |
| 3012 | "0" | "off" | "n" - kexec handover is disabled |
| 3013 | "1" | "on" | "y" - kexec handover is enabled |
| 3014 | |
| 3015 | kho_scratch= [KEXEC,EARLY] |
| 3016 | Format: ll[KMG],mm[KMG],nn[KMG] | nn% |
| 3017 | Defines the size of the KHO scratch region. The KHO |
| 3018 | scratch regions are physically contiguous memory |
| 3019 | ranges that can only be used for non-kernel |
| 3020 | allocations. That way, even when memory is heavily |
| 3021 | fragmented with handed over memory, the kexeced |
| 3022 | kernel will always have enough contiguous ranges to |
| 3023 | bootstrap itself. |
| 3024 | |
| 3025 | It is possible to specify the exact amount of |
| 3026 | memory in the form of "ll[KMG],mm[KMG],nn[KMG]" |
| 3027 | where the first parameter defines the size of a low |
| 3028 | memory scratch area, the second parameter defines |
| 3029 | the size of a global scratch area and the third |
| 3030 | parameter defines the size of additional per-node |
| 3031 | scratch areas. The form "nn%" defines scale factor |
| 3032 | (in percents) of memory that was used during boot. |
| 3033 | |
| 3034 | kmac= [MIPS] Korina ethernet MAC address. |
| 3035 | Configure the RouterBoard 532 series on-chip |
| 3036 | Ethernet adapter MAC address. |
| 3037 | |
| 3038 | kmemleak= [KNL,EARLY] Boot-time kmemleak enable/disable |
| 3039 | Valid arguments: on, off |
| 3040 | Default: on |
| 3041 | Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, |
| 3042 | the default is off. |
| 3043 | |
| 3044 | kprobe_event=[probe-list] |
| 3045 | [FTRACE] Add kprobe events and enable at boot time. |
| 3046 | The probe-list is a semicolon delimited list of probe |
| 3047 | definitions. Each definition is same as kprobe_events |
| 3048 | interface, but the parameters are comma delimited. |
| 3049 | For example, to add a kprobe event on vfs_read with |
| 3050 | arg1 and arg2, add to the command line; |
| 3051 | |
| 3052 | kprobe_event=p,vfs_read,$arg1,$arg2 |
| 3053 | |
| 3054 | See also Documentation/trace/kprobetrace.rst "Kernel |
| 3055 | Boot Parameter" section. |
| 3056 | |
| 3057 | kpti= [ARM64,EARLY] Control page table isolation of |
| 3058 | user and kernel address spaces. |
| 3059 | Default: enabled on cores which need mitigation. |
| 3060 | 0: force disabled |
| 3061 | 1: force enabled |
| 3062 | |
| 3063 | kunit.enable= [KUNIT] Enable executing KUnit tests. Requires |
| 3064 | CONFIG_KUNIT to be set to be fully enabled. The |
| 3065 | default value can be overridden via |
| 3066 | KUNIT_DEFAULT_ENABLED. |
| 3067 | Default is 1 (enabled) |
| 3068 | |
| 3069 | kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. |
| 3070 | Default is 0 (don't ignore, but inject #GP) |
| 3071 | |
| 3072 | kvm.eager_page_split= |
| 3073 | [KVM,X86,RISCV] Controls whether or not KVM will try to |
| 3074 | proactively split all huge pages during dirty logging. |
| 3075 | Eager page splitting reduces interruptions to vCPU |
| 3076 | execution by eliminating the write-protection faults |
| 3077 | and MMU lock contention that would otherwise be |
| 3078 | required to split huge pages lazily. |
| 3079 | |
| 3080 | VM workloads that rarely perform writes or that write |
| 3081 | only to a small region of VM memory may benefit from |
| 3082 | disabling eager page splitting to allow huge pages to |
| 3083 | still be used for reads. |
| 3084 | |
| 3085 | The behavior of eager page splitting depends on whether |
| 3086 | KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If |
| 3087 | disabled, all huge pages in a memslot will be eagerly |
| 3088 | split when dirty logging is enabled on that memslot. If |
| 3089 | enabled, eager page splitting will be performed during |
| 3090 | the KVM_CLEAR_DIRTY ioctl, and only for the pages being |
| 3091 | cleared. |
| 3092 | |
| 3093 | On x86, eager page splitting is only supported when |
| 3094 | kvm.tdp_mmu=Y. |
| 3095 | |
| 3096 | On RISCV, eager page splitting is supported by default. |
| 3097 | |
| 3098 | Default is Y (on). |
| 3099 | |
| 3100 | kvm.enable_pmu=[KVM,X86] |
| 3101 | If enabled, KVM will virtualize PMU functionality based |
| 3102 | on the virtual CPU model defined by userspace. This |
| 3103 | can be overridden on a per-VM basis via |
| 3104 | KVM_CAP_PMU_CAPABILITY. |
| 3105 | |
| 3106 | If disabled, KVM will not virtualize PMU functionality, |
| 3107 | e.g. MSRs, PMCs, PMIs, etc., even if userspace defines |
| 3108 | a virtual CPU model that contains PMU assets. |
| 3109 | |
| 3110 | Note, KVM's vPMU support implicitly requires running |
| 3111 | with an in-kernel local APIC, e.g. to deliver PMIs to |
| 3112 | the guest. Running without an in-kernel local APIC is |
| 3113 | not supported, though KVM will allow such a combination |
| 3114 | (with severely degraded functionality). |
| 3115 | |
| 3116 | See also enable_mediated_pmu. |
| 3117 | |
| 3118 | Default is Y (on). |
| 3119 | |
| 3120 | kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86] |
| 3121 | If enabled, KVM will enable virtualization in hardware |
| 3122 | when KVM is loaded, and disable virtualization when KVM |
| 3123 | is unloaded (if KVM is built as a module). |
| 3124 | |
| 3125 | If disabled, KVM will dynamically enable and disable |
| 3126 | virtualization on-demand when creating and destroying |
| 3127 | VMs, i.e. on the 0=>1 and 1=>0 transitions of the |
| 3128 | number of VMs. |
| 3129 | |
| 3130 | Enabling virtualization at module load avoids potential |
| 3131 | latency for creation of the 0=>1 VM, as KVM serializes |
| 3132 | virtualization enabling across all online CPUs. The |
| 3133 | "cost" of enabling virtualization when KVM is loaded, |
| 3134 | is that doing so may interfere with using out-of-tree |
| 3135 | hypervisors that want to "own" virtualization hardware. |
| 3136 | |
| 3137 | kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. |
| 3138 | Default is false (don't support). |
| 3139 | |
| 3140 | kvm.nx_huge_pages= |
| 3141 | [KVM] Controls the software workaround for the |
| 3142 | X86_BUG_ITLB_MULTIHIT bug. |
| 3143 | force : Always deploy workaround. |
| 3144 | off : Never deploy workaround. |
| 3145 | auto : Deploy workaround based on the presence of |
| 3146 | X86_BUG_ITLB_MULTIHIT. |
| 3147 | |
| 3148 | Default is 'auto'. |
| 3149 | |
| 3150 | If the software workaround is enabled for the host, |
| 3151 | guests do need not to enable it for nested guests. |
| 3152 | |
| 3153 | kvm.nx_huge_pages_recovery_ratio= |
| 3154 | [KVM] Controls how many 4KiB pages are periodically zapped |
| 3155 | back to huge pages. 0 disables the recovery, otherwise if |
| 3156 | the value is N KVM will zap 1/Nth of the 4KiB pages every |
| 3157 | period (see below). The default is 60. |
| 3158 | |
| 3159 | kvm.nx_huge_pages_recovery_period_ms= |
| 3160 | [KVM] Controls the time period at which KVM zaps 4KiB pages |
| 3161 | back to huge pages. If the value is a non-zero N, KVM will |
| 3162 | zap a portion (see ratio above) of the pages every N msecs. |
| 3163 | If the value is 0 (the default), KVM will pick a period based |
| 3164 | on the ratio, such that a page is zapped after 1 hour on average. |
| 3165 | |
| 3166 | kvm-{amd,intel}.enable_mediated_pmu=[KVM,AMD,INTEL] |
| 3167 | If enabled, KVM will provide a mediated virtual PMU, |
| 3168 | instead of the default perf-based virtual PMU (if |
| 3169 | kvm.enable_pmu is true and PMU is enumerated via the |
| 3170 | virtual CPU model). |
| 3171 | |
| 3172 | With a perf-based vPMU, KVM operates as a user of perf, |
| 3173 | i.e. emulates guest PMU counters using perf events. |
| 3174 | KVM-created perf events are managed by perf as regular |
| 3175 | (guest-only) events, e.g. are scheduled in/out, contend |
| 3176 | for hardware resources, etc. Using a perf-based vPMU |
| 3177 | allows guest and host usage of the PMU to co-exist, but |
| 3178 | incurs non-trivial overhead and can result in silently |
| 3179 | dropped guest events (due to resource contention). |
| 3180 | |
| 3181 | With a mediated vPMU, hardware PMU state is context |
| 3182 | switched around the world switch to/from the guest. |
| 3183 | KVM mediates which events the guest can utilize, but |
| 3184 | gives the guest direct access to all other PMU assets |
| 3185 | when possible (KVM may intercept some accesses if the |
| 3186 | virtual CPU model provides a subset of hardware PMU |
| 3187 | functionality). Using a mediated vPMU significantly |
| 3188 | reduces PMU virtualization overhead and eliminates lost |
| 3189 | guest events, but is mutually exclusive with using perf |
| 3190 | to profile KVM guests and adds latency to most VM-Exits |
| 3191 | (to context switch PMU state). |
| 3192 | |
| 3193 | Default is N (off). |
| 3194 | |
| 3195 | kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in |
| 3196 | KVM/SVM. Default is 1 (enabled). |
| 3197 | |
| 3198 | kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, |
| 3199 | a.k.a. Two-Dimensional Page Tables. Default is 1 |
| 3200 | (enabled). Disable by KVM if hardware lacks support |
| 3201 | for NPT. |
| 3202 | |
| 3203 | kvm-amd.ciphertext_hiding_asids= |
| 3204 | [KVM,AMD] Ciphertext hiding prevents disallowed accesses |
| 3205 | to SNP private memory from reading ciphertext. Instead, |
| 3206 | reads will see constant default values (0xff). |
| 3207 | |
| 3208 | If ciphertext hiding is enabled, the joint SEV-ES and |
| 3209 | SEV-SNP ASID space is partitioned into separate SEV-ES |
| 3210 | and SEV-SNP ASID ranges, with the SEV-SNP range being |
| 3211 | [1..max_snp_asid] and the SEV-ES range being |
| 3212 | (max_snp_asid..min_sev_asid), where min_sev_asid is |
| 3213 | enumerated by CPUID.0x.8000_001F[EDX]. |
| 3214 | |
| 3215 | A non-zero value enables SEV-SNP ciphertext hiding and |
| 3216 | adjusts the ASID ranges for SEV-ES and SEV-SNP guests. |
| 3217 | KVM caps the number of SEV-SNP ASIDs at the maximum |
| 3218 | possible value, e.g. specifying -1u will assign all |
| 3219 | joint SEV-ES and SEV-SNP ASIDs to SEV-SNP. Note, |
| 3220 | assigning all joint ASIDs to SEV-SNP, i.e. configuring |
| 3221 | max_snp_asid == min_sev_asid-1, will effectively make |
| 3222 | SEV-ES unusable. |
| 3223 | |
| 3224 | kvm-arm.mode= |
| 3225 | [KVM,ARM,EARLY] Select one of KVM/arm64's modes of |
| 3226 | operation. |
| 3227 | |
| 3228 | none: Forcefully disable KVM. |
| 3229 | |
| 3230 | nvhe: Standard nVHE-based mode, without support for |
| 3231 | protected guests. |
| 3232 | |
| 3233 | protected: Mode with support for guests whose state is |
| 3234 | kept private from the host, using VHE or |
| 3235 | nVHE depending on HW support. |
| 3236 | |
| 3237 | nested: VHE-based mode with support for nested |
| 3238 | virtualization. Requires at least ARMv8.4 |
| 3239 | hardware (with FEAT_NV2). |
| 3240 | |
| 3241 | Defaults to VHE/nVHE based on hardware support. Setting |
| 3242 | mode to "protected" will disable kexec and hibernation |
| 3243 | for the host. To force nVHE on VHE hardware, add |
| 3244 | "arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the |
| 3245 | command-line. |
| 3246 | "nested" and "protected" are experimental and should be |
| 3247 | used with extreme caution. |
| 3248 | |
| 3249 | kvm-arm.vgic_v3_group0_trap= |
| 3250 | [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0 |
| 3251 | system registers |
| 3252 | |
| 3253 | kvm-arm.vgic_v3_group1_trap= |
| 3254 | [KVM,ARM,EARLY] Trap guest accesses to GICv3 group-1 |
| 3255 | system registers |
| 3256 | |
| 3257 | kvm-arm.vgic_v3_common_trap= |
| 3258 | [KVM,ARM,EARLY] Trap guest accesses to GICv3 common |
| 3259 | system registers |
| 3260 | |
| 3261 | kvm-arm.vgic_v4_enable= |
| 3262 | [KVM,ARM,EARLY] Allow use of GICv4 for direct |
| 3263 | injection of LPIs. |
| 3264 | |
| 3265 | kvm-arm.wfe_trap_policy= |
| 3266 | [KVM,ARM] Control when to set WFE instruction trap for |
| 3267 | KVM VMs. Traps are allowed but not guaranteed by the |
| 3268 | CPU architecture. |
| 3269 | |
| 3270 | trap: set WFE instruction trap |
| 3271 | |
| 3272 | notrap: clear WFE instruction trap |
| 3273 | |
| 3274 | kvm-arm.wfi_trap_policy= |
| 3275 | [KVM,ARM] Control when to set WFI instruction trap for |
| 3276 | KVM VMs. Traps are allowed but not guaranteed by the |
| 3277 | CPU architecture. |
| 3278 | |
| 3279 | trap: set WFI instruction trap |
| 3280 | |
| 3281 | notrap: clear WFI instruction trap |
| 3282 | |
| 3283 | kvm_cma_resv_ratio=n [PPC,EARLY] |
| 3284 | Reserves given percentage from system memory area for |
| 3285 | contiguous memory allocation for KVM hash pagetable |
| 3286 | allocation. |
| 3287 | By default it reserves 5% of total system memory. |
| 3288 | Format: <integer> |
| 3289 | Default: 5 |
| 3290 | |
| 3291 | kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, |
| 3292 | a.k.a. Two-Dimensional Page Tables. Default is 1 |
| 3293 | (enabled). Disable by KVM if hardware lacks support |
| 3294 | for EPT. |
| 3295 | |
| 3296 | kvm-intel.emulate_invalid_guest_state= |
| 3297 | [KVM,Intel] Control whether to emulate invalid guest |
| 3298 | state. Ignored if kvm-intel.enable_unrestricted_guest=1, |
| 3299 | as guest state is never invalid for unrestricted |
| 3300 | guests. This param doesn't apply to nested guests (L2), |
| 3301 | as KVM never emulates invalid L2 guest state. |
| 3302 | Default is 1 (enabled). |
| 3303 | |
| 3304 | kvm-intel.flexpriority= |
| 3305 | [KVM,Intel] Control KVM's use of FlexPriority feature |
| 3306 | (TPR shadow). Default is 1 (enabled). Disable by KVM if |
| 3307 | hardware lacks support for it. |
| 3308 | |
| 3309 | kvm-intel.nested= |
| 3310 | [KVM,Intel] Control nested virtualization feature in |
| 3311 | KVM/VMX. Default is 1 (enabled). |
| 3312 | |
| 3313 | kvm-intel.unrestricted_guest= |
| 3314 | [KVM,Intel] Control KVM's use of unrestricted guest |
| 3315 | feature (virtualized real and unpaged mode). Default |
| 3316 | is 1 (enabled). Disable by KVM if EPT is disabled or |
| 3317 | hardware lacks support for it. |
| 3318 | |
| 3319 | kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault |
| 3320 | CVE-2018-3620. |
| 3321 | |
| 3322 | Valid arguments: never, cond, always |
| 3323 | |
| 3324 | always: L1D cache flush on every VMENTER. |
| 3325 | cond: Flush L1D on VMENTER only when the code between |
| 3326 | VMEXIT and VMENTER can leak host memory. |
| 3327 | never: Disables the mitigation |
| 3328 | |
| 3329 | Default is cond (do L1 cache flush in specific instances) |
| 3330 | |
| 3331 | kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor |
| 3332 | Identification feature (tagged TLBs). Default is 1 |
| 3333 | (enabled). Disable by KVM if hardware lacks support |
| 3334 | for it. |
| 3335 | |
| 3336 | l1d_flush= [X86,INTEL,EARLY] |
| 3337 | Control mitigation for L1D based snooping vulnerability. |
| 3338 | |
| 3339 | Certain CPUs are vulnerable to an exploit against CPU |
| 3340 | internal buffers which can forward information to a |
| 3341 | disclosure gadget under certain conditions. |
| 3342 | |
| 3343 | In vulnerable processors, the speculatively |
| 3344 | forwarded data can be used in a cache side channel |
| 3345 | attack, to access data to which the attacker does |
| 3346 | not have direct access. |
| 3347 | |
| 3348 | This parameter controls the mitigation. The |
| 3349 | options are: |
| 3350 | |
| 3351 | on - enable the interface for the mitigation |
| 3352 | |
| 3353 | l1tf= [X86,EARLY] Control mitigation of the L1TF vulnerability on |
| 3354 | affected CPUs |
| 3355 | |
| 3356 | The kernel PTE inversion protection is unconditionally |
| 3357 | enabled and cannot be disabled. |
| 3358 | |
| 3359 | full |
| 3360 | Provides all available mitigations for the |
| 3361 | L1TF vulnerability. Disables SMT and |
| 3362 | enables all mitigations in the |
| 3363 | hypervisors, i.e. unconditional L1D flush. |
| 3364 | |
| 3365 | SMT control and L1D flush control via the |
| 3366 | sysfs interface is still possible after |
| 3367 | boot. Hypervisors will issue a warning |
| 3368 | when the first VM is started in a |
| 3369 | potentially insecure configuration, |
| 3370 | i.e. SMT enabled or L1D flush disabled. |
| 3371 | |
| 3372 | full,force |
| 3373 | Same as 'full', but disables SMT and L1D |
| 3374 | flush runtime control. Implies the |
| 3375 | 'nosmt=force' command line option. |
| 3376 | (i.e. sysfs control of SMT is disabled.) |
| 3377 | |
| 3378 | flush |
| 3379 | Leaves SMT enabled and enables the default |
| 3380 | hypervisor mitigation, i.e. conditional |
| 3381 | L1D flush. |
| 3382 | |
| 3383 | SMT control and L1D flush control via the |
| 3384 | sysfs interface is still possible after |
| 3385 | boot. Hypervisors will issue a warning |
| 3386 | when the first VM is started in a |
| 3387 | potentially insecure configuration, |
| 3388 | i.e. SMT enabled or L1D flush disabled. |
| 3389 | |
| 3390 | flush,nosmt |
| 3391 | |
| 3392 | Disables SMT and enables the default |
| 3393 | hypervisor mitigation. |
| 3394 | |
| 3395 | SMT control and L1D flush control via the |
| 3396 | sysfs interface is still possible after |
| 3397 | boot. Hypervisors will issue a warning |
| 3398 | when the first VM is started in a |
| 3399 | potentially insecure configuration, |
| 3400 | i.e. SMT enabled or L1D flush disabled. |
| 3401 | |
| 3402 | flush,nowarn |
| 3403 | Same as 'flush', but hypervisors will not |
| 3404 | warn when a VM is started in a potentially |
| 3405 | insecure configuration. |
| 3406 | |
| 3407 | off |
| 3408 | Disables hypervisor mitigations and doesn't |
| 3409 | emit any warnings. |
| 3410 | It also drops the swap size and available |
| 3411 | RAM limit restriction on both hypervisor and |
| 3412 | bare metal. |
| 3413 | |
| 3414 | Default is 'flush'. |
| 3415 | |
| 3416 | For details see: Documentation/admin-guide/hw-vuln/l1tf.rst |
| 3417 | |
| 3418 | l2cr= [PPC] |
| 3419 | |
| 3420 | l3cr= [PPC] |
| 3421 | |
| 3422 | lapic [X86-32,APIC,EARLY] Enable the local APIC even if BIOS |
| 3423 | disabled it. |
| 3424 | |
| 3425 | lapic= [X86,APIC] Do not use TSC deadline |
| 3426 | value for LAPIC timer one-shot implementation. Default |
| 3427 | back to the programmable timer unit in the LAPIC. |
| 3428 | Format: notscdeadline |
| 3429 | |
| 3430 | lapic_timer_c2_ok [X86,APIC,EARLY] trust the local apic timer |
| 3431 | in C2 power state. |
| 3432 | |
| 3433 | libata.dma= [LIBATA] DMA control |
| 3434 | libata.dma=0 Disable all PATA and SATA DMA |
| 3435 | libata.dma=1 PATA and SATA Disk DMA only |
| 3436 | libata.dma=2 ATAPI (CDROM) DMA only |
| 3437 | libata.dma=4 Compact Flash DMA only |
| 3438 | Combinations also work, so libata.dma=3 enables DMA |
| 3439 | for disks and CDROMs, but not CFs. |
| 3440 | |
| 3441 | libata.ignore_hpa= [LIBATA] Ignore HPA limit |
| 3442 | libata.ignore_hpa=0 keep BIOS limits (default) |
| 3443 | libata.ignore_hpa=1 ignore limits, using full disk |
| 3444 | |
| 3445 | libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume |
| 3446 | when set. |
| 3447 | Format: <int> |
| 3448 | |
| 3449 | libata.force= [LIBATA] Force configurations. The format is a comma- |
| 3450 | separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. |
| 3451 | PORT and DEVICE are decimal numbers matching port, link |
| 3452 | or device. Basically, it matches the ATA ID string |
| 3453 | printed on console by libata. If the whole ID part is |
| 3454 | omitted, the last PORT and DEVICE values are used. If |
| 3455 | ID hasn't been specified yet, the configuration applies |
| 3456 | to all ports, links and devices. |
| 3457 | |
| 3458 | If only DEVICE is omitted, the parameter applies to |
| 3459 | the port and all links and devices behind it. DEVICE |
| 3460 | number of 0 either selects the first device or the |
| 3461 | first fan-out link behind PMP device. It does not |
| 3462 | select the host link. DEVICE number of 15 selects the |
| 3463 | host link and device attached to it. |
| 3464 | |
| 3465 | The VAL specifies the configuration to force. As long |
| 3466 | as there is no ambiguity, shortcut notation is allowed. |
| 3467 | For example, both 1.5 and 1.5G would work for 1.5Gbps. |
| 3468 | The following configurations can be forced. |
| 3469 | |
| 3470 | * Cable type: 40c, 80c, short40c, unk, ign or sata. |
| 3471 | Any ID with matching PORT is used. |
| 3472 | |
| 3473 | * SATA link speed limit: 1.5Gbps or 3.0Gbps. |
| 3474 | |
| 3475 | * Transfer mode: pio[0-6], mwdma[0-4] and udma[0-7]. |
| 3476 | udma[/][16,25,33,44,66,100,133] notation is also |
| 3477 | allowed. |
| 3478 | |
| 3479 | * nohrst, nosrst, norst: suppress hard, soft and both |
| 3480 | resets. |
| 3481 | |
| 3482 | * rstonce: only attempt one reset during hot-unplug |
| 3483 | link recovery. |
| 3484 | |
| 3485 | * [no]dbdelay: Enable or disable the extra 200ms delay |
| 3486 | before debouncing a link PHY and device presence |
| 3487 | detection. |
| 3488 | |
| 3489 | * [no]ncq: Turn on or off NCQ. |
| 3490 | |
| 3491 | * [no]ncqtrim: Enable or disable queued DSM TRIM. |
| 3492 | |
| 3493 | * [no]ncqati: Enable or disable NCQ on ATI chipsets. |
| 3494 | |
| 3495 | * [no]trim: Enable or disable (unqueued) TRIM. |
| 3496 | |
| 3497 | * trim_zero: Indicate that TRIM command zeroes data. |
| 3498 | |
| 3499 | * max_trim_128m: Set 128M maximum trim size limit. |
| 3500 | |
| 3501 | * [no]dma: Turn on or off DMA transfers. |
| 3502 | |
| 3503 | * atapi_dmadir: Enable ATAPI DMADIR bridge support. |
| 3504 | |
| 3505 | * atapi_mod16_dma: Enable the use of ATAPI DMA for |
| 3506 | commands that are not a multiple of 16 bytes. |
| 3507 | |
| 3508 | * [no]dmalog: Enable or disable the use of the |
| 3509 | READ LOG DMA EXT command to access logs. |
| 3510 | |
| 3511 | * [no]iddevlog: Enable or disable access to the |
| 3512 | identify device data log. |
| 3513 | |
| 3514 | * [no]logdir: Enable or disable access to the general |
| 3515 | purpose log directory. |
| 3516 | |
| 3517 | * max_sec=<sectors>: Set the transfer size limit, in |
| 3518 | number of 512-byte sectors, to the value specified in |
| 3519 | <sectors>. The value specified in <sectors> has to be |
| 3520 | a non-zero positive integer. |
| 3521 | |
| 3522 | * max_sec_128: Set transfer size limit to 128 sectors. |
| 3523 | |
| 3524 | * max_sec_1024: Set transfer size limit to 1024 sectors. |
| 3525 | |
| 3526 | * max_sec_lba48: Set transfer size limit to 65535 sectors. |
| 3527 | |
| 3528 | * external: Mark port as external (hotplug-capable). |
| 3529 | |
| 3530 | * [no]lpm: Enable or disable link power management. |
| 3531 | |
| 3532 | * [no]setxfer: Indicate if transfer speed mode setting |
| 3533 | should be skipped. |
| 3534 | |
| 3535 | * [no]fua: Disable or enable FUA (Force Unit Access) |
| 3536 | support for devices supporting this feature. |
| 3537 | |
| 3538 | * dump_id: Dump IDENTIFY data. |
| 3539 | |
| 3540 | * disable: Disable this device. |
| 3541 | |
| 3542 | If there are multiple matching configurations changing |
| 3543 | the same attribute, the last one is used. |
| 3544 | |
| 3545 | liveupdate= [KNL,EARLY] |
| 3546 | Format: <bool> |
| 3547 | Enable Live Update Orchestrator (LUO). |
| 3548 | Default: off. |
| 3549 | |
| 3550 | lockd.nlm_grace_period=P [NFS] Assign grace period. |
| 3551 | Format: <integer> |
| 3552 | |
| 3553 | lockd.nlm_tcpport=N [NFS] Assign TCP port. |
| 3554 | Format: <integer> |
| 3555 | |
| 3556 | lockd.nlm_timeout=T [NFS] Assign timeout value. |
| 3557 | Format: <integer> |
| 3558 | |
| 3559 | lockd.nlm_udpport=M [NFS] Assign UDP port. |
| 3560 | Format: <integer> |
| 3561 | |
| 3562 | lockdown= [SECURITY,EARLY] |
| 3563 | { integrity | confidentiality } |
| 3564 | Enable the kernel lockdown feature. If set to |
| 3565 | integrity, kernel features that allow userland to |
| 3566 | modify the running kernel are disabled. If set to |
| 3567 | confidentiality, kernel features that allow userland |
| 3568 | to extract confidential information from the kernel |
| 3569 | are also disabled. |
| 3570 | |
| 3571 | locktorture.acq_writer_lim= [KNL] |
| 3572 | Set the time limit in jiffies for a lock |
| 3573 | acquisition. Acquisitions exceeding this limit |
| 3574 | will result in a splat once they do complete. |
| 3575 | |
| 3576 | locktorture.bind_readers= [KNL] |
| 3577 | Specify the list of CPUs to which the readers are |
| 3578 | to be bound. |
| 3579 | |
| 3580 | locktorture.bind_writers= [KNL] |
| 3581 | Specify the list of CPUs to which the writers are |
| 3582 | to be bound. |
| 3583 | |
| 3584 | locktorture.call_rcu_chains= [KNL] |
| 3585 | Specify the number of self-propagating call_rcu() |
| 3586 | chains to set up. These are used to ensure that |
| 3587 | there is a high probability of an RCU grace period |
| 3588 | in progress at any given time. Defaults to 0, |
| 3589 | which disables these call_rcu() chains. |
| 3590 | |
| 3591 | locktorture.long_hold= [KNL] |
| 3592 | Specify the duration in milliseconds for the |
| 3593 | occasional long-duration lock hold time. Defaults |
| 3594 | to 100 milliseconds. Select 0 to disable. |
| 3595 | |
| 3596 | locktorture.nested_locks= [KNL] |
| 3597 | Specify the maximum lock nesting depth that |
| 3598 | locktorture is to exercise, up to a limit of 8 |
| 3599 | (MAX_NESTED_LOCKS). Specify zero to disable. |
| 3600 | Note that this parameter is ineffective on types |
| 3601 | of locks that do not support nested acquisition. |
| 3602 | |
| 3603 | locktorture.nreaders_stress= [KNL] |
| 3604 | Set the number of locking read-acquisition kthreads. |
| 3605 | Defaults to being automatically set based on the |
| 3606 | number of online CPUs. |
| 3607 | |
| 3608 | locktorture.nwriters_stress= [KNL] |
| 3609 | Set the number of locking write-acquisition kthreads. |
| 3610 | |
| 3611 | locktorture.onoff_holdoff= [KNL] |
| 3612 | Set time (s) after boot for CPU-hotplug testing. |
| 3613 | |
| 3614 | locktorture.onoff_interval= [KNL] |
| 3615 | Set time (s) between CPU-hotplug operations, or |
| 3616 | zero to disable CPU-hotplug testing. |
| 3617 | |
| 3618 | locktorture.rt_boost= [KNL] |
| 3619 | Do periodic testing of real-time lock priority |
| 3620 | boosting. Select 0 to disable, 1 to boost |
| 3621 | only rt_mutex, and 2 to boost unconditionally. |
| 3622 | Defaults to 2, which might seem to be an |
| 3623 | odd choice, but which should be harmless for |
| 3624 | non-real-time spinlocks, due to their disabling |
| 3625 | of preemption. Note that non-realtime mutexes |
| 3626 | disable boosting. |
| 3627 | |
| 3628 | locktorture.rt_boost_factor= [KNL] |
| 3629 | Number that determines how often and for how |
| 3630 | long priority boosting is exercised. This is |
| 3631 | scaled down by the number of writers, so that the |
| 3632 | number of boosts per unit time remains roughly |
| 3633 | constant as the number of writers increases. |
| 3634 | On the other hand, the duration of each boost |
| 3635 | increases with the number of writers. |
| 3636 | |
| 3637 | locktorture.shuffle_interval= [KNL] |
| 3638 | Set task-shuffle interval (jiffies). Shuffling |
| 3639 | tasks allows some CPUs to go into dyntick-idle |
| 3640 | mode during the locktorture test. |
| 3641 | |
| 3642 | locktorture.shutdown_secs= [KNL] |
| 3643 | Set time (s) after boot system shutdown. This |
| 3644 | is useful for hands-off automated testing. |
| 3645 | |
| 3646 | locktorture.stat_interval= [KNL] |
| 3647 | Time (s) between statistics printk()s. |
| 3648 | |
| 3649 | locktorture.stutter= [KNL] |
| 3650 | Time (s) to stutter testing, for example, |
| 3651 | specifying five seconds causes the test to run for |
| 3652 | five seconds, wait for five seconds, and so on. |
| 3653 | This tests the locking primitive's ability to |
| 3654 | transition abruptly to and from idle. |
| 3655 | |
| 3656 | locktorture.torture_type= [KNL] |
| 3657 | Specify the locking implementation to test. |
| 3658 | |
| 3659 | locktorture.verbose= [KNL] |
| 3660 | Enable additional printk() statements. |
| 3661 | |
| 3662 | locktorture.writer_fifo= [KNL] |
| 3663 | Run the write-side locktorture kthreads at |
| 3664 | sched_set_fifo() real-time priority. |
| 3665 | |
| 3666 | logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver |
| 3667 | Format: <irq> |
| 3668 | |
| 3669 | loglevel= [KNL,EARLY] |
| 3670 | All Kernel Messages with a loglevel smaller than the |
| 3671 | console loglevel will be printed to the console. It can |
| 3672 | also be changed with klogd or other programs. The |
| 3673 | loglevels are defined as follows: |
| 3674 | |
| 3675 | 0 (KERN_EMERG) system is unusable |
| 3676 | 1 (KERN_ALERT) action must be taken immediately |
| 3677 | 2 (KERN_CRIT) critical conditions |
| 3678 | 3 (KERN_ERR) error conditions |
| 3679 | 4 (KERN_WARNING) warning conditions |
| 3680 | 5 (KERN_NOTICE) normal but significant condition |
| 3681 | 6 (KERN_INFO) informational |
| 3682 | 7 (KERN_DEBUG) debug-level messages |
| 3683 | |
| 3684 | log_buf_len=n[KMG] [KNL,EARLY] |
| 3685 | Sets the size of the printk ring buffer, in bytes. |
| 3686 | n must be a power of two and greater than the |
| 3687 | minimal size. The minimal size is defined by |
| 3688 | LOG_BUF_SHIFT kernel config parameter. There |
| 3689 | is also CONFIG_LOG_CPU_MAX_BUF_SHIFT config |
| 3690 | parameter that allows to increase the default size |
| 3691 | depending on the number of CPUs. See init/Kconfig |
| 3692 | for more details. |
| 3693 | |
| 3694 | logo.nologo [FB] Disables display of the built-in Linux logo. |
| 3695 | This may be used to provide more screen space for |
| 3696 | kernel log messages and is useful when debugging |
| 3697 | kernel boot problems. |
| 3698 | |
| 3699 | lp=0 [LP] Specify parallel ports to use, e.g, |
| 3700 | lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses |
| 3701 | lp=reset first parallel port). 'lp=0' disables the |
| 3702 | lp=auto printer driver. 'lp=reset' (which can be |
| 3703 | specified in addition to the ports) causes |
| 3704 | attached printers to be reset. Using |
| 3705 | lp=port1,port2,... specifies the parallel ports |
| 3706 | to associate lp devices with, starting with |
| 3707 | lp0. A port specification may be 'none' to skip |
| 3708 | that lp device, or a parport name such as |
| 3709 | 'parport0'. Specifying 'lp=auto' instead of a |
| 3710 | port specification list means that device IDs |
| 3711 | from each port should be examined, to see if |
| 3712 | an IEEE 1284-compliant printer is attached; if |
| 3713 | so, the driver will manage that printer. |
| 3714 | See also header of drivers/char/lp.c. |
| 3715 | |
| 3716 | lpj=n [KNL] |
| 3717 | Sets loops_per_jiffy to given constant, thus avoiding |
| 3718 | time-consuming boot-time autodetection (up to 250 ms per |
| 3719 | CPU). 0 enables autodetection (default). To determine |
| 3720 | the correct value for your kernel, boot with normal |
| 3721 | autodetection and see what value is printed. Note that |
| 3722 | on SMP systems the preset will be applied to all CPUs, |
| 3723 | which is likely to cause problems if your CPUs need |
| 3724 | significantly divergent settings. An incorrect value |
| 3725 | will cause delays in the kernel to be wrong, leading to |
| 3726 | unpredictable I/O errors and other breakage. Although |
| 3727 | unlikely, in the extreme case this might damage your |
| 3728 | hardware. |
| 3729 | |
| 3730 | lsm.debug [SECURITY] Enable LSM initialization debugging output. |
| 3731 | |
| 3732 | lsm=lsm1,...,lsmN |
| 3733 | [SECURITY] Choose order of LSM initialization. This |
| 3734 | overrides CONFIG_LSM, and the "security=" parameter. |
| 3735 | |
| 3736 | machtype= [Loongson] Share the same kernel image file between |
| 3737 | different yeeloong laptops. |
| 3738 | Example: machtype=lemote-yeeloong-2f-7inch |
| 3739 | |
| 3740 | maxcpus= [SMP,EARLY] Maximum number of processors that an SMP kernel |
| 3741 | will bring up during bootup. maxcpus=n : n >= 0 limits |
| 3742 | the kernel to bring up 'n' processors. Surely after |
| 3743 | bootup you can bring up the other plugged cpu by executing |
| 3744 | "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus |
| 3745 | only takes effect during system bootup. |
| 3746 | While n=0 is a special case, it is equivalent to "nosmp", |
| 3747 | which also disables the IO APIC. |
| 3748 | |
| 3749 | max_loop= [LOOP] The number of loop block devices that get |
| 3750 | (loop.max_loop) unconditionally pre-created at init time. The default |
| 3751 | number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead |
| 3752 | of statically allocating a predefined number, loop |
| 3753 | devices can be requested on-demand with the |
| 3754 | /dev/loop-control interface. |
| 3755 | |
| 3756 | mce= [X86-{32,64}] |
| 3757 | |
| 3758 | Please see Documentation/arch/x86/x86_64/machinecheck.rst for sysfs runtime tunables. |
| 3759 | |
| 3760 | off |
| 3761 | disable machine check |
| 3762 | |
| 3763 | no_cmci |
| 3764 | disable CMCI(Corrected Machine Check Interrupt) that |
| 3765 | Intel processor supports. Usually this disablement is |
| 3766 | not recommended, but it might be handy if your |
| 3767 | hardware is misbehaving. |
| 3768 | |
| 3769 | Note that you'll get more problems without CMCI than |
| 3770 | with due to the shared banks, i.e. you might get |
| 3771 | duplicated error logs. |
| 3772 | |
| 3773 | dont_log_ce |
| 3774 | don't make logs for corrected errors. All events |
| 3775 | reported as corrected are silently cleared by OS. This |
| 3776 | option will be useful if you have no interest in any |
| 3777 | of corrected errors. |
| 3778 | |
| 3779 | ignore_ce |
| 3780 | disable features for corrected errors, e.g. |
| 3781 | polling timer and CMCI. All events reported as |
| 3782 | corrected are not cleared by OS and remained in its |
| 3783 | error banks. |
| 3784 | |
| 3785 | Usually this disablement is not recommended, however |
| 3786 | if there is an agent checking/clearing corrected |
| 3787 | errors (e.g. BIOS or hardware monitoring |
| 3788 | applications), conflicting with OS's error handling, |
| 3789 | and you cannot deactivate the agent, then this option |
| 3790 | will be a help. |
| 3791 | |
| 3792 | no_lmce |
| 3793 | do not opt-in to Local MCE delivery. Use legacy method |
| 3794 | to broadcast MCEs. |
| 3795 | |
| 3796 | bootlog |
| 3797 | enable logging of machine checks left over from |
| 3798 | booting. Disabled by default on AMD Fam10h and older |
| 3799 | because some BIOS leave bogus ones. |
| 3800 | |
| 3801 | If your BIOS doesn't do that it's a good idea to |
| 3802 | enable though to make sure you log even machine check |
| 3803 | events that result in a reboot. On Intel systems it is |
| 3804 | enabled by default. |
| 3805 | |
| 3806 | nobootlog |
| 3807 | disable boot machine check logging. |
| 3808 | |
| 3809 | print_all |
| 3810 | print all machine check logs to the console. |
| 3811 | |
| 3812 | monarchtimeout (number) |
| 3813 | sets the time in us to wait for other CPUs on machine |
| 3814 | checks. 0 to disable. |
| 3815 | |
| 3816 | bios_cmci_threshold |
| 3817 | don't overwrite the bios-set CMCI threshold. This boot |
| 3818 | option prevents Linux from overwriting the CMCI |
| 3819 | threshold set by the bios. Without this option, Linux |
| 3820 | always sets the CMCI threshold to 1. Enabling this may |
| 3821 | make memory predictive failure analysis less effective |
| 3822 | if the bios sets thresholds for memory errors since we |
| 3823 | will not see details for all errors. |
| 3824 | |
| 3825 | recovery |
| 3826 | force-enable recoverable machine check code paths |
| 3827 | |
| 3828 | Everything else is in sysfs now. |
| 3829 | |
| 3830 | |
| 3831 | md= [HW] RAID subsystems devices and level |
| 3832 | See Documentation/admin-guide/md.rst. |
| 3833 | |
| 3834 | mds= [X86,INTEL,EARLY] |
| 3835 | Control mitigation for the Micro-architectural Data |
| 3836 | Sampling (MDS) vulnerability. |
| 3837 | |
| 3838 | Certain CPUs are vulnerable to an exploit against CPU |
| 3839 | internal buffers which can forward information to a |
| 3840 | disclosure gadget under certain conditions. |
| 3841 | |
| 3842 | In vulnerable processors, the speculatively |
| 3843 | forwarded data can be used in a cache side channel |
| 3844 | attack, to access data to which the attacker does |
| 3845 | not have direct access. |
| 3846 | |
| 3847 | This parameter controls the MDS mitigation. The |
| 3848 | options are: |
| 3849 | |
| 3850 | full - Enable MDS mitigation on vulnerable CPUs |
| 3851 | full,nosmt - Enable MDS mitigation and disable |
| 3852 | SMT on vulnerable CPUs |
| 3853 | off - Unconditionally disable MDS mitigation |
| 3854 | |
| 3855 | On TAA-affected machines, mds=off can be prevented by |
| 3856 | an active TAA mitigation as both vulnerabilities are |
| 3857 | mitigated with the same mechanism so in order to disable |
| 3858 | this mitigation, you need to specify tsx_async_abort=off |
| 3859 | too. |
| 3860 | |
| 3861 | Not specifying this option is equivalent to |
| 3862 | mds=full. |
| 3863 | |
| 3864 | For details see: Documentation/admin-guide/hw-vuln/mds.rst |
| 3865 | |
| 3866 | mem=nn[KMG] [HEXAGON,EARLY] Set the memory size. |
| 3867 | Must be specified, otherwise memory size will be 0. |
| 3868 | |
| 3869 | mem=nn[KMG] [KNL,BOOT,EARLY] Force usage of a specific amount |
| 3870 | of memory Amount of memory to be used in cases |
| 3871 | as follows: |
| 3872 | |
| 3873 | 1 for test; |
| 3874 | 2 when the kernel is not able to see the whole system memory; |
| 3875 | 3 memory that lies after 'mem=' boundary is excluded from |
| 3876 | the hypervisor, then assigned to KVM guests. |
| 3877 | 4 to limit the memory available for kdump kernel. |
| 3878 | |
| 3879 | [ARC,MICROBLAZE] - the limit applies only to low memory, |
| 3880 | high memory is not affected. |
| 3881 | |
| 3882 | [ARM64] - only limits memory covered by the linear |
| 3883 | mapping. The NOMAP regions are not affected. |
| 3884 | |
| 3885 | [X86] Work as limiting max address. Use together |
| 3886 | with memmap= to avoid physical address space collisions. |
| 3887 | Without memmap= PCI devices could be placed at addresses |
| 3888 | belonging to unused RAM. |
| 3889 | |
| 3890 | Note that this only takes effects during boot time since |
| 3891 | in above case 3, memory may need be hot added after boot |
| 3892 | if system memory of hypervisor is not sufficient. |
| 3893 | |
| 3894 | mem=nn[KMG]@ss[KMG] |
| 3895 | [ARM,MIPS,EARLY] - override the memory layout |
| 3896 | reported by firmware. |
| 3897 | Define a memory region of size nn[KMG] starting at |
| 3898 | ss[KMG]. |
| 3899 | Multiple different regions can be specified with |
| 3900 | multiple mem= parameters on the command line. |
| 3901 | |
| 3902 | mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel |
| 3903 | memory. |
| 3904 | |
| 3905 | memblock=debug [KNL,EARLY] Enable memblock debug messages. |
| 3906 | |
| 3907 | memchunk=nn[KMG] |
| 3908 | [KNL,SH] Allow user to override the default size for |
| 3909 | per-device physically contiguous DMA buffers. |
| 3910 | |
| 3911 | memhp_default_state=online/offline/online_kernel/online_movable |
| 3912 | [KNL] Set the initial state for the memory hotplug |
| 3913 | onlining policy. If not specified, the default value is |
| 3914 | set according to the |
| 3915 | CONFIG_MHP_DEFAULT_ONLINE_TYPE kernel config |
| 3916 | options. |
| 3917 | See Documentation/admin-guide/mm/memory-hotplug.rst. |
| 3918 | |
| 3919 | memmap=exactmap [KNL,X86,EARLY] Enable setting of an exact |
| 3920 | E820 memory map, as specified by the user. |
| 3921 | Such memmap=exactmap lines can be constructed based on |
| 3922 | BIOS output or other requirements. See the memmap=nn@ss |
| 3923 | option description. |
| 3924 | |
| 3925 | memmap=nn[KMG]@ss[KMG] |
| 3926 | [KNL, X86,MIPS,XTENSA,EARLY] Force usage of a specific region of memory. |
| 3927 | Region of memory to be used is from ss to ss+nn. |
| 3928 | If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], |
| 3929 | which limits max address to nn[KMG]. |
| 3930 | Multiple different regions can be specified, |
| 3931 | comma delimited. |
| 3932 | Example: |
| 3933 | memmap=100M@2G,100M#3G,1G!1024G |
| 3934 | |
| 3935 | memmap=nn[KMG]#ss[KMG] |
| 3936 | [KNL,ACPI,EARLY] Mark specific memory as ACPI data. |
| 3937 | Region of memory to be marked is from ss to ss+nn. |
| 3938 | |
| 3939 | memmap=nn[KMG]$ss[KMG] |
| 3940 | [KNL,ACPI,EARLY] Mark specific memory as reserved. |
| 3941 | Region of memory to be reserved is from ss to ss+nn. |
| 3942 | Example: Exclude memory from 0x18690000-0x1869ffff |
| 3943 | memmap=64K$0x18690000 |
| 3944 | or |
| 3945 | memmap=0x10000$0x18690000 |
| 3946 | Some bootloaders may need an escape character before '$', |
| 3947 | like Grub2, otherwise '$' and the following number |
| 3948 | will be eaten. |
| 3949 | |
| 3950 | memmap=nn[KMG]!ss[KMG] |
| 3951 | [KNL,X86,EARLY] Mark specific memory as protected. |
| 3952 | Region of memory to be used, from ss to ss+nn. |
| 3953 | The memory region may be marked as e820 type 12 (0xc) |
| 3954 | and is NVDIMM or ADR memory. |
| 3955 | |
| 3956 | memmap=<size>%<offset>-<oldtype>+<newtype> |
| 3957 | [KNL,ACPI,EARLY] Convert memory within the specified region |
| 3958 | from <oldtype> to <newtype>. If "-<oldtype>" is left |
| 3959 | out, the whole region will be marked as <newtype>, |
| 3960 | even if previously unavailable. If "+<newtype>" is left |
| 3961 | out, matching memory will be removed. Types are |
| 3962 | specified as e820 types, e.g., 1 = RAM, 2 = reserved, |
| 3963 | 3 = ACPI, 12 = PRAM. |
| 3964 | |
| 3965 | memory_corruption_check=0/1 [X86,EARLY] |
| 3966 | Some BIOSes seem to corrupt the first 64k of |
| 3967 | memory when doing things like suspend/resume. |
| 3968 | Setting this option will scan the memory |
| 3969 | looking for corruption. Enabling this will |
| 3970 | both detect corruption and prevent the kernel |
| 3971 | from using the memory being corrupted. |
| 3972 | However, it's intended as a diagnostic tool; if |
| 3973 | repeatable BIOS-originated corruption always |
| 3974 | affects the same memory, you can use memmap= |
| 3975 | to prevent the kernel from using that memory. |
| 3976 | |
| 3977 | memory_corruption_check_size=size [X86,EARLY] |
| 3978 | By default it checks for corruption in the low |
| 3979 | 64k, making this memory unavailable for normal |
| 3980 | use. Use this parameter to scan for |
| 3981 | corruption in more or less memory. |
| 3982 | |
| 3983 | memory_corruption_check_period=seconds [X86,EARLY] |
| 3984 | By default it checks for corruption every 60 |
| 3985 | seconds. Use this parameter to check at some |
| 3986 | other rate. 0 disables periodic checking. |
| 3987 | |
| 3988 | memory_hotplug.memmap_on_memory |
| 3989 | [KNL,X86,ARM] Boolean flag to enable this feature. |
| 3990 | Format: {on | off (default)} |
| 3991 | When enabled, runtime hotplugged memory will |
| 3992 | allocate its internal metadata (struct pages, |
| 3993 | those vmemmap pages cannot be optimized even |
| 3994 | if hugetlb_free_vmemmap is enabled) from the |
| 3995 | hotadded memory which will allow to hotadd a |
| 3996 | lot of memory without requiring additional |
| 3997 | memory to do so. |
| 3998 | This feature is disabled by default because it |
| 3999 | has some implication on large (e.g. GB) |
| 4000 | allocations in some configurations (e.g. small |
| 4001 | memory blocks). |
| 4002 | The state of the flag can be read in |
| 4003 | /sys/module/memory_hotplug/parameters/memmap_on_memory. |
| 4004 | Note that even when enabled, there are a few cases where |
| 4005 | the feature is not effective. |
| 4006 | |
| 4007 | mempool_debug [MM] |
| 4008 | Enable mempool debugging. This enables element |
| 4009 | poison checking when freeing elements back to the |
| 4010 | pool. Useful for debugging mempool corruption. |
| 4011 | |
| 4012 | memtest= [KNL,ALPHA,X86,ARM,M68K,PPC,RISCV,EARLY] Enable memtest |
| 4013 | Format: <integer> |
| 4014 | default : 0 <disable> |
| 4015 | Specifies the number of memtest passes to be |
| 4016 | performed. Each pass selects another test |
| 4017 | pattern from a given set of patterns. Memtest |
| 4018 | fills the memory with this pattern, validates |
| 4019 | memory contents and reserves bad memory |
| 4020 | regions that are detected. |
| 4021 | |
| 4022 | mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control |
| 4023 | Valid arguments: on, off |
| 4024 | Default: off |
| 4025 | mem_encrypt=on: Activate SME |
| 4026 | mem_encrypt=off: Do not activate SME |
| 4027 | |
| 4028 | Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst |
| 4029 | for details on when memory encryption can be activated. |
| 4030 | |
| 4031 | mem_sleep_default= [SUSPEND] Default system suspend mode: |
| 4032 | s2idle - Suspend-To-Idle |
| 4033 | shallow - Power-On Suspend or equivalent (if supported) |
| 4034 | deep - Suspend-To-RAM or equivalent (if supported) |
| 4035 | See Documentation/admin-guide/pm/sleep-states.rst. |
| 4036 | |
| 4037 | mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when |
| 4038 | the BIOS has incorrectly applied a workaround. TinyBIOS |
| 4039 | version 0.98 is known to be affected, 0.99 fixes the |
| 4040 | problem by letting the user disable the workaround. |
| 4041 | |
| 4042 | mga= [HW,DRM] |
| 4043 | |
| 4044 | microcode= [X86] Control the behavior of the microcode loader. |
| 4045 | Available options, comma separated: |
| 4046 | |
| 4047 | base_rev=X - with <X> with format: <u32> |
| 4048 | Set the base microcode revision of each thread when in |
| 4049 | debug mode. |
| 4050 | |
| 4051 | dis_ucode_ldr: disable the microcode loader |
| 4052 | |
| 4053 | force_minrev: |
| 4054 | Enable or disable the microcode minimal revision |
| 4055 | enforcement for the runtime microcode loader. |
| 4056 | |
| 4057 | mitigations= |
| 4058 | [X86,PPC,S390,ARM64,EARLY] Control optional mitigations for |
| 4059 | CPU vulnerabilities. This is a set of curated, |
| 4060 | arch-independent options, each of which is an |
| 4061 | aggregation of existing arch-specific options. |
| 4062 | |
| 4063 | Note, "mitigations" is supported if and only if the |
| 4064 | kernel was built with CPU_MITIGATIONS=y. |
| 4065 | |
| 4066 | off |
| 4067 | Disable all optional CPU mitigations. This |
| 4068 | improves system performance, but it may also |
| 4069 | expose users to several CPU vulnerabilities. |
| 4070 | Equivalent to: if nokaslr then kpti=0 [ARM64] |
| 4071 | gather_data_sampling=off [X86] |
| 4072 | indirect_target_selection=off [X86] |
| 4073 | kvm.nx_huge_pages=off [X86] |
| 4074 | l1tf=off [X86] |
| 4075 | mds=off [X86] |
| 4076 | mmio_stale_data=off [X86] |
| 4077 | no_entry_flush [PPC] |
| 4078 | no_uaccess_flush [PPC] |
| 4079 | nobp=0 [S390] |
| 4080 | nopti [X86,PPC] |
| 4081 | nospectre_bhb [ARM64] |
| 4082 | nospectre_v1 [X86,PPC] |
| 4083 | nospectre_v2 [X86,PPC,S390,ARM64] |
| 4084 | reg_file_data_sampling=off [X86] |
| 4085 | retbleed=off [X86] |
| 4086 | spec_rstack_overflow=off [X86] |
| 4087 | spec_store_bypass_disable=off [X86,PPC] |
| 4088 | spectre_bhi=off [X86] |
| 4089 | spectre_v2_user=off [X86] |
| 4090 | srbds=off [X86,INTEL] |
| 4091 | ssbd=force-off [ARM64] |
| 4092 | tsa=off [X86,AMD] |
| 4093 | tsx_async_abort=off [X86] |
| 4094 | vmscape=off [X86] |
| 4095 | |
| 4096 | Exceptions: |
| 4097 | This does not have any effect on |
| 4098 | kvm.nx_huge_pages when |
| 4099 | kvm.nx_huge_pages=force. |
| 4100 | |
| 4101 | auto (default) |
| 4102 | Mitigate all CPU vulnerabilities, but leave SMT |
| 4103 | enabled, even if it's vulnerable. This is for |
| 4104 | users who don't want to be surprised by SMT |
| 4105 | getting disabled across kernel upgrades, or who |
| 4106 | have other ways of avoiding SMT-based attacks. |
| 4107 | Equivalent to: (default behavior) |
| 4108 | |
| 4109 | auto,nosmt |
| 4110 | Mitigate all CPU vulnerabilities, disabling SMT |
| 4111 | if needed. This is for users who always want to |
| 4112 | be fully mitigated, even if it means losing SMT. |
| 4113 | Equivalent to: l1tf=flush,nosmt [X86] |
| 4114 | mds=full,nosmt [X86] |
| 4115 | tsx_async_abort=full,nosmt [X86] |
| 4116 | mmio_stale_data=full,nosmt [X86] |
| 4117 | retbleed=auto,nosmt [X86] |
| 4118 | |
| 4119 | [X86] After one of the above options, additionally |
| 4120 | supports attack-vector based controls as documented in |
| 4121 | Documentation/admin-guide/hw-vuln/attack_vector_controls.rst |
| 4122 | |
| 4123 | mminit_loglevel= |
| 4124 | [KNL,EARLY] When CONFIG_DEBUG_MEMORY_INIT is set, this |
| 4125 | parameter allows control of the logging verbosity for |
| 4126 | the additional memory initialisation checks. A value |
| 4127 | of 0 disables mminit logging and a level of 4 will |
| 4128 | log everything. Information is printed at KERN_DEBUG |
| 4129 | so loglevel=8 may also need to be specified. |
| 4130 | |
| 4131 | mmio_stale_data= |
| 4132 | [X86,INTEL,EARLY] Control mitigation for the Processor |
| 4133 | MMIO Stale Data vulnerabilities. |
| 4134 | |
| 4135 | Processor MMIO Stale Data is a class of |
| 4136 | vulnerabilities that may expose data after an MMIO |
| 4137 | operation. Exposed data could originate or end in |
| 4138 | the same CPU buffers as affected by MDS and TAA. |
| 4139 | Therefore, similar to MDS and TAA, the mitigation |
| 4140 | is to clear the affected CPU buffers. |
| 4141 | |
| 4142 | This parameter controls the mitigation. The |
| 4143 | options are: |
| 4144 | |
| 4145 | full - Enable mitigation on vulnerable CPUs |
| 4146 | |
| 4147 | full,nosmt - Enable mitigation and disable SMT on |
| 4148 | vulnerable CPUs. |
| 4149 | |
| 4150 | off - Unconditionally disable mitigation |
| 4151 | |
| 4152 | On MDS or TAA affected machines, |
| 4153 | mmio_stale_data=off can be prevented by an active |
| 4154 | MDS or TAA mitigation as these vulnerabilities are |
| 4155 | mitigated with the same mechanism so in order to |
| 4156 | disable this mitigation, you need to specify |
| 4157 | mds=off and tsx_async_abort=off too. |
| 4158 | |
| 4159 | Not specifying this option is equivalent to |
| 4160 | mmio_stale_data=full. |
| 4161 | |
| 4162 | For details see: |
| 4163 | Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst |
| 4164 | |
| 4165 | <module>.async_probe[=<bool>] [KNL] |
| 4166 | If no <bool> value is specified or if the value |
| 4167 | specified is not a valid <bool>, enable asynchronous |
| 4168 | probe on this module. Otherwise, enable/disable |
| 4169 | asynchronous probe on this module as indicated by the |
| 4170 | <bool> value. See also: module.async_probe |
| 4171 | |
| 4172 | module.async_probe=<bool> |
| 4173 | [KNL] When set to true, modules will use async probing |
| 4174 | by default. To enable/disable async probing for a |
| 4175 | specific module, use the module specific control that |
| 4176 | is documented under <module>.async_probe. When both |
| 4177 | module.async_probe and <module>.async_probe are |
| 4178 | specified, <module>.async_probe takes precedence for |
| 4179 | the specific module. |
| 4180 | |
| 4181 | module.enable_dups_trace |
| 4182 | [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, |
| 4183 | this means that duplicate request_module() calls will |
| 4184 | trigger a WARN_ON() instead of a pr_warn(). Note that |
| 4185 | if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s |
| 4186 | will always be issued and this option does nothing. |
| 4187 | module.sig_enforce |
| 4188 | [KNL] When CONFIG_MODULE_SIG is set, this means that |
| 4189 | modules without (valid) signatures will fail to load. |
| 4190 | Note that if CONFIG_MODULE_SIG_FORCE is set, that |
| 4191 | is always true, so this option does nothing. |
| 4192 | |
| 4193 | module_blacklist= [KNL] Do not load a comma-separated list of |
| 4194 | modules. Useful for debugging problem modules. |
| 4195 | |
| 4196 | mousedev.tap_time= |
| 4197 | [MOUSE] Maximum time between finger touching and |
| 4198 | leaving touchpad surface for touch to be considered |
| 4199 | a tap and be reported as a left button click (for |
| 4200 | touchpads working in absolute mode only). |
| 4201 | Format: <msecs> |
| 4202 | mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices |
| 4203 | reporting absolute coordinates, such as tablets |
| 4204 | mousedev.yres= [MOUSE] Vertical screen resolution, used for devices |
| 4205 | reporting absolute coordinates, such as tablets |
| 4206 | |
| 4207 | movablecore= [KNL,X86,PPC,EARLY] |
| 4208 | Format: nn[KMGTPE] | nn% |
| 4209 | This parameter is the complement to kernelcore=, it |
| 4210 | specifies the amount of memory used for migratable |
| 4211 | allocations. If both kernelcore and movablecore is |
| 4212 | specified, then kernelcore will be at *least* the |
| 4213 | specified value but may be more. If movablecore on its |
| 4214 | own is specified, the administrator must be careful |
| 4215 | that the amount of memory usable for all allocations |
| 4216 | is not too small. |
| 4217 | |
| 4218 | movable_node [KNL,EARLY] Boot-time switch to make hotplugable memory |
| 4219 | NUMA nodes to be movable. This means that the memory |
| 4220 | of such nodes will be usable only for movable |
| 4221 | allocations which rules out almost all kernel |
| 4222 | allocations. Use with caution! |
| 4223 | |
| 4224 | MTD_Partition= [MTD] |
| 4225 | Format: <name>,<region-number>,<size>,<offset> |
| 4226 | |
| 4227 | MTD_Region= [MTD] Format: |
| 4228 | <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] |
| 4229 | |
| 4230 | mtdparts= [MTD] |
| 4231 | See drivers/mtd/parsers/cmdlinepart.c |
| 4232 | |
| 4233 | mtouchusb.raw_coordinates= |
| 4234 | [HW] Make the MicroTouch USB driver use raw coordinates |
| 4235 | ('y', default) or cooked coordinates ('n') |
| 4236 | |
| 4237 | mtrr=debug [X86,EARLY] |
| 4238 | Enable printing debug information related to MTRR |
| 4239 | registers at boot time. |
| 4240 | |
| 4241 | mtrr_chunk_size=nn[KMG] [X86,EARLY] |
| 4242 | used for mtrr cleanup. It is largest continuous chunk |
| 4243 | that could hold holes aka. UC entries. |
| 4244 | |
| 4245 | mtrr_gran_size=nn[KMG] [X86,EARLY] |
| 4246 | Used for mtrr cleanup. It is granularity of mtrr block. |
| 4247 | Default is 1. |
| 4248 | Large value could prevent small alignment from |
| 4249 | using up MTRRs. |
| 4250 | |
| 4251 | mtrr_spare_reg_nr=n [X86,EARLY] |
| 4252 | Format: <integer> |
| 4253 | Range: 0,7 : spare reg number |
| 4254 | Default : 1 |
| 4255 | Used for mtrr cleanup. It is spare mtrr entries number. |
| 4256 | Set to 2 or more if your graphical card needs more. |
| 4257 | |
| 4258 | multitce=off [PPC] This parameter disables the use of the pSeries |
| 4259 | firmware feature for updating multiple TCE entries |
| 4260 | at a time. |
| 4261 | |
| 4262 | n2= [NET] SDL Inc. RISCom/N2 synchronous serial card |
| 4263 | |
| 4264 | netdev= [NET] NE2000 ISA network devices parameters |
| 4265 | Format: <irq>,<io>,<mem_start>,<mem_end>,<name> |
| 4266 | |
| 4267 | netpoll.carrier_timeout= |
| 4268 | [NET] Specifies amount of time (in seconds) that |
| 4269 | netpoll should wait for a carrier. By default netpoll |
| 4270 | waits 4 seconds. |
| 4271 | |
| 4272 | nf_conntrack.acct= |
| 4273 | [NETFILTER] Enable connection tracking flow accounting |
| 4274 | 0 to disable accounting |
| 4275 | 1 to enable accounting |
| 4276 | Default value is 0. |
| 4277 | |
| 4278 | nfs.cache_getent= |
| 4279 | [NFS] sets the pathname to the program which is used |
| 4280 | to update the NFS client cache entries. |
| 4281 | |
| 4282 | nfs.cache_getent_timeout= |
| 4283 | [NFS] sets the timeout after which an attempt to |
| 4284 | update a cache entry is deemed to have failed. |
| 4285 | |
| 4286 | nfs.callback_nr_threads= |
| 4287 | [NFSv4] set the total number of threads that the |
| 4288 | NFS client will assign to service NFSv4 callback |
| 4289 | requests. |
| 4290 | |
| 4291 | nfs.callback_tcpport= |
| 4292 | [NFS] set the TCP port on which the NFSv4 callback |
| 4293 | channel should listen. |
| 4294 | |
| 4295 | nfs.delay_retrans= |
| 4296 | [NFS] specifies the number of times the NFSv4 client |
| 4297 | retries the request before returning an EAGAIN error, |
| 4298 | after a reply of NFS4ERR_DELAY from the server. |
| 4299 | Only applies if the softerr mount option is enabled, |
| 4300 | and the specified value is >= 0. |
| 4301 | |
| 4302 | nfs.idmap_cache_timeout= |
| 4303 | [NFS] set the maximum lifetime for idmapper cache |
| 4304 | entries. |
| 4305 | |
| 4306 | nfs.max_session_cb_slots= |
| 4307 | [NFSv4.1] Sets the maximum number of session |
| 4308 | slots the client will assign to the callback |
| 4309 | channel. This determines the maximum number of |
| 4310 | callbacks the client will process in parallel for |
| 4311 | a particular server. |
| 4312 | |
| 4313 | nfs.max_session_slots= |
| 4314 | [NFSv4.1] Sets the maximum number of session slots |
| 4315 | the client will attempt to negotiate with the server. |
| 4316 | This limits the number of simultaneous RPC requests |
| 4317 | that the client can send to the NFSv4.1 server. |
| 4318 | Note that there is little point in setting this |
| 4319 | value higher than the max_tcp_slot_table_limit. |
| 4320 | |
| 4321 | nfs.nfs4_disable_idmapping= |
| 4322 | [NFSv4] When set to the default of '1', this option |
| 4323 | ensures that both the RPC level authentication |
| 4324 | scheme and the NFS level operations agree to use |
| 4325 | numeric uids/gids if the mount is using the |
| 4326 | 'sec=sys' security flavour. In effect it is |
| 4327 | disabling idmapping, which can make migration from |
| 4328 | legacy NFSv2/v3 systems to NFSv4 easier. |
| 4329 | Servers that do not support this mode of operation |
| 4330 | will be autodetected by the client, and it will fall |
| 4331 | back to using the idmapper. |
| 4332 | To turn off this behaviour, set the value to '0'. |
| 4333 | |
| 4334 | nfs.nfs4_unique_id= |
| 4335 | [NFS4] Specify an additional fixed unique ident- |
| 4336 | ification string that NFSv4 clients can insert into |
| 4337 | their nfs_client_id4 string. This is typically a |
| 4338 | UUID that is generated at system install time. |
| 4339 | |
| 4340 | nfs.recover_lost_locks= |
| 4341 | [NFSv4] Attempt to recover locks that were lost due |
| 4342 | to a lease timeout on the server. Please note that |
| 4343 | doing this risks data corruption, since there are |
| 4344 | no guarantees that the file will remain unchanged |
| 4345 | after the locks are lost. |
| 4346 | If you want to enable the kernel legacy behaviour of |
| 4347 | attempting to recover these locks, then set this |
| 4348 | parameter to '1'. |
| 4349 | The default parameter value of '0' causes the kernel |
| 4350 | not to attempt recovery of lost locks. |
| 4351 | |
| 4352 | nfs.send_implementation_id= |
| 4353 | [NFSv4.1] Send client implementation identification |
| 4354 | information in exchange_id requests. |
| 4355 | If zero, no implementation identification information |
| 4356 | will be sent. |
| 4357 | The default is to send the implementation identification |
| 4358 | information. |
| 4359 | |
| 4360 | nfs4.layoutstats_timer= |
| 4361 | [NFSv4.2] Change the rate at which the kernel sends |
| 4362 | layoutstats to the pNFS metadata server. |
| 4363 | |
| 4364 | Setting this to value to 0 causes the kernel to use |
| 4365 | whatever value is the default set by the layout |
| 4366 | driver. A non-zero value sets the minimum interval |
| 4367 | in seconds between layoutstats transmissions. |
| 4368 | |
| 4369 | nfsd.inter_copy_offload_enable= |
| 4370 | [NFSv4.2] When set to 1, the server will support |
| 4371 | server-to-server copies for which this server is |
| 4372 | the destination of the copy. |
| 4373 | |
| 4374 | nfsd.nfs4_disable_idmapping= |
| 4375 | [NFSv4] When set to the default of '1', the NFSv4 |
| 4376 | server will return only numeric uids and gids to |
| 4377 | clients using auth_sys, and will accept numeric uids |
| 4378 | and gids from such clients. This is intended to ease |
| 4379 | migration from NFSv2/v3. |
| 4380 | |
| 4381 | nfsd.nfsd4_ssc_umount_timeout= |
| 4382 | [NFSv4.2] When used as the destination of a |
| 4383 | server-to-server copy, knfsd temporarily mounts |
| 4384 | the source server. It caches the mount in case |
| 4385 | it will be needed again, and discards it if not |
| 4386 | used for the number of milliseconds specified by |
| 4387 | this parameter. |
| 4388 | |
| 4389 | nfsaddrs= [NFS] Deprecated. Use ip= instead. |
| 4390 | See Documentation/admin-guide/nfs/nfsroot.rst. |
| 4391 | |
| 4392 | nfsroot= [NFS] nfs root filesystem for disk-less boxes. |
| 4393 | See Documentation/admin-guide/nfs/nfsroot.rst. |
| 4394 | |
| 4395 | nfsrootdebug [NFS] enable nfsroot debugging messages. |
| 4396 | See Documentation/admin-guide/nfs/nfsroot.rst. |
| 4397 | |
| 4398 | nmi_backtrace.backtrace_idle [KNL] |
| 4399 | Dump stacks even of idle CPUs in response to an |
| 4400 | NMI stack-backtrace request. |
| 4401 | |
| 4402 | nmi_debug= [KNL,SH] Specify one or more actions to take |
| 4403 | when a NMI is triggered. |
| 4404 | Format: [state][,regs][,debounce][,die] |
| 4405 | |
| 4406 | nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels |
| 4407 | Format: [panic,][nopanic,][rNNN,][num] |
| 4408 | Valid num: 0 or 1 |
| 4409 | 0 - turn hardlockup detector in nmi_watchdog off |
| 4410 | 1 - turn hardlockup detector in nmi_watchdog on |
| 4411 | rNNN - configure the watchdog with raw perf event 0xNNN |
| 4412 | |
| 4413 | When panic is specified, panic when an NMI watchdog |
| 4414 | timeout occurs (or 'nopanic' to not panic on an NMI |
| 4415 | watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) |
| 4416 | To disable both hard and soft lockup detectors, |
| 4417 | please see 'nowatchdog'. |
| 4418 | This is useful when you use a panic=... timeout and |
| 4419 | need the box quickly up again. |
| 4420 | |
| 4421 | These settings can be accessed at runtime via |
| 4422 | the nmi_watchdog and hardlockup_panic sysctls. |
| 4423 | |
| 4424 | no4lvl [RISCV,EARLY] Disable 4-level and 5-level paging modes. |
| 4425 | Forces kernel to use 3-level paging instead. |
| 4426 | |
| 4427 | no5lvl [X86-64,RISCV,EARLY] Disable 5-level paging mode. Forces |
| 4428 | kernel to use 4-level paging instead. |
| 4429 | |
| 4430 | noalign [KNL,ARM] |
| 4431 | |
| 4432 | noapic [SMP,APIC,EARLY] Tells the kernel to not make use of any |
| 4433 | IOAPICs that may be present in the system. |
| 4434 | |
| 4435 | noapictimer [APIC,X86] Don't set up the APIC timer |
| 4436 | |
| 4437 | noautogroup Disable scheduler automatic task group creation. |
| 4438 | |
| 4439 | nocache [ARM,EARLY] |
| 4440 | |
| 4441 | no_console_suspend |
| 4442 | [HW] Never suspend the console |
| 4443 | Disable suspending of consoles during suspend and |
| 4444 | hibernate operations. Once disabled, debugging |
| 4445 | messages can reach various consoles while the rest |
| 4446 | of the system is being put to sleep (ie, while |
| 4447 | debugging driver suspend/resume hooks). This may |
| 4448 | not work reliably with all consoles, but is known |
| 4449 | to work with serial and VGA consoles. |
| 4450 | To facilitate more flexible debugging, we also add |
| 4451 | console_suspend, a printk module parameter to control |
| 4452 | it. Users could use console_suspend (usually |
| 4453 | /sys/module/printk/parameters/console_suspend) to |
| 4454 | turn on/off it dynamically. |
| 4455 | |
| 4456 | no_debug_objects |
| 4457 | [KNL,EARLY] Disable object debugging |
| 4458 | |
| 4459 | nodsp [SH] Disable hardware DSP at boot time. |
| 4460 | |
| 4461 | noefi [EFI,EARLY] Disable EFI runtime services support. |
| 4462 | |
| 4463 | no_entry_flush [PPC,EARLY] Don't flush the L1-D cache when entering the kernel. |
| 4464 | |
| 4465 | noexec32 [X86-64] |
| 4466 | This affects only 32-bit executables. |
| 4467 | noexec32=on: enable non-executable mappings (default) |
| 4468 | read doesn't imply executable mappings |
| 4469 | noexec32=off: disable non-executable mappings |
| 4470 | read implies executable mappings |
| 4471 | |
| 4472 | no_file_caps Tells the kernel not to honor file capabilities. The |
| 4473 | only way then for a file to be executed with privilege |
| 4474 | is to be setuid root or executed by root. |
| 4475 | |
| 4476 | nofpu [MIPS,SH] Disable hardware FPU at boot time. |
| 4477 | |
| 4478 | nofsgsbase [X86] Disables FSGSBASE instructions. |
| 4479 | |
| 4480 | nofxsr [BUGS=X86-32] Disables x86 floating point extended |
| 4481 | register save and restore. The kernel will only save |
| 4482 | legacy floating-point registers on task switch. |
| 4483 | |
| 4484 | nogbpages [X86] Do not use GB pages for kernel direct mappings. |
| 4485 | |
| 4486 | no_hash_pointers |
| 4487 | [KNL,EARLY] |
| 4488 | Alias for "hash_pointers=never". |
| 4489 | |
| 4490 | nohibernate [HIBERNATION] Disable hibernation and resume. |
| 4491 | |
| 4492 | nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,RISCV,SH] Forces the kernel to |
| 4493 | busy wait in do_idle() and not use the arch_cpu_idle() |
| 4494 | implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP |
| 4495 | to be effective. This is useful on platforms where the |
| 4496 | sleep(SH) or wfi(ARM,ARM64) instructions do not work |
| 4497 | correctly or when doing power measurements to evaluate |
| 4498 | the impact of the sleep instructions. This is also |
| 4499 | useful when using JTAG debugger. |
| 4500 | |
| 4501 | nohpet [X86] Don't use the HPET timer. |
| 4502 | |
| 4503 | nohugeiomap [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge I/O mappings. |
| 4504 | |
| 4505 | nohugevmalloc [KNL,X86,PPC,ARM64,EARLY] Disable kernel huge vmalloc mappings. |
| 4506 | |
| 4507 | nohz= [KNL] Boottime enable/disable dynamic ticks |
| 4508 | Valid arguments: on, off |
| 4509 | Default: on |
| 4510 | |
| 4511 | nohz_full= [KNL,BOOT,SMP,ISOL] |
| 4512 | The argument is a cpu list, as described above. |
| 4513 | In kernels built with CONFIG_NO_HZ_FULL=y, set |
| 4514 | the specified list of CPUs whose tick will be stopped |
| 4515 | whenever possible. The boot CPU will be forced outside |
| 4516 | the range to maintain the timekeeping. Any CPUs |
| 4517 | in this list will have their RCU callbacks offloaded, |
| 4518 | just as if they had also been called out in the |
| 4519 | rcu_nocbs= boot parameter. |
| 4520 | |
| 4521 | Note that this argument takes precedence over |
| 4522 | the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. |
| 4523 | |
| 4524 | noinitrd [Deprecated,RAM] Tells the kernel not to load any configured |
| 4525 | initial RAM disk. Currently this parameter applies to |
| 4526 | initrd only, not to initramfs. But it applies to both |
| 4527 | in EFI mode. |
| 4528 | |
| 4529 | nointremap [X86-64,Intel-IOMMU,EARLY] Do not enable interrupt |
| 4530 | remapping. |
| 4531 | [Deprecated - use intremap=off] |
| 4532 | |
| 4533 | noinvpcid [X86,EARLY] Disable the INVPCID cpu feature. |
| 4534 | |
| 4535 | noiotrap [SH] Disables trapped I/O port accesses. |
| 4536 | |
| 4537 | noirqdebug [X86-32] Disables the code which attempts to detect and |
| 4538 | disable unhandled interrupt sources. |
| 4539 | |
| 4540 | noisapnp [ISAPNP] Disables ISA PnP code. |
| 4541 | |
| 4542 | nokaslr [KNL,EARLY] |
| 4543 | When CONFIG_RANDOMIZE_BASE is set, this disables |
| 4544 | kernel and module base offset ASLR (Address Space |
| 4545 | Layout Randomization). |
| 4546 | |
| 4547 | no-kvmapf [X86,KVM,EARLY] Disable paravirtualized asynchronous page |
| 4548 | fault handling. |
| 4549 | |
| 4550 | no-kvmclock [X86,KVM,EARLY] Disable paravirtualized KVM clock driver |
| 4551 | |
| 4552 | nolapic [X86-32,APIC,EARLY] Do not enable or use the local APIC. |
| 4553 | |
| 4554 | nolapic_timer [X86-32,APIC,EARLY] Do not use the local APIC timer. |
| 4555 | |
| 4556 | nomce [X86-32] Disable Machine Check Exception |
| 4557 | |
| 4558 | nomodeset Disable kernel modesetting. Most systems' firmware |
| 4559 | sets up a display mode and provides framebuffer memory |
| 4560 | for output. With nomodeset, DRM and fbdev drivers will |
| 4561 | not load if they could possibly displace the pre- |
| 4562 | initialized output. Only the system framebuffer will |
| 4563 | be available for use. The respective drivers will not |
| 4564 | perform display-mode changes or accelerated rendering. |
| 4565 | |
| 4566 | Useful as error fallback, or for testing and debugging. |
| 4567 | |
| 4568 | nomodule Disable module load |
| 4569 | |
| 4570 | nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to |
| 4571 | shutdown the other cpus. Instead use the REBOOT_VECTOR |
| 4572 | irq. |
| 4573 | |
| 4574 | nopat [X86,EARLY] Disable PAT (page attribute table extension of |
| 4575 | pagetables) support. |
| 4576 | |
| 4577 | nopcid [X86-64,EARLY] Disable the PCID cpu feature. |
| 4578 | |
| 4579 | nopku [X86] Disable Memory Protection Keys CPU feature found |
| 4580 | in some Intel CPUs. |
| 4581 | |
| 4582 | nopti [X86-64,EARLY] |
| 4583 | Equivalent to pti=off |
| 4584 | |
| 4585 | nopv= [X86,XEN,KVM,HYPER_V,VMWARE,EARLY] |
| 4586 | Disables the PV optimizations forcing the guest to run |
| 4587 | as generic guest with no PV drivers. Currently support |
| 4588 | XEN HVM, KVM, HYPER_V and VMWARE guest. |
| 4589 | |
| 4590 | nopvspin [X86,XEN,KVM,EARLY] |
| 4591 | Disables the qspinlock slow path using PV optimizations |
| 4592 | which allow the hypervisor to 'idle' the guest on lock |
| 4593 | contention. |
| 4594 | |
| 4595 | norandmaps Don't use address space randomization. Equivalent to |
| 4596 | echo 0 > /proc/sys/kernel/randomize_va_space |
| 4597 | |
| 4598 | noresume [SWSUSP] Disables resume and restores original swap |
| 4599 | space. |
| 4600 | |
| 4601 | no-scroll [VGA] Disables scrollback. |
| 4602 | This is required for the Braillex ib80-piezo Braille |
| 4603 | reader made by F.H. Papenmeier (Germany). |
| 4604 | |
| 4605 | nosgx [X86-64,SGX,EARLY] Disables Intel SGX kernel support. |
| 4606 | |
| 4607 | nosmap [PPC,EARLY] |
| 4608 | Disable SMAP (Supervisor Mode Access Prevention) |
| 4609 | even if it is supported by processor. |
| 4610 | |
| 4611 | nosmep [PPC64s,EARLY] |
| 4612 | Disable SMEP (Supervisor Mode Execution Prevention) |
| 4613 | even if it is supported by processor. |
| 4614 | |
| 4615 | nosmp [SMP,EARLY] Tells an SMP kernel to act as a UP kernel, |
| 4616 | and disable the IO APIC. legacy for "maxcpus=0". |
| 4617 | |
| 4618 | nosmt [KNL,MIPS,PPC,EARLY] Disable symmetric multithreading (SMT). |
| 4619 | Equivalent to smt=1. |
| 4620 | |
| 4621 | [KNL,LOONGARCH,X86,ARM64,PPC,S390] Disable symmetric multithreading (SMT). |
| 4622 | nosmt=force: Force disable SMT, cannot be undone |
| 4623 | via the sysfs control file. |
| 4624 | |
| 4625 | nosoftlockup [KNL] Disable the soft-lockup detector. |
| 4626 | |
| 4627 | nospec_store_bypass_disable |
| 4628 | [HW,EARLY] Disable all mitigations for the Speculative |
| 4629 | Store Bypass vulnerability |
| 4630 | |
| 4631 | nospectre_bhb [ARM64,EARLY] Disable all mitigations for Spectre-BHB (branch |
| 4632 | history injection) vulnerability. System may allow data leaks |
| 4633 | with this option. |
| 4634 | |
| 4635 | nospectre_v1 [X86,PPC,EARLY] Disable mitigations for Spectre Variant 1 |
| 4636 | (bounds check bypass). With this option data leaks are |
| 4637 | possible in the system. |
| 4638 | |
| 4639 | nospectre_v2 [X86,PPC_E500,ARM64,EARLY] Disable all mitigations |
| 4640 | for the Spectre variant 2 (indirect branch |
| 4641 | prediction) vulnerability. System may allow data |
| 4642 | leaks with this option. |
| 4643 | |
| 4644 | no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES,RISCV,LOONGARCH,EARLY] |
| 4645 | Disable paravirtualized steal time accounting. steal time |
| 4646 | is computed, but won't influence scheduler behaviour |
| 4647 | |
| 4648 | nosync [HW,M68K] Disables sync negotiation for all devices. |
| 4649 | |
| 4650 | no_timer_check [X86,APIC] Disables the code which tests for broken |
| 4651 | timer IRQ sources, i.e., the IO-APIC timer. This can |
| 4652 | work around problems with incorrect timer |
| 4653 | initialization on some boards. |
| 4654 | |
| 4655 | no_uaccess_flush |
| 4656 | [PPC,EARLY] Don't flush the L1-D cache after accessing user data. |
| 4657 | |
| 4658 | novmcoredd [KNL,KDUMP] |
| 4659 | Disable device dump. Device dump allows drivers to |
| 4660 | append dump data to vmcore so you can collect driver |
| 4661 | specified debug info. Drivers can append the data |
| 4662 | without any limit and this data is stored in memory, |
| 4663 | so this may cause significant memory stress. Disabling |
| 4664 | device dump can help save memory but the driver debug |
| 4665 | data will be no longer available. This parameter |
| 4666 | is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP |
| 4667 | is set. |
| 4668 | |
| 4669 | no-vmw-sched-clock |
| 4670 | [X86,PV_OPS,EARLY] Disable paravirtualized VMware |
| 4671 | scheduler clock and use the default one. |
| 4672 | |
| 4673 | nowatchdog [KNL] Disable both lockup detectors, i.e. |
| 4674 | soft-lockup and NMI watchdog (hard-lockup). |
| 4675 | |
| 4676 | nowb [ARM,EARLY] |
| 4677 | |
| 4678 | nox2apic [X86-64,APIC,EARLY] Do not enable x2APIC mode. |
| 4679 | |
| 4680 | NOTE: this parameter will be ignored on systems with the |
| 4681 | LEGACY_XAPIC_DISABLED bit set in the |
| 4682 | IA32_XAPIC_DISABLE_STATUS MSR. |
| 4683 | |
| 4684 | noxsave [BUGS=X86] Disables x86 extended register state save |
| 4685 | and restore using xsave. The kernel will fallback to |
| 4686 | enabling legacy floating-point and sse state. |
| 4687 | |
| 4688 | noxsaveopt [X86] Disables xsaveopt used in saving x86 extended |
| 4689 | register states. The kernel will fall back to use |
| 4690 | xsave to save the states. By using this parameter, |
| 4691 | performance of saving the states is degraded because |
| 4692 | xsave doesn't support modified optimization while |
| 4693 | xsaveopt supports it on xsaveopt enabled systems. |
| 4694 | |
| 4695 | noxsaves [X86] Disables xsaves and xrstors used in saving and |
| 4696 | restoring x86 extended register state in compacted |
| 4697 | form of xsave area. The kernel will fall back to use |
| 4698 | xsaveopt and xrstor to save and restore the states |
| 4699 | in standard form of xsave area. By using this |
| 4700 | parameter, xsave area per process might occupy more |
| 4701 | memory on xsaves enabled systems. |
| 4702 | |
| 4703 | nr_cpus= [SMP,EARLY] Maximum number of processors that an SMP kernel |
| 4704 | could support. nr_cpus=n : n >= 1 limits the kernel to |
| 4705 | support 'n' processors. It could be larger than the |
| 4706 | number of already plugged CPU during bootup, later in |
| 4707 | runtime you can physically add extra cpu until it reaches |
| 4708 | n. So during boot up some boot time memory for per-cpu |
| 4709 | variables need be pre-allocated for later physical cpu |
| 4710 | hot plugging. |
| 4711 | |
| 4712 | nr_uarts= [SERIAL] maximum number of UARTs to be registered. |
| 4713 | |
| 4714 | numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86, EARLY] |
| 4715 | Disable NUMA, Only set up a single NUMA node |
| 4716 | spanning all memory. |
| 4717 | |
| 4718 | numa=fake=<size>[MG] |
| 4719 | [KNL, ARM64, RISCV, X86, EARLY] |
| 4720 | If given as a memory unit, fills all system RAM with |
| 4721 | nodes of size interleaved over physical nodes. |
| 4722 | |
| 4723 | numa=fake=<N> |
| 4724 | [KNL, ARM64, RISCV, X86, EARLY] |
| 4725 | If given as an integer, fills all system RAM with N |
| 4726 | fake nodes interleaved over physical nodes. |
| 4727 | |
| 4728 | numa=fake=<N>U |
| 4729 | [KNL, ARM64, RISCV, X86, EARLY] |
| 4730 | If given as an integer followed by 'U', it will |
| 4731 | divide each physical node into N emulated nodes. |
| 4732 | |
| 4733 | numa=noacpi [X86] Don't parse the SRAT table for NUMA setup |
| 4734 | |
| 4735 | numa=nohmat [X86] Don't parse the HMAT table for NUMA setup, or |
| 4736 | soft-reserved memory partitioning. |
| 4737 | |
| 4738 | numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic |
| 4739 | NUMA balancing. |
| 4740 | Allowed values are enable and disable |
| 4741 | |
| 4742 | numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. |
| 4743 | 'node', 'default' can be specified |
| 4744 | This can be set from sysctl after boot. |
| 4745 | See Documentation/admin-guide/sysctl/vm.rst for details. |
| 4746 | |
| 4747 | nvme.quirks= [NVME] A list of quirk entries to augment the built-in |
| 4748 | nvme quirk list. List entries are separated by a |
| 4749 | '-' character. |
| 4750 | Each entry has the form VendorID:ProductID:quirk_names. |
| 4751 | The IDs are 4-digits hex numbers and quirk_names is a |
| 4752 | list of quirk names separated by commas. A quirk name |
| 4753 | can be prefixed by '^', meaning that the specified |
| 4754 | quirk must be disabled. |
| 4755 | |
| 4756 | Example: |
| 4757 | nvme.quirks=7710:2267:bogus_nid,^identify_cns-9900:7711:broken_msi |
| 4758 | |
| 4759 | ohci1394_dma=early [HW,EARLY] enable debugging via the ohci1394 driver. |
| 4760 | See Documentation/core-api/debugging-via-ohci1394.rst for more |
| 4761 | info. |
| 4762 | |
| 4763 | olpc_ec_timeout= [OLPC] ms delay when issuing EC commands |
| 4764 | Rather than timing out after 20 ms if an EC |
| 4765 | command is not properly ACKed, override the length |
| 4766 | of the timeout. We have interrupts disabled while |
| 4767 | waiting for the ACK, so if this is set too high |
| 4768 | interrupts *may* be lost! |
| 4769 | |
| 4770 | onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration |
| 4771 | |
| 4772 | Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] |
| 4773 | |
| 4774 | boundary - index of last SLC block on Flex-OneNAND. |
| 4775 | The remaining blocks are configured as MLC blocks. |
| 4776 | lock - Configure if Flex-OneNAND boundary should be locked. |
| 4777 | Once locked, the boundary cannot be changed. |
| 4778 | 1 indicates lock status, 0 indicates unlock status. |
| 4779 | |
| 4780 | oops=panic [KNL,EARLY] |
| 4781 | Always panic on oopses. Default is to just kill the |
| 4782 | process, but there is a small probability of |
| 4783 | deadlocking the machine. |
| 4784 | This will also cause panics on machine check exceptions. |
| 4785 | Useful together with panic=30 to trigger a reboot. |
| 4786 | |
| 4787 | page_alloc.shuffle= |
| 4788 | [KNL] Boolean flag to control whether the page allocator |
| 4789 | should randomize its free lists. This parameter can be |
| 4790 | used to enable/disable page randomization. The state of |
| 4791 | the flag can be read from sysfs at: |
| 4792 | /sys/module/page_alloc/parameters/shuffle. |
| 4793 | This parameter is only available if CONFIG_SHUFFLE_PAGE_ALLOCATOR=y. |
| 4794 | |
| 4795 | page_owner= [KNL,EARLY] Boot-time page_owner enabling option. |
| 4796 | Storage of the information about who allocated |
| 4797 | each page is disabled in default. With this switch, |
| 4798 | we can turn it on. |
| 4799 | on: enable the feature |
| 4800 | |
| 4801 | page_poison= [KNL,EARLY] Boot-time parameter changing the state of |
| 4802 | poisoning on the buddy allocator, available with |
| 4803 | CONFIG_PAGE_POISONING=y. |
| 4804 | off: turn off poisoning (default) |
| 4805 | on: turn on poisoning |
| 4806 | |
| 4807 | page_reporting.page_reporting_order= |
| 4808 | [KNL] Minimal page reporting order |
| 4809 | Format: <integer> |
| 4810 | Adjust the minimal page reporting order. The page |
| 4811 | reporting is disabled when it exceeds MAX_PAGE_ORDER. |
| 4812 | |
| 4813 | page_reporting.page_reporting_delay_ms= |
| 4814 | [KNL] Free page reporting delay in milliseconds |
| 4815 | Format: <unsigned integer> |
| 4816 | Adjust the delay in milliseconds between free page |
| 4817 | reporting intervals. Default is 2000 (2 seconds). |
| 4818 | |
| 4819 | panic= [KNL] Kernel behaviour on panic: delay <timeout> |
| 4820 | timeout > 0: seconds before rebooting |
| 4821 | timeout = 0: wait forever |
| 4822 | timeout < 0: reboot immediately |
| 4823 | Format: <timeout> |
| 4824 | |
| 4825 | panic_on_taint= [KNL,EARLY] |
| 4826 | Bitmask for conditionally calling panic() in add_taint() |
| 4827 | Format: <hex>[,nousertaint] |
| 4828 | Hexadecimal bitmask representing the set of TAINT flags |
| 4829 | that will cause the kernel to panic when add_taint() is |
| 4830 | called with any of the flags in this set. |
| 4831 | The optional switch "nousertaint" can be utilized to |
| 4832 | prevent userspace forced crashes by writing to sysctl |
| 4833 | /proc/sys/kernel/tainted any flagset matching with the |
| 4834 | bitmask set on panic_on_taint. |
| 4835 | See Documentation/admin-guide/tainted-kernels.rst for |
| 4836 | extra details on the taint flags that users can pick |
| 4837 | to compose the bitmask to assign to panic_on_taint. |
| 4838 | |
| 4839 | panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump |
| 4840 | on a WARN(). |
| 4841 | |
| 4842 | panic_force_cpu= |
| 4843 | [KNL,SMP] Force panic handling to execute on a specific CPU. |
| 4844 | Format: <cpu number> |
| 4845 | Some platforms require panic handling to occur on a |
| 4846 | specific CPU for the crash kernel to function correctly. |
| 4847 | This can be due to firmware limitations, interrupt routing |
| 4848 | constraints, or platform-specific requirements where only |
| 4849 | a particular CPU can safely enter the crash kernel. |
| 4850 | When set, panic() will redirect execution to the specified |
| 4851 | CPU before proceeding with the normal panic and kexec flow. |
| 4852 | If the target CPU is offline or unavailable, panic proceeds |
| 4853 | on the current CPU. |
| 4854 | This option should only be used for systems with the above |
| 4855 | constraints as it might cause the panic operation to be less reliable. |
| 4856 | |
| 4857 | panic_print= Bitmask for printing system info when panic happens. |
| 4858 | User can chose combination of the following bits: |
| 4859 | |
| 4860 | bit 0: print all tasks info |
| 4861 | bit 1: print system memory info |
| 4862 | bit 2: print timer info |
| 4863 | bit 3: print locks info if CONFIG_LOCKDEP is on |
| 4864 | bit 4: print ftrace buffer |
| 4865 | bit 5: replay all kernel messages on consoles at the end of panic |
| 4866 | bit 6: print all CPUs backtrace (if available in the arch) |
| 4867 | bit 7: print only tasks in uninterruptible (blocked) state |
| 4868 | |
| 4869 | *Be aware* that this option may print a _lot_ of lines. |
| 4870 | |
| 4871 | There is an increased risk of losing older messages in |
| 4872 | the log. Maybe worth to setup a bigger log buffer with |
| 4873 | "log_buf_len" along with this. |
| 4874 | |
| 4875 | Also consider using "printk.console_no_auto_verbose=Y" |
| 4876 | if using this along with pstore, to avoid extra delays |
| 4877 | due to increased console verbosity during panic. |
| 4878 | |
| 4879 | panic_sys_info= A comma separated list of extra information to be dumped |
| 4880 | on panic. |
| 4881 | Format: val[,val...] |
| 4882 | Where @val can be any of the following: |
| 4883 | |
| 4884 | tasks: print all tasks info |
| 4885 | mem: print system memory info |
| 4886 | timers: print timers info |
| 4887 | locks: print locks info if CONFIG_LOCKDEP is on |
| 4888 | ftrace: print ftrace buffer |
| 4889 | all_bt: print all CPUs backtrace (if available in the arch) |
| 4890 | blocked_tasks: print only tasks in uninterruptible (blocked) state |
| 4891 | |
| 4892 | This is a human readable alternative to the 'panic_print' |
| 4893 | option above. Please be aware of the caveats using this, |
| 4894 | check the 'panic_print' parameter for more details. |
| 4895 | |
| 4896 | panic_console_replay |
| 4897 | When panic happens, replay all kernel messages on |
| 4898 | consoles at the end of panic. |
| 4899 | |
| 4900 | parkbd.port= [HW] Parallel port number the keyboard adapter is |
| 4901 | connected to, default is 0. |
| 4902 | Format: <parport#> |
| 4903 | parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, |
| 4904 | 0 for XT, 1 for AT (default is AT). |
| 4905 | Format: <mode> |
| 4906 | |
| 4907 | parport= [HW,PPT] Specify parallel ports. 0 disables. |
| 4908 | Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } |
| 4909 | Use 'auto' to force the driver to use any |
| 4910 | IRQ/DMA settings detected (the default is to |
| 4911 | ignore detected IRQ/DMA settings because of |
| 4912 | possible conflicts). You can specify the base |
| 4913 | address, IRQ, and DMA settings; IRQ and DMA |
| 4914 | should be numbers, or 'auto' (for using detected |
| 4915 | settings on that particular port), or 'nofifo' |
| 4916 | (to avoid using a FIFO even if it is detected). |
| 4917 | Parallel ports are assigned in the order they |
| 4918 | are specified on the command line, starting |
| 4919 | with parport0. |
| 4920 | |
| 4921 | parport_init_mode= [HW,PPT] |
| 4922 | Configure VIA parallel port to operate in |
| 4923 | a specific mode. This is necessary on Pegasos |
| 4924 | computer where firmware has no options for setting |
| 4925 | up parallel port mode and sets it to spp. |
| 4926 | Currently this function knows 686a and 8231 chips. |
| 4927 | Format: [spp|ps2|epp|ecp|ecpepp] |
| 4928 | |
| 4929 | pata_legacy.all= [HW,LIBATA] |
| 4930 | Format: <int> |
| 4931 | Set to non-zero to probe primary and secondary ISA |
| 4932 | port ranges on PCI systems where no PCI PATA device |
| 4933 | has been found at either range. Disabled by default. |
| 4934 | |
| 4935 | pata_legacy.autospeed= [HW,LIBATA] |
| 4936 | Format: <int> |
| 4937 | Set to non-zero if a chip is present that snoops speed |
| 4938 | changes. Disabled by default. |
| 4939 | |
| 4940 | pata_legacy.iordy_mask= [HW,LIBATA] |
| 4941 | Format: <int> |
| 4942 | IORDY enable mask. Set individual bits to allow IORDY |
| 4943 | for the respective channel. Bit 0 is for the first |
| 4944 | legacy channel handled by this driver, bit 1 is for |
| 4945 | the second channel, and so on. The sequence will often |
| 4946 | correspond to the primary legacy channel, the secondary |
| 4947 | legacy channel, and so on, but the handling of a PCI |
| 4948 | bus and the use of other driver options may interfere |
| 4949 | with the sequence. By default IORDY is allowed across |
| 4950 | all channels. |
| 4951 | |
| 4952 | pata_legacy.pio_mask= [HW,LIBATA] |
| 4953 | Format: <int> |
| 4954 | PIO mode mask for autospeed devices. Set individual |
| 4955 | bits to allow the use of the respective PIO modes. |
| 4956 | Bit 0 is for mode 0, bit 1 is for mode 1, and so on. |
| 4957 | All modes allowed by default. |
| 4958 | |
| 4959 | pata_legacy.probe_all= [HW,LIBATA] |
| 4960 | Format: <int> |
| 4961 | Set to non-zero to probe tertiary and further ISA |
| 4962 | port ranges on PCI systems. Disabled by default. |
| 4963 | |
| 4964 | pata_legacy.probe_mask= [HW,LIBATA] |
| 4965 | Format: <int> |
| 4966 | Probe mask for legacy ISA PATA ports. Depending on |
| 4967 | platform configuration and the use of other driver |
| 4968 | options up to 6 legacy ports are supported: 0x1f0, |
| 4969 | 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing |
| 4970 | of individual ports can be disabled by setting the |
| 4971 | corresponding bits in the mask to 1. Bit 0 is for |
| 4972 | the first port in the list above (0x1f0), and so on. |
| 4973 | By default all supported ports are probed. |
| 4974 | |
| 4975 | pata_platform.pio_mask= [HW,LIBATA] |
| 4976 | Format: <int> |
| 4977 | Supported PIO mode mask. Set individual bits to allow |
| 4978 | the use of the respective PIO modes. Bit 0 is for |
| 4979 | mode 0, bit 1 is for mode 1, and so on. Mode 0 only |
| 4980 | allowed by default. |
| 4981 | |
| 4982 | pause_on_oops=<int> |
| 4983 | Halt all CPUs after the first oops has been printed for |
| 4984 | the specified number of seconds. This is to be used if |
| 4985 | your oopses keep scrolling off the screen. |
| 4986 | |
| 4987 | pci=option[,option...] [PCI,EARLY] various PCI subsystem options. |
| 4988 | |
| 4989 | Some options herein operate on a specific device |
| 4990 | or a set of devices (<pci_dev>). These are |
| 4991 | specified in one of the following formats: |
| 4992 | |
| 4993 | [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* |
| 4994 | pci:<vendor>:<device>[:<subvendor>:<subdevice>] |
| 4995 | |
| 4996 | Note: the first format specifies a PCI |
| 4997 | bus/device/function address which may change |
| 4998 | if new hardware is inserted, if motherboard |
| 4999 | firmware changes, or due to changes caused |
| 5000 | by other kernel parameters. If the |
| 5001 | domain is left unspecified, it is |
| 5002 | taken to be zero. Optionally, a path |
| 5003 | to a device through multiple device/function |
| 5004 | addresses can be specified after the base |
| 5005 | address (this is more robust against |
| 5006 | renumbering issues). The second format |
| 5007 | selects devices using IDs from the |
| 5008 | configuration space which may match multiple |
| 5009 | devices in the system. |
| 5010 | |
| 5011 | earlydump dump PCI config space before the kernel |
| 5012 | changes anything |
| 5013 | off [X86] don't probe for the PCI bus |
| 5014 | bios [X86-32] force use of PCI BIOS, don't access |
| 5015 | the hardware directly. Use this if your machine |
| 5016 | has a non-standard PCI host bridge. |
| 5017 | nobios [X86-32] disallow use of PCI BIOS, only direct |
| 5018 | hardware access methods are allowed. Use this |
| 5019 | if you experience crashes upon bootup and you |
| 5020 | suspect they are caused by the BIOS. |
| 5021 | conf1 [X86] Force use of PCI Configuration Access |
| 5022 | Mechanism 1 (config address in IO port 0xCF8, |
| 5023 | data in IO port 0xCFC, both 32-bit). |
| 5024 | conf2 [X86] Force use of PCI Configuration Access |
| 5025 | Mechanism 2 (IO port 0xCF8 is an 8-bit port for |
| 5026 | the function, IO port 0xCFA, also 8-bit, sets |
| 5027 | bus number. The config space is then accessed |
| 5028 | through ports 0xC000-0xCFFF). |
| 5029 | See http://wiki.osdev.org/PCI for more info |
| 5030 | on the configuration access mechanisms. |
| 5031 | noaer [PCIE] If the PCIEAER kernel config parameter is |
| 5032 | enabled, this kernel boot option can be used to |
| 5033 | disable the use of PCIE advanced error reporting. |
| 5034 | nodomains [PCI] Disable support for multiple PCI |
| 5035 | root domains (aka PCI segments, in ACPI-speak). |
| 5036 | nommconf [X86] Disable use of MMCONFIG for PCI |
| 5037 | Configuration |
| 5038 | check_enable_amd_mmconf [X86] check for and enable |
| 5039 | properly configured MMIO access to PCI |
| 5040 | config space on AMD family 10h CPU |
| 5041 | nomsi [MSI] If the PCI_MSI kernel config parameter is |
| 5042 | enabled, this kernel boot option can be used to |
| 5043 | disable the use of MSI interrupts system-wide. |
| 5044 | noioapicquirk [APIC] Disable all boot interrupt quirks. |
| 5045 | Safety option to keep boot IRQs enabled. This |
| 5046 | should never be necessary. |
| 5047 | ioapicreroute [APIC] Enable rerouting of boot IRQs to the |
| 5048 | primary IO-APIC for bridges that cannot disable |
| 5049 | boot IRQs. This fixes a source of spurious IRQs |
| 5050 | when the system masks IRQs. |
| 5051 | noioapicreroute [APIC] Disable workaround that uses the |
| 5052 | boot IRQ equivalent of an IRQ that connects to |
| 5053 | a chipset where boot IRQs cannot be disabled. |
| 5054 | The opposite of ioapicreroute. |
| 5055 | biosirq [X86-32] Use PCI BIOS calls to get the interrupt |
| 5056 | routing table. These calls are known to be buggy |
| 5057 | on several machines and they hang the machine |
| 5058 | when used, but on other computers it's the only |
| 5059 | way to get the interrupt routing table. Try |
| 5060 | this option if the kernel is unable to allocate |
| 5061 | IRQs or discover secondary PCI buses on your |
| 5062 | motherboard. |
| 5063 | rom [X86] Assign address space to expansion ROMs. |
| 5064 | Use with caution as certain devices share |
| 5065 | address decoders between ROMs and other |
| 5066 | resources. |
| 5067 | norom [X86] Do not assign address space to |
| 5068 | expansion ROMs that do not already have |
| 5069 | BIOS assigned address ranges. |
| 5070 | nobar [X86] Do not assign address space to the |
| 5071 | BARs that weren't assigned by the BIOS. |
| 5072 | irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be |
| 5073 | assigned automatically to PCI devices. You can |
| 5074 | make the kernel exclude IRQs of your ISA cards |
| 5075 | this way. |
| 5076 | pirqaddr=0xAAAAA [X86] Specify the physical address |
| 5077 | of the PIRQ table (normally generated |
| 5078 | by the BIOS) if it is outside the |
| 5079 | F0000h-100000h range. |
| 5080 | lastbus=N [X86] Scan all buses thru bus #N. Can be |
| 5081 | useful if the kernel is unable to find your |
| 5082 | secondary buses and you want to tell it |
| 5083 | explicitly which ones they are. |
| 5084 | assign-busses [X86] Always assign all PCI bus |
| 5085 | numbers ourselves, overriding |
| 5086 | whatever the firmware may have done. |
| 5087 | usepirqmask [X86] Honor the possible IRQ mask stored |
| 5088 | in the BIOS $PIR table. This is needed on |
| 5089 | some systems with broken BIOSes, notably |
| 5090 | some HP Pavilion N5400 and Omnibook XE3 |
| 5091 | notebooks. This will have no effect if ACPI |
| 5092 | IRQ routing is enabled. |
| 5093 | noacpi [X86] Do not use ACPI for IRQ routing |
| 5094 | or for PCI scanning. |
| 5095 | use_crs [X86] Use PCI host bridge window information |
| 5096 | from ACPI. On BIOSes from 2008 or later, this |
| 5097 | is enabled by default. If you need to use this, |
| 5098 | please report a bug. |
| 5099 | nocrs [X86] Ignore PCI host bridge windows from ACPI. |
| 5100 | If you need to use this, please report a bug. |
| 5101 | use_e820 [X86] Use E820 reservations to exclude parts of |
| 5102 | PCI host bridge windows. This is a workaround |
| 5103 | for BIOS defects in host bridge _CRS methods. |
| 5104 | If you need to use this, please report a bug to |
| 5105 | <[email protected]>. |
| 5106 | no_e820 [X86] Ignore E820 reservations for PCI host |
| 5107 | bridge windows. This is the default on modern |
| 5108 | hardware. If you need to use this, please report |
| 5109 | a bug to <[email protected]>. |
| 5110 | routeirq Do IRQ routing for all PCI devices. |
| 5111 | This is normally done in pci_enable_device(), |
| 5112 | so this option is a temporary workaround |
| 5113 | for broken drivers that don't call it. |
| 5114 | skip_isa_align [X86] do not align io start addr, so can |
| 5115 | handle more pci cards |
| 5116 | noearly [X86] Don't do any early type 1 scanning. |
| 5117 | This might help on some broken boards which |
| 5118 | machine check when some devices' config space |
| 5119 | is read. But various workarounds are disabled |
| 5120 | and some IOMMU drivers will not work. |
| 5121 | bfsort Sort PCI devices into breadth-first order. |
| 5122 | This sorting is done to get a device |
| 5123 | order compatible with older (<= 2.4) kernels. |
| 5124 | nobfsort Don't sort PCI devices into breadth-first order. |
| 5125 | pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) |
| 5126 | tuning and use the BIOS-configured MPS defaults. |
| 5127 | pcie_bus_safe Set every device's MPS to the largest value |
| 5128 | supported by all devices below the root complex. |
| 5129 | pcie_bus_perf Set device MPS to the largest allowable MPS |
| 5130 | based on its parent bus. Also set MRRS (Max |
| 5131 | Read Request Size) to the largest supported |
| 5132 | value (no larger than the MPS that the device |
| 5133 | or bus can support) for best performance. |
| 5134 | pcie_bus_peer2peer Set every device's MPS to 128B, which |
| 5135 | every device is guaranteed to support. This |
| 5136 | configuration allows peer-to-peer DMA between |
| 5137 | any pair of devices, possibly at the cost of |
| 5138 | reduced performance. This also guarantees |
| 5139 | that hot-added devices will work. |
| 5140 | cbiosize=nn[KMG] The fixed amount of bus space which is |
| 5141 | reserved for the CardBus bridge's IO window. |
| 5142 | The default value is 256 bytes. |
| 5143 | cbmemsize=nn[KMG] The fixed amount of bus space which is |
| 5144 | reserved for the CardBus bridge's memory |
| 5145 | window. The default value is 64 megabytes. |
| 5146 | resource_alignment= |
| 5147 | Format: |
| 5148 | [<order of align>@]<pci_dev>[; ...] |
| 5149 | Specifies alignment and device to reassign |
| 5150 | aligned memory resources. How to |
| 5151 | specify the device is described above. |
| 5152 | If <order of align> is not specified, |
| 5153 | PAGE_SIZE is used as alignment. |
| 5154 | A PCI-PCI bridge can be specified if resource |
| 5155 | windows need to be expanded. |
| 5156 | To specify the alignment for several |
| 5157 | instances of a device, the PCI vendor, |
| 5158 | device, subvendor, and subdevice may be |
| 5159 | specified, e.g., 12@pci:8086:9c22:103c:198f |
| 5160 | for 4096-byte alignment. |
| 5161 | ecrc= Enable/disable PCIe ECRC (transaction layer |
| 5162 | end-to-end CRC checking). Only effective if |
| 5163 | OS has native AER control (either granted by |
| 5164 | ACPI _OSC or forced via "pcie_ports=native") |
| 5165 | bios: Use BIOS/firmware settings. This is the |
| 5166 | the default. |
| 5167 | off: Turn ECRC off |
| 5168 | on: Turn ECRC on. |
| 5169 | hpiosize=nn[KMG] The fixed amount of bus space which is |
| 5170 | reserved for hotplug bridge's IO window. |
| 5171 | Default size is 256 bytes. |
| 5172 | hpmmiosize=nn[KMG] The fixed amount of bus space which is |
| 5173 | reserved for hotplug bridge's MMIO window. |
| 5174 | Default size is 2 megabytes. |
| 5175 | hpmmioprefsize=nn[KMG] The fixed amount of bus space which is |
| 5176 | reserved for hotplug bridge's MMIO_PREF window. |
| 5177 | Default size is 2 megabytes. |
| 5178 | hpmemsize=nn[KMG] The fixed amount of bus space which is |
| 5179 | reserved for hotplug bridge's MMIO and |
| 5180 | MMIO_PREF window. |
| 5181 | Default size is 2 megabytes. |
| 5182 | hpbussize=nn The minimum amount of additional bus numbers |
| 5183 | reserved for buses below a hotplug bridge. |
| 5184 | Default is 1. |
| 5185 | realloc= Enable/disable reallocating PCI bridge resources |
| 5186 | if allocations done by BIOS are too small to |
| 5187 | accommodate resources required by all child |
| 5188 | devices. |
| 5189 | off: Turn realloc off |
| 5190 | on: Turn realloc on |
| 5191 | realloc same as realloc=on |
| 5192 | noari do not use PCIe ARI. |
| 5193 | noats [PCIE, Intel-IOMMU, AMD-IOMMU] |
| 5194 | do not use PCIe ATS (and IOMMU device IOTLB). |
| 5195 | pcie_scan_all Scan all possible PCIe devices. Otherwise we |
| 5196 | only look for one device below a PCIe downstream |
| 5197 | port. |
| 5198 | big_root_window Try to add a big 64bit memory window to the PCIe |
| 5199 | root complex on AMD CPUs. Some GFX hardware |
| 5200 | can resize a BAR to allow access to all VRAM. |
| 5201 | Adding the window is slightly risky (it may |
| 5202 | conflict with unreported devices), so this |
| 5203 | taints the kernel. |
| 5204 | disable_acs_redir=<pci_dev>[; ...] |
| 5205 | Specify one or more PCI devices (in the format |
| 5206 | specified above) separated by semicolons. |
| 5207 | Each device specified will have the PCI ACS |
| 5208 | redirect capabilities forced off which will |
| 5209 | allow P2P traffic between devices through |
| 5210 | bridges without forcing it upstream. Note: |
| 5211 | this removes isolation between devices and |
| 5212 | may put more devices in an IOMMU group. |
| 5213 | config_acs= |
| 5214 | Format: |
| 5215 | <ACS flags>@<pci_dev>[; ...] |
| 5216 | Specify one or more PCI devices (in the format |
| 5217 | specified above) optionally prepended with flags |
| 5218 | and separated by semicolons. The respective |
| 5219 | capabilities will be enabled, disabled or |
| 5220 | unchanged based on what is specified in |
| 5221 | flags. |
| 5222 | |
| 5223 | ACS Flags is defined as follows: |
| 5224 | bit-0 : ACS Source Validation |
| 5225 | bit-1 : ACS Translation Blocking |
| 5226 | bit-2 : ACS P2P Request Redirect |
| 5227 | bit-3 : ACS P2P Completion Redirect |
| 5228 | bit-4 : ACS Upstream Forwarding |
| 5229 | bit-5 : ACS P2P Egress Control |
| 5230 | bit-6 : ACS Direct Translated P2P |
| 5231 | Each bit can be marked as: |
| 5232 | '0' – force disabled |
| 5233 | '1' – force enabled |
| 5234 | 'x' – unchanged |
| 5235 | For example, |
| 5236 | pci=config_acs=10x@pci:0:0 |
| 5237 | would configure all devices that support |
| 5238 | ACS to enable P2P Request Redirect, disable |
| 5239 | Translation Blocking, and leave Source |
| 5240 | Validation unchanged from whatever power-up |
| 5241 | or firmware set it to. |
| 5242 | |
| 5243 | Note: this may remove isolation between devices |
| 5244 | and may put more devices in an IOMMU group. |
| 5245 | force_floating [S390] Force usage of floating interrupts. |
| 5246 | nomio [S390] Do not use MIO instructions. |
| 5247 | norid [S390] ignore the RID field and force use of |
| 5248 | one PCI domain per PCI function |
| 5249 | notph [PCIE] If the PCIE_TPH kernel config parameter |
| 5250 | is enabled, this kernel boot option can be used |
| 5251 | to disable PCIe TLP Processing Hints support |
| 5252 | system-wide. |
| 5253 | |
| 5254 | pcie_aspm= [PCIE] Forcibly enable or ignore PCIe Active State Power |
| 5255 | Management. |
| 5256 | off Don't touch ASPM configuration at all. Leave any |
| 5257 | configuration done by firmware unchanged. |
| 5258 | force Enable ASPM even on devices that claim not to support it. |
| 5259 | WARNING: Forcing ASPM on may cause system lockups. |
| 5260 | |
| 5261 | pcie_ports= [PCIE] PCIe port services handling: |
| 5262 | native Use native PCIe services (PME, AER, DPC, PCIe hotplug) |
| 5263 | even if the platform doesn't give the OS permission to |
| 5264 | use them. This may cause conflicts if the platform |
| 5265 | also tries to use these services. |
| 5266 | dpc-native Use native PCIe service for DPC only. May |
| 5267 | cause conflicts if firmware uses AER or DPC. |
| 5268 | compat Disable native PCIe services (PME, AER, DPC, PCIe |
| 5269 | hotplug). |
| 5270 | |
| 5271 | pcie_port_pm= [PCIE] PCIe port power management handling: |
| 5272 | off Disable power management of all PCIe ports |
| 5273 | force Forcibly enable power management of all PCIe ports |
| 5274 | |
| 5275 | pcie_pme= [PCIE,PM] Native PCIe PME signaling options: |
| 5276 | nomsi Do not use MSI for native PCIe PME signaling (this makes |
| 5277 | all PCIe root ports use INTx for all services). |
| 5278 | |
| 5279 | pd_ignore_unused |
| 5280 | [PM] |
| 5281 | Keep all power-domains already enabled by bootloader on, |
| 5282 | even if no driver has claimed them. This is useful |
| 5283 | for debug and development, but should not be |
| 5284 | needed on a platform with proper driver support. |
| 5285 | |
| 5286 | pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at |
| 5287 | boot time. |
| 5288 | Format: { 0 | 1 } |
| 5289 | See arch/parisc/kernel/pdc_chassis.c |
| 5290 | |
| 5291 | percpu_alloc= [MM,EARLY] |
| 5292 | Select which percpu first chunk allocator to use. |
| 5293 | Currently supported values are "embed" and "page". |
| 5294 | Archs may support subset or none of the selections. |
| 5295 | See comments in mm/percpu.c for details on each |
| 5296 | allocator. This parameter is primarily for debugging |
| 5297 | and performance comparison. |
| 5298 | |
| 5299 | pirq= [SMP,APIC] Manual mp-table setup |
| 5300 | See Documentation/arch/x86/i386/IO-APIC.rst. |
| 5301 | |
| 5302 | plip= [PPT,NET] Parallel port network link |
| 5303 | Format: { parport<nr> | timid | 0 } |
| 5304 | See also Documentation/admin-guide/parport.rst. |
| 5305 | |
| 5306 | pmtmr= [X86] Manual setup of pmtmr I/O Port. |
| 5307 | Override pmtimer IOPort with a hex value. |
| 5308 | e.g. pmtmr=0x508 |
| 5309 | |
| 5310 | pmu_override= [PPC] Override the PMU. |
| 5311 | This option takes over the PMU facility, so it is no |
| 5312 | longer usable by perf. Setting this option starts the |
| 5313 | PMU counters by setting MMCR0 to 0 (the FC bit is |
| 5314 | cleared). If a number is given, then MMCR1 is set to |
| 5315 | that number, otherwise (e.g., 'pmu_override=on'), MMCR1 |
| 5316 | remains 0. |
| 5317 | |
| 5318 | pm_async= [PM] |
| 5319 | Format: off |
| 5320 | This parameter sets the initial value of the |
| 5321 | /sys/power/pm_async sysfs knob at boot time. |
| 5322 | If set to "off", disables asynchronous suspend and |
| 5323 | resume of devices during system-wide power transitions. |
| 5324 | This can be useful on platforms where device |
| 5325 | dependencies are not well-defined, or for debugging |
| 5326 | power management issues. Asynchronous operations are |
| 5327 | enabled by default. |
| 5328 | |
| 5329 | |
| 5330 | pm_debug_messages [SUSPEND,KNL] |
| 5331 | Enable suspend/resume debug messages during boot up. |
| 5332 | |
| 5333 | pm.dpm_watchdog_enabled= |
| 5334 | [PM] Enable or disable the DPM watchdog. Requires |
| 5335 | CONFIG_PM_SLEEP and CONFIG_DPM_WATCHDOG enabled. |
| 5336 | Format: <bool> |
| 5337 | Default value is set by CONFIG_DPM_WATCHDOG_ENABLED. |
| 5338 | |
| 5339 | pnp.debug=1 [PNP] |
| 5340 | Enable PNP debug messages (depends on the |
| 5341 | CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time |
| 5342 | via /sys/module/pnp/parameters/debug. We always show |
| 5343 | current resource usage; turning this on also shows |
| 5344 | possible settings and some assignment information. |
| 5345 | |
| 5346 | pnpacpi= [ACPI] |
| 5347 | { off } |
| 5348 | |
| 5349 | pnpbios= [ISAPNP] |
| 5350 | { on | off | curr | res | no-curr | no-res } |
| 5351 | |
| 5352 | pnp_reserve_irq= |
| 5353 | [ISAPNP] Exclude IRQs for the autoconfiguration |
| 5354 | |
| 5355 | pnp_reserve_dma= |
| 5356 | [ISAPNP] Exclude DMAs for the autoconfiguration |
| 5357 | |
| 5358 | pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration |
| 5359 | Ranges are in pairs (I/O port base and size). |
| 5360 | |
| 5361 | pnp_reserve_mem= |
| 5362 | [ISAPNP] Exclude memory regions for the |
| 5363 | autoconfiguration. |
| 5364 | Ranges are in pairs (memory base and size). |
| 5365 | |
| 5366 | ports= [IP_VS_FTP] IPVS ftp helper module |
| 5367 | Default is 21. |
| 5368 | Up to 8 (IP_VS_APP_MAX_PORTS) ports |
| 5369 | may be specified. |
| 5370 | Format: <port>,<port>.... |
| 5371 | |
| 5372 | possible_cpus= [SMP,S390,X86] |
| 5373 | Format: <unsigned int> |
| 5374 | Set the number of possible CPUs, overriding the |
| 5375 | regular discovery mechanisms (such as ACPI/FW, etc). |
| 5376 | |
| 5377 | powersave=off [PPC] This option disables power saving features. |
| 5378 | It specifically disables cpuidle and sets the |
| 5379 | platform machine description specific power_save |
| 5380 | function to NULL. On Idle the CPU just reduces |
| 5381 | execution priority. |
| 5382 | |
| 5383 | ppc_strict_facility_enable |
| 5384 | [PPC,ENABLE] This option catches any kernel floating point, |
| 5385 | Altivec, VSX and SPE outside of regions specifically |
| 5386 | allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). |
| 5387 | There is some performance impact when enabling this. |
| 5388 | |
| 5389 | ppc_tm= [PPC,EARLY] |
| 5390 | Format: {"off"} |
| 5391 | Disable Hardware Transactional Memory |
| 5392 | |
| 5393 | preempt= [KNL] |
| 5394 | Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC |
| 5395 | none - Limited to cond_resched() calls |
| 5396 | voluntary - Limited to cond_resched() and might_sleep() calls |
| 5397 | full - Any section that isn't explicitly preempt disabled |
| 5398 | can be preempted anytime. Tasks will also yield |
| 5399 | contended spinlocks (if the critical section isn't |
| 5400 | explicitly preempt disabled beyond the lock itself). |
| 5401 | lazy - Scheduler controlled. Similar to full but instead |
| 5402 | of preempting the task immediately, the task gets |
| 5403 | one HZ tick time to yield itself before the |
| 5404 | preemption will be forced. One such preemption |
| 5405 | point is when the task returns to user space. |
| 5406 | |
| 5407 | print-fatal-signals= |
| 5408 | [KNL] debug: print fatal signals |
| 5409 | |
| 5410 | If enabled, warn about various signal handling |
| 5411 | related application anomalies: too many signals, |
| 5412 | too many POSIX.1 timers, fatal signals causing a |
| 5413 | coredump - etc. |
| 5414 | |
| 5415 | If you hit the warning due to signal overflow, |
| 5416 | you might want to try "ulimit -i unlimited". |
| 5417 | |
| 5418 | default: off. |
| 5419 | |
| 5420 | printk.always_kmsg_dump= |
| 5421 | Trigger kmsg_dump for cases other than kernel oops or |
| 5422 | panics |
| 5423 | Format: <bool> (1/Y/y=enable, 0/N/n=disable) |
| 5424 | default: disabled |
| 5425 | |
| 5426 | printk.console_no_auto_verbose= |
| 5427 | Disable console loglevel raise on oops, panic |
| 5428 | or lockdep-detected issues (only if lock debug is on). |
| 5429 | With an exception to setups with low baudrate on |
| 5430 | serial console, keeping this 0 is a good choice |
| 5431 | in order to provide more debug information. |
| 5432 | Format: <bool> |
| 5433 | default: 0 (auto_verbose is enabled) |
| 5434 | |
| 5435 | printk.debug_non_panic_cpus= |
| 5436 | Allows storing messages from non-panic CPUs into |
| 5437 | the printk log buffer during panic(). They are |
| 5438 | flushed to consoles by the panic-CPU on |
| 5439 | a best-effort basis. |
| 5440 | Format: <bool> (1/Y/y=enable, 0/N/n=disable) |
| 5441 | Default: disabled |
| 5442 | |
| 5443 | printk.devkmsg={on,off,ratelimit} |
| 5444 | Control writing to /dev/kmsg. |
| 5445 | on - unlimited logging to /dev/kmsg from userspace |
| 5446 | off - logging to /dev/kmsg disabled |
| 5447 | ratelimit - ratelimit the logging |
| 5448 | Default: ratelimit |
| 5449 | |
| 5450 | printk.time= Show timing data prefixed to each printk message line |
| 5451 | Format: <bool> (1/Y/y=enable, 0/N/n=disable) |
| 5452 | |
| 5453 | proc_mem.force_override= [KNL] |
| 5454 | Format: {always | ptrace | never} |
| 5455 | Traditionally /proc/pid/mem allows memory permissions to be |
| 5456 | overridden without restrictions. This option may be set to |
| 5457 | restrict that. Can be one of: |
| 5458 | - 'always': traditional behavior always allows mem overrides. |
| 5459 | - 'ptrace': only allow mem overrides for active ptracers. |
| 5460 | - 'never': never allow mem overrides. |
| 5461 | If not specified, default is the CONFIG_PROC_MEM_* choice. |
| 5462 | |
| 5463 | processor.max_cstate= [HW,ACPI] |
| 5464 | Limit processor to maximum C-state |
| 5465 | max_cstate=9 overrides any DMI blacklist limit. |
| 5466 | |
| 5467 | processor.nocst [HW,ACPI] |
| 5468 | Ignore the _CST method to determine C-states, |
| 5469 | instead using the legacy FADT method |
| 5470 | |
| 5471 | profile= [KNL] Enable kernel profiling via /proc/profile |
| 5472 | Format: [<profiletype>,]<number> |
| 5473 | Param: <profiletype>: "schedule" or "kvm" |
| 5474 | [defaults to kernel profiling] |
| 5475 | Param: "schedule" - profile schedule points. |
| 5476 | Param: "kvm" - profile VM exits. |
| 5477 | Param: <number> - step/bucket size as a power of 2 for |
| 5478 | statistical time based profiling. |
| 5479 | |
| 5480 | prot_virt= [S390] enable hosting protected virtual machines |
| 5481 | isolated from the hypervisor (if hardware supports |
| 5482 | that). If enabled, the default kernel base address |
| 5483 | might be overridden even when Kernel Address Space |
| 5484 | Layout Randomization is disabled. |
| 5485 | Format: <bool> |
| 5486 | |
| 5487 | psi= [KNL] Enable or disable pressure stall information |
| 5488 | tracking. |
| 5489 | Format: <bool> |
| 5490 | |
| 5491 | psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to |
| 5492 | probe for; one of (bare|imps|exps|lifebook|any). |
| 5493 | psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports |
| 5494 | per second. |
| 5495 | psmouse.resetafter= [HW,MOUSE] |
| 5496 | Try to reset the device after so many bad packets |
| 5497 | (0 = never). |
| 5498 | psmouse.resolution= |
| 5499 | [HW,MOUSE] Set desired mouse resolution, in dpi. |
| 5500 | psmouse.smartscroll= |
| 5501 | [HW,MOUSE] Controls Logitech smartscroll autorepeat. |
| 5502 | 0 = disabled, 1 = enabled (default). |
| 5503 | |
| 5504 | pstore.backend= Specify the name of the pstore backend to use |
| 5505 | |
| 5506 | pti= [X86-64] Control Page Table Isolation of user and |
| 5507 | kernel address spaces. Disabling this feature |
| 5508 | removes hardening, but improves performance of |
| 5509 | system calls and interrupts. |
| 5510 | |
| 5511 | on - unconditionally enable |
| 5512 | off - unconditionally disable |
| 5513 | auto - kernel detects whether your CPU model is |
| 5514 | vulnerable to issues that PTI mitigates |
| 5515 | |
| 5516 | Not specifying this option is equivalent to pti=auto. |
| 5517 | |
| 5518 | pty.legacy_count= |
| 5519 | [KNL] Number of legacy pty's. Overwrites compiled-in |
| 5520 | default number. |
| 5521 | |
| 5522 | quiet [KNL,EARLY] Disable most log messages |
| 5523 | |
| 5524 | r128= [HW,DRM] |
| 5525 | |
| 5526 | radix_hcall_invalidate=on [PPC/PSERIES] |
| 5527 | Disable RADIX GTSE feature and use hcall for TLB |
| 5528 | invalidate. |
| 5529 | |
| 5530 | raid= [HW,RAID] |
| 5531 | See Documentation/admin-guide/md.rst. |
| 5532 | |
| 5533 | ramdisk_size= [RAM] Sizes of RAM disks in kilobytes |
| 5534 | See Documentation/admin-guide/blockdev/ramdisk.rst. |
| 5535 | |
| 5536 | ramdisk_start= [Deprecated,RAM] RAM disk image start address |
| 5537 | |
| 5538 | random.trust_cpu=off |
| 5539 | [KNL,EARLY] Disable trusting the use of the CPU's |
| 5540 | random number generator (if available) to |
| 5541 | initialize the kernel's RNG. |
| 5542 | |
| 5543 | random.trust_bootloader=off |
| 5544 | [KNL,EARLY] Disable trusting the use of the a seed |
| 5545 | passed by the bootloader (if available) to |
| 5546 | initialize the kernel's RNG. |
| 5547 | |
| 5548 | randomize_kstack_offset= |
| 5549 | [KNL,EARLY] Enable or disable kernel stack offset |
| 5550 | randomization, which provides roughly 5 bits of |
| 5551 | entropy, frustrating memory corruption attacks |
| 5552 | that depend on stack address determinism or |
| 5553 | cross-syscall address exposures. This is only |
| 5554 | available on architectures that have defined |
| 5555 | CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. |
| 5556 | Format: <bool> (1/Y/y=enable, 0/N/n=disable) |
| 5557 | Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. |
| 5558 | |
| 5559 | ras=option[,option,...] [KNL] RAS-specific options |
| 5560 | |
| 5561 | cec_disable [X86] |
| 5562 | Disable the Correctable Errors Collector, |
| 5563 | see CONFIG_RAS_CEC help text. |
| 5564 | |
| 5565 | rcu_nocbs[=cpu-list] |
| 5566 | [KNL] The optional argument is a cpu list, |
| 5567 | as described above. |
| 5568 | |
| 5569 | In kernels built with CONFIG_RCU_NOCB_CPU=y, |
| 5570 | enable the no-callback CPU mode, which prevents |
| 5571 | such CPUs' callbacks from being invoked in |
| 5572 | softirq context. Invocation of such CPUs' RCU |
| 5573 | callbacks will instead be offloaded to "rcuox/N" |
| 5574 | kthreads created for that purpose, where "x" is |
| 5575 | "p" for RCU-preempt, "s" for RCU-sched, and "g" |
| 5576 | for the kthreads that mediate grace periods; and |
| 5577 | "N" is the CPU number. This reduces OS jitter on |
| 5578 | the offloaded CPUs, which can be useful for HPC |
| 5579 | and real-time workloads. It can also improve |
| 5580 | energy efficiency for asymmetric multiprocessors. |
| 5581 | |
| 5582 | If a cpulist is passed as an argument, the specified |
| 5583 | list of CPUs is set to no-callback mode from boot. |
| 5584 | |
| 5585 | Otherwise, if the '=' sign and the cpulist |
| 5586 | arguments are omitted, no CPU will be set to |
| 5587 | no-callback mode from boot but the mode may be |
| 5588 | toggled at runtime via cpusets. |
| 5589 | |
| 5590 | Note that this argument takes precedence over |
| 5591 | the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. |
| 5592 | |
| 5593 | rcu_nocb_poll [KNL] |
| 5594 | Rather than requiring that offloaded CPUs |
| 5595 | (specified by rcu_nocbs= above) explicitly |
| 5596 | awaken the corresponding "rcuoN" kthreads, |
| 5597 | make these kthreads poll for callbacks. |
| 5598 | This improves the real-time response for the |
| 5599 | offloaded CPUs by relieving them of the need to |
| 5600 | wake up the corresponding kthread, but degrades |
| 5601 | energy efficiency by requiring that the kthreads |
| 5602 | periodically wake up to do the polling. |
| 5603 | |
| 5604 | rcutree.blimit= [KNL] |
| 5605 | Set maximum number of finished RCU callbacks to |
| 5606 | process in one batch. |
| 5607 | |
| 5608 | rcutree.csd_lock_suppress_rcu_stall= [KNL] |
| 5609 | Do only a one-line RCU CPU stall warning when |
| 5610 | there is an ongoing too-long CSD-lock wait. |
| 5611 | |
| 5612 | rcutree.do_rcu_barrier= [KNL] |
| 5613 | Request a call to rcu_barrier(). This is |
| 5614 | throttled so that userspace tests can safely |
| 5615 | hammer on the sysfs variable if they so choose. |
| 5616 | If triggered before the RCU grace-period machinery |
| 5617 | is fully active, this will error out with EAGAIN. |
| 5618 | |
| 5619 | rcutree.dump_tree= [KNL] |
| 5620 | Dump the structure of the rcu_node combining tree |
| 5621 | out at early boot. This is used for diagnostic |
| 5622 | purposes, to verify correct tree setup. |
| 5623 | |
| 5624 | rcutree.gp_cleanup_delay= [KNL] |
| 5625 | Set the number of jiffies to delay each step of |
| 5626 | RCU grace-period cleanup. |
| 5627 | |
| 5628 | rcutree.gp_init_delay= [KNL] |
| 5629 | Set the number of jiffies to delay each step of |
| 5630 | RCU grace-period initialization. |
| 5631 | |
| 5632 | rcutree.gp_preinit_delay= [KNL] |
| 5633 | Set the number of jiffies to delay each step of |
| 5634 | RCU grace-period pre-initialization, that is, |
| 5635 | the propagation of recent CPU-hotplug changes up |
| 5636 | the rcu_node combining tree. |
| 5637 | |
| 5638 | rcutree.jiffies_till_first_fqs= [KNL] |
| 5639 | Set delay from grace-period initialization to |
| 5640 | first attempt to force quiescent states. |
| 5641 | Units are jiffies, minimum value is zero, |
| 5642 | and maximum value is HZ. |
| 5643 | |
| 5644 | rcutree.jiffies_till_next_fqs= [KNL] |
| 5645 | Set delay between subsequent attempts to force |
| 5646 | quiescent states. Units are jiffies, minimum |
| 5647 | value is one, and maximum value is HZ. |
| 5648 | |
| 5649 | rcutree.jiffies_till_sched_qs= [KNL] |
| 5650 | Set required age in jiffies for a |
| 5651 | given grace period before RCU starts |
| 5652 | soliciting quiescent-state help from |
| 5653 | rcu_note_context_switch() and cond_resched(). |
| 5654 | If not specified, the kernel will calculate |
| 5655 | a value based on the most recent settings |
| 5656 | of rcutree.jiffies_till_first_fqs |
| 5657 | and rcutree.jiffies_till_next_fqs. |
| 5658 | This calculated value may be viewed in |
| 5659 | rcutree.jiffies_to_sched_qs. Any attempt to set |
| 5660 | rcutree.jiffies_to_sched_qs will be cheerfully |
| 5661 | overwritten. |
| 5662 | |
| 5663 | rcutree.kthread_prio= [KNL,BOOT] |
| 5664 | Set the SCHED_FIFO priority of the RCU per-CPU |
| 5665 | kthreads (rcuc/N). This value is also used for |
| 5666 | the priority of the RCU boost threads (rcub/N) |
| 5667 | and for the RCU grace-period kthreads (rcu_bh, |
| 5668 | rcu_preempt, and rcu_sched). If RCU_BOOST is |
| 5669 | set, valid values are 1-99 and the default is 1 |
| 5670 | (the least-favored priority). Otherwise, when |
| 5671 | RCU_BOOST is not set, valid values are 0-99 and |
| 5672 | the default is zero (non-realtime operation). |
| 5673 | When RCU_NOCB_CPU is set, also adjust the |
| 5674 | priority of NOCB callback kthreads. |
| 5675 | |
| 5676 | rcutree.nocb_nobypass_lim_per_jiffy= [KNL] |
| 5677 | On callback-offloaded (rcu_nocbs) CPUs, |
| 5678 | RCU reduces the lock contention that would |
| 5679 | otherwise be caused by callback floods through |
| 5680 | use of the ->nocb_bypass list. However, in the |
| 5681 | common non-flooded case, RCU queues directly to |
| 5682 | the main ->cblist in order to avoid the extra |
| 5683 | overhead of the ->nocb_bypass list and its lock. |
| 5684 | But if there are too many callbacks queued during |
| 5685 | a single jiffy, RCU pre-queues the callbacks into |
| 5686 | the ->nocb_bypass queue. The definition of "too |
| 5687 | many" is supplied by this kernel boot parameter. |
| 5688 | |
| 5689 | rcutree.nohz_full_patience_delay= [KNL] |
| 5690 | On callback-offloaded (rcu_nocbs) CPUs, avoid |
| 5691 | disturbing RCU unless the grace period has |
| 5692 | reached the specified age in milliseconds. |
| 5693 | Defaults to zero. Large values will be capped |
| 5694 | at five seconds. All values will be rounded down |
| 5695 | to the nearest value representable by jiffies. |
| 5696 | |
| 5697 | rcutree.qhimark= [KNL] |
| 5698 | Set threshold of queued RCU callbacks beyond which |
| 5699 | batch limiting is disabled. |
| 5700 | |
| 5701 | rcutree.qlowmark= [KNL] |
| 5702 | Set threshold of queued RCU callbacks below which |
| 5703 | batch limiting is re-enabled. |
| 5704 | |
| 5705 | rcutree.qovld= [KNL] |
| 5706 | Set threshold of queued RCU callbacks beyond which |
| 5707 | RCU's force-quiescent-state scan will aggressively |
| 5708 | enlist help from cond_resched() and sched IPIs to |
| 5709 | help CPUs more quickly reach quiescent states. |
| 5710 | Set to less than zero to make this be set based |
| 5711 | on rcutree.qhimark at boot time and to zero to |
| 5712 | disable more aggressive help enlistment. |
| 5713 | |
| 5714 | rcutree.rcu_delay_page_cache_fill_msec= [KNL] |
| 5715 | Set the page-cache refill delay (in milliseconds) |
| 5716 | in response to low-memory conditions. The range |
| 5717 | of permitted values is in the range 0:100000. |
| 5718 | |
| 5719 | rcutree.rcu_divisor= [KNL] |
| 5720 | Set the shift-right count to use to compute |
| 5721 | the callback-invocation batch limit bl from |
| 5722 | the number of callbacks queued on this CPU. |
| 5723 | The result will be bounded below by the value of |
| 5724 | the rcutree.blimit kernel parameter. Every bl |
| 5725 | callbacks, the softirq handler will exit in |
| 5726 | order to allow the CPU to do other work. |
| 5727 | |
| 5728 | Please note that this callback-invocation batch |
| 5729 | limit applies only to non-offloaded callback |
| 5730 | invocation. Offloaded callbacks are instead |
| 5731 | invoked in the context of an rcuoc kthread, which |
| 5732 | scheduler will preempt as it does any other task. |
| 5733 | |
| 5734 | rcutree.rcu_fanout_exact= [KNL] |
| 5735 | Disable autobalancing of the rcu_node combining |
| 5736 | tree. This is used by rcutorture, and might |
| 5737 | possibly be useful for architectures having high |
| 5738 | cache-to-cache transfer latencies. |
| 5739 | |
| 5740 | rcutree.rcu_fanout_leaf= [KNL] |
| 5741 | Change the number of CPUs assigned to each |
| 5742 | leaf rcu_node structure. Useful for very |
| 5743 | large systems, which will choose the value 64, |
| 5744 | and for NUMA systems with large remote-access |
| 5745 | latencies, which will choose a value aligned |
| 5746 | with the appropriate hardware boundaries. |
| 5747 | |
| 5748 | rcutree.rcu_min_cached_objs= [KNL] |
| 5749 | Minimum number of objects which are cached and |
| 5750 | maintained per one CPU. Object size is equal |
| 5751 | to PAGE_SIZE. The cache allows to reduce the |
| 5752 | pressure to page allocator, also it makes the |
| 5753 | whole algorithm to behave better in low memory |
| 5754 | condition. |
| 5755 | |
| 5756 | rcutree.rcu_nocb_gp_stride= [KNL] |
| 5757 | Set the number of NOCB callback kthreads in |
| 5758 | each group, which defaults to the square root |
| 5759 | of the number of CPUs. Larger numbers reduce |
| 5760 | the wakeup overhead on the global grace-period |
| 5761 | kthread, but increases that same overhead on |
| 5762 | each group's NOCB grace-period kthread. |
| 5763 | |
| 5764 | rcutree.rcu_kick_kthreads= [KNL] |
| 5765 | Cause the grace-period kthread to get an extra |
| 5766 | wake_up() if it sleeps three times longer than |
| 5767 | it should at force-quiescent-state time. |
| 5768 | This wake_up() will be accompanied by a |
| 5769 | WARN_ONCE() splat and an ftrace_dump(). |
| 5770 | |
| 5771 | rcutree.rcu_resched_ns= [KNL] |
| 5772 | Limit the time spend invoking a batch of RCU |
| 5773 | callbacks to the specified number of nanoseconds. |
| 5774 | By default, this limit is checked only once |
| 5775 | every 32 callbacks in order to limit the pain |
| 5776 | inflicted by local_clock() overhead. |
| 5777 | |
| 5778 | rcutree.rcu_unlock_delay= [KNL] |
| 5779 | In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, |
| 5780 | this specifies an rcu_read_unlock()-time delay |
| 5781 | in microseconds. This defaults to zero. |
| 5782 | Larger delays increase the probability of |
| 5783 | catching RCU pointer leaks, that is, buggy use |
| 5784 | of RCU-protected pointers after the relevant |
| 5785 | rcu_read_unlock() has completed. |
| 5786 | |
| 5787 | rcutree.sysrq_rcu= [KNL] |
| 5788 | Commandeer a sysrq key to dump out Tree RCU's |
| 5789 | rcu_node tree with an eye towards determining |
| 5790 | why a new grace period has not yet started. |
| 5791 | |
| 5792 | rcutree.use_softirq= [KNL] |
| 5793 | If set to zero, move all RCU_SOFTIRQ processing to |
| 5794 | per-CPU rcuc kthreads. Defaults to a non-zero |
| 5795 | value, meaning that RCU_SOFTIRQ is used by default. |
| 5796 | Specify rcutree.use_softirq=0 to use rcuc kthreads. |
| 5797 | |
| 5798 | But note that CONFIG_PREEMPT_RT=y kernels disable |
| 5799 | this kernel boot parameter, forcibly setting it |
| 5800 | to zero. |
| 5801 | |
| 5802 | rcutree.enable_rcu_lazy= [KNL] |
| 5803 | To save power, batch RCU callbacks and flush after |
| 5804 | delay, memory pressure or callback list growing too |
| 5805 | big. |
| 5806 | |
| 5807 | rcutree.rcu_normal_wake_from_gp= [KNL] |
| 5808 | Reduces a latency of synchronize_rcu() call. This approach |
| 5809 | maintains its own track of synchronize_rcu() callers, so it |
| 5810 | does not interact with regular callbacks because it does not |
| 5811 | use a call_rcu[_hurry]() path. Please note, this is for a |
| 5812 | normal grace period. |
| 5813 | |
| 5814 | How to disable it: |
| 5815 | |
| 5816 | echo 0 > /sys/module/rcutree/parameters/rcu_normal_wake_from_gp |
| 5817 | or pass a boot parameter "rcutree.rcu_normal_wake_from_gp=0" |
| 5818 | |
| 5819 | Default is 1 if it is not explicitly disabled by the boot parameter |
| 5820 | passing 0. |
| 5821 | |
| 5822 | rcuscale.gp_async= [KNL] |
| 5823 | Measure performance of asynchronous |
| 5824 | grace-period primitives such as call_rcu(). |
| 5825 | |
| 5826 | rcuscale.gp_async_max= [KNL] |
| 5827 | Specify the maximum number of outstanding |
| 5828 | callbacks per writer thread. When a writer |
| 5829 | thread exceeds this limit, it invokes the |
| 5830 | corresponding flavor of rcu_barrier() to allow |
| 5831 | previously posted callbacks to drain. |
| 5832 | |
| 5833 | rcuscale.gp_exp= [KNL] |
| 5834 | Measure performance of expedited synchronous |
| 5835 | grace-period primitives. |
| 5836 | |
| 5837 | rcuscale.holdoff= [KNL] |
| 5838 | Set test-start holdoff period. The purpose of |
| 5839 | this parameter is to delay the start of the |
| 5840 | test until boot completes in order to avoid |
| 5841 | interference. |
| 5842 | |
| 5843 | rcuscale.kfree_by_call_rcu= [KNL] |
| 5844 | In kernels built with CONFIG_RCU_LAZY=y, test |
| 5845 | call_rcu() instead of kfree_rcu(). |
| 5846 | |
| 5847 | rcuscale.kfree_mult= [KNL] |
| 5848 | Instead of allocating an object of size kfree_obj, |
| 5849 | allocate one of kfree_mult * sizeof(kfree_obj). |
| 5850 | Defaults to 1. |
| 5851 | |
| 5852 | rcuscale.kfree_rcu_test= [KNL] |
| 5853 | Set to measure performance of kfree_rcu() flooding. |
| 5854 | |
| 5855 | rcuscale.kfree_rcu_test_double= [KNL] |
| 5856 | Test the double-argument variant of kfree_rcu(). |
| 5857 | If this parameter has the same value as |
| 5858 | rcuscale.kfree_rcu_test_single, both the single- |
| 5859 | and double-argument variants are tested. |
| 5860 | |
| 5861 | rcuscale.kfree_rcu_test_single= [KNL] |
| 5862 | Test the single-argument variant of kfree_rcu(). |
| 5863 | If this parameter has the same value as |
| 5864 | rcuscale.kfree_rcu_test_double, both the single- |
| 5865 | and double-argument variants are tested. |
| 5866 | |
| 5867 | rcuscale.kfree_nthreads= [KNL] |
| 5868 | The number of threads running loops of kfree_rcu(). |
| 5869 | |
| 5870 | rcuscale.kfree_alloc_num= [KNL] |
| 5871 | Number of allocations and frees done in an iteration. |
| 5872 | |
| 5873 | rcuscale.kfree_loops= [KNL] |
| 5874 | Number of loops doing rcuscale.kfree_alloc_num number |
| 5875 | of allocations and frees. |
| 5876 | |
| 5877 | rcuscale.minruntime= [KNL] |
| 5878 | Set the minimum test run time in seconds. This |
| 5879 | does not affect the data-collection interval, |
| 5880 | but instead allows better measurement of things |
| 5881 | like CPU consumption. |
| 5882 | |
| 5883 | rcuscale.nreaders= [KNL] |
| 5884 | Set number of RCU readers. The value -1 selects |
| 5885 | N, where N is the number of CPUs. A value |
| 5886 | "n" less than -1 selects N-n+1, where N is again |
| 5887 | the number of CPUs. For example, -2 selects N |
| 5888 | (the number of CPUs), -3 selects N+1, and so on. |
| 5889 | A value of "n" less than or equal to -N selects |
| 5890 | a single reader. |
| 5891 | |
| 5892 | rcuscale.nwriters= [KNL] |
| 5893 | Set number of RCU writers. The values operate |
| 5894 | the same as for rcuscale.nreaders. |
| 5895 | N, where N is the number of CPUs |
| 5896 | |
| 5897 | rcuscale.scale_type= [KNL] |
| 5898 | Specify the RCU implementation to test. |
| 5899 | |
| 5900 | rcuscale.shutdown= [KNL] |
| 5901 | Shut the system down after performance tests |
| 5902 | complete. This is useful for hands-off automated |
| 5903 | testing. |
| 5904 | |
| 5905 | rcuscale.verbose= [KNL] |
| 5906 | Enable additional printk() statements. |
| 5907 | |
| 5908 | rcuscale.writer_holdoff= [KNL] |
| 5909 | Write-side holdoff between grace periods, |
| 5910 | in microseconds. The default of zero says |
| 5911 | no holdoff. |
| 5912 | |
| 5913 | rcuscale.writer_holdoff_jiffies= [KNL] |
| 5914 | Additional write-side holdoff between grace |
| 5915 | periods, but in jiffies. The default of zero |
| 5916 | says no holdoff. |
| 5917 | |
| 5918 | rcutorture.fqs_duration= [KNL] |
| 5919 | Set duration of force_quiescent_state bursts |
| 5920 | in microseconds. |
| 5921 | |
| 5922 | rcutorture.fqs_holdoff= [KNL] |
| 5923 | Set holdoff time within force_quiescent_state bursts |
| 5924 | in microseconds. |
| 5925 | |
| 5926 | rcutorture.fqs_stutter= [KNL] |
| 5927 | Set wait time between force_quiescent_state bursts |
| 5928 | in seconds. |
| 5929 | |
| 5930 | rcutorture.fwd_progress= [KNL] |
| 5931 | Specifies the number of kthreads to be used |
| 5932 | for RCU grace-period forward-progress testing |
| 5933 | for the types of RCU supporting this notion. |
| 5934 | Defaults to 1 kthread, values less than zero or |
| 5935 | greater than the number of CPUs cause the number |
| 5936 | of CPUs to be used. |
| 5937 | |
| 5938 | rcutorture.fwd_progress_div= [KNL] |
| 5939 | Specify the fraction of a CPU-stall-warning |
| 5940 | period to do tight-loop forward-progress testing. |
| 5941 | |
| 5942 | rcutorture.fwd_progress_holdoff= [KNL] |
| 5943 | Number of seconds to wait between successive |
| 5944 | forward-progress tests. |
| 5945 | |
| 5946 | rcutorture.fwd_progress_need_resched= [KNL] |
| 5947 | Enclose cond_resched() calls within checks for |
| 5948 | need_resched() during tight-loop forward-progress |
| 5949 | testing. |
| 5950 | |
| 5951 | rcutorture.gp_cond= [KNL] |
| 5952 | Use conditional/asynchronous update-side |
| 5953 | normal-grace-period primitives, if available. |
| 5954 | |
| 5955 | rcutorture.gp_cond_exp= [KNL] |
| 5956 | Use conditional/asynchronous update-side |
| 5957 | expedited-grace-period primitives, if available. |
| 5958 | |
| 5959 | rcutorture.gp_cond_full= [KNL] |
| 5960 | Use conditional/asynchronous update-side |
| 5961 | normal-grace-period primitives that also take |
| 5962 | concurrent expedited grace periods into account, |
| 5963 | if available. |
| 5964 | |
| 5965 | rcutorture.gp_cond_exp_full= [KNL] |
| 5966 | Use conditional/asynchronous update-side |
| 5967 | expedited-grace-period primitives that also take |
| 5968 | concurrent normal grace periods into account, |
| 5969 | if available. |
| 5970 | |
| 5971 | rcutorture.gp_cond_wi= [KNL] |
| 5972 | Nominal wait interval for normal conditional |
| 5973 | grace periods (specified by rcutorture's |
| 5974 | gp_cond and gp_cond_full module parameters), |
| 5975 | in microseconds. The actual wait interval will |
| 5976 | be randomly selected to nanosecond granularity up |
| 5977 | to this wait interval. Defaults to 16 jiffies, |
| 5978 | for example, 16,000 microseconds on a system |
| 5979 | with HZ=1000. |
| 5980 | |
| 5981 | rcutorture.gp_cond_wi_exp= [KNL] |
| 5982 | Nominal wait interval for expedited conditional |
| 5983 | grace periods (specified by rcutorture's |
| 5984 | gp_cond_exp and gp_cond_exp_full module |
| 5985 | parameters), in microseconds. The actual wait |
| 5986 | interval will be randomly selected to nanosecond |
| 5987 | granularity up to this wait interval. Defaults to |
| 5988 | 128 microseconds. |
| 5989 | |
| 5990 | rcutorture.gp_exp= [KNL] |
| 5991 | Use expedited update-side primitives, if available. |
| 5992 | |
| 5993 | rcutorture.gp_normal= [KNL] |
| 5994 | Use normal (non-expedited) asynchronous |
| 5995 | update-side primitives, if available. |
| 5996 | |
| 5997 | rcutorture.gp_poll= [KNL] |
| 5998 | Use polled update-side normal-grace-period |
| 5999 | primitives, if available. |
| 6000 | |
| 6001 | rcutorture.gp_poll_exp= [KNL] |
| 6002 | Use polled update-side expedited-grace-period |
| 6003 | primitives, if available. |
| 6004 | |
| 6005 | rcutorture.gp_poll_full= [KNL] |
| 6006 | Use polled update-side normal-grace-period |
| 6007 | primitives that also take concurrent expedited |
| 6008 | grace periods into account, if available. |
| 6009 | |
| 6010 | rcutorture.gp_poll_exp_full= [KNL] |
| 6011 | Use polled update-side expedited-grace-period |
| 6012 | primitives that also take concurrent normal |
| 6013 | grace periods into account, if available. |
| 6014 | |
| 6015 | rcutorture.gp_poll_wi= [KNL] |
| 6016 | Nominal wait interval for normal conditional |
| 6017 | grace periods (specified by rcutorture's |
| 6018 | gp_poll and gp_poll_full module parameters), |
| 6019 | in microseconds. The actual wait interval will |
| 6020 | be randomly selected to nanosecond granularity up |
| 6021 | to this wait interval. Defaults to 16 jiffies, |
| 6022 | for example, 16,000 microseconds on a system |
| 6023 | with HZ=1000. |
| 6024 | |
| 6025 | rcutorture.gp_poll_wi_exp= [KNL] |
| 6026 | Nominal wait interval for expedited conditional |
| 6027 | grace periods (specified by rcutorture's |
| 6028 | gp_poll_exp and gp_poll_exp_full module |
| 6029 | parameters), in microseconds. The actual wait |
| 6030 | interval will be randomly selected to nanosecond |
| 6031 | granularity up to this wait interval. Defaults to |
| 6032 | 128 microseconds. |
| 6033 | |
| 6034 | rcutorture.gp_sync= [KNL] |
| 6035 | Use normal (non-expedited) synchronous |
| 6036 | update-side primitives, if available. If all |
| 6037 | of rcutorture.gp_cond=, rcutorture.gp_exp=, |
| 6038 | rcutorture.gp_normal=, and rcutorture.gp_sync= |
| 6039 | are zero, rcutorture acts as if is interpreted |
| 6040 | they are all non-zero. |
| 6041 | |
| 6042 | rcutorture.gpwrap_lag= [KNL] |
| 6043 | Enable grace-period wrap lag testing. Setting |
| 6044 | to false prevents the gpwrap lag test from |
| 6045 | running. Default is true. |
| 6046 | |
| 6047 | rcutorture.gpwrap_lag_gps= [KNL] |
| 6048 | Set the value for grace-period wrap lag during |
| 6049 | active lag testing periods. This controls how many |
| 6050 | grace periods differences we tolerate between |
| 6051 | rdp and rnp's gp_seq before setting overflow flag. |
| 6052 | The default is always set to 8. |
| 6053 | |
| 6054 | rcutorture.gpwrap_lag_cycle_mins= [KNL] |
| 6055 | Set the total cycle duration for gpwrap lag |
| 6056 | testing in minutes. This is the total time for |
| 6057 | one complete cycle of active and inactive |
| 6058 | testing periods. Default is 30 minutes. |
| 6059 | |
| 6060 | rcutorture.gpwrap_lag_active_mins= [KNL] |
| 6061 | Set the duration for which gpwrap lag is active |
| 6062 | within each cycle, in minutes. During this time, |
| 6063 | the grace-period wrap lag will be set to the |
| 6064 | value specified by gpwrap_lag_gps. Default is |
| 6065 | 5 minutes. |
| 6066 | |
| 6067 | rcutorture.irqreader= [KNL] |
| 6068 | Run RCU readers from irq handlers, or, more |
| 6069 | accurately, from a timer handler. Not all RCU |
| 6070 | flavors take kindly to this sort of thing. |
| 6071 | |
| 6072 | rcutorture.leakpointer= [KNL] |
| 6073 | Leak an RCU-protected pointer out of the reader. |
| 6074 | This can of course result in splats, and is |
| 6075 | intended to test the ability of things like |
| 6076 | CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect |
| 6077 | such leaks. |
| 6078 | |
| 6079 | rcutorture.n_barrier_cbs= [KNL] |
| 6080 | Set callbacks/threads for rcu_barrier() testing. |
| 6081 | |
| 6082 | rcutorture.nfakewriters= [KNL] |
| 6083 | Set number of concurrent RCU writers. These just |
| 6084 | stress RCU, they don't participate in the actual |
| 6085 | test, hence the "fake". |
| 6086 | |
| 6087 | rcutorture.nocbs_nthreads= [KNL] |
| 6088 | Set number of RCU callback-offload togglers. |
| 6089 | Zero (the default) disables toggling. |
| 6090 | |
| 6091 | rcutorture.nocbs_toggle= [KNL] |
| 6092 | Set the delay in milliseconds between successive |
| 6093 | callback-offload toggling attempts. |
| 6094 | |
| 6095 | rcutorture.nreaders= [KNL] |
| 6096 | Set number of RCU readers. The value -1 selects |
| 6097 | N-1, where N is the number of CPUs. A value |
| 6098 | "n" less than -1 selects N-n-2, where N is again |
| 6099 | the number of CPUs. For example, -2 selects N |
| 6100 | (the number of CPUs), -3 selects N+1, and so on. |
| 6101 | |
| 6102 | rcutorture.nwriters= [KNL] |
| 6103 | Set number of RCU writers, which must be either |
| 6104 | zero or one. For additional writers, use instead |
| 6105 | the rcutorture.nfakewriters parameter. |
| 6106 | |
| 6107 | rcutorture.object_debug= [KNL] |
| 6108 | Enable debug-object double-call_rcu() testing. |
| 6109 | |
| 6110 | rcutorture.onoff_holdoff= [KNL] |
| 6111 | Set time (s) after boot for CPU-hotplug testing. |
| 6112 | |
| 6113 | rcutorture.onoff_interval= [KNL] |
| 6114 | Set time (jiffies) between CPU-hotplug operations, |
| 6115 | or zero to disable CPU-hotplug testing. |
| 6116 | |
| 6117 | rcutorture.preempt_duration= [KNL] |
| 6118 | Set duration (in milliseconds) of preemptions |
| 6119 | by a high-priority FIFO real-time task. Set to |
| 6120 | zero (the default) to disable. The CPUs to |
| 6121 | preempt are selected randomly from the set that |
| 6122 | are online at a given point in time. Races with |
| 6123 | CPUs going offline are ignored, with that attempt |
| 6124 | at preemption skipped. |
| 6125 | |
| 6126 | rcutorture.preempt_interval= [KNL] |
| 6127 | Set interval (in milliseconds, defaulting to one |
| 6128 | second) between preemptions by a high-priority |
| 6129 | FIFO real-time task. This delay is mediated |
| 6130 | by an hrtimer and is further fuzzed to avoid |
| 6131 | inadvertent synchronizations. |
| 6132 | |
| 6133 | rcutorture.read_exit_burst= [KNL] |
| 6134 | The number of times in a given read-then-exit |
| 6135 | episode that a set of read-then-exit kthreads |
| 6136 | is spawned. |
| 6137 | |
| 6138 | rcutorture.read_exit_delay= [KNL] |
| 6139 | The delay, in seconds, between successive |
| 6140 | read-then-exit testing episodes. |
| 6141 | |
| 6142 | rcutorture.reader_flavor= [KNL] |
| 6143 | A bit mask indicating which readers to use. |
| 6144 | If there is more than one bit set, the readers |
| 6145 | are entered from low-order bit up, and are |
| 6146 | exited in the opposite order. For SRCU, the |
| 6147 | 0x1 bit is normal readers, 0x2 NMI-safe readers, |
| 6148 | and 0x4 light-weight readers. |
| 6149 | |
| 6150 | rcutorture.shuffle_interval= [KNL] |
| 6151 | Set task-shuffle interval (s). Shuffling tasks |
| 6152 | allows some CPUs to go into dyntick-idle mode |
| 6153 | during the rcutorture test. |
| 6154 | |
| 6155 | rcutorture.shutdown_secs= [KNL] |
| 6156 | Set time (s) after boot system shutdown. This |
| 6157 | is useful for hands-off automated testing. |
| 6158 | |
| 6159 | rcutorture.stall_cpu= [KNL] |
| 6160 | Duration of CPU stall (s) to test RCU CPU stall |
| 6161 | warnings, zero to disable. |
| 6162 | |
| 6163 | rcutorture.stall_cpu_block= [KNL] |
| 6164 | Sleep while stalling if set. This will result |
| 6165 | in warnings from preemptible RCU in addition to |
| 6166 | any other stall-related activity. Note that |
| 6167 | in kernels built with CONFIG_PREEMPTION=n and |
| 6168 | CONFIG_PREEMPT_COUNT=y, this parameter will |
| 6169 | cause the CPU to pass through a quiescent state. |
| 6170 | Given CONFIG_PREEMPTION=n, this will suppress |
| 6171 | RCU CPU stall warnings, but will instead result |
| 6172 | in scheduling-while-atomic splats. |
| 6173 | |
| 6174 | Use of this module parameter results in splats. |
| 6175 | |
| 6176 | |
| 6177 | rcutorture.stall_cpu_holdoff= [KNL] |
| 6178 | Time to wait (s) after boot before inducing stall. |
| 6179 | |
| 6180 | rcutorture.stall_cpu_irqsoff= [KNL] |
| 6181 | Disable interrupts while stalling if set, but only |
| 6182 | on the first stall in the set. |
| 6183 | |
| 6184 | rcutorture.stall_cpu_repeat= [KNL] |
| 6185 | Number of times to repeat the stall sequence, |
| 6186 | so that rcutorture.stall_cpu_repeat=3 will result |
| 6187 | in four stall sequences. |
| 6188 | |
| 6189 | rcutorture.stall_gp_kthread= [KNL] |
| 6190 | Duration (s) of forced sleep within RCU |
| 6191 | grace-period kthread to test RCU CPU stall |
| 6192 | warnings, zero to disable. If both stall_cpu |
| 6193 | and stall_gp_kthread are specified, the |
| 6194 | kthread is starved first, then the CPU. |
| 6195 | |
| 6196 | rcutorture.stall_only= [KNL] |
| 6197 | Shut off all rcutorture kthreads other than the |
| 6198 | RCU CPU stall-warning test kthreads. The purpose |
| 6199 | of this is to test production applictions' |
| 6200 | reactions to CPU stalls, and with minimal |
| 6201 | additional overhead. Or you can omit the |
| 6202 | stall-warning tests as well and get a heavy |
| 6203 | no-op, your choice! |
| 6204 | |
| 6205 | rcutorture.stat_interval= [KNL] |
| 6206 | Time (s) between statistics printk()s. |
| 6207 | |
| 6208 | rcutorture.stutter= [KNL] |
| 6209 | Time (s) to stutter testing, for example, specifying |
| 6210 | five seconds causes the test to run for five seconds, |
| 6211 | wait for five seconds, and so on. This tests RCU's |
| 6212 | ability to transition abruptly to and from idle. |
| 6213 | |
| 6214 | rcutorture.test_boost= [KNL] |
| 6215 | Test RCU priority boosting? 0=no, 1=maybe, 2=yes. |
| 6216 | "Maybe" means test if the RCU implementation |
| 6217 | under test support RCU priority boosting. |
| 6218 | |
| 6219 | rcutorture.test_boost_duration= [KNL] |
| 6220 | Duration (s) of each individual boost test. |
| 6221 | |
| 6222 | rcutorture.test_boost_holdoff= [KNL] |
| 6223 | Holdoff time (s) from start of test to the start |
| 6224 | of RCU priority-boost testing. Defaults to zero, |
| 6225 | that is, no holdoff. |
| 6226 | |
| 6227 | rcutorture.test_boost_interval= [KNL] |
| 6228 | Interval (s) between each boost test. |
| 6229 | |
| 6230 | rcutorture.test_no_idle_hz= [KNL] |
| 6231 | Test RCU's dyntick-idle handling. See also the |
| 6232 | rcutorture.shuffle_interval parameter. |
| 6233 | |
| 6234 | rcutorture.torture_type= [KNL] |
| 6235 | Specify the RCU implementation to test. |
| 6236 | |
| 6237 | rcutorture.verbose= [KNL] |
| 6238 | Enable additional printk() statements. |
| 6239 | |
| 6240 | rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] |
| 6241 | Dump ftrace buffer after reporting RCU CPU |
| 6242 | stall warning. |
| 6243 | |
| 6244 | rcupdate.rcu_cpu_stall_notifiers= [KNL] |
| 6245 | Provide RCU CPU stall notifiers, but see the |
| 6246 | warnings in the RCU_CPU_STALL_NOTIFIER Kconfig |
| 6247 | option's help text. TL;DR: You almost certainly |
| 6248 | do not want rcupdate.rcu_cpu_stall_notifiers. |
| 6249 | |
| 6250 | rcupdate.rcu_cpu_stall_suppress= [KNL] |
| 6251 | Suppress RCU CPU stall warning messages. |
| 6252 | |
| 6253 | rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] |
| 6254 | Suppress RCU CPU stall warning messages and |
| 6255 | rcutorture writer stall warnings that occur |
| 6256 | during early boot, that is, during the time |
| 6257 | before the init task is spawned. |
| 6258 | |
| 6259 | rcupdate.rcu_cpu_stall_timeout= [KNL] |
| 6260 | Set timeout for RCU CPU stall warning messages. |
| 6261 | The value is in seconds and the maximum allowed |
| 6262 | value is 300 seconds. |
| 6263 | |
| 6264 | rcupdate.rcu_exp_cpu_stall_timeout= [KNL] |
| 6265 | Set timeout for expedited RCU CPU stall warning |
| 6266 | messages. The value is in milliseconds |
| 6267 | and the maximum allowed value is 21000 |
| 6268 | milliseconds. Please note that this value is |
| 6269 | adjusted to an arch timer tick resolution. |
| 6270 | Setting this to zero causes the value from |
| 6271 | rcupdate.rcu_cpu_stall_timeout to be used (after |
| 6272 | conversion from seconds to milliseconds). |
| 6273 | |
| 6274 | rcupdate.rcu_cpu_stall_cputime= [KNL] |
| 6275 | Provide statistics on the cputime and count of |
| 6276 | interrupts and tasks during the sampling period. For |
| 6277 | multiple continuous RCU stalls, all sampling periods |
| 6278 | begin at half of the first RCU stall timeout. |
| 6279 | |
| 6280 | rcupdate.rcu_exp_stall_task_details= [KNL] |
| 6281 | Print stack dumps of any tasks blocking the |
| 6282 | current expedited RCU grace period during an |
| 6283 | expedited RCU CPU stall warning. |
| 6284 | |
| 6285 | rcupdate.rcu_expedited= [KNL] |
| 6286 | Use expedited grace-period primitives, for |
| 6287 | example, synchronize_rcu_expedited() instead |
| 6288 | of synchronize_rcu(). This reduces latency, |
| 6289 | but can increase CPU utilization, degrade |
| 6290 | real-time latency, and degrade energy efficiency. |
| 6291 | No effect on CONFIG_TINY_RCU kernels. |
| 6292 | |
| 6293 | rcupdate.rcu_normal= [KNL] |
| 6294 | Use only normal grace-period primitives, |
| 6295 | for example, synchronize_rcu() instead of |
| 6296 | synchronize_rcu_expedited(). This improves |
| 6297 | real-time latency, CPU utilization, and |
| 6298 | energy efficiency, but can expose users to |
| 6299 | increased grace-period latency. This parameter |
| 6300 | overrides rcupdate.rcu_expedited. No effect on |
| 6301 | CONFIG_TINY_RCU kernels. |
| 6302 | |
| 6303 | rcupdate.rcu_normal_after_boot= [KNL] |
| 6304 | Once boot has completed (that is, after |
| 6305 | rcu_end_inkernel_boot() has been invoked), use |
| 6306 | only normal grace-period primitives. No effect |
| 6307 | on CONFIG_TINY_RCU kernels. |
| 6308 | |
| 6309 | But note that CONFIG_PREEMPT_RT=y kernels enables |
| 6310 | this kernel boot parameter, forcibly setting |
| 6311 | it to the value one, that is, converting any |
| 6312 | post-boot attempt at an expedited RCU grace |
| 6313 | period to instead use normal non-expedited |
| 6314 | grace-period processing. |
| 6315 | |
| 6316 | rcupdate.rcu_task_collapse_lim= [KNL] |
| 6317 | Set the maximum number of callbacks present |
| 6318 | at the beginning of a grace period that allows |
| 6319 | the RCU Tasks flavors to collapse back to using |
| 6320 | a single callback queue. This switching only |
| 6321 | occurs when rcupdate.rcu_task_enqueue_lim is |
| 6322 | set to the default value of -1. |
| 6323 | |
| 6324 | rcupdate.rcu_task_contend_lim= [KNL] |
| 6325 | Set the minimum number of callback-queuing-time |
| 6326 | lock-contention events per jiffy required to |
| 6327 | cause the RCU Tasks flavors to switch to per-CPU |
| 6328 | callback queuing. This switching only occurs |
| 6329 | when rcupdate.rcu_task_enqueue_lim is set to |
| 6330 | the default value of -1. |
| 6331 | |
| 6332 | rcupdate.rcu_task_enqueue_lim= [KNL] |
| 6333 | Set the number of callback queues to use for the |
| 6334 | RCU Tasks family of RCU flavors. The default |
| 6335 | of -1 allows this to be automatically (and |
| 6336 | dynamically) adjusted. This parameter is intended |
| 6337 | for use in testing. |
| 6338 | |
| 6339 | rcupdate.rcu_task_lazy_lim= [KNL] |
| 6340 | Number of callbacks on a given CPU that will |
| 6341 | cancel laziness on that CPU. Use -1 to disable |
| 6342 | cancellation of laziness, but be advised that |
| 6343 | doing so increases the danger of OOM due to |
| 6344 | callback flooding. |
| 6345 | |
| 6346 | rcupdate.rcu_task_stall_info= [KNL] |
| 6347 | Set initial timeout in jiffies for RCU task stall |
| 6348 | informational messages, which give some indication |
| 6349 | of the problem for those not patient enough to |
| 6350 | wait for ten minutes. Informational messages are |
| 6351 | only printed prior to the stall-warning message |
| 6352 | for a given grace period. Disable with a value |
| 6353 | less than or equal to zero. Defaults to ten |
| 6354 | seconds. A change in value does not take effect |
| 6355 | until the beginning of the next grace period. |
| 6356 | |
| 6357 | rcupdate.rcu_task_stall_info_mult= [KNL] |
| 6358 | Multiplier for time interval between successive |
| 6359 | RCU task stall informational messages for a given |
| 6360 | RCU tasks grace period. This value is clamped |
| 6361 | to one through ten, inclusive. It defaults to |
| 6362 | the value three, so that the first informational |
| 6363 | message is printed 10 seconds into the grace |
| 6364 | period, the second at 40 seconds, the third at |
| 6365 | 160 seconds, and then the stall warning at 600 |
| 6366 | seconds would prevent a fourth at 640 seconds. |
| 6367 | |
| 6368 | rcupdate.rcu_task_stall_timeout= [KNL] |
| 6369 | Set timeout in jiffies for RCU task stall |
| 6370 | warning messages. Disable with a value less |
| 6371 | than or equal to zero. Defaults to ten minutes. |
| 6372 | A change in value does not take effect until |
| 6373 | the beginning of the next grace period. |
| 6374 | |
| 6375 | rcupdate.rcu_tasks_lazy_ms= [KNL] |
| 6376 | Set timeout in milliseconds RCU Tasks asynchronous |
| 6377 | callback batching for call_rcu_tasks(). |
| 6378 | A negative value will take the default. A value |
| 6379 | of zero will disable batching. Batching is |
| 6380 | always disabled for synchronize_rcu_tasks(). |
| 6381 | |
| 6382 | rcupdate.rcu_self_test= [KNL] |
| 6383 | Run the RCU early boot self tests |
| 6384 | |
| 6385 | rdinit= [KNL] |
| 6386 | Format: <full_path> |
| 6387 | Run specified binary instead of /init from the ramdisk, |
| 6388 | used for early userspace startup. See initrd. |
| 6389 | |
| 6390 | rdrand= [X86,EARLY] |
| 6391 | force - Override the decision by the kernel to hide the |
| 6392 | advertisement of RDRAND support (this affects |
| 6393 | certain AMD processors because of buggy BIOS |
| 6394 | support, specifically around the suspend/resume |
| 6395 | path). |
| 6396 | |
| 6397 | rdt= [HW,X86,RDT] |
| 6398 | Turn on/off individual RDT features. List is: |
| 6399 | cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, |
| 6400 | mba, smba, bmec, abmc, sdciae, energy[:guid], |
| 6401 | perf[:guid]. |
| 6402 | E.g. to turn on cmt and turn off mba use: |
| 6403 | rdt=cmt,!mba |
| 6404 | To turn off all energy telemetry monitoring and ensure that |
| 6405 | perf telemetry monitoring associated with guid 0x12345 |
| 6406 | is enabled use: |
| 6407 | rdt=!energy,perf:0x12345 |
| 6408 | |
| 6409 | reboot= [KNL] |
| 6410 | Format (x86 or x86_64): |
| 6411 | [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ |
| 6412 | [[,]s[mp]####] \ |
| 6413 | [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ |
| 6414 | [[,]f[orce]] |
| 6415 | Where reboot_mode is one of warm (soft) or cold (hard) or gpio |
| 6416 | (prefix with 'panic_' to set mode for panic |
| 6417 | reboot only), |
| 6418 | reboot_type is one of bios, acpi, kbd, triple, efi, or pci, |
| 6419 | reboot_force is either force or not specified, |
| 6420 | reboot_cpu is s[mp]#### with #### being the processor |
| 6421 | to be used for rebooting. |
| 6422 | |
| 6423 | acpi |
| 6424 | Use the ACPI RESET_REG in the FADT. If ACPI is not |
| 6425 | configured or the ACPI reset does not work, the reboot |
| 6426 | path attempts the reset using the keyboard controller. |
| 6427 | |
| 6428 | bios |
| 6429 | Use the CPU reboot vector for warm reset |
| 6430 | |
| 6431 | cold |
| 6432 | Set the cold reboot flag |
| 6433 | |
| 6434 | default |
| 6435 | There are some built-in platform specific "quirks" |
| 6436 | - you may see: "reboot: <name> series board detected. |
| 6437 | Selecting <type> for reboots." In the case where you |
| 6438 | think the quirk is in error (e.g. you have newer BIOS, |
| 6439 | or newer board) using this option will ignore the |
| 6440 | built-in quirk table, and use the generic default |
| 6441 | reboot actions. |
| 6442 | |
| 6443 | efi |
| 6444 | Use efi reset_system runtime service. If EFI is not |
| 6445 | configured or the EFI reset does not work, the reboot |
| 6446 | path attempts the reset using the keyboard controller. |
| 6447 | |
| 6448 | force |
| 6449 | Don't stop other CPUs on reboot. This can make reboot |
| 6450 | more reliable in some cases. |
| 6451 | |
| 6452 | kbd |
| 6453 | Use the keyboard controller. cold reset (default) |
| 6454 | |
| 6455 | pci |
| 6456 | Use a write to the PCI config space register 0xcf9 to |
| 6457 | trigger reboot. |
| 6458 | |
| 6459 | triple |
| 6460 | Force a triple fault (init) |
| 6461 | |
| 6462 | warm |
| 6463 | Don't set the cold reboot flag |
| 6464 | |
| 6465 | Using warm reset will be much faster especially on big |
| 6466 | memory systems because the BIOS will not go through |
| 6467 | the memory check. Disadvantage is that not all |
| 6468 | hardware will be completely reinitialized on reboot so |
| 6469 | there may be boot problems on some systems. |
| 6470 | |
| 6471 | |
| 6472 | refscale.holdoff= [KNL] |
| 6473 | Set test-start holdoff period. The purpose of |
| 6474 | this parameter is to delay the start of the |
| 6475 | test until boot completes in order to avoid |
| 6476 | interference. |
| 6477 | |
| 6478 | refscale.lookup_instances= [KNL] |
| 6479 | Number of data elements to use for the forms of |
| 6480 | SLAB_TYPESAFE_BY_RCU testing. A negative number |
| 6481 | is negated and multiplied by nr_cpu_ids, while |
| 6482 | zero specifies nr_cpu_ids. |
| 6483 | |
| 6484 | refscale.loops= [KNL] |
| 6485 | Set the number of loops over the synchronization |
| 6486 | primitive under test. Increasing this number |
| 6487 | reduces noise due to loop start/end overhead, |
| 6488 | but the default has already reduced the per-pass |
| 6489 | noise to a handful of picoseconds on ca. 2020 |
| 6490 | x86 laptops. |
| 6491 | |
| 6492 | refscale.nreaders= [KNL] |
| 6493 | Set number of readers. The default value of -1 |
| 6494 | selects N, where N is roughly 75% of the number |
| 6495 | of CPUs. A value of zero is an interesting choice. |
| 6496 | |
| 6497 | refscale.nruns= [KNL] |
| 6498 | Set number of runs, each of which is dumped onto |
| 6499 | the console log. |
| 6500 | |
| 6501 | refscale.readdelay= [KNL] |
| 6502 | Set the read-side critical-section duration, |
| 6503 | measured in microseconds. |
| 6504 | |
| 6505 | refscale.scale_type= [KNL] |
| 6506 | Specify the read-protection implementation to test. |
| 6507 | |
| 6508 | refscale.shutdown= [KNL] |
| 6509 | Shut down the system at the end of the performance |
| 6510 | test. This defaults to 1 (shut it down) when |
| 6511 | refscale is built into the kernel and to 0 (leave |
| 6512 | it running) when refscale is built as a module. |
| 6513 | |
| 6514 | refscale.verbose= [KNL] |
| 6515 | Enable additional printk() statements. |
| 6516 | |
| 6517 | refscale.verbose_batched= [KNL] |
| 6518 | Batch the additional printk() statements. If zero |
| 6519 | (the default) or negative, print everything. Otherwise, |
| 6520 | print every Nth verbose statement, where N is the value |
| 6521 | specified. |
| 6522 | |
| 6523 | regulator_ignore_unused |
| 6524 | [REGULATOR] |
| 6525 | Prevents regulator framework from disabling regulators |
| 6526 | that are unused, due no driver claiming them. This may |
| 6527 | be useful for debug and development, but should not be |
| 6528 | needed on a platform with proper driver support. |
| 6529 | |
| 6530 | relax_domain_level= |
| 6531 | [KNL, SMP] Set scheduler's default relax_domain_level. |
| 6532 | See Documentation/admin-guide/cgroup-v1/cpusets.rst. |
| 6533 | |
| 6534 | reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory |
| 6535 | Format: <base1>,<size1>[,<base2>,<size2>,...] |
| 6536 | Reserve I/O ports or memory so the kernel won't use |
| 6537 | them. If <base> is less than 0x10000, the region |
| 6538 | is assumed to be I/O ports; otherwise it is memory. |
| 6539 | |
| 6540 | reserve_mem= [MM] |
| 6541 | Format: nn[KMG]:<align>:<label> |
| 6542 | Reserve physical memory and label it with a name that |
| 6543 | other subsystems can use to access it. This is typically |
| 6544 | used for systems that do not wipe the RAM, and this command |
| 6545 | line will try to reserve the same physical memory on |
| 6546 | soft reboots. Note, it is not guaranteed to be the same |
| 6547 | location. For example, if anything about the system changes |
| 6548 | or if booting a different kernel. It can also fail if KASLR |
| 6549 | places the kernel at the location of where the RAM reservation |
| 6550 | was from a previous boot, the new reservation will be at a |
| 6551 | different location. |
| 6552 | Any subsystem using this feature must add a way to verify |
| 6553 | that the contents of the physical memory is from a previous |
| 6554 | boot, as there may be cases where the memory will not be |
| 6555 | located at the same location. |
| 6556 | |
| 6557 | The format is size:align:label for example, to request |
| 6558 | 12 megabytes of 4096 alignment for ramoops: |
| 6559 | |
| 6560 | reserve_mem=12M:4096:oops ramoops.mem_name=oops |
| 6561 | |
| 6562 | reservetop= [X86-32,EARLY] |
| 6563 | Format: nn[KMG] |
| 6564 | Reserves a hole at the top of the kernel virtual |
| 6565 | address space. |
| 6566 | |
| 6567 | reset_devices [KNL] Force drivers to reset the underlying device |
| 6568 | during initialization. |
| 6569 | |
| 6570 | resume= [SWSUSP] |
| 6571 | Specify the partition device for software suspend |
| 6572 | Format: |
| 6573 | {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} |
| 6574 | |
| 6575 | resume_offset= [SWSUSP] |
| 6576 | Specify the offset from the beginning of the partition |
| 6577 | given by "resume=" at which the swap header is located, |
| 6578 | in <PAGE_SIZE> units (needed only for swap files). |
| 6579 | See Documentation/power/swsusp-and-swap-files.rst |
| 6580 | |
| 6581 | resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to |
| 6582 | read the resume files |
| 6583 | |
| 6584 | resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. |
| 6585 | Useful for devices that are detected asynchronously |
| 6586 | (e.g. USB and MMC devices). |
| 6587 | |
| 6588 | retain_initrd [RAM] Keep initrd memory after extraction. After boot, it will |
| 6589 | be accessible via /sys/firmware/initrd. |
| 6590 | |
| 6591 | retbleed= [X86] Control mitigation of RETBleed (Arbitrary |
| 6592 | Speculative Code Execution with Return Instructions) |
| 6593 | vulnerability. |
| 6594 | |
| 6595 | AMD-based UNRET and IBPB mitigations alone do not stop |
| 6596 | sibling threads from influencing the predictions of other |
| 6597 | sibling threads. For that reason, STIBP is used on pro- |
| 6598 | cessors that support it, and mitigate SMT on processors |
| 6599 | that don't. |
| 6600 | |
| 6601 | off - no mitigation |
| 6602 | auto - automatically select a mitigation |
| 6603 | auto,nosmt - automatically select a mitigation, |
| 6604 | disabling SMT if necessary for |
| 6605 | the full mitigation (only on Zen1 |
| 6606 | and older without STIBP). |
| 6607 | ibpb - On AMD, mitigate short speculation |
| 6608 | windows on basic block boundaries too. |
| 6609 | Safe, highest perf impact. It also |
| 6610 | enables STIBP if present. Not suitable |
| 6611 | on Intel. |
| 6612 | ibpb,nosmt - Like "ibpb" above but will disable SMT |
| 6613 | when STIBP is not available. This is |
| 6614 | the alternative for systems which do not |
| 6615 | have STIBP. |
| 6616 | unret - Force enable untrained return thunks, |
| 6617 | only effective on AMD f15h-f17h based |
| 6618 | systems. |
| 6619 | unret,nosmt - Like unret, but will disable SMT when STIBP |
| 6620 | is not available. This is the alternative for |
| 6621 | systems which do not have STIBP. |
| 6622 | |
| 6623 | Selecting 'auto' will choose a mitigation method at run |
| 6624 | time according to the CPU. |
| 6625 | |
| 6626 | Not specifying this option is equivalent to retbleed=auto. |
| 6627 | |
| 6628 | rfkill.default_state= |
| 6629 | 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, |
| 6630 | etc. communication is blocked by default. |
| 6631 | 1 Unblocked. |
| 6632 | |
| 6633 | rfkill.master_switch_mode= |
| 6634 | 0 The "airplane mode" button does nothing. |
| 6635 | 1 The "airplane mode" button toggles between everything |
| 6636 | blocked and the previous configuration. |
| 6637 | 2 The "airplane mode" button toggles between everything |
| 6638 | blocked and everything unblocked. |
| 6639 | |
| 6640 | ring3mwait=disable |
| 6641 | [KNL] Disable ring 3 MONITOR/MWAIT feature on supported |
| 6642 | CPUs. |
| 6643 | |
| 6644 | riscv_isa_fallback [RISCV,EARLY] |
| 6645 | When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit |
| 6646 | falling back to detecting extension support by parsing |
| 6647 | "riscv,isa" property on devicetree systems when the |
| 6648 | replacement properties are not found. See the Kconfig |
| 6649 | entry for RISCV_ISA_FALLBACK. |
| 6650 | |
| 6651 | riscv_nousercfi= |
| 6652 | all Disable user CFI ABI to userspace even if cpu extension |
| 6653 | are available. |
| 6654 | bcfi Disable user backward CFI ABI to userspace even if |
| 6655 | the shadow stack extension is available. |
| 6656 | fcfi Disable user forward CFI ABI to userspace even if the |
| 6657 | landing pad extension is available. |
| 6658 | |
| 6659 | ro [KNL] Mount root device read-only on boot |
| 6660 | |
| 6661 | rodata= [KNL,EARLY] |
| 6662 | on Mark read-only kernel memory as read-only (default). |
| 6663 | off Leave read-only kernel memory writable for debugging. |
| 6664 | noalias Mark read-only kernel memory as read-only but retain |
| 6665 | writable aliases in the direct map for regions outside |
| 6666 | of the kernel image. [arm64] |
| 6667 | |
| 6668 | rockchip.usb_uart |
| 6669 | [EARLY] |
| 6670 | Enable the uart passthrough on the designated usb port |
| 6671 | on Rockchip SoCs. When active, the signals of the |
| 6672 | debug-uart get routed to the D+ and D- pins of the usb |
| 6673 | port and the regular usb controller gets disabled. |
| 6674 | |
| 6675 | root= [KNL] Root filesystem |
| 6676 | Usually this is a block device specifier of some kind, |
| 6677 | see the early_lookup_bdev comment in |
| 6678 | block/early-lookup.c for details. |
| 6679 | Alternatively this can be "ram" for the legacy initial |
| 6680 | ramdisk, "nfs" and "cifs" for root on a network file |
| 6681 | system, or "mtd" and "ubi" for mounting from raw flash. |
| 6682 | |
| 6683 | rootdelay= [KNL] Delay (in seconds) to pause before attempting to |
| 6684 | mount the root filesystem |
| 6685 | |
| 6686 | rootflags= [KNL] Set root filesystem mount option string |
| 6687 | |
| 6688 | rseq_slice_ext= [KNL] RSEQ based time slice extension |
| 6689 | Format: boolean |
| 6690 | Control enablement of RSEQ based time slice extension. |
| 6691 | Default is 'on'. |
| 6692 | |
| 6693 | initramfs_options= [KNL] |
| 6694 | Specify mount options for the initramfs mount. |
| 6695 | |
| 6696 | rootfstype= [KNL] Set root filesystem type |
| 6697 | |
| 6698 | rootwait [KNL] Wait (indefinitely) for root device to show up. |
| 6699 | Useful for devices that are detected asynchronously |
| 6700 | (e.g. USB and MMC devices). |
| 6701 | |
| 6702 | rootwait= [KNL] Maximum time (in seconds) to wait for root device |
| 6703 | to show up before attempting to mount the root |
| 6704 | filesystem. |
| 6705 | |
| 6706 | rproc_mem=nn[KMG][@address] |
| 6707 | [KNL,ARM,CMA] Remoteproc physical memory block. |
| 6708 | Memory area to be used by remote processor image, |
| 6709 | managed by CMA. |
| 6710 | |
| 6711 | rseq_debug= [KNL] Enable or disable restartable sequence |
| 6712 | debug mode. Defaults to CONFIG_RSEQ_DEBUG_DEFAULT_ENABLE. |
| 6713 | Format: <bool> |
| 6714 | |
| 6715 | rt_group_sched= [KNL] Enable or disable SCHED_RR/FIFO group scheduling |
| 6716 | when CONFIG_RT_GROUP_SCHED=y. Defaults to |
| 6717 | !CONFIG_RT_GROUP_SCHED_DEFAULT_DISABLED. |
| 6718 | Format: <bool> |
| 6719 | |
| 6720 | rw [KNL] Mount root device read-write on boot |
| 6721 | |
| 6722 | S [KNL] Run init in single mode |
| 6723 | |
| 6724 | s390_iommu= [HW,S390] |
| 6725 | Set s390 IOTLB flushing mode |
| 6726 | strict |
| 6727 | With strict flushing every unmap operation will result |
| 6728 | in an IOTLB flush. Default is lazy flushing before |
| 6729 | reuse, which is faster. Deprecated, equivalent to |
| 6730 | iommu.strict=1. |
| 6731 | |
| 6732 | s390_iommu_aperture= [KNL,S390] |
| 6733 | Specifies the size of the per device DMA address space |
| 6734 | accessible through the DMA and IOMMU APIs as a decimal |
| 6735 | factor of the size of main memory. |
| 6736 | The default is 1 meaning that one can concurrently use |
| 6737 | as many DMA addresses as physical memory is installed, |
| 6738 | if supported by hardware, and thus map all of memory |
| 6739 | once. With a value of 2 one can map all of memory twice |
| 6740 | and so on. As a special case a factor of 0 imposes no |
| 6741 | restrictions other than those given by hardware at the |
| 6742 | cost of significant additional memory use for tables. |
| 6743 | |
| 6744 | |
| 6745 | sched_proxy_exec= [KNL] |
| 6746 | Enables or disables "proxy execution" style |
| 6747 | solution to mutex-based priority inversion. |
| 6748 | Format: <bool> |
| 6749 | |
| 6750 | sched_smt_asym_packing= [KNL,SMP] |
| 6751 | Format: on |
| 6752 | Force asymmetric packing at the SMT scheduling domain. |
| 6753 | Idle CPU selection prefers siblings with a higher |
| 6754 | arch_asym_cpu_priority(). The default implementation |
| 6755 | prefers lower-numbered logical CPUs, but architecture |
| 6756 | overrides remain authoritative. Equal priorities do |
| 6757 | not establish a sibling preference. For example, x86 |
| 6758 | with CONFIG_SCHED_MC_PRIO normally assigns the same |
| 6759 | core priority to both SMT siblings, so this option |
| 6760 | does not favor either sibling. |
| 6761 | |
| 6762 | sched_verbose [KNL,EARLY] Enables verbose scheduler debug messages. |
| 6763 | |
| 6764 | schedstats= [KNL,X86] Enable or disable scheduled statistics. |
| 6765 | Allowed values are enable and disable. This feature |
| 6766 | incurs a small amount of overhead in the scheduler |
| 6767 | but is useful for debugging and performance tuning. |
| 6768 | |
| 6769 | sched_thermal_decay_shift= |
| 6770 | [Deprecated] |
| 6771 | [KNL, SMP] Set a decay shift for scheduler thermal |
| 6772 | pressure signal. Thermal pressure signal follows the |
| 6773 | default decay period of other scheduler pelt |
| 6774 | signals(usually 32 ms but configurable). Setting |
| 6775 | sched_thermal_decay_shift will left shift the decay |
| 6776 | period for the thermal pressure signal by the shift |
| 6777 | value. |
| 6778 | i.e. with the default pelt decay period of 32 ms |
| 6779 | sched_thermal_decay_shift thermal pressure decay pr |
| 6780 | 1 64 ms |
| 6781 | 2 128 ms |
| 6782 | and so on. |
| 6783 | Format: integer between 0 and 10 |
| 6784 | Default is 0. |
| 6785 | |
| 6786 | scftorture.holdoff= [KNL] |
| 6787 | Number of seconds to hold off before starting |
| 6788 | test. Defaults to zero for module insertion and |
| 6789 | to 10 seconds for built-in smp_call_function() |
| 6790 | tests. |
| 6791 | |
| 6792 | scftorture.longwait= [KNL] |
| 6793 | Request ridiculously long waits randomly selected |
| 6794 | up to the chosen limit in seconds. Zero (the |
| 6795 | default) disables this feature. Please note |
| 6796 | that requesting even small non-zero numbers of |
| 6797 | seconds can result in RCU CPU stall warnings, |
| 6798 | softlockup complaints, and so on. |
| 6799 | |
| 6800 | scftorture.nthreads= [KNL] |
| 6801 | Number of kthreads to spawn to invoke the |
| 6802 | smp_call_function() family of functions. |
| 6803 | The default of -1 specifies a number of kthreads |
| 6804 | equal to the number of CPUs. |
| 6805 | |
| 6806 | scftorture.onoff_holdoff= [KNL] |
| 6807 | Number seconds to wait after the start of the |
| 6808 | test before initiating CPU-hotplug operations. |
| 6809 | |
| 6810 | scftorture.onoff_interval= [KNL] |
| 6811 | Number seconds to wait between successive |
| 6812 | CPU-hotplug operations. Specifying zero (which |
| 6813 | is the default) disables CPU-hotplug operations. |
| 6814 | |
| 6815 | scftorture.shutdown_secs= [KNL] |
| 6816 | The number of seconds following the start of the |
| 6817 | test after which to shut down the system. The |
| 6818 | default of zero avoids shutting down the system. |
| 6819 | Non-zero values are useful for automated tests. |
| 6820 | |
| 6821 | scftorture.stat_interval= [KNL] |
| 6822 | The number of seconds between outputting the |
| 6823 | current test statistics to the console. A value |
| 6824 | of zero disables statistics output. |
| 6825 | |
| 6826 | scftorture.stutter_cpus= [KNL] |
| 6827 | The number of jiffies to wait between each change |
| 6828 | to the set of CPUs under test. |
| 6829 | |
| 6830 | scftorture.use_cpus_read_lock= [KNL] |
| 6831 | Use use_cpus_read_lock() instead of the default |
| 6832 | preempt_disable() to disable CPU hotplug |
| 6833 | while invoking one of the smp_call_function*() |
| 6834 | functions. |
| 6835 | |
| 6836 | scftorture.verbose= [KNL] |
| 6837 | Enable additional printk() statements. |
| 6838 | |
| 6839 | scftorture.weight_single= [KNL] |
| 6840 | The probability weighting to use for the |
| 6841 | smp_call_function_single() function with a zero |
| 6842 | "wait" parameter. A value of -1 selects the |
| 6843 | default if all other weights are -1. However, |
| 6844 | if at least one weight has some other value, a |
| 6845 | value of -1 will instead select a weight of zero. |
| 6846 | |
| 6847 | scftorture.weight_single_wait= [KNL] |
| 6848 | The probability weighting to use for the |
| 6849 | smp_call_function_single() function with a |
| 6850 | non-zero "wait" parameter. See weight_single. |
| 6851 | |
| 6852 | scftorture.weight_many= [KNL] |
| 6853 | The probability weighting to use for the |
| 6854 | smp_call_function_many() function with a zero |
| 6855 | "wait" parameter. See weight_single. |
| 6856 | Note well that setting a high probability for |
| 6857 | this weighting can place serious IPI load |
| 6858 | on the system. |
| 6859 | |
| 6860 | scftorture.weight_many_wait= [KNL] |
| 6861 | The probability weighting to use for the |
| 6862 | smp_call_function_many() function with a |
| 6863 | non-zero "wait" parameter. See weight_single |
| 6864 | and weight_many. |
| 6865 | |
| 6866 | scftorture.weight_all= [KNL] |
| 6867 | The probability weighting to use for the |
| 6868 | smp_call_function_all() function with a zero |
| 6869 | "wait" parameter. See weight_single and |
| 6870 | weight_many. |
| 6871 | |
| 6872 | scftorture.weight_all_wait= [KNL] |
| 6873 | The probability weighting to use for the |
| 6874 | smp_call_function_all() function with a |
| 6875 | non-zero "wait" parameter. See weight_single |
| 6876 | and weight_many. |
| 6877 | |
| 6878 | sdw_mclk_divider=[SDW] |
| 6879 | Specify the MCLK divider for Intel SoundWire buses in |
| 6880 | case the BIOS does not provide the clock rate properly. |
| 6881 | |
| 6882 | skew_tick= [KNL,EARLY] Offset the periodic timer tick per cpu to mitigate |
| 6883 | xtime_lock contention on larger systems, and/or RCU lock |
| 6884 | contention on all systems with CONFIG_MAXSMP set. |
| 6885 | Format: { "0" | "1" } |
| 6886 | 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1") |
| 6887 | 1 -- enable. |
| 6888 | Note: increases power consumption, thus should only be |
| 6889 | enabled if running jitter sensitive (HPC/RT) workloads. |
| 6890 | |
| 6891 | security= [SECURITY] Choose a legacy "major" security module to |
| 6892 | enable at boot. This has been deprecated by the |
| 6893 | "lsm=" parameter. |
| 6894 | |
| 6895 | selinux= [SELINUX] Disable or enable SELinux at boot time. |
| 6896 | Format: { "0" | "1" } |
| 6897 | See security/selinux/Kconfig help text. |
| 6898 | 0 -- disable. |
| 6899 | 1 -- enable. |
| 6900 | Default value is 1. |
| 6901 | |
| 6902 | serialnumber [BUGS=X86-32] |
| 6903 | |
| 6904 | sev=option[,option...] [X86-64] |
| 6905 | |
| 6906 | debug |
| 6907 | Enable debug messages. |
| 6908 | |
| 6909 | nosnp |
| 6910 | Do not enable SEV-SNP (applies to host/hypervisor |
| 6911 | only). Setting 'nosnp' avoids the RMP check overhead |
| 6912 | in memory accesses when users do not want to run |
| 6913 | SEV-SNP guests. |
| 6914 | |
| 6915 | shapers= [NET] |
| 6916 | Maximal number of shapers. |
| 6917 | |
| 6918 | show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller |
| 6919 | Limit apic dumping. The parameter defines the maximal |
| 6920 | number of local apics being dumped. Also it is possible |
| 6921 | to set it to "all" by meaning -- no limit here. |
| 6922 | Format: { 1 (default) | 2 | ... | all }. |
| 6923 | The parameter valid if only apic=debug or |
| 6924 | apic=verbose is specified. |
| 6925 | Example: apic=debug show_lapic=all |
| 6926 | |
| 6927 | slab_debug[=options[,slabs][;[options[,slabs]]...]] [MM] |
| 6928 | Enabling slab_debug allows one to determine the |
| 6929 | culprit if slab objects become corrupted. Enabling |
| 6930 | slab_debug can create guard zones around objects and |
| 6931 | may poison objects when not in use. Also tracks the |
| 6932 | last alloc / free. For more information see |
| 6933 | Documentation/admin-guide/mm/slab.rst. |
| 6934 | (slub_debug legacy name also accepted for now) |
| 6935 | |
| 6936 | Using this option implies the "no_hash_pointers" |
| 6937 | option which can be undone by adding the |
| 6938 | "hash_pointers=always" option. |
| 6939 | |
| 6940 | slab_max_order= [MM] |
| 6941 | Determines the maximum allowed order for slabs. |
| 6942 | A high setting may cause OOMs due to memory |
| 6943 | fragmentation. For more information see |
| 6944 | Documentation/admin-guide/mm/slab.rst. |
| 6945 | (slub_max_order legacy name also accepted for now) |
| 6946 | |
| 6947 | slab_merge [MM] |
| 6948 | Enable merging of slabs with similar size when the |
| 6949 | kernel is built without CONFIG_SLAB_MERGE_DEFAULT. |
| 6950 | (slub_merge legacy name also accepted for now) |
| 6951 | |
| 6952 | slab_min_objects= [MM] |
| 6953 | The minimum number of objects per slab. SLUB will |
| 6954 | increase the slab order up to slab_max_order to |
| 6955 | generate a sufficiently large slab able to contain |
| 6956 | the number of objects indicated. The higher the number |
| 6957 | of objects the smaller the overhead of tracking slabs |
| 6958 | and the less frequently locks need to be acquired. |
| 6959 | For more information see |
| 6960 | Documentation/admin-guide/mm/slab.rst. |
| 6961 | (slub_min_objects legacy name also accepted for now) |
| 6962 | |
| 6963 | slab_min_order= [MM] |
| 6964 | Determines the minimum page order for slabs. Must be |
| 6965 | lower or equal to slab_max_order. For more information see |
| 6966 | Documentation/admin-guide/mm/slab.rst. |
| 6967 | (slub_min_order legacy name also accepted for now) |
| 6968 | |
| 6969 | slab_nomerge [MM] |
| 6970 | Disable merging of slabs with similar size. May be |
| 6971 | necessary if there is some reason to distinguish |
| 6972 | allocs to different slabs, especially in hardened |
| 6973 | environments where the risk of heap overflows and |
| 6974 | layout control by attackers can usually be |
| 6975 | frustrated by disabling merging. This will reduce |
| 6976 | most of the exposure of a heap attack to a single |
| 6977 | cache (risks via metadata attacks are mostly |
| 6978 | unchanged). Debug options disable merging on their |
| 6979 | own. |
| 6980 | For more information see |
| 6981 | Documentation/admin-guide/mm/slab.rst. |
| 6982 | (slub_nomerge legacy name also accepted for now) |
| 6983 | |
| 6984 | slab_strict_numa [MM] |
| 6985 | Support memory policies on a per object level |
| 6986 | in the slab allocator. The default is for memory |
| 6987 | policies to be applied at the folio level when |
| 6988 | a new folio is needed or a partial folio is |
| 6989 | retrieved from the lists. Increases overhead |
| 6990 | in the slab fastpaths but gains more accurate |
| 6991 | NUMA kernel object placement which helps with slow |
| 6992 | interconnects in NUMA systems. |
| 6993 | |
| 6994 | slram= [HW,MTD] |
| 6995 | |
| 6996 | smart2= [HW] |
| 6997 | Format: <io1>[,<io2>[,...,<io8>]] |
| 6998 | |
| 6999 | smp.csd_lock_timeout= [KNL] |
| 7000 | Specify the period of time in milliseconds |
| 7001 | that smp_call_function() and friends will wait |
| 7002 | for a CPU to release the CSD lock. This is |
| 7003 | useful when diagnosing bugs involving CPUs |
| 7004 | disabling interrupts for extended periods |
| 7005 | of time. Defaults to 5,000 milliseconds, and |
| 7006 | setting a value of zero disables this feature. |
| 7007 | This feature may be more efficiently disabled |
| 7008 | using the csdlock_debug- kernel parameter. |
| 7009 | |
| 7010 | smp.panic_on_ipistall= [KNL] |
| 7011 | If a csd_lock_timeout extends for more than |
| 7012 | the specified number of milliseconds, panic the |
| 7013 | system. By default, let CSD-lock acquisition |
| 7014 | take as long as they take. Specifying 300,000 |
| 7015 | for this value provides a 5-minute timeout. |
| 7016 | |
| 7017 | smt= [KNL,MIPS,S390,EARLY] Set the maximum number of threads |
| 7018 | (logical CPUs) to use per physical CPU on systems |
| 7019 | capable of symmetric multithreading (SMT). Will |
| 7020 | be capped to the actual hardware limit. |
| 7021 | Format: <integer> |
| 7022 | Default: -1 (no limit) |
| 7023 | |
| 7024 | softlockup_panic= |
| 7025 | [KNL] Should the soft-lockup detector generate panics. |
| 7026 | Format: <int> |
| 7027 | |
| 7028 | A value of non-zero instructs the soft-lockup detector |
| 7029 | to panic the machine when a soft-lockup duration exceeds |
| 7030 | N thresholds. It is also controlled by the kernel.softlockup_panic |
| 7031 | sysctl and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the |
| 7032 | respective build-time switch to that functionality. |
| 7033 | |
| 7034 | softlockup_all_cpu_backtrace= |
| 7035 | [KNL] Should the soft-lockup detector generate |
| 7036 | backtraces on all cpus. |
| 7037 | Format: 0 | 1 |
| 7038 | |
| 7039 | sonypi.*= [HW] Sony Programmable I/O Control Device driver |
| 7040 | See Documentation/admin-guide/laptops/sonypi.rst |
| 7041 | |
| 7042 | spectre_bhi= [X86] Control mitigation of Branch History Injection |
| 7043 | (BHI) vulnerability. This setting affects the |
| 7044 | deployment of the HW BHI control and the SW BHB |
| 7045 | clearing sequence. |
| 7046 | |
| 7047 | on - (default) Enable the HW or SW mitigation as |
| 7048 | needed. This protects the kernel from |
| 7049 | both syscalls and VMs. |
| 7050 | vmexit - On systems which don't have the HW mitigation |
| 7051 | available, enable the SW mitigation on vmexit |
| 7052 | ONLY. On such systems, the host kernel is |
| 7053 | protected from VM-originated BHI attacks, but |
| 7054 | may still be vulnerable to syscall attacks. |
| 7055 | off - Disable the mitigation. |
| 7056 | |
| 7057 | spectre_v2= [X86,EARLY] Control mitigation of Spectre variant 2 |
| 7058 | (indirect branch speculation) vulnerability. |
| 7059 | The default operation protects the kernel from |
| 7060 | user space attacks. |
| 7061 | |
| 7062 | on - unconditionally enable, implies |
| 7063 | spectre_v2_user=on |
| 7064 | off - unconditionally disable, implies |
| 7065 | spectre_v2_user=off |
| 7066 | auto - kernel detects whether your CPU model is |
| 7067 | vulnerable |
| 7068 | |
| 7069 | Selecting 'on' will, and 'auto' may, choose a |
| 7070 | mitigation method at run time according to the |
| 7071 | CPU, the available microcode, the setting of the |
| 7072 | CONFIG_MITIGATION_RETPOLINE configuration option, |
| 7073 | and the compiler with which the kernel was built. |
| 7074 | |
| 7075 | Selecting 'on' will also enable the mitigation |
| 7076 | against user space to user space task attacks. |
| 7077 | Selecting specific mitigation does not force enable |
| 7078 | user mitigations. |
| 7079 | |
| 7080 | Selecting 'off' will disable both the kernel and |
| 7081 | the user space protections. |
| 7082 | |
| 7083 | Specific mitigations can also be selected manually: |
| 7084 | |
| 7085 | retpoline - replace indirect branches |
| 7086 | retpoline,generic - Retpolines |
| 7087 | retpoline,lfence - LFENCE; indirect branch |
| 7088 | retpoline,amd - alias for retpoline,lfence |
| 7089 | eibrs - Enhanced/Auto IBRS |
| 7090 | eibrs,retpoline - Enhanced/Auto IBRS + Retpolines |
| 7091 | eibrs,lfence - Enhanced/Auto IBRS + LFENCE |
| 7092 | ibrs - use IBRS to protect kernel |
| 7093 | |
| 7094 | Not specifying this option is equivalent to |
| 7095 | spectre_v2=auto. |
| 7096 | |
| 7097 | spectre_v2_user= |
| 7098 | [X86] Control mitigation of Spectre variant 2 |
| 7099 | (indirect branch speculation) vulnerability between |
| 7100 | user space tasks |
| 7101 | |
| 7102 | on - Unconditionally enable mitigations. Is |
| 7103 | enforced by spectre_v2=on |
| 7104 | |
| 7105 | off - Unconditionally disable mitigations. Is |
| 7106 | enforced by spectre_v2=off |
| 7107 | |
| 7108 | prctl - Indirect branch speculation is enabled, |
| 7109 | but mitigation can be enabled via prctl |
| 7110 | per thread. The mitigation control state |
| 7111 | is inherited on fork. |
| 7112 | |
| 7113 | prctl,ibpb |
| 7114 | - Like "prctl" above, but only STIBP is |
| 7115 | controlled per thread. IBPB is issued |
| 7116 | always when switching between different user |
| 7117 | space processes. |
| 7118 | |
| 7119 | seccomp |
| 7120 | - Same as "prctl" above, but all seccomp |
| 7121 | threads will enable the mitigation unless |
| 7122 | they explicitly opt out. |
| 7123 | |
| 7124 | seccomp,ibpb |
| 7125 | - Like "seccomp" above, but only STIBP is |
| 7126 | controlled per thread. IBPB is issued |
| 7127 | always when switching between different |
| 7128 | user space processes. |
| 7129 | |
| 7130 | auto - Kernel selects the mitigation depending on |
| 7131 | the available CPU features and vulnerability. |
| 7132 | |
| 7133 | Default mitigation: "prctl" |
| 7134 | |
| 7135 | Not specifying this option is equivalent to |
| 7136 | spectre_v2_user=auto. |
| 7137 | |
| 7138 | spec_rstack_overflow= |
| 7139 | [X86,EARLY] Control RAS overflow mitigation on AMD Zen CPUs |
| 7140 | |
| 7141 | off - Disable mitigation |
| 7142 | microcode - Enable microcode mitigation only |
| 7143 | safe-ret - Enable sw-only safe RET mitigation (default) |
| 7144 | ibpb - Enable mitigation by issuing IBPB on |
| 7145 | kernel entry |
| 7146 | ibpb-vmexit - Issue IBPB only on VMEXIT |
| 7147 | (cloud-specific mitigation) |
| 7148 | |
| 7149 | spec_store_bypass_disable= |
| 7150 | [HW,EARLY] Control Speculative Store Bypass (SSB) Disable mitigation |
| 7151 | (Speculative Store Bypass vulnerability) |
| 7152 | |
| 7153 | Certain CPUs are vulnerable to an exploit against a |
| 7154 | a common industry wide performance optimization known |
| 7155 | as "Speculative Store Bypass" in which recent stores |
| 7156 | to the same memory location may not be observed by |
| 7157 | later loads during speculative execution. The idea |
| 7158 | is that such stores are unlikely and that they can |
| 7159 | be detected prior to instruction retirement at the |
| 7160 | end of a particular speculation execution window. |
| 7161 | |
| 7162 | In vulnerable processors, the speculatively forwarded |
| 7163 | store can be used in a cache side channel attack, for |
| 7164 | example to read memory to which the attacker does not |
| 7165 | directly have access (e.g. inside sandboxed code). |
| 7166 | |
| 7167 | This parameter controls whether the Speculative Store |
| 7168 | Bypass optimization is used. |
| 7169 | |
| 7170 | On x86 the options are: |
| 7171 | |
| 7172 | on - Unconditionally disable Speculative Store Bypass |
| 7173 | off - Unconditionally enable Speculative Store Bypass |
| 7174 | auto - Kernel detects whether the CPU model contains an |
| 7175 | implementation of Speculative Store Bypass and |
| 7176 | picks the most appropriate mitigation. If the |
| 7177 | CPU is not vulnerable, "off" is selected. If the |
| 7178 | CPU is vulnerable the default mitigation is |
| 7179 | architecture and Kconfig dependent. See below. |
| 7180 | prctl - Control Speculative Store Bypass per thread |
| 7181 | via prctl. Speculative Store Bypass is enabled |
| 7182 | for a process by default. The state of the control |
| 7183 | is inherited on fork. |
| 7184 | seccomp - Same as "prctl" above, but all seccomp threads |
| 7185 | will disable SSB unless they explicitly opt out. |
| 7186 | |
| 7187 | Default mitigations: |
| 7188 | X86: "prctl" |
| 7189 | |
| 7190 | On powerpc the options are: |
| 7191 | |
| 7192 | on,auto - On Power8 and Power9 insert a store-forwarding |
| 7193 | barrier on kernel entry and exit. On Power7 |
| 7194 | perform a software flush on kernel entry and |
| 7195 | exit. |
| 7196 | off - No action. |
| 7197 | |
| 7198 | Not specifying this option is equivalent to |
| 7199 | spec_store_bypass_disable=auto. |
| 7200 | |
| 7201 | split_llc= |
| 7202 | [X86,EARLY] Split the LLC N-ways |
| 7203 | |
| 7204 | When set, the LLC is split this many ways by matching |
| 7205 | 'core_id % n'. This is setup before SMP bringup and |
| 7206 | used during SMP bringup before it knows the full |
| 7207 | topology. If your core count doesn't nicely divide by |
| 7208 | the number given, you get to keep the pieces. |
| 7209 | |
| 7210 | This is mostly a debug feature to emulate multiple LLCs |
| 7211 | on hardware that only have a single LLC. |
| 7212 | |
| 7213 | split_lock_detect= |
| 7214 | [X86] Enable split lock detection or bus lock detection |
| 7215 | |
| 7216 | When enabled (and if hardware support is present), atomic |
| 7217 | instructions that access data across cache line |
| 7218 | boundaries will result in an alignment check exception |
| 7219 | for split lock detection or a debug exception for |
| 7220 | bus lock detection. |
| 7221 | |
| 7222 | off - not enabled |
| 7223 | |
| 7224 | warn - the kernel will emit rate-limited warnings |
| 7225 | about applications triggering the #AC |
| 7226 | exception or the #DB exception. This mode is |
| 7227 | the default on CPUs that support split lock |
| 7228 | detection or bus lock detection. Default |
| 7229 | behavior is by #AC if both features are |
| 7230 | enabled in hardware. |
| 7231 | |
| 7232 | fatal - the kernel will send SIGBUS to applications |
| 7233 | that trigger the #AC exception or the #DB |
| 7234 | exception. Default behavior is by #AC if |
| 7235 | both features are enabled in hardware. |
| 7236 | |
| 7237 | ratelimit:N - |
| 7238 | Set system wide rate limit to N bus locks |
| 7239 | per second for bus lock detection. |
| 7240 | 0 < N <= 1000. |
| 7241 | |
| 7242 | N/A for split lock detection. |
| 7243 | |
| 7244 | |
| 7245 | If an #AC exception is hit in the kernel or in |
| 7246 | firmware (i.e. not while executing in user mode) |
| 7247 | the kernel will oops in either "warn" or "fatal" |
| 7248 | mode. |
| 7249 | |
| 7250 | #DB exception for bus lock is triggered only when |
| 7251 | CPL > 0. |
| 7252 | |
| 7253 | srbds= [X86,INTEL,EARLY] |
| 7254 | Control the Special Register Buffer Data Sampling |
| 7255 | (SRBDS) mitigation. |
| 7256 | |
| 7257 | Certain CPUs are vulnerable to an MDS-like |
| 7258 | exploit which can leak bits from the random |
| 7259 | number generator. |
| 7260 | |
| 7261 | By default, this issue is mitigated by |
| 7262 | microcode. However, the microcode fix can cause |
| 7263 | the RDRAND and RDSEED instructions to become |
| 7264 | much slower. Among other effects, this will |
| 7265 | result in reduced throughput from /dev/urandom. |
| 7266 | |
| 7267 | The microcode mitigation can be disabled with |
| 7268 | the following option: |
| 7269 | |
| 7270 | off: Disable mitigation and remove |
| 7271 | performance impact to RDRAND and RDSEED |
| 7272 | |
| 7273 | srcutree.big_cpu_lim [KNL] |
| 7274 | Specifies the number of CPUs constituting a |
| 7275 | large system, such that srcu_struct structures |
| 7276 | should immediately allocate an srcu_node array. |
| 7277 | This kernel-boot parameter defaults to 128, |
| 7278 | but takes effect only when the low-order four |
| 7279 | bits of srcutree.convert_to_big is equal to 3 |
| 7280 | (decide at boot). |
| 7281 | |
| 7282 | srcutree.convert_to_big [KNL] |
| 7283 | Specifies under what conditions an SRCU tree |
| 7284 | srcu_struct structure will be converted to big |
| 7285 | form, that is, with an rcu_node tree: |
| 7286 | |
| 7287 | 0: Never. |
| 7288 | 1: At init_srcu_struct() time. |
| 7289 | 2: When rcutorture decides to. |
| 7290 | 3: Decide at boot time (default). |
| 7291 | 0x1X: Above plus if high contention. |
| 7292 | |
| 7293 | Either way, the srcu_node tree will be sized based |
| 7294 | on the actual runtime number of CPUs (nr_cpu_ids) |
| 7295 | instead of the compile-time CONFIG_NR_CPUS. |
| 7296 | |
| 7297 | srcutree.counter_wrap_check [KNL] |
| 7298 | Specifies how frequently to check for |
| 7299 | grace-period sequence counter wrap for the |
| 7300 | srcu_data structure's ->srcu_gp_seq_needed field. |
| 7301 | The greater the number of bits set in this kernel |
| 7302 | parameter, the less frequently counter wrap will |
| 7303 | be checked for. Note that the bottom two bits |
| 7304 | are ignored. |
| 7305 | |
| 7306 | srcutree.exp_holdoff [KNL] |
| 7307 | Specifies how many nanoseconds must elapse |
| 7308 | since the end of the last SRCU grace period for |
| 7309 | a given srcu_struct until the next normal SRCU |
| 7310 | grace period will be considered for automatic |
| 7311 | expediting. Set to zero to disable automatic |
| 7312 | expediting. |
| 7313 | |
| 7314 | srcutree.srcu_max_nodelay [KNL] |
| 7315 | Specifies the number of no-delay instances |
| 7316 | per jiffy for which the SRCU grace period |
| 7317 | worker thread will be rescheduled with zero |
| 7318 | delay. Beyond this limit, worker thread will |
| 7319 | be rescheduled with a sleep delay of one jiffy. |
| 7320 | |
| 7321 | srcutree.srcu_max_nodelay_phase [KNL] |
| 7322 | Specifies the per-grace-period phase, number of |
| 7323 | non-sleeping polls of readers. Beyond this limit, |
| 7324 | grace period worker thread will be rescheduled |
| 7325 | with a sleep delay of one jiffy, between each |
| 7326 | rescan of the readers, for a grace period phase. |
| 7327 | |
| 7328 | srcutree.srcu_retry_check_delay [KNL] |
| 7329 | Specifies number of microseconds of non-sleeping |
| 7330 | delay between each non-sleeping poll of readers. |
| 7331 | |
| 7332 | srcutree.small_contention_lim [KNL] |
| 7333 | Specifies the number of update-side contention |
| 7334 | events per jiffy will be tolerated before |
| 7335 | initiating a conversion of an srcu_struct |
| 7336 | structure to big form. Note that the value of |
| 7337 | srcutree.convert_to_big must have the 0x10 bit |
| 7338 | set for contention-based conversions to occur. |
| 7339 | |
| 7340 | ssbd= [ARM64,HW,EARLY] |
| 7341 | Speculative Store Bypass Disable control |
| 7342 | |
| 7343 | On CPUs that are vulnerable to the Speculative |
| 7344 | Store Bypass vulnerability and offer a |
| 7345 | firmware based mitigation, this parameter |
| 7346 | indicates how the mitigation should be used: |
| 7347 | |
| 7348 | force-on: Unconditionally enable mitigation for |
| 7349 | for both kernel and userspace |
| 7350 | force-off: Unconditionally disable mitigation for |
| 7351 | for both kernel and userspace |
| 7352 | kernel: Always enable mitigation in the |
| 7353 | kernel, and offer a prctl interface |
| 7354 | to allow userspace to register its |
| 7355 | interest in being mitigated too. |
| 7356 | |
| 7357 | stack_guard_gap= [MM] |
| 7358 | override the default stack gap protection. The value |
| 7359 | is in page units and it defines how many pages prior |
| 7360 | to (for stacks growing down) resp. after (for stacks |
| 7361 | growing up) the main stack are reserved for no other |
| 7362 | mapping. Default value is 256 pages. |
| 7363 | |
| 7364 | stack_depot_disable= [KNL,EARLY] |
| 7365 | Setting this to true through kernel command line will |
| 7366 | disable the stack depot thereby saving the static memory |
| 7367 | consumed by the stack hash table. By default this is set |
| 7368 | to false. |
| 7369 | |
| 7370 | stack_depot_max_pools= [KNL,EARLY] |
| 7371 | Specify the maximum number of pools to use for storing |
| 7372 | stack traces. Pools are allocated on-demand up to this |
| 7373 | limit. Default value is 8191 pools. |
| 7374 | |
| 7375 | stacktrace [FTRACE] |
| 7376 | Enable the stack tracer on boot up. |
| 7377 | |
| 7378 | stacktrace_filter=[function-list] |
| 7379 | [FTRACE] Limit the functions that the stack tracer |
| 7380 | will trace at boot up. function-list is a comma-separated |
| 7381 | list of functions. This list can be changed at run |
| 7382 | time by the stack_trace_filter file in the debugfs |
| 7383 | tracing directory. Note, this enables stack tracing |
| 7384 | and the stacktrace above is not needed. |
| 7385 | |
| 7386 | sti= [PARISC,HW] |
| 7387 | Format: <num> |
| 7388 | Set the STI (builtin display/keyboard on the HP-PARISC |
| 7389 | machines) console (graphic card) which should be used |
| 7390 | as the initial boot-console. |
| 7391 | See also comment in drivers/video/sticore.c. |
| 7392 | |
| 7393 | sti_font= [HW] |
| 7394 | See comment in drivers/video/sticore.c. |
| 7395 | |
| 7396 | stifb= [HW] |
| 7397 | Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] |
| 7398 | |
| 7399 | strict_sas_size= |
| 7400 | [X86] |
| 7401 | Format: <bool> |
| 7402 | Enable or disable strict sigaltstack size checks |
| 7403 | against the required signal frame size which |
| 7404 | depends on the supported FPU features. This can |
| 7405 | be used to filter out binaries which have |
| 7406 | not yet been made aware of AT_MINSIGSTKSZ. |
| 7407 | |
| 7408 | stress_hpt [PPC,EARLY] |
| 7409 | Limits the number of kernel HPT entries in the hash |
| 7410 | page table to increase the rate of hash page table |
| 7411 | faults on kernel addresses. |
| 7412 | |
| 7413 | stress_slb [PPC,EARLY] |
| 7414 | Limits the number of kernel SLB entries, and flushes |
| 7415 | them frequently to increase the rate of SLB faults |
| 7416 | on kernel addresses. |
| 7417 | |
| 7418 | no_slb_preload [PPC,EARLY] |
| 7419 | Disables slb preloading for userspace. |
| 7420 | |
| 7421 | sunrpc.min_resvport= |
| 7422 | sunrpc.max_resvport= |
| 7423 | [NFS,SUNRPC] |
| 7424 | SunRPC servers often require that client requests |
| 7425 | originate from a privileged port (i.e. a port in the |
| 7426 | range 0 < portnr < 1024). |
| 7427 | An administrator who wishes to reserve some of these |
| 7428 | ports for other uses may adjust the range that the |
| 7429 | kernel's sunrpc client considers to be privileged |
| 7430 | using these two parameters to set the minimum and |
| 7431 | maximum port values. |
| 7432 | |
| 7433 | sunrpc.svc_rpc_per_connection_limit= |
| 7434 | [NFS,SUNRPC] |
| 7435 | Limit the number of requests that the server will |
| 7436 | process in parallel from a single connection. |
| 7437 | The default value is 0 (no limit). |
| 7438 | |
| 7439 | sunrpc.pool_mode= |
| 7440 | [NFS] |
| 7441 | Deprecated. The NFS server now always uses one |
| 7442 | service thread pool per NUMA node (equivalent to a |
| 7443 | single global pool on non-NUMA machines). All of |
| 7444 | the previously accepted values (auto, global, |
| 7445 | percpu, pernode) are still accepted for backward |
| 7446 | compatibility but are ignored: the mode is always |
| 7447 | pernode, and reads always return "pernode". |
| 7448 | |
| 7449 | sunrpc.tcp_slot_table_entries= |
| 7450 | sunrpc.udp_slot_table_entries= |
| 7451 | [NFS,SUNRPC] |
| 7452 | Sets the upper limit on the number of simultaneous |
| 7453 | RPC calls that can be sent from the client to a |
| 7454 | server. Increasing these values may allow you to |
| 7455 | improve throughput, but will also increase the |
| 7456 | amount of memory reserved for use by the client. |
| 7457 | |
| 7458 | suspend.pm_test_delay= |
| 7459 | [SUSPEND] |
| 7460 | Sets the number of seconds to remain in a suspend test |
| 7461 | mode before resuming the system (see |
| 7462 | /sys/power/pm_test). Only available when CONFIG_PM_DEBUG |
| 7463 | is set. Default value is 5. |
| 7464 | |
| 7465 | svm= [PPC] |
| 7466 | Format: { on | off | y | n | 1 | 0 } |
| 7467 | This parameter controls use of the Protected |
| 7468 | Execution Facility on pSeries. |
| 7469 | |
| 7470 | swiotlb= [ARM,PPC,MIPS,X86,S390,EARLY] |
| 7471 | Format: { <int> [,<int>] | force | noforce | track_hiwater} |
| 7472 | <int> -- Number of I/O TLB slabs |
| 7473 | <int> -- Second integer after comma. Number of swiotlb |
| 7474 | areas with their own lock. Will be rounded up |
| 7475 | to a power of 2. |
| 7476 | force -- force using of bounce buffers even if they |
| 7477 | wouldn't be automatically used by the kernel |
| 7478 | noforce -- Never use bounce buffers (for debugging) |
| 7479 | track_hiwater -- Track high watermark of swiotlb buffers. |
| 7480 | Only available when CONFIG_DEBUG_FS is set. |
| 7481 | |
| 7482 | switches= [HW,M68k,EARLY] |
| 7483 | |
| 7484 | sysctl.*= [KNL] |
| 7485 | Set a sysctl parameter, right before loading the init |
| 7486 | process, as if the value was written to the respective |
| 7487 | /proc/sys/... file. Both '.' and '/' are recognized as |
| 7488 | separators. Unrecognized parameters and invalid values |
| 7489 | are reported in the kernel log. Sysctls registered |
| 7490 | later by a loaded module cannot be set this way. |
| 7491 | Example: sysctl.vm.swappiness=40 |
| 7492 | |
| 7493 | sysrq_always_enabled |
| 7494 | [KNL] |
| 7495 | Ignore sysrq setting - this boot parameter will |
| 7496 | neutralize any effect of /proc/sys/kernel/sysrq. |
| 7497 | Useful for debugging. |
| 7498 | |
| 7499 | tcpmhash_entries= [KNL,NET] |
| 7500 | Set the number of tcp_metrics_hash slots. |
| 7501 | Default value is 8192 or 16384 depending on total |
| 7502 | ram pages. This is used to specify the TCP metrics |
| 7503 | cache size. See Documentation/networking/ip-sysctl.rst |
| 7504 | "tcp_no_metrics_save" section for more details. |
| 7505 | |
| 7506 | tdfx= [HW,DRM] |
| 7507 | |
| 7508 | test_suspend= [SUSPEND] |
| 7509 | Format: { "mem" | "standby" | "freeze" }[,N] |
| 7510 | Specify "mem" (for Suspend-to-RAM) or "standby" (for |
| 7511 | standby suspend) or "freeze" (for suspend type freeze) |
| 7512 | as the system sleep state during system startup with |
| 7513 | the optional capability to repeat N number of times. |
| 7514 | The system is woken from this state using a |
| 7515 | wakeup-capable RTC alarm. |
| 7516 | |
| 7517 | thash_entries= [KNL,NET] |
| 7518 | Set number of hash buckets for TCP connection |
| 7519 | |
| 7520 | thermal.act= [HW,ACPI] |
| 7521 | -1: disable all active trip points in all thermal zones |
| 7522 | <degrees C>: override all lowest active trip points |
| 7523 | |
| 7524 | thermal.crt= [HW,ACPI] |
| 7525 | -1: disable all critical trip points in all thermal zones |
| 7526 | <degrees C>: override all critical trip points |
| 7527 | |
| 7528 | thermal.off= [HW,ACPI] |
| 7529 | 1: disable ACPI thermal control |
| 7530 | |
| 7531 | thermal.psv= [HW,ACPI] |
| 7532 | -1: disable all passive trip points |
| 7533 | <degrees C>: override all passive trip points to this |
| 7534 | value |
| 7535 | |
| 7536 | thermal.tzp= [HW,ACPI] |
| 7537 | Specify global default ACPI thermal zone polling rate |
| 7538 | <deci-seconds>: poll all this frequency |
| 7539 | 0: no polling (default) |
| 7540 | |
| 7541 | thp_anon= [KNL] |
| 7542 | Format: <size>[KMG],<size>[KMG]:<state>;<size>[KMG]-<size>[KMG]:<state> |
| 7543 | state is one of "always", "madvise", "never" or "inherit". |
| 7544 | Control the default behavior of the system with respect |
| 7545 | to anonymous transparent hugepages. |
| 7546 | Can be used multiple times for multiple anon THP sizes. |
| 7547 | See Documentation/admin-guide/mm/transhuge.rst for more |
| 7548 | details. |
| 7549 | |
| 7550 | threadirqs [KNL,EARLY] |
| 7551 | Force threading of all interrupt handlers except those |
| 7552 | marked explicitly IRQF_NO_THREAD. |
| 7553 | |
| 7554 | thp_shmem= [KNL] |
| 7555 | Format: <size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-<size>[KMG]:<policy> |
| 7556 | Control the default policy of each hugepage size for the |
| 7557 | internal shmem mount. <policy> is one of policies available |
| 7558 | for the shmem mount ("always", "inherit", "never", "within_size", |
| 7559 | and "advise"). |
| 7560 | It can be used multiple times for multiple shmem THP sizes. |
| 7561 | See Documentation/admin-guide/mm/transhuge.rst for more |
| 7562 | details. |
| 7563 | |
| 7564 | tlbi= [X86-64] |
| 7565 | Format: {ipi} |
| 7566 | ipi: switch to IPI-based TLB flushing |
| 7567 | |
| 7568 | topology= [S390,EARLY] |
| 7569 | Format: {off | on} |
| 7570 | Specify if the kernel should make use of the cpu |
| 7571 | topology information if the hardware supports this. |
| 7572 | The scheduler will make use of this information and |
| 7573 | e.g. base its process migration decisions on it. |
| 7574 | Default is on. |
| 7575 | |
| 7576 | torture.disable_onoff_at_boot= [KNL] |
| 7577 | Prevent the CPU-hotplug component of torturing |
| 7578 | until after init has spawned. |
| 7579 | |
| 7580 | torture.ftrace_dump_at_shutdown= [KNL] |
| 7581 | Dump the ftrace buffer at torture-test shutdown, |
| 7582 | even if there were no errors. This can be a |
| 7583 | very costly operation when many torture tests |
| 7584 | are running concurrently, especially on systems |
| 7585 | with rotating-rust storage. |
| 7586 | |
| 7587 | torture.verbose_sleep_frequency= [KNL] |
| 7588 | Specifies how many verbose printk()s should be |
| 7589 | emitted between each sleep. The default of zero |
| 7590 | disables verbose-printk() sleeping. |
| 7591 | |
| 7592 | torture.verbose_sleep_duration= [KNL] |
| 7593 | Duration of each verbose-printk() sleep in jiffies. |
| 7594 | |
| 7595 | tpm.disable_pcr_integrity= [HW,TPM] |
| 7596 | Do not protect PCR registers from unintended physical |
| 7597 | access, or interposers in the bus by the means of |
| 7598 | having an integrity protected session wrapped around |
| 7599 | TPM2_PCR_Extend command. Consider this in a situation |
| 7600 | where TPM is heavily utilized by IMA, thus protection |
| 7601 | causing a major performance hit, and the space where |
| 7602 | machines are deployed is by other means guarded. |
| 7603 | |
| 7604 | tpm_crb_ffa.busy_timeout_ms= [ARM64,TPM] |
| 7605 | Maximum time in milliseconds to retry sending a message |
| 7606 | to the TPM service before giving up. This parameter controls |
| 7607 | how long the system will continue retrying when the TPM |
| 7608 | service is busy. |
| 7609 | Format: <unsigned int> |
| 7610 | Default: 2000 (2 seconds) |
| 7611 | |
| 7612 | tpm_suspend_pcr=[HW,TPM] |
| 7613 | Format: integer pcr id |
| 7614 | Specify that at suspend time, the tpm driver |
| 7615 | should extend the specified pcr with zeros, |
| 7616 | as a workaround for some chips which fail to |
| 7617 | flush the last written pcr on TPM_SaveState. |
| 7618 | This will guarantee that all the other pcrs |
| 7619 | are saved. |
| 7620 | |
| 7621 | tpm_tis.interrupts= [HW,TPM] |
| 7622 | Enable interrupts for the MMIO based physical layer |
| 7623 | for the FIFO interface. By default it is set to false |
| 7624 | (0). For more information about TPM hardware interfaces |
| 7625 | defined by Trusted Computing Group (TCG) see |
| 7626 | https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ |
| 7627 | |
| 7628 | tp_printk [FTRACE] |
| 7629 | Have the tracepoints sent to printk as well as the |
| 7630 | tracing ring buffer. This is useful for early boot up |
| 7631 | where the system hangs or reboots and does not give the |
| 7632 | option for reading the tracing buffer or performing a |
| 7633 | ftrace_dump_on_oops. |
| 7634 | |
| 7635 | To turn off having tracepoints sent to printk, |
| 7636 | echo 0 > /proc/sys/kernel/tracepoint_printk |
| 7637 | Note, echoing 1 into this file without the |
| 7638 | tp_printk kernel cmdline option has no effect. |
| 7639 | |
| 7640 | The tp_printk_stop_on_boot (see below) can also be used |
| 7641 | to stop the printing of events to console at |
| 7642 | late_initcall_sync. |
| 7643 | |
| 7644 | ** CAUTION ** |
| 7645 | |
| 7646 | Having tracepoints sent to printk() and activating high |
| 7647 | frequency tracepoints such as irq or sched, can cause |
| 7648 | the system to live lock. |
| 7649 | |
| 7650 | tp_printk_stop_on_boot [FTRACE] |
| 7651 | When tp_printk (above) is set, it can cause a lot of noise |
| 7652 | on the console. It may be useful to only include the |
| 7653 | printing of events during boot up, as user space may |
| 7654 | make the system inoperable. |
| 7655 | |
| 7656 | This command line option will stop the printing of events |
| 7657 | to console at the late_initcall_sync() time frame. |
| 7658 | |
| 7659 | trace_buf_size=nn[KMG] |
| 7660 | [FTRACE] will set tracing buffer size on each cpu. |
| 7661 | |
| 7662 | trace_clock= [FTRACE] Set the clock used for tracing events |
| 7663 | at boot up. |
| 7664 | local - Use the per CPU time stamp counter |
| 7665 | (converted into nanoseconds). Fast, but |
| 7666 | depending on the architecture, may not be |
| 7667 | in sync between CPUs. |
| 7668 | global - Event time stamps are synchronized across |
| 7669 | CPUs. May be slower than the local clock, |
| 7670 | but better for some race conditions. |
| 7671 | counter - Simple counting of events (1, 2, ..) |
| 7672 | note, some counts may be skipped due to the |
| 7673 | infrastructure grabbing the clock more than |
| 7674 | once per event. |
| 7675 | uptime - Use jiffies as the time stamp. |
| 7676 | perf - Use the same clock that perf uses. |
| 7677 | mono - Use ktime_get_mono_fast_ns() for time stamps. |
| 7678 | mono_raw - Use ktime_get_raw_fast_ns() for time |
| 7679 | stamps. |
| 7680 | boot - Use ktime_get_boot_fast_ns() for time stamps. |
| 7681 | Architectures may add more clocks. See |
| 7682 | Documentation/trace/ftrace.rst for more details. |
| 7683 | |
| 7684 | trace_event=[event-list] |
| 7685 | [FTRACE] Set and start specified trace events in order |
| 7686 | to facilitate early boot debugging. The event-list is a |
| 7687 | comma-separated list of trace events to enable. See |
| 7688 | also Documentation/trace/events.rst |
| 7689 | |
| 7690 | To enable modules, use :mod: keyword: |
| 7691 | |
| 7692 | trace_event=:mod:<module> |
| 7693 | |
| 7694 | The value before :mod: will only enable specific events |
| 7695 | that are part of the module. See the above mentioned |
| 7696 | document for more information. |
| 7697 | |
| 7698 | trace_instance=[instance-info] |
| 7699 | [FTRACE] Create a ring buffer instance early in boot up. |
| 7700 | This will be listed in: |
| 7701 | |
| 7702 | /sys/kernel/tracing/instances |
| 7703 | |
| 7704 | Events can be enabled at the time the instance is created |
| 7705 | via: |
| 7706 | |
| 7707 | trace_instance=<name>,<system1>:<event1>,<system2>:<event2> |
| 7708 | |
| 7709 | Note, the "<system*>:" portion is optional if the event is |
| 7710 | unique. |
| 7711 | |
| 7712 | trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall |
| 7713 | |
| 7714 | will enable the "sched_switch" event (note, the "sched:" is optional, and |
| 7715 | the same thing would happen if it was left off). The irq_handler_entry |
| 7716 | event, and all events under the "initcall" system. |
| 7717 | |
| 7718 | Flags can be added to the instance to modify its behavior when it is |
| 7719 | created. The flags are separated by '^'. |
| 7720 | |
| 7721 | The available flags are: |
| 7722 | |
| 7723 | traceoff - Have the tracing instance tracing disabled after it is created. |
| 7724 | traceprintk - Have trace_printk() write into this trace instance |
| 7725 | (note, "printk" and "trace_printk" can also be used) |
| 7726 | |
| 7727 | trace_instance=foo^traceoff^traceprintk,sched,irq |
| 7728 | |
| 7729 | The flags must come before the defined events. |
| 7730 | |
| 7731 | If memory has been reserved (see memmap for x86), the instance |
| 7732 | can use that memory: |
| 7733 | |
| 7734 | memmap=12M$0x284500000 trace_instance=boot_map@0x284500000:12M |
| 7735 | |
| 7736 | The above will create a "boot_map" instance that uses the physical |
| 7737 | memory at 0x284500000 that is 12Megs. The per CPU buffers of that |
| 7738 | instance will be split up accordingly. |
| 7739 | |
| 7740 | Alternatively, the memory can be reserved by the reserve_mem option: |
| 7741 | |
| 7742 | reserve_mem=12M:4096:trace trace_instance=boot_map@trace |
| 7743 | |
| 7744 | This will reserve 12 megabytes at boot up with a 4096 byte alignment |
| 7745 | and place the ring buffer in this memory. Note that due to KASLR, the |
| 7746 | memory may not be the same location each time, which will not preserve |
| 7747 | the buffer content. |
| 7748 | |
| 7749 | Also note that the layout of the ring buffer data may change between |
| 7750 | kernel versions where the validator will fail and reset the ring buffer |
| 7751 | if the layout is not the same as the previous kernel. |
| 7752 | |
| 7753 | If the ring buffer is used for persistent bootups and has events enabled, |
| 7754 | it is recommend to disable tracing so that events from a previous boot do not |
| 7755 | mix with events of the current boot (unless you are debugging a random crash |
| 7756 | at boot up). |
| 7757 | |
| 7758 | reserve_mem=12M:4096:trace trace_instance=boot_map^traceoff^traceprintk@trace,sched,irq |
| 7759 | |
| 7760 | Note, saving the trace buffer across reboots does require that the system |
| 7761 | is set up to not wipe memory. For instance, CONFIG_RESET_ATTACK_MITIGATION |
| 7762 | can force a memory reset on boot which will clear any trace that was stored. |
| 7763 | This is just one of many ways that can clear memory. Make sure your system |
| 7764 | keeps the content of memory across reboots before relying on this option. |
| 7765 | |
| 7766 | NB: Both the mapped address and size must be page aligned for the architecture. |
| 7767 | |
| 7768 | See also Documentation/trace/debugging.rst |
| 7769 | |
| 7770 | |
| 7771 | trace_options=[option-list] |
| 7772 | [FTRACE] Enable or disable tracer options at boot. |
| 7773 | The option-list is a comma delimited list of options |
| 7774 | that can be enabled or disabled just as if you were |
| 7775 | to echo the option name into |
| 7776 | |
| 7777 | /sys/kernel/tracing/trace_options |
| 7778 | |
| 7779 | For example, to enable stacktrace option (to dump the |
| 7780 | stack trace of each event), add to the command line: |
| 7781 | |
| 7782 | trace_options=stacktrace |
| 7783 | |
| 7784 | See also Documentation/trace/ftrace.rst "trace options" |
| 7785 | section. |
| 7786 | |
| 7787 | trace_trigger=[trigger-list] |
| 7788 | [FTRACE] Add an event trigger on specific events. |
| 7789 | Set a trigger on top of a specific event, with an optional |
| 7790 | filter. |
| 7791 | |
| 7792 | The format is "trace_trigger=<event>.<trigger>[ if <filter>],..." |
| 7793 | Where more than one trigger may be specified that are comma delimited. |
| 7794 | |
| 7795 | For example: |
| 7796 | |
| 7797 | trace_trigger="sched_switch.stacktrace if prev_state == 2" |
| 7798 | |
| 7799 | The above will enable the "stacktrace" trigger on the "sched_switch" |
| 7800 | event but only trigger it if the "prev_state" of the "sched_switch" |
| 7801 | event is "2" (TASK_UNINTERRUPTIBLE). |
| 7802 | |
| 7803 | See also "Event triggers" in Documentation/trace/events.rst |
| 7804 | |
| 7805 | |
| 7806 | traceoff_after_boot |
| 7807 | [FTRACE] Sometimes tracing is used to debug issues |
| 7808 | during the boot process. Since the trace buffer has a |
| 7809 | limited amount of storage, it may be prudent to |
| 7810 | disable tracing after the boot is finished, otherwise |
| 7811 | the critical information may be overwritten. With this |
| 7812 | option, the main tracing buffer will be turned off at |
| 7813 | the end of the boot process. |
| 7814 | |
| 7815 | traceoff_on_warning |
| 7816 | [FTRACE] enable this option to disable tracing when a |
| 7817 | warning is hit. This turns off "tracing_on". Tracing can |
| 7818 | be enabled again by echoing '1' into the "tracing_on" |
| 7819 | file located in /sys/kernel/tracing/ |
| 7820 | |
| 7821 | This option is useful, as it disables the trace before |
| 7822 | the WARNING dump is called, which prevents the trace to |
| 7823 | be filled with content caused by the warning output. |
| 7824 | |
| 7825 | This option can also be set at run time via the sysctl |
| 7826 | option: kernel/traceoff_on_warning |
| 7827 | |
| 7828 | transparent_hugepage= |
| 7829 | [KNL] |
| 7830 | Format: [always|madvise|never] |
| 7831 | Can be used to control the default behavior of the system |
| 7832 | with respect to transparent hugepages. |
| 7833 | See Documentation/admin-guide/mm/transhuge.rst |
| 7834 | for more details. |
| 7835 | |
| 7836 | transparent_hugepage_shmem= [KNL] |
| 7837 | Format: [always|within_size|advise|never|deny|force] |
| 7838 | Can be used to control the hugepage allocation policy for |
| 7839 | the internal shmem mount. |
| 7840 | See Documentation/admin-guide/mm/transhuge.rst |
| 7841 | for more details. |
| 7842 | |
| 7843 | transparent_hugepage_tmpfs= [KNL] |
| 7844 | Format: [always|within_size|advise|never] |
| 7845 | Can be used to control the default hugepage allocation policy |
| 7846 | for the tmpfs mount. |
| 7847 | See Documentation/admin-guide/mm/transhuge.rst |
| 7848 | for more details. |
| 7849 | |
| 7850 | trusted.source= [KEYS] |
| 7851 | Format: <string> |
| 7852 | This parameter identifies the trust source as a backend |
| 7853 | for trusted keys implementation. Supported trust |
| 7854 | sources: |
| 7855 | - "tpm" |
| 7856 | - "tee" |
| 7857 | - "caam" |
| 7858 | - "dcp" |
| 7859 | - "pkwm" |
| 7860 | If not specified then it defaults to iterating through |
| 7861 | the trust source list starting with TPM and assigns the |
| 7862 | first trust source as a backend which is initialized |
| 7863 | successfully during iteration. |
| 7864 | |
| 7865 | trusted.debug= [KEYS] |
| 7866 | Format: <bool> |
| 7867 | Enable trusted keys debug traces at runtime when |
| 7868 | CONFIG_TRUSTED_KEYS_DEBUG=y. |
| 7869 | |
| 7870 | To make the traces visible after enabling the option, |
| 7871 | use trusted.dyndbg='+p' as needed. By convention, |
| 7872 | the subsystem uses pr_debug() for these traces. |
| 7873 | |
| 7874 | SAFETY: The traces can leak sensitive data, so be |
| 7875 | cautious before enabling this. They remain inactive |
| 7876 | unless this parameter is set this option to a true |
| 7877 | value. |
| 7878 | |
| 7879 | Default: false |
| 7880 | |
| 7881 | trusted.rng= [KEYS] |
| 7882 | Format: <string> |
| 7883 | The RNG used to generate key material for trusted keys. |
| 7884 | Can be one of: |
| 7885 | - "kernel" |
| 7886 | - the same value as trusted.source: "tpm" or "tee" |
| 7887 | - "default" |
| 7888 | If not specified, "default" is used. In this case, |
| 7889 | the RNG's choice is left to each individual trust source. |
| 7890 | |
| 7891 | trusted.dcp_use_otp_key |
| 7892 | This is intended to be used in combination with |
| 7893 | trusted.source=dcp and will select the DCP OTP key |
| 7894 | instead of the DCP UNIQUE key blob encryption. |
| 7895 | |
| 7896 | trusted.dcp_skip_zk_test |
| 7897 | This is intended to be used in combination with |
| 7898 | trusted.source=dcp and will disable the check if the |
| 7899 | blob key is all zeros. This is helpful for situations where |
| 7900 | having this key zero'ed is acceptable. E.g. in testing |
| 7901 | scenarios. |
| 7902 | |
| 7903 | tsa= [X86] Control mitigation for Transient Scheduler |
| 7904 | Attacks on AMD CPUs. Search the following in your |
| 7905 | favourite search engine for more details: |
| 7906 | |
| 7907 | "Technical guidance for mitigating transient scheduler |
| 7908 | attacks". |
| 7909 | |
| 7910 | off - disable the mitigation |
| 7911 | on - enable the mitigation (default) |
| 7912 | user - mitigate only user/kernel transitions |
| 7913 | vm - mitigate only guest/host transitions |
| 7914 | |
| 7915 | |
| 7916 | tsc= Disable clocksource stability checks for TSC. |
| 7917 | Format: <string> |
| 7918 | [x86] reliable: mark tsc clocksource as reliable, this |
| 7919 | disables clocksource verification at runtime, as well |
| 7920 | as the stability checks done at bootup. Used to enable |
| 7921 | high-resolution timer mode on older hardware, and in |
| 7922 | virtualized environment. |
| 7923 | [x86] noirqtime: Do not use TSC to do irq accounting. |
| 7924 | Used to run time disable IRQ_TIME_ACCOUNTING on any |
| 7925 | platforms where RDTSC is slow and this accounting |
| 7926 | can add overhead. |
| 7927 | [x86] unstable: mark the TSC clocksource as unstable, this |
| 7928 | marks the TSC unconditionally unstable at bootup and |
| 7929 | avoids any further wobbles once the TSC watchdog notices. |
| 7930 | [x86] nowatchdog: disable clocksource watchdog. Used |
| 7931 | in situations with strict latency requirements (where |
| 7932 | interruptions from clocksource watchdog are not |
| 7933 | acceptable). |
| 7934 | [x86] recalibrate: force recalibration against a HW timer |
| 7935 | (HPET or PM timer) on systems whose TSC frequency was |
| 7936 | obtained from HW or FW using either an MSR or CPUID(0x15). |
| 7937 | Warn if the difference is more than 500 ppm. |
| 7938 | [x86] watchdog: Enforce the clocksource watchdog on TSC |
| 7939 | |
| 7940 | tsc_early_khz= [X86,EARLY] Skip early TSC calibration and use the given |
| 7941 | value instead. Useful when the early TSC frequency discovery |
| 7942 | procedure is not reliable, such as on overclocked systems |
| 7943 | with CPUID.16h support and partial CPUID.15h support. |
| 7944 | Format: <unsigned int> |
| 7945 | |
| 7946 | tsx= [X86] Control Transactional Synchronization |
| 7947 | Extensions (TSX) feature in Intel processors that |
| 7948 | support TSX control. |
| 7949 | |
| 7950 | This parameter controls the TSX feature. The options are: |
| 7951 | |
| 7952 | on - Enable TSX on the system. Although there are |
| 7953 | mitigations for all known security vulnerabilities, |
| 7954 | TSX has been known to be an accelerator for |
| 7955 | several previous speculation-related CVEs, and |
| 7956 | so there may be unknown security risks associated |
| 7957 | with leaving it enabled. |
| 7958 | |
| 7959 | off - Disable TSX on the system. (Note that this |
| 7960 | option takes effect only on newer CPUs which are |
| 7961 | not vulnerable to MDS, i.e., have |
| 7962 | MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get |
| 7963 | the new IA32_TSX_CTRL MSR through a microcode |
| 7964 | update. This new MSR allows for the reliable |
| 7965 | deactivation of the TSX functionality.) |
| 7966 | |
| 7967 | auto - Disable TSX if X86_BUG_TAA is present, |
| 7968 | otherwise enable TSX on the system. |
| 7969 | |
| 7970 | Not specifying this option is equivalent to tsx=off. |
| 7971 | |
| 7972 | See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst |
| 7973 | for more details. |
| 7974 | |
| 7975 | tsx_async_abort= [X86,INTEL,EARLY] Control mitigation for the TSX Async |
| 7976 | Abort (TAA) vulnerability. |
| 7977 | |
| 7978 | Similar to Micro-architectural Data Sampling (MDS) |
| 7979 | certain CPUs that support Transactional |
| 7980 | Synchronization Extensions (TSX) are vulnerable to an |
| 7981 | exploit against CPU internal buffers which can forward |
| 7982 | information to a disclosure gadget under certain |
| 7983 | conditions. |
| 7984 | |
| 7985 | In vulnerable processors, the speculatively forwarded |
| 7986 | data can be used in a cache side channel attack, to |
| 7987 | access data to which the attacker does not have direct |
| 7988 | access. |
| 7989 | |
| 7990 | This parameter controls the TAA mitigation. The |
| 7991 | options are: |
| 7992 | |
| 7993 | full - Enable TAA mitigation on vulnerable CPUs |
| 7994 | if TSX is enabled. |
| 7995 | |
| 7996 | full,nosmt - Enable TAA mitigation and disable SMT on |
| 7997 | vulnerable CPUs. If TSX is disabled, SMT |
| 7998 | is not disabled because CPU is not |
| 7999 | vulnerable to cross-thread TAA attacks. |
| 8000 | off - Unconditionally disable TAA mitigation |
| 8001 | |
| 8002 | On MDS-affected machines, tsx_async_abort=off can be |
| 8003 | prevented by an active MDS mitigation as both vulnerabilities |
| 8004 | are mitigated with the same mechanism so in order to disable |
| 8005 | this mitigation, you need to specify mds=off too. |
| 8006 | |
| 8007 | Not specifying this option is equivalent to |
| 8008 | tsx_async_abort=full. On CPUs which are MDS affected |
| 8009 | and deploy MDS mitigation, TAA mitigation is not |
| 8010 | required and doesn't provide any additional |
| 8011 | mitigation. |
| 8012 | |
| 8013 | For details see: |
| 8014 | Documentation/admin-guide/hw-vuln/tsx_async_abort.rst |
| 8015 | |
| 8016 | turbografx.map[2|3]= [HW,JOY] |
| 8017 | TurboGraFX parallel port interface |
| 8018 | Format: |
| 8019 | <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> |
| 8020 | See also Documentation/input/devices/joystick-parport.rst |
| 8021 | |
| 8022 | udbg-immortal [PPC] When debugging early kernel crashes that |
| 8023 | happen after console_init() and before a proper |
| 8024 | console driver takes over, this boot options might |
| 8025 | help "seeing" what's going on. |
| 8026 | |
| 8027 | uhash_entries= [KNL,NET] |
| 8028 | Set number of hash buckets for UDP/UDP-Lite connections |
| 8029 | |
| 8030 | uhci-hcd.ignore_oc= |
| 8031 | [USB] Ignore overcurrent events (default N). |
| 8032 | Some badly-designed motherboards generate lots of |
| 8033 | bogus events, for ports that aren't wired to |
| 8034 | anything. Set this parameter to avoid log spamming. |
| 8035 | Note that genuine overcurrent events won't be |
| 8036 | reported either. |
| 8037 | |
| 8038 | unaligned_scalar_speed= |
| 8039 | [RISCV] |
| 8040 | Format: {slow | fast | unsupported} |
| 8041 | Allow skipping scalar unaligned access speed tests. This |
| 8042 | is useful for testing alternative code paths and to skip |
| 8043 | the tests in environments where they run too slowly. All |
| 8044 | CPUs must have the same scalar unaligned access speed. |
| 8045 | |
| 8046 | unaligned_vector_speed= |
| 8047 | [RISCV] |
| 8048 | Format: {slow | fast | unsupported} |
| 8049 | Allow skipping vector unaligned access speed tests. This |
| 8050 | is useful for testing alternative code paths and to skip |
| 8051 | the tests in environments where they run too slowly. All |
| 8052 | CPUs must have the same vector unaligned access speed. |
| 8053 | |
| 8054 | unknown_nmi_panic |
| 8055 | [X86] Cause panic on unknown NMI. |
| 8056 | |
| 8057 | unwind_debug [X86-64,EARLY] |
| 8058 | Enable unwinder debug output. This can be |
| 8059 | useful for debugging certain unwinder error |
| 8060 | conditions, including corrupt stacks and |
| 8061 | bad/missing unwinder metadata. |
| 8062 | |
| 8063 | usbcore.authorized_default= |
| 8064 | [USB] Default USB device authorization: |
| 8065 | (default -1 = authorized (same as 1), |
| 8066 | 0 = not authorized, 1 = authorized, 2 = authorized |
| 8067 | if device connected to internal port) |
| 8068 | |
| 8069 | usbcore.autosuspend= |
| 8070 | [USB] The autosuspend time delay (in seconds) used |
| 8071 | for newly-detected USB devices (default 2). This |
| 8072 | is the time required before an idle device will be |
| 8073 | autosuspended. Devices for which the delay is set |
| 8074 | to a negative value won't be autosuspended at all. |
| 8075 | |
| 8076 | usbcore.usbfs_snoop= |
| 8077 | [USB] Set to log all usbfs traffic (default 0 = off). |
| 8078 | |
| 8079 | usbcore.usbfs_snoop_max= |
| 8080 | [USB] Maximum number of bytes to snoop in each URB |
| 8081 | (default = 65536). |
| 8082 | |
| 8083 | usbcore.blinkenlights= |
| 8084 | [USB] Set to cycle leds on hubs (default 0 = off). |
| 8085 | |
| 8086 | usbcore.old_scheme_first= |
| 8087 | [USB] Start with the old device initialization |
| 8088 | scheme (default 0 = off). |
| 8089 | |
| 8090 | usbcore.usbfs_memory_mb= |
| 8091 | [USB] Memory limit (in MB) for buffers allocated by |
| 8092 | usbfs (default = 16, 0 = max = 2047). |
| 8093 | |
| 8094 | usbcore.use_both_schemes= |
| 8095 | [USB] Try the other device initialization scheme |
| 8096 | if the first one fails (default 1 = enabled). |
| 8097 | |
| 8098 | usbcore.initial_descriptor_timeout= |
| 8099 | [USB] Specifies timeout for the initial 64-byte |
| 8100 | USB_REQ_GET_DESCRIPTOR request in milliseconds |
| 8101 | (default 5000 = 5.0 seconds). |
| 8102 | |
| 8103 | usbcore.nousb [USB] Disable the USB subsystem |
| 8104 | |
| 8105 | usbcore.quirks= |
| 8106 | [USB] A list of quirk entries to augment the built-in |
| 8107 | usb core quirk list. List entries are separated by |
| 8108 | commas. Each entry has the form |
| 8109 | VendorID:ProductID:Flags. The IDs are 4-digit hex |
| 8110 | numbers and Flags is a set of letters. Each letter |
| 8111 | will change the built-in quirk; setting it if it is |
| 8112 | clear and clearing it if it is set. The letters have |
| 8113 | the following meanings: |
| 8114 | a = USB_QUIRK_STRING_FETCH_255 (string |
| 8115 | descriptors must not be fetched using |
| 8116 | a 255-byte read); |
| 8117 | b = USB_QUIRK_RESET_RESUME (device can't resume |
| 8118 | correctly so reset it instead); |
| 8119 | c = USB_QUIRK_NO_SET_INTF (device can't handle |
| 8120 | Set-Interface requests); |
| 8121 | d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't |
| 8122 | handle its Configuration or Interface |
| 8123 | strings); |
| 8124 | e = USB_QUIRK_RESET (device can't be reset |
| 8125 | (e.g morph devices), don't use reset); |
| 8126 | f = USB_QUIRK_HONOR_BNUMINTERFACES (device has |
| 8127 | more interface descriptions than the |
| 8128 | bNumInterfaces count, and can't handle |
| 8129 | talking to these interfaces); |
| 8130 | g = USB_QUIRK_DELAY_INIT (device needs a pause |
| 8131 | during initialization, after we read |
| 8132 | the device descriptor); |
| 8133 | h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For |
| 8134 | high speed and super speed interrupt |
| 8135 | endpoints, the USB 2.0 and USB 3.0 spec |
| 8136 | require the interval in microframes (1 |
| 8137 | microframe = 125 microseconds) to be |
| 8138 | calculated as interval = 2 ^ |
| 8139 | (bInterval-1). |
| 8140 | Devices with this quirk report their |
| 8141 | bInterval as the result of this |
| 8142 | calculation instead of the exponent |
| 8143 | variable used in the calculation); |
| 8144 | i = USB_QUIRK_DEVICE_QUALIFIER (device can't |
| 8145 | handle device_qualifier descriptor |
| 8146 | requests); |
| 8147 | j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device |
| 8148 | generates spurious wakeup, ignore |
| 8149 | remote wakeup capability); |
| 8150 | k = USB_QUIRK_NO_LPM (device can't handle Link |
| 8151 | Power Management); |
| 8152 | l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL |
| 8153 | (Device reports its bInterval as linear |
| 8154 | frames instead of the USB 2.0 |
| 8155 | calculation); |
| 8156 | m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs |
| 8157 | to be disconnected before suspend to |
| 8158 | prevent spurious wakeup); |
| 8159 | n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a |
| 8160 | pause after every control message); |
| 8161 | o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra |
| 8162 | delay after resetting its port); |
| 8163 | p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT |
| 8164 | (Reduce timeout of the SET_ADDRESS |
| 8165 | request from 5000 ms to 500 ms); |
| 8166 | q = USB_QUIRK_FORCE_ONE_CONFIG (Device |
| 8167 | claims zero configurations, |
| 8168 | forcing to 1); |
| 8169 | r = USB_QUIRK_WINDOWS_CONFIG_REQ_SIZE (Device |
| 8170 | fails during initialization when asked for |
| 8171 | 9-bytes configuration descriptor request. |
| 8172 | Ask for 255-bytes request instead to mirror |
| 8173 | Windows' behavior); |
| 8174 | Example: quirks=0781:5580:bk,0a5c:5834:gij |
| 8175 | |
| 8176 | usbhid.mousepoll= |
| 8177 | [USBHID] The interval which mice are to be polled at. |
| 8178 | |
| 8179 | usbhid.jspoll= |
| 8180 | [USBHID] The interval which joysticks are to be polled at. |
| 8181 | |
| 8182 | usbhid.kbpoll= |
| 8183 | [USBHID] The interval which keyboards are to be polled at. |
| 8184 | |
| 8185 | usb-storage.delay_use= |
| 8186 | [UMS] The delay in seconds before a new device is |
| 8187 | scanned for Logical Units (default 1). |
| 8188 | Optionally the delay in milliseconds if the value has |
| 8189 | suffix with "ms". |
| 8190 | Example: delay_use=2567ms |
| 8191 | |
| 8192 | usb-storage.quirks= |
| 8193 | [UMS] A list of quirks entries to supplement or |
| 8194 | override the built-in unusual_devs list. List |
| 8195 | entries are separated by commas. Each entry has |
| 8196 | the form VID:PID:Flags where VID and PID are Vendor |
| 8197 | and Product ID values (4-digit hex numbers) and |
| 8198 | Flags is a set of characters, each corresponding |
| 8199 | to a common usb-storage quirk flag as follows: |
| 8200 | a = SANE_SENSE (collect more than 18 bytes |
| 8201 | of sense data, not on uas); |
| 8202 | b = BAD_SENSE (don't collect more than 18 |
| 8203 | bytes of sense data, not on uas); |
| 8204 | c = FIX_CAPACITY (decrease the reported |
| 8205 | device capacity by one sector); |
| 8206 | d = NO_READ_DISC_INFO (don't use |
| 8207 | READ_DISC_INFO command, not on uas); |
| 8208 | e = NO_READ_CAPACITY_16 (don't use |
| 8209 | READ_CAPACITY_16 command); |
| 8210 | f = NO_REPORT_OPCODES (don't use report opcodes |
| 8211 | command, uas only); |
| 8212 | g = MAX_SECTORS_240 (don't transfer more than |
| 8213 | 240 sectors at a time, uas only); |
| 8214 | h = CAPACITY_HEURISTICS (decrease the |
| 8215 | reported device capacity by one |
| 8216 | sector if the number is odd); |
| 8217 | i = IGNORE_DEVICE (don't bind to this |
| 8218 | device); |
| 8219 | j = NO_REPORT_LUNS (don't use report luns |
| 8220 | command, uas only); |
| 8221 | k = NO_SAME (do not use WRITE_SAME, uas only) |
| 8222 | l = NOT_LOCKABLE (don't try to lock and |
| 8223 | unlock ejectable media, not on uas); |
| 8224 | m = MAX_SECTORS_64 (don't transfer more |
| 8225 | than 64 sectors = 32 KB at a time, |
| 8226 | not on uas); |
| 8227 | n = INITIAL_READ10 (force a retry of the |
| 8228 | initial READ(10) command, not on uas); |
| 8229 | o = CAPACITY_OK (accept the capacity |
| 8230 | reported by the device, not on uas); |
| 8231 | p = WRITE_CACHE (the device cache is ON |
| 8232 | by default, not on uas); |
| 8233 | r = IGNORE_RESIDUE (the device reports |
| 8234 | bogus residue values, not on uas); |
| 8235 | s = SINGLE_LUN (the device has only one |
| 8236 | Logical Unit); |
| 8237 | t = NO_ATA_1X (don't allow ATA(12) and ATA(16) |
| 8238 | commands, uas only); |
| 8239 | u = IGNORE_UAS (don't bind to the uas driver); |
| 8240 | w = NO_WP_DETECT (don't test whether the |
| 8241 | medium is write-protected). |
| 8242 | y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE |
| 8243 | even if the device claims no cache, |
| 8244 | not on uas) |
| 8245 | Example: quirks=0419:aaf5:rl,0421:0433:rc |
| 8246 | |
| 8247 | user_debug= [KNL,ARM] |
| 8248 | Format: <int> |
| 8249 | See arch/arm/Kconfig.debug help text. |
| 8250 | 1 - undefined instruction events |
| 8251 | 2 - system calls |
| 8252 | 4 - invalid data aborts |
| 8253 | 8 - SIGSEGV faults |
| 8254 | 16 - SIGBUS faults |
| 8255 | Example: user_debug=31 |
| 8256 | |
| 8257 | vdso= [X86,SH,SPARC] |
| 8258 | On X86_32, this is an alias for vdso32=. Otherwise: |
| 8259 | |
| 8260 | vdso=1: enable VDSO (the default) |
| 8261 | vdso=0: disable VDSO mapping |
| 8262 | |
| 8263 | vdso32= [X86] Control the 32-bit vDSO |
| 8264 | vdso32=1: enable 32-bit VDSO |
| 8265 | vdso32=0 or vdso32=2: disable 32-bit VDSO |
| 8266 | |
| 8267 | See the help text for CONFIG_COMPAT_VDSO for more |
| 8268 | details. If CONFIG_COMPAT_VDSO is set, the default is |
| 8269 | vdso32=0; otherwise, the default is vdso32=1. |
| 8270 | |
| 8271 | For compatibility with older kernels, vdso32=2 is an |
| 8272 | alias for vdso32=0. |
| 8273 | |
| 8274 | Try vdso32=0 if you encounter an error that says: |
| 8275 | dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! |
| 8276 | |
| 8277 | video= [FB,EARLY] Frame buffer configuration |
| 8278 | See Documentation/fb/modedb.rst. |
| 8279 | |
| 8280 | video.brightness_switch_enabled= [ACPI] |
| 8281 | Format: [0|1] |
| 8282 | If set to 1, on receiving an ACPI notify event |
| 8283 | generated by hotkey, video driver will adjust brightness |
| 8284 | level and then send out the event to user space through |
| 8285 | the allocated input device. If set to 0, video driver |
| 8286 | will only send out the event without touching backlight |
| 8287 | brightness level. |
| 8288 | default: 1 |
| 8289 | |
| 8290 | virtio_mmio.device= |
| 8291 | [VMMIO] Memory mapped virtio (platform) device. |
| 8292 | |
| 8293 | <size>@<baseaddr>:<irq>[:<id>] |
| 8294 | where: |
| 8295 | <size> := size (can use standard suffixes |
| 8296 | like K, M and G) |
| 8297 | <baseaddr> := physical base address |
| 8298 | <irq> := interrupt number (as passed to |
| 8299 | request_irq()) |
| 8300 | <id> := (optional) platform device id |
| 8301 | example: |
| 8302 | virtio_mmio.device=1K@0x100b0000:48:7 |
| 8303 | |
| 8304 | Can be used multiple times for multiple devices. |
| 8305 | |
| 8306 | vga= [BOOT,X86-32] Select a particular video mode |
| 8307 | See Documentation/arch/x86/boot.rst and |
| 8308 | Documentation/admin-guide/svga.rst. |
| 8309 | Use vga=ask for menu. |
| 8310 | This is actually a boot loader parameter; the value is |
| 8311 | passed to the kernel using a special protocol. |
| 8312 | |
| 8313 | vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. |
| 8314 | May slow down system boot speed, especially when |
| 8315 | enabled on systems with a large amount of memory. |
| 8316 | All options are enabled by default, and this |
| 8317 | interface is meant to allow for selectively |
| 8318 | enabling or disabling specific virtual memory |
| 8319 | debugging features. |
| 8320 | |
| 8321 | Available options are: |
| 8322 | P Enable page structure init time poisoning |
| 8323 | - Disable all of the above options |
| 8324 | |
| 8325 | vmalloc=nn[KMG] [KNL,BOOT,EARLY] Forces the vmalloc area to have an |
| 8326 | exact size of <nn>. This can be used to increase |
| 8327 | the minimum size (128MB on x86, arm32 platforms). |
| 8328 | It can also be used to decrease the size and leave more room |
| 8329 | for directly mapped kernel RAM. Note that this parameter does |
| 8330 | not exist on many other platforms (including arm64, alpha, |
| 8331 | loongarch, arc, csky, hexagon, microblaze, mips, nios2, openrisc, |
| 8332 | parisc, m64k, powerpc, riscv, sh, um, xtensa, s390, sparc). |
| 8333 | |
| 8334 | vmcp_cma=nn[MG] [KNL,S390,EARLY] |
| 8335 | Sets the memory size reserved for contiguous memory |
| 8336 | allocations for the vmcp device driver. |
| 8337 | |
| 8338 | vmhalt= [KNL,S390] Perform z/VM CP command after system halt. |
| 8339 | Format: <command> |
| 8340 | |
| 8341 | vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. |
| 8342 | Format: <command> |
| 8343 | |
| 8344 | vmpoff= [KNL,S390] Perform z/VM CP command after power off. |
| 8345 | Format: <command> |
| 8346 | |
| 8347 | vmscape= [X86] Controls mitigation for VMscape attacks. |
| 8348 | VMscape attacks can leak information from a userspace |
| 8349 | hypervisor to a guest via speculative side-channels. |
| 8350 | |
| 8351 | off - disable the mitigation |
| 8352 | ibpb - use Indirect Branch Prediction Barrier |
| 8353 | (IBPB) mitigation (default) |
| 8354 | force - force vulnerability detection even on |
| 8355 | unaffected processors |
| 8356 | |
| 8357 | vsyscall= [X86-64,EARLY] |
| 8358 | Controls the behavior of vsyscalls (i.e. calls to |
| 8359 | fixed addresses of 0xffffffffff600x00 from legacy |
| 8360 | code). Most statically-linked binaries and older |
| 8361 | versions of glibc use these calls. Because these |
| 8362 | functions are at fixed addresses, they make nice |
| 8363 | targets for exploits that can control RIP. |
| 8364 | |
| 8365 | emulate Vsyscalls turn into traps and are emulated |
| 8366 | reasonably safely. The vsyscall page is |
| 8367 | readable. This disables the Linear |
| 8368 | Address Space Separation (LASS) security |
| 8369 | feature and makes the system less secure. |
| 8370 | |
| 8371 | xonly [default] Vsyscalls turn into traps and are |
| 8372 | emulated reasonably safely. The vsyscall |
| 8373 | page is not readable. |
| 8374 | |
| 8375 | none Vsyscalls don't work at all. This makes |
| 8376 | them quite hard to use for exploits but |
| 8377 | might break your system. |
| 8378 | |
| 8379 | vt.color= [VT] Default text color. |
| 8380 | Format: 0xYX, X = foreground, Y = background. |
| 8381 | Default: 0x07 = light gray on black. |
| 8382 | |
| 8383 | vt.cur_default= [VT] Default cursor shape. |
| 8384 | Format: 0xCCBBAA, where AA, BB, and CC are the same as |
| 8385 | the parameters of the <Esc>[?A;B;Cc escape sequence; |
| 8386 | see vga-softcursor.rst. Default: 2 = underline. |
| 8387 | |
| 8388 | vt.default_blu= [VT] |
| 8389 | Format: <blue0>,<blue1>,<blue2>,...,<blue15> |
| 8390 | Change the default blue palette of the console. |
| 8391 | This is a 16-member array composed of values |
| 8392 | ranging from 0-255. |
| 8393 | |
| 8394 | vt.default_grn= [VT] |
| 8395 | Format: <green0>,<green1>,<green2>,...,<green15> |
| 8396 | Change the default green palette of the console. |
| 8397 | This is a 16-member array composed of values |
| 8398 | ranging from 0-255. |
| 8399 | |
| 8400 | vt.default_red= [VT] |
| 8401 | Format: <red0>,<red1>,<red2>,...,<red15> |
| 8402 | Change the default red palette of the console. |
| 8403 | This is a 16-member array composed of values |
| 8404 | ranging from 0-255. |
| 8405 | |
| 8406 | vt.default_utf8= |
| 8407 | [VT] |
| 8408 | Format=<0|1> |
| 8409 | Set system-wide default UTF-8 mode for all tty's. |
| 8410 | Default is 1, i.e. UTF-8 mode is enabled for all |
| 8411 | newly opened terminals. |
| 8412 | |
| 8413 | vt.global_cursor_default= |
| 8414 | [VT] |
| 8415 | Format=<-1|0|1> |
| 8416 | Set system-wide default for whether a cursor |
| 8417 | is shown on new VTs. Default is -1, |
| 8418 | i.e. cursors will be created by default unless |
| 8419 | overridden by individual drivers. 0 will hide |
| 8420 | cursors, 1 will display them. |
| 8421 | |
| 8422 | vt.italic= [VT] Default color for italic text; 0-15. |
| 8423 | Default: 2 = green. |
| 8424 | |
| 8425 | vt.underline= [VT] Default color for underlined text; 0-15. |
| 8426 | Default: 3 = cyan. |
| 8427 | |
| 8428 | watchdog timers [HW,WDT] For information on watchdog timers, |
| 8429 | see Documentation/watchdog/watchdog-parameters.rst |
| 8430 | or other driver-specific files in the |
| 8431 | Documentation/watchdog/ directory. |
| 8432 | |
| 8433 | watchdog_thresh= |
| 8434 | [KNL] |
| 8435 | Set the hard lockup detector stall duration |
| 8436 | threshold in seconds. The soft lockup detector |
| 8437 | threshold is set to twice the value. A value of 0 |
| 8438 | disables both lockup detectors. Default is 10 |
| 8439 | seconds. |
| 8440 | |
| 8441 | workqueue.unbound_cpus= |
| 8442 | [KNL,SMP] Specify to constrain one or some CPUs |
| 8443 | to use in unbound workqueues. |
| 8444 | Format: <cpu-list> |
| 8445 | By default, all online CPUs are available for |
| 8446 | unbound workqueues. |
| 8447 | |
| 8448 | workqueue.watchdog_thresh= |
| 8449 | If CONFIG_WQ_WATCHDOG is configured, workqueue can |
| 8450 | warn stall conditions and dump internal state to |
| 8451 | help debugging. 0 disables workqueue stall |
| 8452 | detection; otherwise, it's the stall threshold |
| 8453 | duration in seconds. The default value is 30 and |
| 8454 | it can be updated at runtime by writing to the |
| 8455 | corresponding sysfs file. |
| 8456 | |
| 8457 | workqueue.panic_on_stall=<uint> |
| 8458 | Panic when workqueue stall is detected by |
| 8459 | CONFIG_WQ_WATCHDOG. It sets the number times of the |
| 8460 | stall to trigger panic. |
| 8461 | |
| 8462 | The default is set by CONFIG_BOOTPARAM_WQ_STALL_PANIC, |
| 8463 | which is 0 (disabled) if not configured. |
| 8464 | |
| 8465 | workqueue.panic_on_stall_time=<uint> |
| 8466 | Panic when a workqueue stall has been continuous for |
| 8467 | the specified number of seconds. Unlike panic_on_stall |
| 8468 | which counts accumulated stall events, this triggers |
| 8469 | based on the duration of a single continuous stall. |
| 8470 | |
| 8471 | The default is 0, which disables the time-based panic. |
| 8472 | |
| 8473 | workqueue.cpu_intensive_thresh_us= |
| 8474 | Per-cpu work items which run for longer than this |
| 8475 | threshold are automatically considered CPU intensive |
| 8476 | and excluded from concurrency management to prevent |
| 8477 | them from noticeably delaying other per-cpu work |
| 8478 | items. Default is 10000 (10ms). |
| 8479 | |
| 8480 | If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel |
| 8481 | will report the work functions which violate this |
| 8482 | threshold repeatedly. They are likely good |
| 8483 | candidates for using WQ_UNBOUND workqueues instead. |
| 8484 | |
| 8485 | workqueue.cpu_intensive_warning_thresh=<uint> |
| 8486 | If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel |
| 8487 | will report the work functions which violate the |
| 8488 | intensive_threshold_us repeatedly. In order to prevent |
| 8489 | spurious warnings, start printing only after a work |
| 8490 | function has violated this threshold number of times. |
| 8491 | |
| 8492 | The default is 4 times. 0 disables the warning. |
| 8493 | |
| 8494 | workqueue.power_efficient |
| 8495 | Per-cpu workqueues are generally preferred because |
| 8496 | they show better performance thanks to cache |
| 8497 | locality; unfortunately, per-cpu workqueues tend to |
| 8498 | be more power hungry than unbound workqueues. |
| 8499 | |
| 8500 | Enabling this makes the per-cpu workqueues which |
| 8501 | were observed to contribute significantly to power |
| 8502 | consumption unbound, leading to measurably lower |
| 8503 | power usage at the cost of small performance |
| 8504 | overhead. |
| 8505 | |
| 8506 | The default value of this parameter is determined by |
| 8507 | the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. |
| 8508 | |
| 8509 | workqueue.default_affinity_scope= |
| 8510 | Select the default affinity scope to use for unbound |
| 8511 | workqueues. Can be one of "cpu", "smt", "cache", |
| 8512 | "cache_shard", "numa" and "system". Default is |
| 8513 | "cache_shard". For more |
| 8514 | information, see the Affinity Scopes section in |
| 8515 | Documentation/core-api/workqueue.rst. |
| 8516 | |
| 8517 | This can be changed after boot by writing to the |
| 8518 | matching /sys/module/workqueue/parameters file. All |
| 8519 | workqueues with the "default" affinity scope will be |
| 8520 | updated accordingly. |
| 8521 | |
| 8522 | workqueue.debug_force_rr_cpu |
| 8523 | Workqueue used to implicitly guarantee that work |
| 8524 | items queued without explicit CPU specified are put |
| 8525 | on the local CPU. This guarantee is no longer true |
| 8526 | and while local CPU is still preferred work items |
| 8527 | may be put on foreign CPUs. This debug option |
| 8528 | forces round-robin CPU selection to flush out |
| 8529 | usages which depend on the now broken guarantee. |
| 8530 | When enabled, memory and cache locality will be |
| 8531 | impacted. |
| 8532 | |
| 8533 | writecombine= [LOONGARCH,EARLY] Control the MAT (Memory Access |
| 8534 | Type) of ioremap_wc(). |
| 8535 | |
| 8536 | on - Enable writecombine, use WUC for ioremap_wc() |
| 8537 | off - Disable writecombine, use SUC for ioremap_wc() |
| 8538 | |
| 8539 | x2apic_phys [X86-64,APIC,EARLY] Use x2apic physical mode instead of |
| 8540 | default x2apic cluster mode on platforms |
| 8541 | supporting x2apic. |
| 8542 | |
| 8543 | xen_512gb_limit [KNL,X86-64,XEN] |
| 8544 | Restricts the kernel running paravirtualized under Xen |
| 8545 | to use only up to 512 GB of RAM. The reason to do so is |
| 8546 | crash analysis tools and Xen tools for doing domain |
| 8547 | save/restore/migration must be enabled to handle larger |
| 8548 | domains. |
| 8549 | |
| 8550 | xen_console_io [XEN,EARLY] |
| 8551 | Boolean option to enable/disable the usage of the Xen |
| 8552 | console_io hypercalls to read and write to the console. |
| 8553 | Mostly useful for debugging and development. |
| 8554 | |
| 8555 | xen_emul_unplug= [HW,X86,XEN,EARLY] |
| 8556 | Unplug Xen emulated devices |
| 8557 | Format: [unplug0,][unplug1] |
| 8558 | ide-disks -- unplug primary master IDE devices |
| 8559 | aux-ide-disks -- unplug non-primary-master IDE devices |
| 8560 | nics -- unplug network devices |
| 8561 | all -- unplug all emulated devices (NICs and IDE disks) |
| 8562 | unnecessary -- unplugging emulated devices is |
| 8563 | unnecessary even if the host did not respond to |
| 8564 | the unplug protocol |
| 8565 | never -- do not unplug even if version check succeeds |
| 8566 | |
| 8567 | xen_legacy_crash [X86,XEN,EARLY] |
| 8568 | Crash from Xen panic notifier, without executing late |
| 8569 | panic() code such as dumping handler. |
| 8570 | |
| 8571 | xen_mc_debug [X86,XEN,EARLY] |
| 8572 | Enable multicall debugging when running as a Xen PV guest. |
| 8573 | Enabling this feature will reduce performance a little |
| 8574 | bit, so it should only be enabled for obtaining extended |
| 8575 | debug data in case of multicall errors. |
| 8576 | |
| 8577 | xen_msr_safe= [X86,XEN,EARLY] |
| 8578 | Format: <bool> |
| 8579 | Select whether to always use non-faulting (safe) MSR |
| 8580 | access functions when running as Xen PV guest. The |
| 8581 | default value is controlled by CONFIG_XEN_PV_MSR_SAFE. |
| 8582 | |
| 8583 | xen_nopv [X86] |
| 8584 | Disables the PV optimizations forcing the HVM guest to |
| 8585 | run as generic HVM guest with no PV drivers. |
| 8586 | This option is obsoleted by the "nopv" option, which |
| 8587 | has equivalent effect for XEN platform. |
| 8588 | |
| 8589 | xen_no_vector_callback |
| 8590 | [KNL,X86,XEN,EARLY] Disable the vector callback for Xen |
| 8591 | event channel interrupts. |
| 8592 | |
| 8593 | xen_scrub_pages= [XEN] |
| 8594 | Boolean option to control scrubbing pages before giving them back |
| 8595 | to Xen, for use by other domains. Can be also changed at runtime |
| 8596 | with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. |
| 8597 | Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. |
| 8598 | |
| 8599 | xen_timer_slop= [X86-64,XEN,EARLY] |
| 8600 | Set the timer slop (in nanoseconds) for the virtual Xen |
| 8601 | timers (default is 100000). This adjusts the minimum |
| 8602 | delta of virtualized Xen timers, where lower values |
| 8603 | improve timer resolution at the expense of processing |
| 8604 | more timer interrupts. |
| 8605 | |
| 8606 | xen.balloon_boot_timeout= [XEN] |
| 8607 | The time (in seconds) to wait before giving up to boot |
| 8608 | in case initial ballooning fails to free enough memory. |
| 8609 | Applies only when running as HVM or PVH guest and |
| 8610 | started with less memory configured than allowed at |
| 8611 | max. Default is 180. |
| 8612 | |
| 8613 | xen.event_eoi_delay= [XEN] |
| 8614 | How long to delay EOI handling in case of event |
| 8615 | storms (jiffies). Default is 10. |
| 8616 | |
| 8617 | xen.event_loop_timeout= [XEN] |
| 8618 | After which time (jiffies) the event handling loop |
| 8619 | should start to delay EOI handling. Default is 2. |
| 8620 | |
| 8621 | xen.fifo_events= [XEN] |
| 8622 | Boolean parameter to disable using fifo event handling |
| 8623 | even if available. Normally fifo event handling is |
| 8624 | preferred over the 2-level event handling, as it is |
| 8625 | fairer and the number of possible event channels is |
| 8626 | much higher. Default is on (use fifo events). |
| 8627 | |
| 8628 | xirc2ps_cs= [NET,PCMCIA] |
| 8629 | Format: |
| 8630 | <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] |
| 8631 | |
| 8632 | xive= [PPC] |
| 8633 | By default on POWER9 and above, the kernel will |
| 8634 | natively use the XIVE interrupt controller. This option |
| 8635 | allows the fallback firmware mode to be used: |
| 8636 | |
| 8637 | off Fallback to firmware control of XIVE interrupt |
| 8638 | controller on both pseries and powernv |
| 8639 | platforms. Only useful on POWER9 and above. |
| 8640 | |
| 8641 | xive.store-eoi=off [PPC] |
| 8642 | By default on POWER10 and above, the kernel will use |
| 8643 | stores for EOI handling when the XIVE interrupt mode |
| 8644 | is active. This option allows the XIVE driver to use |
| 8645 | loads instead, as on POWER9. |
| 8646 | |
| 8647 | xhci-hcd.quirks [USB,KNL] |
| 8648 | A hex value specifying bitmask with supplemental xhci |
| 8649 | host controller quirks. Meaning of each bit can be |
| 8650 | consulted in header drivers/usb/host/xhci.h. |
| 8651 | |
| 8652 | xmon [PPC,EARLY] |
| 8653 | Format: { early | on | rw | ro | off } |
| 8654 | Controls if xmon debugger is enabled. Default is off. |
| 8655 | Passing only "xmon" is equivalent to "xmon=early". |
| 8656 | early Call xmon as early as possible on boot; xmon |
| 8657 | debugger is called from setup_arch(). |
| 8658 | on xmon debugger hooks will be installed so xmon |
| 8659 | is only called on a kernel crash. Default mode, |
| 8660 | i.e. either "ro" or "rw" mode, is controlled |
| 8661 | with CONFIG_XMON_DEFAULT_RO_MODE. |
| 8662 | rw xmon debugger hooks will be installed so xmon |
| 8663 | is called only on a kernel crash, mode is write, |
| 8664 | meaning SPR registers, memory and, other data |
| 8665 | can be written using xmon commands. |
| 8666 | ro same as "rw" option above but SPR registers, |
| 8667 | memory, and other data can't be written using |
| 8668 | xmon commands. |
| 8669 | off xmon is disabled. |