| 1 | Overview |
| 2 | ======== |
| 3 | |
| 4 | For general security related questions of perf_event_open() syscall usage, |
| 5 | performance monitoring and observability operations by Perf see here: |
| 6 | https://www.kernel.org/doc/html/latest/admin-guide/perf-security.html |
| 7 | |
| 8 | Enabling LSM based mandatory access control (MAC) to perf_event_open() syscall |
| 9 | ============================================================================== |
| 10 | |
| 11 | LSM hooks for mandatory access control for perf_event_open() syscall can be |
| 12 | used starting from Linux v5.3. Below are the steps to extend Fedora (v31) with |
| 13 | Targeted policy with perf_event_open() access control capabilities: |
| 14 | |
| 15 | 1. Download selinux-policy SRPM package (e.g. selinux-policy-3.14.4-48.fc31.src.rpm on FC31) |
| 16 | and install it so rpmbuild directory would exist in the current working directory: |
| 17 | |
| 18 | # rpm -Uhv selinux-policy-3.14.4-48.fc31.src.rpm |
| 19 | |
| 20 | 2. Get into rpmbuild/SPECS directory and unpack the source code: |
| 21 | |
| 22 | # rpmbuild -bp selinux-policy.spec |
| 23 | |
| 24 | 3. Place patch below at rpmbuild/BUILD/selinux-policy-b86eaaf4dbcf2d51dd4432df7185c0eaf3cbcc02 |
| 25 | directory and apply it: |
| 26 | |
| 27 | # patch -p1 < selinux-policy-perf-events-perfmon.patch |
| 28 | patching file policy/flask/access_vectors |
| 29 | patching file policy/flask/security_classes |
| 30 | # cat selinux-policy-perf-events-perfmon.patch |
| 31 | diff -Nura a/policy/flask/access_vectors b/policy/flask/access_vectors |
| 32 | --- a/policy/flask/access_vectors 2020-02-04 18:19:53.000000000 +0300 |
| 33 | +++ b/policy/flask/access_vectors 2020-02-28 23:37:25.000000000 +0300 |
| 34 | @@ -174,6 +174,7 @@ |
| 35 | wake_alarm |
| 36 | block_suspend |
| 37 | audit_read |
| 38 | + perfmon |
| 39 | } |
| 40 | |
| 41 | # |
| 42 | @@ -1099,3 +1100,15 @@ |
| 43 | |
| 44 | class xdp_socket |
| 45 | inherits socket |
| 46 | + |
| 47 | +class perf_event |
| 48 | +{ |
| 49 | + open |
| 50 | + cpu |
| 51 | + kernel |
| 52 | + tracepoint |
| 53 | + read |
| 54 | + write |
| 55 | +} |
| 56 | + |
| 57 | + |
| 58 | diff -Nura a/policy/flask/security_classes b/policy/flask/security_classes |
| 59 | --- a/policy/flask/security_classes 2020-02-04 18:19:53.000000000 +0300 |
| 60 | +++ b/policy/flask/security_classes 2020-02-28 21:35:17.000000000 +0300 |
| 61 | @@ -200,4 +200,6 @@ |
| 62 | |
| 63 | class xdp_socket |
| 64 | |
| 65 | +class perf_event |
| 66 | + |
| 67 | # FLASK |
| 68 | |
| 69 | 4. Get into rpmbuild/SPECS directory and build policy packages from patched sources: |
| 70 | |
| 71 | # rpmbuild --noclean --noprep -ba selinux-policy.spec |
| 72 | |
| 73 | so you have this: |
| 74 | |
| 75 | # ls -alh rpmbuild/RPMS/noarch/ |
| 76 | total 33M |
| 77 | drwxr-xr-x. 2 root root 4.0K Mar 20 12:16 . |
| 78 | drwxr-xr-x. 3 root root 4.0K Mar 20 12:16 .. |
| 79 | -rw-r--r--. 1 root root 112K Mar 20 12:16 selinux-policy-3.14.4-48.fc31.noarch.rpm |
| 80 | -rw-r--r--. 1 root root 1.2M Mar 20 12:17 selinux-policy-devel-3.14.4-48.fc31.noarch.rpm |
| 81 | -rw-r--r--. 1 root root 2.3M Mar 20 12:17 selinux-policy-doc-3.14.4-48.fc31.noarch.rpm |
| 82 | -rw-r--r--. 1 root root 12M Mar 20 12:17 selinux-policy-minimum-3.14.4-48.fc31.noarch.rpm |
| 83 | -rw-r--r--. 1 root root 4.5M Mar 20 12:16 selinux-policy-mls-3.14.4-48.fc31.noarch.rpm |
| 84 | -rw-r--r--. 1 root root 111K Mar 20 12:16 selinux-policy-sandbox-3.14.4-48.fc31.noarch.rpm |
| 85 | -rw-r--r--. 1 root root 14M Mar 20 12:17 selinux-policy-targeted-3.14.4-48.fc31.noarch.rpm |
| 86 | |
| 87 | 5. Install SELinux packages from Fedora repo, if not already done so, and |
| 88 | update with the patched rpms above: |
| 89 | |
| 90 | # rpm -Uhv rpmbuild/RPMS/noarch/selinux-policy-* |
| 91 | |
| 92 | 6. Enable SELinux Permissive mode for Targeted policy, if not already done so: |
| 93 | |
| 94 | # cat /etc/selinux/config |
| 95 | |
| 96 | # This file controls the state of SELinux on the system. |
| 97 | # SELINUX= can take one of these three values: |
| 98 | # enforcing - SELinux security policy is enforced. |
| 99 | # permissive - SELinux prints warnings instead of enforcing. |
| 100 | # disabled - No SELinux policy is loaded. |
| 101 | SELINUX=permissive |
| 102 | # SELINUXTYPE= can take one of these three values: |
| 103 | # targeted - Targeted processes are protected, |
| 104 | # minimum - Modification of targeted policy. Only selected processes are protected. |
| 105 | # mls - Multi Level Security protection. |
| 106 | SELINUXTYPE=targeted |
| 107 | |
| 108 | 7. Enable filesystem SELinux labeling at the next reboot: |
| 109 | |
| 110 | # touch /.autorelabel |
| 111 | |
| 112 | 8. Reboot machine and it will label filesystems and load Targeted policy into the kernel; |
| 113 | |
| 114 | 9. Login and check that dmesg output doesn't mention that perf_event class is unknown to SELinux subsystem; |
| 115 | |
| 116 | 10. Check that SELinux is enabled and in Permissive mode |
| 117 | |
| 118 | # getenforce |
| 119 | Permissive |
| 120 | |
| 121 | 11. Turn SELinux into Enforcing mode: |
| 122 | |
| 123 | # setenforce 1 |
| 124 | # getenforce |
| 125 | Enforcing |
| 126 | |
| 127 | Opening access to perf_event_open() syscall on Fedora with SELinux |
| 128 | ================================================================== |
| 129 | |
| 130 | Access to performance monitoring and observability operations by Perf |
| 131 | can be limited for superuser or CAP_PERFMON or CAP_SYS_ADMIN privileged |
| 132 | processes. MAC policy settings (e.g. SELinux) can be loaded into the kernel |
| 133 | and prevent unauthorized access to perf_event_open() syscall. In such case |
| 134 | Perf tool provides a message similar to the one below: |
| 135 | |
| 136 | # perf stat |
| 137 | Error: |
| 138 | Access to performance monitoring and observability operations is limited. |
| 139 | Enforced MAC policy settings (SELinux) can limit access to performance |
| 140 | monitoring and observability operations. Inspect system audit records for |
| 141 | more perf_event access control information and adjusting the policy. |
| 142 | Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open |
| 143 | access to performance monitoring and observability operations for users |
| 144 | without CAP_PERFMON or CAP_SYS_ADMIN Linux capability. |
| 145 | perf_event_paranoid setting is -1: |
| 146 | -1: Allow use of (almost) all events by all users |
| 147 | Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK |
| 148 | >= 0: Disallow raw and ftrace function tracepoint access |
| 149 | >= 1: Disallow CPU event access |
| 150 | >= 2: Disallow kernel profiling |
| 151 | To make the adjusted perf_event_paranoid setting permanent preserve it |
| 152 | in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>) |
| 153 | |
| 154 | To make sure that access is limited by MAC policy settings inspect system |
| 155 | audit records using journalctl command or /var/log/audit/audit.log so the |
| 156 | output would contain AVC denied records related to perf_event: |
| 157 | |
| 158 | # journalctl --reverse --no-pager | grep perf_event |
| 159 | |
| 160 | python3[1318099]: SELinux is preventing perf from open access on the perf_event labeled unconfined_t. |
| 161 | If you believe that perf should be allowed open access on perf_event labeled unconfined_t by default. |
| 162 | setroubleshoot[1318099]: SELinux is preventing perf from open access on the perf_event labeled unconfined_t. For complete SELinux messages run: sealert -l 4595ce5b-e58f-462c-9d86-3bc2074935de |
| 163 | audit[1318098]: AVC avc: denied { open } for pid=1318098 comm="perf" scontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 tcontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 tclass=perf_event permissive=0 |
| 164 | |
| 165 | In order to open access to perf_event_open() syscall MAC policy settings can |
| 166 | require to be extended. On SELinux system this can be done by loading a special |
| 167 | policy module extending base policy settings. Perf related policy module can |
| 168 | be generated using the system audit records about blocking perf_event access. |
| 169 | Run the command below to generate my-perf.te policy extension file with |
| 170 | perf_event related rules: |
| 171 | |
| 172 | # ausearch -c 'perf' --raw | audit2allow -M my-perf && cat my-perf.te |
| 173 | |
| 174 | module my-perf 1.0; |
| 175 | |
| 176 | require { |
| 177 | type unconfined_t; |
| 178 | class perf_event { cpu kernel open read tracepoint write }; |
| 179 | } |
| 180 | |
| 181 | #============= unconfined_t ============== |
| 182 | allow unconfined_t self:perf_event { cpu kernel open read tracepoint write }; |
| 183 | |
| 184 | Now compile, pack and load my-perf.pp extension module into the kernel: |
| 185 | |
| 186 | # checkmodule -M -m -o my-perf.mod my-perf.te |
| 187 | # semodule_package -o my-perf.pp -m my-perf.mod |
| 188 | # semodule -X 300 -i my-perf.pp |
| 189 | |
| 190 | After all those taken steps above access to perf_event_open() syscall should |
| 191 | now be allowed by the policy settings. Check access running Perf like this: |
| 192 | |
| 193 | # perf stat |
| 194 | ^C |
| 195 | Performance counter stats for 'system wide': |
| 196 | |
| 197 | 36,387.41 msec cpu-clock # 7.999 CPUs utilized |
| 198 | 2,629 context-switches # 0.072 K/sec |
| 199 | 57 cpu-migrations # 0.002 K/sec |
| 200 | 1 page-faults # 0.000 K/sec |
| 201 | 263,721,559 cycles # 0.007 GHz |
| 202 | 175,746,713 instructions # 0.67 insn per cycle |
| 203 | 19,628,798 branches # 0.539 M/sec |
| 204 | 1,259,201 branch-misses # 6.42% of all branches |
| 205 | |
| 206 | 4.549061439 seconds time elapsed |
| 207 | |
| 208 | The generated perf-event.pp related policy extension module can be removed |
| 209 | from the kernel using this command: |
| 210 | |
| 211 | # semodule -X 300 -r my-perf |
| 212 | |
| 213 | Alternatively the module can be temporarily disabled and enabled back using |
| 214 | these two commands: |
| 215 | |
| 216 | # semodule -d my-perf |
| 217 | # semodule -e my-perf |
| 218 | |
| 219 | If something went wrong |
| 220 | ======================= |
| 221 | |
| 222 | To turn SELinux into Permissive mode: |
| 223 | # setenforce 0 |
| 224 | |
| 225 | To fully disable SELinux during kernel boot [3] set kernel command line parameter selinux=0 |
| 226 | |
| 227 | To remove SELinux labeling from local filesystems: |
| 228 | # find / -mount -print0 | xargs -0 setfattr -h -x security.selinux |
| 229 | |
| 230 | To fully turn SELinux off a machine set SELINUX=disabled at /etc/selinux/config file and reboot; |
| 231 | |
| 232 | Links |
| 233 | ===== |
| 234 | |
| 235 | [1] https://download-ib01.fedoraproject.org/pub/fedora/linux/updates/31/Everything/SRPMS/Packages/s/selinux-policy-3.14.4-49.fc31.src.rpm |
| 236 | [2] https://docs.fedoraproject.org/en-US/Fedora/11/html/Security-Enhanced_Linux/sect-Security-Enhanced_Linux-Working_with_SELinux-Enabling_and_Disabling_SELinux.html |
| 237 | [3] https://danwalsh.livejournal.com/10972.html |