Extend Perf tool with the check of /sys/fs/selinux/enforce value and notify
in case access to perf_event_open() syscall is restricted by the enforced
SELinux policy settings.
Testing and evaluation (Fedora 31 x86_64 with enforced Targeted policy extended
by perf_event class (see refpolicy [1] master branch)):
[root@host ~]# ps -Z
LABEL PID TTY TIME CMD
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 3960 pts/1 00:00:00 bash
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4167 pts/1 00:00:00 ps
[root@host ~]# ls -alhZ /usr/local/bin/
total 56M
drwxr-xr-x. 2 root root system_u:object_r:bin_t:s0 4.0K Mar 4 12:27 .
drwxr-xr-x. 12 root root system_u:object_r:usr_t:s0 4.0K Jul 25 2019 ..
-rwxr-xr-x. 1 root root system_u:object_r:bin_t:s0 4.1M Jan 23 2017 bash
-rwxr-xr-x. 1 root root system_u:object_r:bin_t:s0 4.1M Jan 23 2017 bash.before_shellshock_patch
...
-rwxr-xr-x. 1 root root system_u:object_r:bin_t:s0 372 May 14 2019 flask
-rwxr-xr-x. 1 root root unconfined_u:object_r:bin_t:s0 24M Mar 4 12:15 perf <== unprivileged users (perf_event_paranoid)
-rwxr-x---. 1 root perf_users unconfined_u:object_r:bin_t:s0 24M Mar 4 12:19 perf.cap <== perf_users (CAP_SYS_ADMIN)
-rwxr-xr-x. 1 root root system_u:object_r:bin_t:s0 44K Dec 8 2016 spiff
...
lrwxrwxrwx. 1 root root system_u:object_r:bin_t:s0 4 Aug 21 2018 zstdmt -> zstd
[root@host ~]# getenforce
Enforcing
=== Access by unprivileged user ===
[user@host ~]$ ps -Z
LABEL PID TTY TIME CMD
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4043 pts/2 00:00:00 bash
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4168 pts/2 00:00:00 ps
[user@host ~]$ /usr/local/bin/perf stat -- ls
Error:
Access to performance monitoring and observability operations is limited.
SELinux Enforcing mode is enabled and can limit access to performance
monitoring and observability operations. Inspect system audit records
for more perf_event access control information and adjusting the policy.
Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open
access to performance monitoring and observability operations for users
without CAP_SYS_ADMIN capability. perf_event_paranoid setting is -1:
-1: Allow use of (almost) all events by all users
Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
>= 0: Disallow raw and ftrace function tracepoint access
>= 1: Disallow CPU event access
>= 2: Disallow kernel profiling
To make the adjusted perf_event_paranoid setting permanent preserve it
in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>)
[root@host ~]# journalctl --follow
... audit[4186]: AVC avc: denied { open } for pid=4186 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
... audit[4186]: AVC avc: denied { open } for pid=4186 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
... setroubleshoot[4194]: SELinux is preventing perf from open access on the perf_event labeled unconfined_t. For complete SELinux messages run: sealert -l 9a6f3db2-3d8f-461e-afad-0b5c3a9c3b9d
... python3[4194]: SELinux is preventing perf from open access on the perf_event labeled unconfined_t.
***** Plugin catchall (100. confidence) suggests **************************
If you believe that perf should be allowed open access on perf_event labeled unconfined_t by default.
Then you should report this as a bug.
You can generate a local policy module to allow this access.
Do
allow this access for now by executing:
# ausearch -c 'perf' --raw | audit2allow -M my-perf
# semodule -X 300 -i my-perf.pp
=== Access by perf privileged user ===
[user@host ~]$ ps -Z
LABEL PID TTY TIME CMD
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4043 pts/2 00:00:00 bash
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4168 pts/2 00:00:00 ps
[user@host ~]$ libcap/progs/getcap /usr/local/bin/perf.cap
/usr/local/bin/perf.cap = cap_sys_ptrace,cap_syslog,cap_sys_admin+ep
[user@host ~]$ /usr/local/bin/perf.cap stat -- ls
Error:
Access to performance monitoring and observability operations is limited.
SELinux Enforcing mode is enabled and can limit access to performance
monitoring and observability operations. Inspect system audit records
for more perf_event access control information and adjusting the policy.
Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open
access to performance monitoring and observability operations for users
without CAP_SYS_ADMIN capability. perf_event_paranoid setting is -1:
-1: Allow use of (almost) all events by all users
Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
>= 0: Disallow raw and ftrace function tracepoint access
>= 1: Disallow CPU event access
>= 2: Disallow kernel profiling
To make the adjusted perf_event_paranoid setting permanent preserve it
in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>)
[root@host ~]# journalctl --follow
... audit[3926]: AVC avc: denied { open } for pid=3926 comm="perf.cap" 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
... audit[3926]: AVC avc: denied { open } for pid=3926 comm="perf.cap" 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
... setroubleshoot[3934]: SELinux is preventing perf from open access on the perf_event labeled unconfined_t. For complete SELinux messages run: sealert -l 9a6f3db2-3d8f-461e-afad-0b5c3a9c3b9d
... python3[3934]: SELinux is preventing perf from open access on the perf_event labeled unconfined_t.
***** Plugin catchall (100. confidence) suggests **************************
If you believe that perf should be allowed open access on perf_event labeled unconfined_t by default.
Then you should report this as a bug.
You can generate a local policy module to allow this access.
Do
allow this access for now by executing:
# ausearch -c 'perf' --raw | audit2allow -M my-perf
# semodule -X 300 -i my-perf.pp
=== Open access to performance monitoring and observability operations in unconfined_t domain ===
[root@host ~]# ausearch -c 'perf' --raw | audit2allow -M my-perf && cat my-perf.te
module my-perf 1.0;
require {
type unconfined_t;
class perf_event { cpu kernel open read tracepoint write };
}
#============= unconfined_t ==============
allow unconfined_t self:perf_event { cpu kernel open read tracepoint write };
[root@host ~]# semodule -X 300 -i my-perf.pp
[user@host ~]$ ps -Z
LABEL PID TTY TIME CMD
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4043 pts/2 00:00:00 bash
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 4168 pts/2 00:00:00 ps
[user@host ~]$ /usr/local/bin/perf stat -- ls
Desktop Documents Downloads intel Music perf.data perf.data.old Pictures Public Templates Videos
Performance counter stats for 'ls':
0.72 msec task-clock:u # 0.655 CPUs utilized
0 context-switches:u # 0.000 K/sec
0 cpu-migrations:u # 0.000 K/sec
98 page-faults:u # 0.137 M/sec
908,356 cycles:u # 1.266 GHz
729,984 instructions:u # 0.80 insn per cycle
142,774 branches:u # 198.968 M/sec
8,238 branch-misses:u # 5.77% of all branches
0.001095239 seconds time elapsed
0.001147000 seconds user
0.000000000 seconds sys
[user@host ~]$ /usr/local/bin/perf stat -a
Error:
Access to performance monitoring and observability operations is limited.
SELinux Enforcing mode is enabled and can limit access to performance
monitoring and observability operations. Inspect system audit records
for more perf_event access control information and adjusting the policy.
Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open
access to performance monitoring and observability operations for users
without CAP_SYS_ADMIN capability. perf_event_paranoid setting is -1:
-1: Allow use of (almost) all events by all users
Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
>= 0: Disallow raw and ftrace function tracepoint access
>= 1: Disallow CPU event access
>= 2: Disallow kernel profiling
To make the adjusted perf_event_paranoid setting permanent preserve it
in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>)
[user@host ~]$ /usr/local/bin/perf.cap stat -a
^C
Performance counter stats for 'system wide':
13,427.05 msec cpu-clock # 7.997 CPUs utilized
783 context-switches # 0.058 K/sec
29 cpu-migrations # 0.002 K/sec
6 page-faults # 0.000 K/sec
161,084,874 cycles # 0.012 GHz
146,823,131 instructions # 0.91 insn per cycle
12,164,802 branches # 0.906 M/sec
380,350 branch-misses # 3.13% of all branches
1.678938906 seconds time elapsed
[1] https://github.com/SELinuxProject/refpolicy
Signed-off-by: Alexey Budankov <alexey.budankov@xxxxxxxxxxxxxxx>
---
tools/perf/util/cloexec.c | 4 ++--
tools/perf/util/evsel.c | 40 +++++++++++++++++++++++----------------
2 files changed, 26 insertions(+), 18 deletions(-)
diff --git a/tools/perf/util/cloexec.c b/tools/perf/util/cloexec.c
index a12872f2856a..9c8ec816261b 100644
--- a/tools/perf/util/cloexec.c
+++ b/tools/perf/util/cloexec.c
@@ -65,7 +65,7 @@ static int perf_flag_probe(void)
return 1;
}
- WARN_ONCE(err != EINVAL && err != EBUSY,
+ WARN_ONCE(err != EINVAL && err != EBUSY && err != EACCES,
"perf_event_open(..., PERF_FLAG_FD_CLOEXEC) failed with unexpected error %d (%s)\n",
err, str_error_r(err, sbuf, sizeof(sbuf)));
@@ -83,7 +83,7 @@ static int perf_flag_probe(void)
if (fd >= 0)
close(fd);
- if (WARN_ONCE(fd < 0 && err != EBUSY,
+ if (WARN_ONCE(fd < 0 && err != EBUSY && err != EACCES,
"perf_event_open(..., 0) failed unexpectedly with error %d (%s)\n",
err, str_error_r(err, sbuf, sizeof(sbuf))))
return -1;
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 816d930d774e..f03ce1d362d3 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -2493,32 +2493,40 @@ int perf_evsel__open_strerror(struct evsel *evsel, struct target *target,
int err, char *msg, size_t size)
{
char sbuf[STRERR_BUFSIZE];
- int printed = 0;
+ int printed = 0, enforced = 0;
switch (err) {
case EPERM:
case EACCES:
+ printed += scnprintf(msg + printed, size - printed,
+ "Access to performance monitoring and observability operations is limited.\n");
+
+ if (!sysfs__read_int("fs/selinux/enforce", &enforced)) {
+ if (enforced) {
+ printed += scnprintf(msg + printed, size - printed,
+ "SELinux Enforcing mode is enabled and can limit access to performance\n"
+ "monitoring and observability operations. Inspect system audit records\n"
+ "for more perf_event access control information and adjusting the policy.\n");
+ }
+ }
+
if (err == EPERM)
- printed = scnprintf(msg, size,
- "No permission to enable %s event.\n\n",
+ printed += scnprintf(msg, size,
+ "No permission to enable %s event.\n",
perf_evsel__name(evsel));
return scnprintf(msg + printed, size - printed,
- "You may not have permission to collect %sstats.\n\n"
- "Consider tweaking /proc/sys/kernel/perf_event_paranoid,\n"
- "which controls use of the performance events system by\n"
- "unprivileged users (without CAP_SYS_ADMIN).\n\n"
- "The current value is %d:\n\n"
+ "Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open\n"
+ "access to performance monitoring and observability operations for users\n"
+ "without CAP_SYS_ADMIN capability. perf_event_paranoid setting is %d:\n"
" -1: Allow use of (almost) all events by all users\n"
" Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK\n"
- ">= 0: Disallow ftrace function tracepoint by users without CAP_SYS_ADMIN\n"
- " Disallow raw tracepoint access by users without CAP_SYS_ADMIN\n"
- ">= 1: Disallow CPU event access by users without CAP_SYS_ADMIN\n"
- ">= 2: Disallow kernel profiling by users without CAP_SYS_ADMIN\n\n"
- "To make this setting permanent, edit /etc/sysctl.conf too, e.g.:\n\n"
- " kernel.perf_event_paranoid = -1\n" ,
- target->system_wide ? "system-wide " : "",
- perf_event_paranoid());
+ ">= 0: Disallow raw and ftrace function tracepoint access\n"
+ ">= 1: Disallow CPU event access\n"
+ ">= 2: Disallow kernel profiling\n"
+ "To make the adjusted perf_event_paranoid setting permanent preserve it\n"
+ "in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>)",
+ perf_event_paranoid());
case ENOENT:
return scnprintf(msg, size, "The %s event is not supported.",
perf_evsel__name(evsel));
--
2.24.1
