It seems perf sets the exclude_guest bit because of Intel PEBS
implementation which uses a virtual address. IIUC now kernel disables
PEBS when it goes to the guest mode regardless of this bit so we don't
need to set it explicitly. At least for the other archs/vendors.
I found the commit 1342798cc1 set the exclude_guest for precise_ip
in the tool and the commit 20b279ddb3 added kernel side enforcement
which was reverted by commit a706d965dc later.
Actually it doesn't set the exclude_guest for the default event
(cycles:P) already.
$ grep -m1 vendor /proc/cpuinfo
vendor_id : GenuineIntel
$ perf record -e cycles:P true
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.002 MB perf.data (9 samples) ]
$ perf evlist -v | tr ',' '\n' | grep -e exclude -e precise
precise_ip: 3
But having lower 'p' modifier set the bit for some reason.
$ perf record -e cycles:pp true
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.002 MB perf.data (9 samples) ]
$ perf evlist -v | tr ',' '\n' | grep -e exclude -e precise
precise_ip: 2
exclude_guest: 1
Actually AMD IBS suffers from this because it doesn't support excludes
and having this bit effectively disables new features in the current
implementation (due to the missing feature check).
$ grep -m1 vendor /proc/cpuinfo
vendor_id : AuthenticAMD
$ perf record -W -e cycles:p -vv true 2>&1 | grep switching
switching off PERF_FORMAT_LOST support
switching off weight struct support
switching off bpf_event
switching off ksymbol
switching off cloexec flag
switching off mmap2
switching off exclude_guest, exclude_host
By not setting exclude_guest, we can fix this inconsistency and the
troubles.
Reviewed-by: Ian Rogers <irogers@google.com>
Reviewed-by: James Clark <james.clark@linaro.org>
Reviewed-by: Ravi Bangoria <ravi.bangoria@amd.com>
Acked-by: Kan Liang <kan.liang@linux.intel.com>
Cc: James Clark <james.clark@arm.com>
Cc: Atish Patra <atishp@atishpatra.org>
Cc: Mingwei Zhang <mizhang@google.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Palmer Dabbelt <palmer@rivosinc.com>
Link: https://lore.kernel.org/r/20241016062359.264929-5-namhyung@kernel.org
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The exclude_guest in the event attribute is to limit profiling in the
host environment. But I'm not sure why we want to set it by default
cause we don't care about it in most cases and I feel like it just
makes new PMU implementation complicated.
Of course it's useful for perf kvm command so I added the
exclude_GH_default variable to preserve the old behavior for perf kvm
and other commands like perf record and stat won't set the exclude bit.
This is helpful for AMD IBS case since having exclude_guest bit will
clear new feature bit due to the missing feature check logic.
$ sysctl kernel.perf_event_paranoid
kernel.perf_event_paranoid = 0
$ perf record -W -e ibs_op// -vv true 2>&1 | grep switching
switching off PERF_FORMAT_LOST support
switching off weight struct support
switching off bpf_event
switching off ksymbol
switching off cloexec flag
switching off mmap2
switching off exclude_guest, exclude_host
Intestingly, I found it sets the exclude_bit if "u" modifier is used.
I don't know why but it's neither intuitive nor consistent. Let's
remove the bit there too.
Reviewed-by: Ian Rogers <irogers@google.com>
Reviewed-by: James Clark <james.clark@linaro.org>
Reviewed-by: Ravi Bangoria <ravi.bangoria@amd.com>
Acked-by: Kan Liang <kan.liang@linux.intel.com>
Cc: James Clark <james.clark@arm.com>
Cc: Atish Patra <atishp@atishpatra.org>
Cc: Mingwei Zhang <mizhang@google.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Palmer Dabbelt <palmer@rivosinc.com>
Link: https://lore.kernel.org/r/20241016062359.264929-3-namhyung@kernel.org
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Using it:
$ perf test -w noplop
No workload found: noplop
$
$ perf test -w
Error: switch `w' requires a value
Usage: perf test [<options>] [{list <test-name-fragment>|[<test-name-fragments>|<test-numbers>]}]
-w, --workload <work>
workload to run for testing, use '--list-workloads' to list the available ones.
$
$ perf test --list-workloads
noploop
thloop
leafloop
sqrtloop
brstack
datasym
landlock
$
Would be good at some point to have a description in 'struct test_workload'.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Reviewed-by: James Clark <james.clark@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Clark Williams <williams@redhat.com>
Link: https://lore.kernel.org/r/20241020021842.1752770-3-acme@kernel.org
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Perf test case 84 'perf pipe recording and injection test'
sometime fails on s390, especially on z/VM virtual machines.
This is caused by a very short run time of workload
# perf test -w noploop
which runs for 1 second. Occasionally this is not long
enough and the perf report has no samples for symbol noploop.
Fix this and enlarge the runtime for the perf work load
to 3 seconds. This ensures the symbol noploop is always
present. Since only s390 is affected, make this loop
architecture dependend.
Output before:
Inject -b build-ids test
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.195 MB - ]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.277 MB - ]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.195 MB - ]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.160 MB
/tmp/perf.data.ELzRdq (4031 samples) ]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.195 MB - ]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.195 MB - ]
Inject -b build-ids test [Success]
Inject --buildid-all build-ids test
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.195 MB - ]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.014 MB - ]
Inject --buildid-all build-ids test [Failed - cannot find
noploop function in pipe #2]
Output after:
Successful execution for over 10 times in a loop.
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Suggested-by: Namhyung Kim <namhyung@kernel.org>
Cc: agordeev@linux.ibm.com
Cc: gor@linux.ibm.com
Cc: hca@linux.ibm.com
Link: https://lore.kernel.org/r/20241018081732.1391060-1-tmricht@linux.ibm.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Like in the metricgroup tests, it should check the permission first and
then skip relevant failures accordingly.
Also it needs to try again with the system wide flag properly. On the
second round, check if the result has the metric name because other
failure cases are checked in the first round already.
Reviewed-by: Ian Rogers <irogers@google.com>
Link: https://lore.kernel.org/r/20241018204306.741972-1-namhyung@kernel.org
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
On my system, perf list is very slow to print the whole events. I think
there's a performance issue in SDT and uprobes event listing. I noticed
this issue while running perf test on x86 but it takes long to check
some CoreSight event which should be skipped quickly.
Anyway, some test uses perf list to check whether the required event is
available before running the test. The perf list command can take an
argument to specify event class or (glob) pattern. But glob pattern is
only to suppress output for unmatched ones after checking all events.
In this case, specifying event class is better to reduce the number of
events it checks and to avoid buggy subsystems entirely.
No functional changes intended.
Reviewed-by: James Clark <james.clark@linaro.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: German Gomez <german.gomez@arm.com>
Cc: Carsten Haitzler <carsten.haitzler@arm.com>
Cc: Leo Yan <leo.yan@arm.com>
Link: https://lore.kernel.org/r/20241016065654.269994-1-namhyung@kernel.org
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
perf fails to compile on systems with GCC version11
as below:
In file included from /usr/include/string.h:519,
from /home/athir/perf-tools-next/tools/include/linux/bitmap.h:5,
from /home/athir/perf-tools-next/tools/perf/util/pmu.h:5,
from /home/athir/perf-tools-next/tools/perf/util/evsel.h:14,
from /home/athir/perf-tools-next/tools/perf/util/evlist.h:14,
from tests/tool_pmu.c:3:
In function ‘strncpy’,
inlined from ‘do_test’ at tests/tool_pmu.c:25:3:
/usr/include/bits/string_fortified.h:95:10: error: ‘__builtin_strncpy’ specified bound 128 equals destination size [-Werror=stringop-truncation]
95 | return __builtin___strncpy_chk (__dest, __src, __len,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
96 | __glibc_objsize (__dest));
| ~~~~~~~~~~~~~~~~~~~~~~~~~
The compile error is from strncpy refernce in do_test:
strncpy(str, tool_pmu__event_to_str(ev), sizeof(str));
This behaviour is not observed with GCC version 8, but observed
with GCC version 11 . This is message from gcc for detecting
truncation while using strncpu. Use snprintf instead of strncpy
here to be safe.
Signed-off-by: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: akanksha@linux.ibm.com
Cc: hbathini@linux.ibm.com
Cc: kjain@linux.ibm.com
Cc: maddy@linux.ibm.com
Cc: disgoel@linux.vnet.ibm.com
Cc: linuxppc-dev@lists.ozlabs.org
Link: https://lore.kernel.org/r/20241013173742.71882-1-atrajeev@linux.vnet.ibm.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Hard coded terms like "config=10" are skipped by perf_pmu__config
assuming they were already applied to a perf_event_attr by parse
event's config_attr function. When doing a reverse number to name
lookup in perf_pmu__name_from_config, as the hardcoded terms aren't
applied the config value is incorrect leading to misses or false
matches. Fix this by adding a parameter to have perf_pmu__config apply
hardcoded terms too (not just in parse event's config_term_common).
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20241002032016.333748-3-irogers@google.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
With the patch 0b6c5371c0 "Add missing topdown metrics events" eight
topdown metric events with numbers ranging from 0x8000 to 0x8700 were
added to the test since they were added as 'perf stat' default events.
Later the patch 951efb9976 "Update no event/metric expectations" kept
only 4 of those events(0x8000-0x8300).
Currently, the topdown events with numbers 0x8400 to 0x8700 are missing
from the list of expected events resulting in a failure. Add back the
missing topdown events.
Fixes: 951efb9976 ("perf test attr: Update no event/metric expectations")
Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com>
Tested-by: Ian Rogers <irogers@google.com>
Cc: mpetlan@redhat.com
Link: https://lore.kernel.org/r/20240311081611.7835-1-vmolnaro@redhat.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Test "Setup struct perf_event_attr" consists of multiple test cases that
can affect the max sample rate value for perf events. Some test cases
check this value as it should not be lowered under the set minimum for
the given test. Currently, it is possible for the test cases to affect
each other as the previous tests can lower the sample rate, leading to
a possible failure of some of the future test cases as the value is not
restored at any point.
# 10: Setup struct perf_event_attr:
--- start ---
test child forked, pid 104220
Using CPUID 0x00000000413fd0c1
running './tests/attr/test-record-C0'
Current sample rate: 10000
running './tests/attr/test-record-basic'
Current sample rate: 900
running './tests/attr/test-record-branch-any'
Current sample rate: 600
running './tests/attr/test-record-dummy-C0'
Current sample rate: 600
expected sample_period=4000, got 600
FAILED './tests/attr/test-record-dummy-C0' - match failure
Restore the max sample rate value for perf events to a reasonable value
before each test case if its value was lowered too much to ensure the
same conditions for each test case.
# 10: Setup struct perf_event_attr:
--- start ---
test child forked, pid 107222
Using CPUID 0x00000000413fd0c1
running './tests/attr/test-record-C0'
Current sample rate: 10000
running './tests/attr/test-record-basic'
Current sample rate: 800
running './tests/attr/test-record-branch-any'
Current sample rate: 700
unsupp './tests/attr/test-record-branch-any'
running './tests/attr/test-record-branch-filter-any'
Current sample rate: 10000
running './tests/attr/test-record-count'
Current sample rate: 10000
running './tests/attr/test-record-data'
Current sample rate: 600
running './tests/attr/test-record-dummy-C0'
Current sample rate: 800
running './tests/attr/test-record-freq'
Current sample rate: 10000
...
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Radostin Stoyanov <rstoyano@redhat.com>
Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com>
Link: https://lore.kernel.org/r/20241003125136.15918-1-vmolnaro@redhat.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Command perf test 86 fails on s390:
# perf test -F 86
ping 868299 [007] 28248.013596: probe_libc:inet_pton_1: (3ff95948020)
3ff95948020 inet_pton+0x0 (inlined)
3ff9595e6e7 text_to_binary_address+0x1007 (inlined)
3ff9595e6e7 gaih_inet+0x1007 (inlined)
FAIL: expected backtrace entry \
"main\+0x[[:xdigit:]]+[[:space:]]\(.*/bin/ping.*\)$"
got "3ff9595e6e7 gaih_inet+0x1007 (inlined)"
86: probe libc's inet_pton & backtrace it with ping : FAILED!
#
The root cause is a new stack layout, two functions have been added
as seen below.
# perf script | tac | grep -m1 '^ping' -B9 | tac
ping 866856 [007] 25979.494921: probe_libc:inet_pton: (3ff8ec48020)
3ff8ec48020 inet_pton+0x0 (inlined)
new --> 3ff8ec5e6e7 text_to_binary_address+0x1007 (inlined)
new --> 3ff8ec5e6e7 gaih_inet+0x1007 (inlined)
3ff8ec5e6e7 getaddrinfo+0x1007 (/usr/lib64/libc.so.6)
2aa3fe04bf5 main+0xff5 (/usr/bin/ping)
3ff8eb34a5b __libc_start_call_main+0x8b (/usr/lib64/libc.so.6)
3ff8eb34b5d __libc_start_main@GLIBC_2.2+0xad (inlined)
2aa3fe06a1f [unknown] (/usr/bin/ping)
#
The new functions in the call chain are:
- text_to_binary_address()
- gaih_inet().
Both functions are inlined and do not show up in the output
of the nm command:
# nm -a /usr/lib64/libc.so.6 | \
grep -E '(text_to_binary_address|gaih_inet)$'
#
There is no possibility to add these 2 functions depending on their
existance in the C library.
Add text_to_binary_address() and gaih_inet() to the list of
expected functions in an compatible way and extend the regular
expression. On s390 the backtrace can now be
Before After
probe_libc:inet_pton probe_libc:inet_pton
inet_pton inet_pton
getaddrinfo getaddrinfo | text_to_binary_address
main main | gaih_inet
Output after:
# perf test -F 86
86: probe libc's inet_pton & backtrace it with ping : Ok
#
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Cc: agordeev@linux.ibm.com
Cc: gor@linux.ibm.com
Cc: hca@linux.ibm.com
Cc: sumanthk@linux.ibm.com
Link: https://lore.kernel.org/r/20241001124224.3370306-1-tmricht@linux.ibm.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The "perf record" tool will now default to this new mode if the user
specifies a sampling group when not in system-wide mode, and when
"--no-inherit" is not specified.
This change updates evsel to allow the combination of inherit
and PERF_SAMPLE_READ.
A fallback is implemented for kernel versions where this feature is not
supported.
Signed-off-by: Ben Gainey <ben.gainey@arm.com>
Cc: james.clark@arm.com
Link: https://lore.kernel.org/r/20241001121505.1009685-3-ben.gainey@arm.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
perf test 70 takes a long time. One culprit is the output of command
perf annotate. Per default enabled are
- demangle symbol names
- interleave source code with assembly code.
Disable demangle of symbols and abort the annotation
after the first 250 lines.
This speeds up the test case considerable, for example
on s390:
Output before:
# time perf test 70
70: perf annotate basic tests : Ok
.....
real 2m7.467s
user 1m26.869s
sys 0m34.086s
#
Output after:
# time perf test 70
70: perf annotate basic tests : Ok
real 0m3.341s
user 0m1.606s
sys 0m0.362s
#
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: James Clark <james.clark@linaro.org>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: sumanthk@linux.ibm.com
Link: https://lore.kernel.org/r/20240917085706.249691-1-tmricht@linux.ibm.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
In addition to the existing support for libbfd and calling out to
an external addr2line command, add support for using libllvm directly.
This is both faster than libbfd, and can be enabled in distro builds
(the LLVM license has an explicit provision for GPLv2 compatibility).
Thus, it is set as the primary choice if available.
As an example, running 'perf report' on a medium-size profile with
DWARF-based backtraces took 58 seconds with LLVM, 78 seconds with
libbfd, 153 seconds with external llvm-addr2line, and I got tired and
aborted the test after waiting for 55 minutes with external bfd
addr2line (which is the default for perf as compiled by distributions
today).
Evidently, for this case, the bfd addr2line process needs 18 seconds (on
a 5.2 GHz Zen 3) to load the .debug ELF in question, hits the 1-second
timeout and gets killed during initialization, getting restarted anew
every time. Having an in-process addr2line makes this much more robust.
As future extensions, libllvm can be used in many other places where
we currently use libbfd or other libraries:
- Symbol enumeration (in particular, for PE binaries).
- Demangling (including non-Itanium demangling, e.g. Microsoft
or Rust).
- Disassembling (perf annotate).
However, these are much less pressing; most people don't profile PE
binaries, and perf has non-bfd paths for ELF. The same with demangling;
the default _cxa_demangle path works fine for most users, and while bfd
objdump can be slow on large binaries, it is possible to use
--objdump=llvm-objdump to get the speed benefits. (It appears
LLVM-based demangling is very simple, should we want that.)
Tested with LLVM 14, 15, 16, 18 and 19. For some reason, LLVM 12 was not
correctly detected using feature_check, and thus was not tested.
Committer notes:
Added the name and a __maybe_unused to address:
1 13.50 almalinux:8 : FAIL gcc version 8.5.0 20210514 (Red Hat 8.5.0-22) (GCC)
util/srcline.c: In function 'dso__free_a2l':
util/srcline.c:184:20: error: parameter name omitted
void dso__free_a2l(struct dso *)
^~~~~~~~~~~~
make[3]: *** [/git/perf-6.11.0-rc3/tools/build/Makefile.build:158: util] Error 2
Signed-off-by: Steinar H. Gunderson <sesse@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ian Rogers <irogers@google.com>
Link: https://lore.kernel.org/r/20240803152008.2818485-1-sesse@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
