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>
Prior to this patch the first and the last error encountered during
parsing are printed. To see other errors verbose needs
enabling. Unfortunately this can drop useful errors, in particular on
terms. This patch changes the errors so that instead of the first and
last all errors are recorded and printed, the underlying data
structure is changed to a list.
Before:
```
$ perf stat -e 'slots/edge=2/' true
event syntax error: 'slots/edge=2/'
\___ Bad event or PMU
Unable to find PMU or event on a PMU of 'slots'
Initial error:
event syntax error: 'slots/edge=2/'
\___ Cannot find PMU `slots'. Missing kernel support?
Run 'perf list' for a list of valid events
Usage: perf stat [<options>] [<command>]
-e, --event <event> event selector. use 'perf list' to list available events
```
After:
```
$ perf stat -e 'slots/edge=2/' true
event syntax error: 'slots/edge=2/'
\___ Bad event or PMU
Unable to find PMU or event on a PMU of 'slots'
event syntax error: 'slots/edge=2/'
\___ value too big for format (edge), maximum is 1
event syntax error: 'slots/edge=2/'
\___ Cannot find PMU `slots'. Missing kernel support?
Run 'perf list' for a list of valid events
Usage: perf stat [<options>] [<command>]
-e, --event <event> event selector. use 'perf list' to list available events
```
Signed-off-by: Ian Rogers <irogers@google.com>
Reviewed-by: James Clark <james.clark@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: tchen168@asu.edu
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20240131134940.593788-3-irogers@google.com
The perf tool has previously made legacy events the priority so with
or without a PMU the legacy event would be opened:
$ perf stat -e cpu-cycles,cpu/cpu-cycles/ true
Using CPUID GenuineIntel-6-8D-1
intel_pt default config: tsc,mtc,mtc_period=3,psb_period=3,pt,branch
Attempting to add event pmu 'cpu' with 'cpu-cycles,' that may result in non-fatal errors
After aliases, add event pmu 'cpu' with 'cpu-cycles,' that may result in non-fatal errors
Control descriptor is not initialized
------------------------------------------------------------
perf_event_attr:
type 0 (PERF_TYPE_HARDWARE)
size 136
config 0 (PERF_COUNT_HW_CPU_CYCLES)
sample_type IDENTIFIER
read_format TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
disabled 1
inherit 1
enable_on_exec 1
exclude_guest 1
------------------------------------------------------------
sys_perf_event_open: pid 833967 cpu -1 group_fd -1 flags 0x8 = 3
------------------------------------------------------------
perf_event_attr:
type 0 (PERF_TYPE_HARDWARE)
size 136
config 0 (PERF_COUNT_HW_CPU_CYCLES)
sample_type IDENTIFIER
read_format TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
disabled 1
inherit 1
enable_on_exec 1
exclude_guest 1
------------------------------------------------------------
...
Fixes to make hybrid/BIG.little PMUs behave correctly, ie as core PMUs
capable of opening legacy events on each, removing hard coded "cpu_core"
and "cpu_atom" Intel PMU names, etc. caused a behavioral difference on
Apple/ARM due to latent issues in the PMU driver reported in:
https://lore.kernel.org/lkml/08f1f185-e259-4014-9ca4-6411d5c1bc65@marcan.st/
As part of that report Mark Rutland <mark.rutland@arm.com> requested
that legacy events not be higher in priority when a PMU is specified
reversing what has until this change been perf's default behavior. With
this change the above becomes:
$ perf stat -e cpu-cycles,cpu/cpu-cycles/ true
Using CPUID GenuineIntel-6-8D-1
Attempt to add: cpu/cpu-cycles=0/
..after resolving event: cpu/event=0x3c/
Control descriptor is not initialized
------------------------------------------------------------
perf_event_attr:
type 0 (PERF_TYPE_HARDWARE)
size 136
config 0 (PERF_COUNT_HW_CPU_CYCLES)
sample_type IDENTIFIER
read_format TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
disabled 1
inherit 1
enable_on_exec 1
exclude_guest 1
------------------------------------------------------------
sys_perf_event_open: pid 827628 cpu -1 group_fd -1 flags 0x8 = 3
------------------------------------------------------------
perf_event_attr:
type 4 (PERF_TYPE_RAW)
size 136
config 0x3c
sample_type IDENTIFIER
read_format TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
disabled 1
inherit 1
enable_on_exec 1
exclude_guest 1
------------------------------------------------------------
...
So the second event has become a raw event as
/sys/devices/cpu/events/cpu-cycles exists.
A fix was necessary to config_term_pmu in parse-events.c as check_alias
expansion needs to happen after config_term_pmu, and config_term_pmu may
need calling a second time because of this.
config_term_pmu is updated to not use the legacy event when the PMU has
such a named event (either from JSON or sysfs).
The bulk of this change is updating all of the parse-events test
expectations so that if a sysfs/JSON event exists for a PMU the test
doesn't fail - a further sign, if it were needed, that the legacy event
priority was a known and tested behavior of the perf tool.
Reported-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Hector Martin <marcan@marcan.st>
Tested-by: Marc Zyngier <maz@kernel.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20231123042922.834425-1-irogers@google.com
[ Initialize the 'alias_rewrote_terms' variable to false to address a clang warning ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Testcase "Parsing of all PMU events from sysfs" parse events for
all PMUs, and not just cpu. In case of powerpc, the PowerVM
environment supports events from hv_24x7 and hv_gpci PMU which
is of example format like below:
- hv_24x7/CPM_ADJUNCT_INST,domain=?,core=?/
- hv_gpci/event,partition_id=?/
The value for "?" needs to be filled in depending on system
configuration. It is better to skip these parametrized events
in this test as it is done in:
'commit b50d691e50 ("perf test: Fix "all PMU test" to skip
parametrized events")' which handled a simialr instance with
"all PMU test".
Fix parse-events test to skip parametrized events since
it needs proper setup of the parameters.
Signed-off-by: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Tested-by: Ian Rogers <irogers@google.com>
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Reviewed-by: Kajol Jain <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/20230927181703.80936-1-atrajeev@linux.vnet.ibm.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The event parser needs to handle two special cases:
1) legacy events like L1-dcache-load-miss. These event names don't
appear in JSON or sysfs, and lookup tables are used for the config
value.
2) raw events where 'r0xead' is the same as 'read' unless the PMU has
an event called 'read' in which case the event has priority.
The previous parser to handle these cases would scan all PMUs for
components of event names. These components would then be used to
classify in the lexer whether the token should be part of a legacy
event, a raw event or an event. The grammar would handle legacy event
tokens or recombining the tokens back into a regular event name. The
code wasn't PMU specific and had issues around events like AMD's
branch-brs that would fail to parse as it expects brs to be a suffix
on a legacy event style name:
$ perf stat -e branch-brs true
event syntax error: 'branch-brs'
\___ parser error
This change removes processing all PMUs by using the lexer in the form
of a regular expression matcher. The lexer will return the token for
the longest matched sequence of characters, and in the event of a tie
the first. The legacy events are a fixed number of regular
expressions, and by matching these before a name token its possible to
generate an accurate legacy event token with everything else matching
as a name. Because of the lexer change the handling of hyphens in the
grammar can be removed as hyphens just become a part of the name.
To handle raw events and terms the parser is changed to defer trying
to evaluate whether something is a raw event until the PMU is known in
the grammar. Once the PMU is known, the events of the PMU can be
scanned for the 'read' style problem. A new term type is added for
these raw terms, used to enable deferring the evaluation.
While this change is large, it has stats of:
170 insertions(+), 436 deletions(-)
the bulk of the change is deleting the old approach. It isn't possible
to break apart the code added due to the dependencies on how the parts
of the parsing work.
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ahmad Yasin <ahmad.yasin@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Florian Fischer <florian.fischer@muhq.space>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kang Minchul <tegongkang@gmail.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Sumanth Korikkar <sumanthk@linux.ibm.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
Cc: Weilin Wang <weilin.wang@intel.com>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20230502223851.2234828-19-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Remove the LIBTRACEEVENT_DYNAMIC and LIBTRACEFS_DYNAMIC make command
line variables.
If libtraceevent isn't installed or NO_LIBTRACEEVENT=1 is passed to the
build, don't compile in libtraceevent and libtracefs support.
This also disables CONFIG_TRACE that controls "perf trace".
CONFIG_LIBTRACEEVENT is used to control enablement in Build/Makefiles,
HAVE_LIBTRACEEVENT is used in C code.
Without HAVE_LIBTRACEEVENT tracepoints are disabled and as such the
commands kmem, kwork, lock, sched and timechart are removed. The
majority of commands continue to work including "perf test".
Committer notes:
Fixed up a tools/perf/util/Build reject and added:
#include <traceevent/event-parse.h>
to tools/perf/util/scripting-engines/trace-event-perl.c.
Committer testing:
$ rpm -qi libtraceevent-devel
Name : libtraceevent-devel
Version : 1.5.3
Release : 2.fc36
Architecture: x86_64
Install Date: Mon 25 Jul 2022 03:20:19 PM -03
Group : Unspecified
Size : 27728
License : LGPLv2+ and GPLv2+
Signature : RSA/SHA256, Fri 15 Apr 2022 02:11:58 PM -03, Key ID 999f7cbf38ab71f4
Source RPM : libtraceevent-1.5.3-2.fc36.src.rpm
Build Date : Fri 15 Apr 2022 10:57:01 AM -03
Build Host : buildvm-x86-05.iad2.fedoraproject.org
Packager : Fedora Project
Vendor : Fedora Project
URL : https://git.kernel.org/pub/scm/libs/libtrace/libtraceevent.git/
Bug URL : https://bugz.fedoraproject.org/libtraceevent
Summary : Development headers of libtraceevent
Description :
Development headers of libtraceevent-libs
$
Default build:
$ ldd ~/bin/perf | grep tracee
libtraceevent.so.1 => /lib64/libtraceevent.so.1 (0x00007f1dcaf8f000)
$
# perf trace -e sched:* --max-events 10
0.000 migration/0/17 sched:sched_migrate_task(comm: "", pid: 1603763 (perf), prio: 120, dest_cpu: 1)
0.005 migration/0/17 sched:sched_wake_idle_without_ipi(cpu: 1)
0.011 migration/0/17 sched:sched_switch(prev_comm: "", prev_pid: 17 (migration/0), prev_state: 1, next_comm: "", next_prio: 120)
1.173 :0/0 sched:sched_wakeup(comm: "", pid: 3138 (gnome-terminal-), prio: 120)
1.180 :0/0 sched:sched_switch(prev_comm: "", prev_prio: 120, next_comm: "", next_pid: 3138 (gnome-terminal-), next_prio: 120)
0.156 migration/1/21 sched:sched_migrate_task(comm: "", pid: 1603763 (perf), prio: 120, orig_cpu: 1, dest_cpu: 2)
0.160 migration/1/21 sched:sched_wake_idle_without_ipi(cpu: 2)
0.166 migration/1/21 sched:sched_switch(prev_comm: "", prev_pid: 21 (migration/1), prev_state: 1, next_comm: "", next_prio: 120)
1.183 :0/0 sched:sched_wakeup(comm: "", pid: 1602985 (kworker/u16:0-f), prio: 120, target_cpu: 1)
1.186 :0/0 sched:sched_switch(prev_comm: "", prev_prio: 120, next_comm: "", next_pid: 1602985 (kworker/u16:0-f), next_prio: 120)
#
Had to tweak tools/perf/util/setup.py to make sure the python binding
shared object links with libtraceevent if -DHAVE_LIBTRACEEVENT is
present in CFLAGS.
Building with NO_LIBTRACEEVENT=1 uncovered some more build failures:
- Make building of data-convert-bt.c to CONFIG_LIBTRACEEVENT=y
- perf-$(CONFIG_LIBTRACEEVENT) += scripts/
- bpf_kwork.o needs also to be dependent on CONFIG_LIBTRACEEVENT=y
- The python binding needed some fixups and util/trace-event.c can't be
built and linked with the python binding shared object, so remove it
in tools/perf/util/setup.py and exclude it from the list of
dependencies in the python/perf.so Makefile.perf target.
Building without libtraceevent-devel installed uncovered more build
failures:
- The python binding tools/perf/util/python.c was assuming that
traceevent/parse-events.h was always available, which was the case
when we defaulted to using the in-kernel tools/lib/traceevent/ files,
now we need to enclose it under ifdef HAVE_LIBTRACEEVENT, just like
the other parts of it that deal with tracepoints.
- We have to ifdef the rules in the Build files with
CONFIG_LIBTRACEEVENT=y to build builtin-trace.c and
tools/perf/trace/beauty/ as we only ifdef setting CONFIG_TRACE=y when
setting NO_LIBTRACEEVENT=1 in the make command line, not when we don't
detect libtraceevent-devel installed in the system. Simplification here
to avoid these two ways of disabling builtin-trace.c and not having
CONFIG_TRACE=y when libtraceevent-devel isn't installed is the clean
way.
From Athira:
<quote>
tools/perf/arch/powerpc/util/Build
-perf-y += kvm-stat.o
+perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
</quote>
Then, ditto for arm64 and s390, detected by container cross build tests.
- s/390 uses test__checkevent_tracepoint() that is now only available if
HAVE_LIBTRACEEVENT is defined, enclose the callsite with ifder HAVE_LIBTRACEEVENT.
Also from Athira:
<quote>
With this change, I could successfully compile in these environment:
- Without libtraceevent-devel installed
- With libtraceevent-devel installed
- With “make NO_LIBTRACEEVENT=1”
</quote>
Then, finally rename CONFIG_TRACEEVENT to CONFIG_LIBTRACEEVENT for
consistency with other libraries detected in tools/perf/.
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: bpf@vger.kernel.org
Link: http://lore.kernel.org/lkml/20221205225940.3079667-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Skip an event configuration for event names with a dash/minus in them.
Events with a dash/minus in their name cause parsing issues as legacy
encoding of events would use a dash/minus as a separator.
The parser separates events with dashes into prefixes and suffixes and
then recombines them. Unfortunately if an event has part of its name
that matches a legacy token then the recombining fails.
This is seen for branch-brs where branch is a legacy token. branch-brs
was introduced to sysfs in:
https://lore.kernel.org/all/20220322221517.2510440-5-eranian@google.com/
The failure is shown below as well as the workaround to use a config
where the dash/minus isn't treated specially:
```
$ perf stat -e branch-brs true
event syntax error: 'branch-brs'
\___ parser error
$ perf stat -e cpu/branch-brs/ true
Performance counter stats for 'true':
46,179 cpu/branch-brs/
```
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.garry@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20221013011205.3151391-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
