- 15 May, 2023 27 commits
-
-
Ian Rogers authored
An event like "cpu/instructions/" typically parses due to there being a sysfs event called instructions. On hybrid recursive parsing means that the hardware event is encoded in the attribute, with the PMU being placed in the high bits of the config: ''' $ perf stat -vv -e 'cpu_core/cycles/' true ... ------------------------------------------------------------ perf_event_attr: size 136 config 0x400000000 sample_type IDENTIFIER read_format TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING disabled 1 inherit 1 enable_on_exec 1 exclude_guest 1 ------------------------------------------------------------ ''' Make this behavior the default by adding a new term type and token for hardware events. The token gathers both the numeric config and the parsed name, so that if the token appears like "cycles/name=cycles/" then the token can be handled like a name. The numeric value isn't sufficient to distinguish say "cpu-cycles" from "cycles". Extend the parse-events test so that all current non-PMU hardware parsing tests, also test with the PMU cpu - more than half the change. 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/20230502223851.2234828-34-irogers@google.comSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Ian Rogers authored
Remove assumptions of just 1 core PMU. Signed-off-by:
-
Ian Rogers authored
Rather than limit the --cputype argument for "perf list" and "perf stat" to hybrid PMUs of just cpu_atom and cpu_core, allow any PMU. Note, that if cpu_atom isn't mounted but a filter of cpu_atom is requested, then this will now fail. As such a filter would never succeed, no events can come from that unmounted PMU, then this behavior could never have been useful and failing is clearer. Signed-off-by:
-
Ian Rogers authored
To support the cputype argument added to "perf stat" for hybrid it is necessary to filter events during wildcard matching. Add a scanner argument for the filter and checking it when wildcard matching. Signed-off-by:
-
Ian Rogers authored
Use the typed parse_state rather than void* _parse_state when available. Signed-off-by:
-
Ian Rogers authored
The event parser no longer needs to recurse in case of a legacy cache event in a PMU, the necessary wild card logic has moved to perf_pmu__supports_legacy_cache and perf_pmu__supports_wildcard_numeric. Signed-off-by:
-
Ian Rogers authored
Legacy raw events like r1a open as PERF_TYPE_RAW on non-hybrid systems and on each hybrid PMU on hybrid systems. Rather than iterate hybrid PMUs add a perf_pmu__supports_wildcard_numeric function that says when a numeric event should be opened upon it. If the parsed event specifies the type of the PMU then don't wildcard match PMUs, use the specified PMU type. Signed-off-by:
-
Ian Rogers authored
Mirroring parse_events_add_cache, list the legacy name alongside its alias with the PMU. Remove the now unnecessary hybrid logic. Note, the alias output removes the event type descriptor, so: L1-dcache-loads [Hardware cache event] becomes: L1-dcache-loads OR cpu/L1-dcache-loads/ Signed-off-by:
-
Ian Rogers authored
It is inconsistent that "perf stat -e instructions-retired" wildcard opens on all PMUs while legacy cache events like "perf stat -e L1-dcache-load-miss" do not. A behavior introduced by hybrid is that a legacy cache event like L1-dcache-load-miss should wildcard open on all hybrid PMUs. Previously hybrid would call to is_event_supported for each PMU, a failure of which results in the event not being added. This isn't done in this case as the parser should just create perf_event_attr and the later open should fail, or the counter give "<not counted>". If this wants to be avoided then the PMU can be named with the event. Signed-off-by:
-
Ian Rogers authored
Allow a legacy cache event to be both, for example, "L1-dcache-load-miss" and "cpu/L1-dcache-load-miss/" by introducing a new legacy cache term type. The term type is processed in config_term_pmu, setting both the type in perf_event_attr and the config. The code to determine the config is factored out of parse_events_add_cache and shared. If the PMU doesn't support legacy events, currently just core/hybrid PMUs do, then the term is treated like a PE_NAME term - as before. If only terms are being parsed, such as for perf_pmu__new_alias, then the PE_LEGACY_CACHE token is always parsed as PE_NAME. Signed-off-by:
-
Ian Rogers authored
Assert hybrid extended types are as expected. Signed-off-by:
-
Ian Rogers authored
Don't assume evlist order. Switch to a loop rather than depend on evlist order for raw events test. Update hybrid event expectations. Previous values were based on parsing legacy hardware events from sysfs, update to the correct PMU specific legacy values. Signed-off-by:
-
Ian Rogers authored
The tests use x86 hybrid specific PMUs. Signed-off-by:
-
Ian Rogers authored
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 authored
The strlist in print_hwcache_events holds the event names as they are generated, and then it is iterated and printed. This is unnecessary and each event can just be printed as it is processed. Rename the variable i to res, to be more intention revealing and consistent with other code. Signed-off-by:
-
Ian Rogers authored
Change add_event to always set pmu_name when possible as not all code checks both pmu->name and evsel->pmu_name, for example, uniquify_counter in stat-display.c. Signed-off-by:
-
Ian Rogers authored
Set attr.type to PMU type early so that later terms can override the value. Setting the value in perf_pmu__config means that earlier steps, like config_term_pmu, can override the value. Signed-off-by:
-
Ian Rogers authored
Opening hardware names and a legacy cache event on a hybrid PMU opens it on each PMU. Parsing and checking indexes fails, as the parsed index is double the expected. Avoid checking the index by just comparing the names immediately after the parse. This change removes hard coded hybrid logic and removes assumptions about the expansion of an event. On hybrid the PMUs may or may not support an event and so using a distance isn't a consistent solution. Reviewed-by:
-
Ian Rogers authored
test__checkevent_config_cache checks the parsing of "L1-dcache-misses/name=cachepmu/". Don't just check that the name is set correctly, also validate the rest of the perf_event_attr for L1-dcache-misses. Reviewed-by:
-
Ian Rogers authored
Add helper to test the config of an evsel. Dependent on the type of the evsel, mask the config so that high-bits containing the extended PMU type are ignored. Signed-off-by:
-
Ian Rogers authored
Rather than skip all tests in test__events_pmu if PMU cpu isn't present, use the per-test valid test. This allows the running of software PMU tests on hybrid and arm systems. Reviewed-by:
-
Ian Rogers authored
Parse events for all PMUs, and not just cpu, in test "Parsing of all PMU events from sysfs". Signed-off-by:
-
Ian Rogers authored
Previously these constraints were disabled as they contained topdown events. Since: https://lore.kernel.org/all/20230312021543.3060328-9-irogers@google.com/ the topdown events are correctly grouped even if no group exists. This change was created by PR: https://github.com/intel/perfmon/pull/71Signed-off-by:
-
Ian Rogers authored
Previously these constraints were disabled as they contained topdown events. Since: https://lore.kernel.org/all/20230312021543.3060328-9-irogers@google.com/ the topdown events are correctly grouped even if no group exists. This change was created by PR: https://github.com/intel/perfmon/pull/71Signed-off-by:
-
Ian Rogers authored
Previously these constraints were disabled as they contained topdown events. Since: https://lore.kernel.org/all/20230312021543.3060328-9-irogers@google.com/ the topdown events are correctly grouped even if no group exists. This change was created by PR: https://github.com/intel/perfmon/pull/71Signed-off-by:
-
Ian Rogers authored
Previously these constraints were disabled as they contained topdown events. Since: https://lore.kernel.org/all/20230312021543.3060328-9-irogers@google.com/ the topdown events are correctly grouped even if no group exists. This change was created by PR: https://github.com/intel/perfmon/pull/71Signed-off-by:
-
Ian Rogers authored
Previously these constraints were disabled as they contained topdown events. Since: https://lore.kernel.org/all/20230312021543.3060328-9-irogers@google.com/ the topdown events are correctly grouped even if no group exists. This change was created by PR: https://github.com/intel/perfmon/pull/71Signed-off-by:
-
- 12 May, 2023 5 commits
-
-
Adrian Hunter authored
Print dso offset only for object files, and in those cases force using the dso->long_name if the dso->name starts with '[' or the dso is kcore, in order to avoid special names such as [vdso], or mixing up kcore with vmlinux. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Cc: Changbin Du <changbin.du@huawei.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20230424055107.12105-3-adrian.hunter@intel.comSigned-off-by:
-
Adrian Hunter authored
Declare dso const, so that functions can be called with const struct *dso. Signed-off-by:
-
Changbin Du authored
This adds a new 'dsoff' field to print dso offset for resolved symbols, and the offset is appended to dso name. Default output: $ perf script ls 2695501 3011030.487017: 500000 cycles: 152cc73ef4b5 get_common_indices.constprop.0+0x155 (/usr/lib/x86_64-linux-gnu/ld-2.31.so) ls 2695501 3011030.487018: 500000 cycles: ffffffff99045b3e [unknown] ([unknown]) ls 2695501 3011030.487018: 500000 cycles: ffffffff9968e107 [unknown] ([unknown]) ls 2695501 3011030.487018: 500000 cycles: ffffffffc1f54afb [unknown] ([unknown]) ls 2695501 3011030.487018: 500000 cycles: ffffffff9968382f [unknown] ([unknown]) ls 2695501 3011030.487019: 500000 cycles: ffffffff99e00094 [unknown] ([unknown]) ls 2695501 3011030.487019: 500000 cycles: 152cc718a8d0 __errno_location@plt+0x0 (/usr/lib/x86_64-linux-gnu/libselinux.so.1) Display 'dsoff' field: $ perf script -F +dsoff ls 2695501 3011030.487017: 500000 cycles: 152cc73ef4b5 get_common_indices.constprop.0+0x155 (/usr/lib/x86_64-linux-gnu/ld-2.31.so+0x1c4b5) ls 2695501 3011030.487018: 500000 cycles: ffffffff99045b3e [unknown] ([unknown]) ls 2695501 3011030.487018: 500000 cycles: ffffffff9968e107 [unknown] ([unknown]) ls 2695501 3011030.487018: 500000 cycles: ffffffffc1f54afb [unknown] ([unknown]) ls 2695501 3011030.487018: 500000 cycles: ffffffff9968382f [unknown] ([unknown]) ls 2695501 3011030.487019: 500000 cycles: ffffffff99e00094 [unknown] ([unknown]) ls 2695501 3011030.487019: 500000 cycles: 152cc718a8d0 __errno_location@plt+0x0 (/usr/lib/x86_64-linux-gnu/libselinux.so.1+0x68d0) ls 2695501 3011030.487019: 500000 cycles: ffffffff992a6db0 [unknown] ([unknown]) Signed-off-by:Changbin Du <changbin.du@huawei.com> Acked-by:
-
Changbin Du authored
This adds a helper function map__fprintf_dsoname_dsoff() to print dsoname with optional dso offset. Suggested-by:
-
Paran Lee authored
Add RISC-V binutils path on lookup triplets. Reviewed-by:
Palmer Dabbelt <palmer@rivosinc.com> Signed-off-by:
Paran Lee <p4ranlee@gmail.com> Acked-by:
-
- 10 May, 2023 8 commits
-
-
Jiri Olsa authored
If 'struct rq' isn't defined in lock_contention.bpf.c then the type for the 'runqueue' variable ends up being a forward declaration (BTF_KIND_FWD) while the kernel has it defined (BTF_KIND_STRUCT). This makes libbpf decide it has incompatible types and then fails to load the BPF skeleton: # perf lock con -ab sleep 1 libbpf: extern (var ksym) 'runqueues': incompatible types, expected [95] fwd rq, but kernel has [55509] struct rq libbpf: failed to load object 'lock_contention_bpf' libbpf: failed to load BPF skeleton 'lock_contention_bpf': -22 Failed to load lock-contention BPF skeleton lock contention BPF setup failed # Add it as an empty struct to satisfy that type verification: # perf lock con -ab sleep 1 contended total wait max wait avg wait type caller 2 50.64 us 25.38 us 25.32 us spinlock tick_do_update_jiffies64+0x25 1 26.18 us 26.18 us 26.18 us spinlock tick_do_update_jiffies64+0x25 # Committer notes: Extracted from a larger patch as Namhyung had already fixed the other issues in e53de7b6 ("perf lock contention: Fix struct rq lock access"). Signed-off-by:Jiri Olsa <jolsa@kernel.org> Tested-by:
Namhyung Kim <namhyung@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Andrii Nakryiko <andrii@kernel.org> Cc: Changbin Du <changbin.du@huawei.com> Cc: Hao Luo <haoluo@google.com> Cc: Ian Rogers <irogers@google.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Song Liu <song@kernel.org> Cc: Stephane Eranian <eranian@google.com> Cc: Yang Jihong <yangjihong1@huawei.com> Link: https://lore.kernel.org/lkml/ZFVqeKLssg7uzxzI@kravaSigned-off-by:
-
James Clark authored
Pre 5.11 kernels don't support 'contextid1' and 'contextid2' so validation would be skipped. By adding an additional check for 'contextid', old kernels will still have validation done even though contextid would either be contextid1 or contextid2. Additionally now that it's possible to override options, an existing bug in the validation is revealed. 'val' is overwritten by the contextid1 validation, and re-used for contextid2 validation causing it to always fail. '!val || val != 0x4' is the same as 'val != 0x4' because 0 is also != 4, so that expression can be simplified and the temp variable not overwritten. Fixes: 35c51f83 ("perf cs-etm: Validate options after applying them") Reviewed-by:
Leo Yan <leo.yan@linaro.org> Signed-off-by:
James Clark <james.clark@arm.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ian Rogers <irogers@google.com> 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: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Poirier <mathieu.poirier@linaro.org> Cc: Mike Leach <mike.leach@linaro.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Suzuki Poulouse <suzuki.poulose@arm.com> Cc: Will Deacon <will@kernel.org> Cc: coresight@lists.linaro.org Cc: linux-arm-kernel@lists.infradead.org Link: https://lore.kernel.org/all/20230501073452.GA4660@leoy-yangtze.lan Link: https://lore.kernel.org/r/20230504144822.1938717-1-james.clark@arm.comSigned-off-by:
-
James Clark authored
With EXTRA_CFLAGS=-DREFCNT_CHECKING=1 and build-test, some unwrapped map accesses appear. Wrap it in the new accessor to fix the error: error: 'struct perf_cpu_map' has no member named 'map' Signed-off-by:
-
Sandipan Das authored
When using the global aggregation mode, running perf script after perf stat record can result in a segmentation fault as seen with commit 8b76a318 ("perf stat: Remove unused perf_counts.aggr field"). Add a basic test to the existing suite of stat-related tests for checking if that workflow runs without erroring out. Signed-off-by:
Sandipan Das <sandipan.das@amd.com> Acked-by:
-
Sandipan Das authored
The script command does not support aggregation modes by itself although that can be achieved using post-processing scripts. Because of this, it does not allocate memory for aggregated event values. Upon running perf stat record, the aggregation mode is set in the perf data file. If the mode is AGGR_GLOBAL, the aggregated event values are accessed and this leads to a segmentation fault since these were never allocated to begin with. Set the mode to AGGR_NONE explicitly to avoid this. E.g. $ perf stat record -e cycles true $ perf script Before: Segmentation fault (core dumped) After: CPU THREAD VAL ENA RUN TIME EVENT -1 231919 162831 362069 362069 935289 cycles:u Fixes: 8b76a318 ("perf stat: Remove unused perf_counts.aggr field") Signed-off-by:
-
Ian Rogers authored
There are insufficient headers in tools/include to satisfy building BPF programs and their header dependencies. Add the system include paths from the non-BPF clang compile so that these headers can be found. This code was taken from: tools/testing/selftests/bpf/Makefile Committer notes: Had to adjust the '#ifndef NO_BPF_SKEL' to '#ifdef BUILD_BPF_SKEL' as reverted that build BPF skels by default. Also cope with the addition of -I$(srctree)/tools/include/uapi done by Yang Jihong so that we prefer using the kernel sources headers instead of older ones in the system. Signed-off-by:
-
Yang Jihong authored
Currently, vmlinux.h uses the bpf.h and perf_event.h header files in the system path. If the header files in compilation environment are old, compilation may fail. For example: /home/yangjihong/linux/tools/perf/util/bpf_skel/.tmp/../vmlinux.h:151:27: error: field has incomplete type 'union perf_sample_weight' union perf_sample_weight weight; Use the bpf.h and perf_event.h files in the source code directory to avoid compilation compatibility problems. Signed-off-by:Yang Jihong <yangjihong1@huawei.com> Tested-by:
-
Adrian Hunter authored
Do not assume which events may have a PMU name, allowing the logic to keep an AUX event group together. Example: Before: $ perf record --no-bpf-event -c 10 -e '{intel_pt//,tlb_flush.stlb_any/aux-sample-size=8192/pp}:u' -- sleep 0.1 WARNING: events were regrouped to match PMUs Cannot add AUX area sampling to a group leader $ After: $ perf record --no-bpf-event -c 10 -e '{intel_pt//,tlb_flush.stlb_any/aux-sample-size=8192/pp}:u' -- sleep 0.1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.078 MB perf.data ] $ perf script -F-dso,+addr | grep -C5 tlb_flush.stlb_any | head -11 sleep 20444 [003] 7939.510243: 1 branches:uH: 7f5350cc82a2 dl_main+0x9a2 => 7f5350cb38f0 _dl_add_to_namespace_list+0x0 sleep 20444 [003] 7939.510243: 1 branches:uH: 7f5350cb3908 _dl_add_to_namespace_list+0x18 => 7f5350cbb080 rtld_mutex_dummy+0x0 sleep 20444 [003] 7939.510243: 1 branches:uH: 7f5350cc8350 dl_main+0xa50 => 0 [unknown] sleep 20444 [003] 7939.510244: 1 branches:uH: 7f5350cc83ca dl_main+0xaca => 7f5350caeb60 _dl_process_pt_gnu_property+0x0 sleep 20444 [003] 7939.510245: 1 branches:uH: 7f5350caeb60 _dl_process_pt_gnu_property+0x0 => 0 [unknown] sleep 20444 7939.510245: 10 tlb_flush.stlb_any/aux-sample-size=8192/pp: 0 7f5350caeb60 _dl_process_pt_gnu_property+0x0 sleep 20444 [003] 7939.510254: 1 branches:uH: 7f5350cc87fe dl_main+0xefe => 7f5350ccd240 strcmp+0x0 sleep 20444 [003] 7939.510254: 1 branches:uH: 7f5350cc8862 dl_main+0xf62 => 0 [unknown] sleep 20444 [003] 7939.510255: 1 branches:uH: 7f5350cc9cdc dl_main+0x23dc => 0 [unknown] sleep 20444 [003] 7939.510257: 1 branches:uH: 7f5350cc89f6 dl_main+0x10f6 => 7f5350cb9530 _dl_setup_hash+0x0 sleep 20444 [003] 7939.510257: 1 branches:uH: 7f5350cc8a2d dl_main+0x112d => 7f5350cb3990 _dl_new_object+0x0 $ Fixes: 347c2f0a ("perf parse-events: Sort and group parsed events") Suggested-by:
-
