- 17 Mar, 2020 2 commits
-
-
Jin Yao authored
Previously we could get the report of branch type statistics. For example: # perf record -j any,save_type ... # t perf report --stdio # # Branch Statistics: # COND_FWD: 40.6% COND_BWD: 4.1% CROSS_4K: 24.7% CROSS_2M: 12.3% COND: 44.7% UNCOND: 0.0% IND: 6.1% CALL: 24.5% RET: 24.7% But now for the recent perf, it can't report the branch type statistics. It's a regression issue caused by commit 40c39e30 ("perf report: Fix a no annotate browser displayed issue"), which only counts the branch type statistics for browser mode. This patch moves the branch_type_count() outside of ui__has_annotation() checking, then branch type statistics can work for stdio mode. Fixes: 40c39e30 ("perf report: Fix a no annotate browser displayed issue") Signed-off-by:Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200313134607.12873-1-yao.jin@linux.intel.comSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Ian Rogers authored
When mmap2 events are synthesized the ino_generation field isn't being set leading to uninitialized memory being compared. Caught with clang's -fsanitize=memory: ==124733==WARNING: MemorySanitizer: use-of-uninitialized-value #0 0x55a96a6a65cc in __dso_id__cmp tools/perf/util/dsos.c:23:6 #1 0x55a96a6a81d5 in dso_id__cmp tools/perf/util/dsos.c:38:9 #2 0x55a96a6a717f in __dso__cmp_long_name tools/perf/util/dsos.c:74:15 #3 0x55a96a6a6c4c in __dsos__findnew_link_by_longname_id tools/perf/util/dsos.c:106:12 #4 0x55a96a6a851e in __dsos__findnew_by_longname_id tools/perf/util/dsos.c:178:9 #5 0x55a96a6a7798 in __dsos__find_id tools/perf/util/dsos.c:191:9 #6 0x55a96a6a7b57 in __dsos__findnew_id tools/perf/util/dsos.c:251:20 #7 0x55a96a6a7a57 in dsos__findnew_id tools/perf/util/dsos.c:259:17 #8 0x55a96a7776ae in machine__findnew_dso_id tools/perf/util/machine.c:2709:9 #9 0x55a96a77dfcf in map__new tools/perf/util/map.c:193:10 #10 0x55a96a77240a in machine__process_mmap2_event tools/perf/util/machine.c:1670:8 #11 0x55a96a7741a3 in machine__process_event tools/perf/util/machine.c:1882:9 #12 0x55a96a6aee39 in perf_event__process tools/perf/util/event.c:454:9 #13 0x55a96a87d633 in perf_tool__process_synth_event tools/perf/util/synthetic-events.c:63:9 #14 0x55a96a87f131 in perf_event__synthesize_mmap_events tools/perf/util/synthetic-events.c:403:7 #15 0x55a96a8815d6 in __event__synthesize_thread tools/perf/util/synthetic-events.c:548:9 #16 0x55a96a882bff in __perf_event__synthesize_threads tools/perf/util/synthetic-events.c:681:3 #17 0x55a96a881ec2 in perf_event__synthesize_threads tools/perf/util/synthetic-events.c:750:9 #18 0x55a96a562b26 in synth_all tools/perf/tests/mmap-thread-lookup.c:136:9 #19 0x55a96a5623b1 in mmap_events tools/perf/tests/mmap-thread-lookup.c:174:8 #20 0x55a96a561fa0 in test__mmap_thread_lookup tools/perf/tests/mmap-thread-lookup.c:230:2 #21 0x55a96a52c182 in run_test tools/perf/tests/builtin-test.c:378:9 #22 0x55a96a52afc1 in test_and_print tools/perf/tests/builtin-test.c:408:9 #23 0x55a96a52966e in __cmd_test tools/perf/tests/builtin-test.c:603:4 #24 0x55a96a52855d in cmd_test tools/perf/tests/builtin-test.c:747:9 #25 0x55a96a2844d4 in run_builtin tools/perf/perf.c:312:11 #26 0x55a96a282bd0 in handle_internal_command tools/perf/perf.c:364:8 #27 0x55a96a284097 in run_argv tools/perf/perf.c:408:2 #28 0x55a96a282223 in main tools/perf/perf.c:538:3 Uninitialized value was stored to memory at #1 0x55a96a6a18f7 in dso__new_id tools/perf/util/dso.c:1230:14 #2 0x55a96a6a78ee in __dsos__addnew_id tools/perf/util/dsos.c:233:20 #3 0x55a96a6a7bcc in __dsos__findnew_id tools/perf/util/dsos.c:252:21 #4 0x55a96a6a7a57 in dsos__findnew_id tools/perf/util/dsos.c:259:17 #5 0x55a96a7776ae in machine__findnew_dso_id tools/perf/util/machine.c:2709:9 #6 0x55a96a77dfcf in map__new tools/perf/util/map.c:193:10 #7 0x55a96a77240a in machine__process_mmap2_event tools/perf/util/machine.c:1670:8 #8 0x55a96a7741a3 in machine__process_event tools/perf/util/machine.c:1882:9 #9 0x55a96a6aee39 in perf_event__process tools/perf/util/event.c:454:9 #10 0x55a96a87d633 in perf_tool__process_synth_event tools/perf/util/synthetic-events.c:63:9 #11 0x55a96a87f131 in perf_event__synthesize_mmap_events tools/perf/util/synthetic-events.c:403:7 #12 0x55a96a8815d6 in __event__synthesize_thread tools/perf/util/synthetic-events.c:548:9 #13 0x55a96a882bff in __perf_event__synthesize_threads tools/perf/util/synthetic-events.c:681:3 #14 0x55a96a881ec2 in perf_event__synthesize_threads tools/perf/util/synthetic-events.c:750:9 #15 0x55a96a562b26 in synth_all tools/perf/tests/mmap-thread-lookup.c:136:9 #16 0x55a96a5623b1 in mmap_events tools/perf/tests/mmap-thread-lookup.c:174:8 #17 0x55a96a561fa0 in test__mmap_thread_lookup tools/perf/tests/mmap-thread-lookup.c:230:2 #18 0x55a96a52c182 in run_test tools/perf/tests/builtin-test.c:378:9 #19 0x55a96a52afc1 in test_and_print tools/perf/tests/builtin-test.c:408:9 Uninitialized value was stored to memory at #0 0x55a96a7725af in machine__process_mmap2_event tools/perf/util/machine.c:1646:25 #1 0x55a96a7741a3 in machine__process_event tools/perf/util/machine.c:1882:9 #2 0x55a96a6aee39 in perf_event__process tools/perf/util/event.c:454:9 #3 0x55a96a87d633 in perf_tool__process_synth_event tools/perf/util/synthetic-events.c:63:9 #4 0x55a96a87f131 in perf_event__synthesize_mmap_events tools/perf/util/synthetic-events.c:403:7 #5 0x55a96a8815d6 in __event__synthesize_thread tools/perf/util/synthetic-events.c:548:9 #6 0x55a96a882bff in __perf_event__synthesize_threads tools/perf/util/synthetic-events.c:681:3 #7 0x55a96a881ec2 in perf_event__synthesize_threads tools/perf/util/synthetic-events.c:750:9 #8 0x55a96a562b26 in synth_all tools/perf/tests/mmap-thread-lookup.c:136:9 #9 0x55a96a5623b1 in mmap_events tools/perf/tests/mmap-thread-lookup.c:174:8 #10 0x55a96a561fa0 in test__mmap_thread_lookup tools/perf/tests/mmap-thread-lookup.c:230:2 #11 0x55a96a52c182 in run_test tools/perf/tests/builtin-test.c:378:9 #12 0x55a96a52afc1 in test_and_print tools/perf/tests/builtin-test.c:408:9 #13 0x55a96a52966e in __cmd_test tools/perf/tests/builtin-test.c:603:4 #14 0x55a96a52855d in cmd_test tools/perf/tests/builtin-test.c:747:9 #15 0x55a96a2844d4 in run_builtin tools/perf/perf.c:312:11 #16 0x55a96a282bd0 in handle_internal_command tools/perf/perf.c:364:8 #17 0x55a96a284097 in run_argv tools/perf/perf.c:408:2 #18 0x55a96a282223 in main tools/perf/perf.c:538:3 Uninitialized value was created by a heap allocation #0 0x55a96a22f60d in malloc llvm/llvm-project/compiler-rt/lib/msan/msan_interceptors.cpp:925:3 #1 0x55a96a882948 in __perf_event__synthesize_threads tools/perf/util/synthetic-events.c:655:15 #2 0x55a96a881ec2 in perf_event__synthesize_threads tools/perf/util/synthetic-events.c:750:9 #3 0x55a96a562b26 in synth_all tools/perf/tests/mmap-thread-lookup.c:136:9 #4 0x55a96a5623b1 in mmap_events tools/perf/tests/mmap-thread-lookup.c:174:8 #5 0x55a96a561fa0 in test__mmap_thread_lookup tools/perf/tests/mmap-thread-lookup.c:230:2 #6 0x55a96a52c182 in run_test tools/perf/tests/builtin-test.c:378:9 #7 0x55a96a52afc1 in test_and_print tools/perf/tests/builtin-test.c:408:9 #8 0x55a96a52966e in __cmd_test tools/perf/tests/builtin-test.c:603:4 #9 0x55a96a52855d in cmd_test tools/perf/tests/builtin-test.c:747:9 #10 0x55a96a2844d4 in run_builtin tools/perf/perf.c:312:11 #11 0x55a96a282bd0 in handle_internal_command tools/perf/perf.c:364:8 #12 0x55a96a284097 in run_argv tools/perf/perf.c:408:2 #13 0x55a96a282223 in main tools/perf/perf.c:538:3 SUMMARY: MemorySanitizer: use-of-uninitialized-value tools/perf/util/dsos.c:23:6 in __dso_id__cmp Signed-off-by:Ian Rogers <irogers@google.com> Acked-by:
Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: clang-built-linux@googlegroups.com Link: http://lore.kernel.org/lkml/20200313053129.131264-1-irogers@google.comSigned-off-by:
-
- 13 Mar, 2020 1 commit
-
-
Ian Rogers authored
If the shell test directory isn't present the exit code will be 255 but with no error messages printed. Add an error message. Signed-off-by:
-
- 12 Mar, 2020 1 commit
-
-
Alexey Budankov authored
Correct maxnode parameter value passed to mbind() syscall to be the amount of node mask bits to analyze plus 1. Dynamically allocate node mask memory depending on the index of node of cpu being profiled. Fixes: c44a8b44 ("perf record: Bind the AIO user space buffers to nodes") Signed-off-by:
Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/c7ea8ffe-1357-bf9e-3a89-1da1d8e9b75b@linux.intel.com [ Remove leftover nr_bits + 1 comment in mbind() call ] Signed-off-by:
-
- 11 Mar, 2020 12 commits
-
-
Michael Petlan authored
Since common_callchain has been added to the argument array, we need to reflect it in perl-based scripts, because otherwise the following args would be shifted and thus incorrect. E.g. rw-by-pid and calculation of read and written bytes: Before: read counts by pid: pid comm # reads bytes_requested bytes_read ------ -------------------- ----------- ---------- ---------- 19301 dd 4 424510450039736 0 After: read counts by pid: pid comm # reads bytes_requested bytes_read ------ -------------------- ----------- ---------- ---------- 19301 dd 4 9536 4341 Committer testing: To see before after first do: # perf script record rw-by-pid ^C Now you'll have a perf.data file to report on, then do before and after using: # perf script report rw-by-pid Anbd notice the bytes_request/bytes_read, as above. Signed-off-by:Michael Petlan <mpetlan@redhat.com> Tested-by:
-
Adrian Hunter authored
Make it easy for people looking in intel-pt.txt to find the new file. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Tested-by:
-
Adrian Hunter authored
Add references to Intel PT man page in man pages of associated tools. Signed-off-by:
-
Adrian Hunter authored
Make the Intel PT documentation into a man page. Signed-off-by:
-
Ian Rogers authored
Currently the man page dates reflect the date the man pages were built. This patch adjusts the date so that the date is when then man page last had a commit against it. The date is generated using 'git log'. Committer testing: $ git log -1 --pretty="format:%cd" --date=short tools/perf/Documentation/perf-top.txt 2020-01-14 Before: rm -rf /tmp/build/perf mkdir -p /tmp/build/perf make -C tools/perf O=/tmp/build/perf/ install $ date Wed 11 Mar 2020 10:21:19 AM -03 $ man perf-top | tail -1 perf 03/11/2020 PERF-TOP(1) $ After: rm -rf /tmp/build/perf mkdir -p /tmp/build/perf make -C tools/perf O=/tmp/build/perf/ install $ date $ date Wed 11 Mar 2020 10:24:06 AM -03 $ man perf-top | tail -1 perf 2020-01-14 PERF-TOP(1) $ Signed-off-by:
-
Leo Yan authored
The variable 'offset' in function cs_etm__sample() is u64 type, it's not appropriate to check it with 'while (offset > 0)'; this patch changes to 'while (offset)'. Signed-off-by:
Leo Yan <leo.yan@linaro.org> Reviewed-by:
Mathieu Poirier <mathieu.poirier@linaro.org> Reviewed-by:
Mike Leach <mike.leach@linaro.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Robert Walker <robert.walker@arm.com> Cc: Suzuki Poulouse <suzuki.poulose@arm.com> Cc: coresight ml <coresight@lists.linaro.org> Cc: linux-arm-kernel@lists.infradead.org Link: http://lore.kernel.org/lkml/20200219021811.20067-6-leo.yan@linaro.orgSigned-off-by:
-
Leo Yan authored
If an instruction range packet can generate multiple instruction samples, these samples share the same last branches; it's not necessary to copy the same last branches repeatedly for these samples within the same packet. This patch moves out the last branches copying from function cs_etm__synth_instruction_sample(), and execute it prior to generating instruction samples. Signed-off-by:
-
Leo Yan authored
When 'etm->instructions_sample_period' is less than 'tidq->period_instructions', the function cs_etm__sample() cannot handle this case properly with its logic. Let's see below flow as an example: - If we set itrace option '--itrace=i4', then function cs_etm__sample() has variables with initialized values: tidq->period_instructions = 0 etm->instructions_sample_period = 4 - When the first packet is coming: packet->instr_count = 10; the number of instructions executed in this packet is 10, thus update period_instructions as below: tidq->period_instructions = 0 + 10 = 10 instrs_over = 10 - 4 = 6 offset = 10 - 6 - 1 = 3 tidq->period_instructions = instrs_over = 6 - When the second packet is coming: packet->instr_count = 10; in the second pass, assume 10 instructions in the trace sample again: tidq->period_instructions = 6 + 10 = 16 instrs_over = 16 - 4 = 12 offset = 10 - 12 - 1 = -3 -> the negative value tidq->period_instructions = instrs_over = 12 So after handle these two packets, there have below issues: The first issue is that cs_etm__instr_addr() returns the address within the current trace sample of the instruction related to offset, so the offset is supposed to be always unsigned value. But in fact, function cs_etm__sample() might calculate a negative offset value (in handling the second packet, the offset is -3) and pass to cs_etm__instr_addr() with u64 type with a big positive integer. The second issue is it only synthesizes 2 samples for sample period = 4. In theory, every packet has 10 instructions so the two packets have total 20 instructions, 20 instructions should generate 5 samples (4 x 5 = 20). This is because cs_etm__sample() only calls once cs_etm__synth_instruction_sample() to generate instruction sample per range packet. This patch fixes the logic in function cs_etm__sample(); the basic idea for handling coming packet is: - To synthesize the first instruction sample, it combines the left instructions from the previous packet and the head of the new packet; then generate continuous samples with sample period; - At the tail of the new packet, if it has the rest instructions, these instructions will be left for the sequential sample. Suggested-by:
-
Leo Yan authored
Every time synthesize instruction sample, the last branch recording will be reset. This is fine if the instruction period is big enough, for example if use the option '--itrace=i100000', the last branch array is reset for every sample with 100000 instructions per period; before generate the next instruction sample, there has the sufficient packets coming to fill the last branch array. On the other hand, if set a very small period, the packets will be significantly reduced between two continuous instruction samples, thus the last branch array is almost empty for new instruction sample by frequently resetting. To allow the last branches to work properly for any instruction periods, this patch avoids to reset the last branch for every instruction sample and only reset it when flush the trace data. The last branches will be reset only for two cases, one is for trace starting, another case is for discontinuous trace; other cases can keep recording last branches for continuous instruction samples. Signed-off-by:
-
Leo Yan authored
If use option '--itrace=iNNN' with Arm CoreSight trace data, perf tool fails inject instruction samples; the root cause is the packets are only swapped for branch samples and last branches but not for instruction samples, so the new coming packets cannot be properly handled for only synthesizing instruction samples. To fix this issue, this patch refactors the code with a new function cs_etm__packet_swap() which is used to swap packets and adds the condition for instruction samples. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
And add the '/' to avoid looking at things like "/system/libsomething", when all we want to know if it is like "/system/lib/something", i.e. if it is in that system library dir. Using strstarts() avoids off-by-one errors like recently fixed in this file. Since this adds the '/' I separated this patch, another patch will make this consistent by removing other strncmp(str, prefix, manually calculated prefix length) usage. Reported-by:
Dominik Czarnota <dominik.b.czarnota@gmail.com> Acked-by:
-
disconnect3d authored
This patch fixes an off-by-one error in strncpy size argument in tools/perf/util/map.c. The issue is that in: strncmp(filename, "/system/lib/", 11) the passed string literal: "/system/lib/" has 12 bytes (without the NULL byte) and the passed size argument is 11. As a result, the logic won't match the ending "/" byte and will pass filepaths that are stored in other directories e.g. "/system/libmalicious/bin" or just "/system/libmalicious". This functionality seems to be present only on Android. I assume the /system/ directory is only writable by the root user, so I don't think this bug has much (or any) security impact. Fixes: eca81836 ("perf tools: Add automatic remapping of Android libraries") Signed-off-by:
-
- 10 Mar, 2020 18 commits
-
-
Kan Liang authored
Add NO_NMI_WATCHDOG metric constraint to Page_Walks_Utilization for Sky Lake and Cascade Lake. Committer testing: On a Lenovo T480S, Intel(R) Core(TM) i7-8650U Kaby Lake, that looking at x86's mapfile.csv file is a: $ grep -w skylake tools/perf/pmu-events/arch/x86/mapfile.csv GenuineIntel-6-[4589]E,v24,skylake,core $ So uses the constraint added in this patch in this file: tools/perf/pmu-events/arch/x86/skylake/skl-metrics.json Before: # perf stat -a -M Page_Walks_Utilization sleep 2 Performance counter stats for 'system wide': <not counted> itlb_misses.walk_pending (0.00%) <not counted> dtlb_load_misses.walk_pending (0.00%) <not counted> dtlb_store_misses.walk_pending (0.00%) <not counted> ept.walk_pending (0.00%) <not counted> cycles (0.00%) 2.001750514 seconds time elapsed Some events weren't counted. Try disabling the NMI watchdog: echo 0 > /proc/sys/kernel/nmi_watchdog perf stat ... echo 1 > /proc/sys/kernel/nmi_watchdog The events in group usually have to be from the same PMU. Try reorganizing the group. # After: # perf stat -a -M Page_Walks_Utilization sleep 2 Splitting metric group Page_Walks_Utilization into standalone metrics. Try disabling the NMI watchdog to comply NO_NMI_WATCHDOG metric constraint: echo 0 > /proc/sys/kernel/nmi_watchdog perf stat ... echo 1 > /proc/sys/kernel/nmi_watchdog , Performance counter stats for 'system wide': 36,883,102 itlb_misses.walk_pending # 0.1 Page_Walks_Utilization (79.99%) 123,104,146 dtlb_load_misses.walk_pending (80.02%) 13,720,795 dtlb_store_misses.walk_pending (79.99%) 0 ept.walk_pending (79.99%) 1,519,948,400 cycles (80.01%) 2.002170780 seconds time elapsed # Before and after, if we disable the nmi_watchdog we get: # echo 0 > /proc/sys/kernel/nmi_watchdog # perf stat -a -M Page_Walks_Utilization sleep 2 Performance counter stats for 'system wide': 33,721,658 itlb_misses.walk_pending # 0.1 Page_Walks_Utilization 84,070,996 dtlb_load_misses.walk_pending 9,816,071 dtlb_store_misses.walk_pending 0 ept.walk_pending 704,920,899 cycles 2.002331670 seconds time elapsed # More information about the metric expressions: # perf stat -v -a -M Page_Walks_Utilization sleep 2 Using CPUID GenuineIntel-6-8E-A metric expr ( itlb_misses.walk_pending + dtlb_load_misses.walk_pending + dtlb_store_misses.walk_pending + ept.walk_pending ) / ( 2 * cycles ) for Page_Walks_Utilization found event itlb_misses.walk_pending found event dtlb_load_misses.walk_pending found event dtlb_store_misses.walk_pending found event ept.walk_pending found event cycles adding {itlb_misses.walk_pending,dtlb_load_misses.walk_pending,dtlb_store_misses.walk_pending,ept.walk_pending,cycles}:W -> cpu/umask=0x10,(null)=0x186a3,event=0x85/ -> cpu/umask=0x10,(null)=0x1e8483,event=0x8/ -> cpu/umask=0x10,(null)=0x1e8483,event=0x49/ -> cpu/umask=0x10,(null)=0x1e8483,event=0x4f/ itlb_misses.walk_pending: 8085772 16010162799 16010162799 dtlb_load_misses.walk_pending: 28134579 16010162799 16010162799 dtlb_store_misses.walk_pending: 7276535 16010162799 16010162799 ept.walk_pending: 2 16010162799 16010162799 cycles: 315140605 16010162799 16010162799 Performance counter stats for 'system wide': 8,085,772 itlb_misses.walk_pending # 0.1 Page_Walks_Utilization 28,134,579 dtlb_load_misses.walk_pending 7,276,535 dtlb_store_misses.walk_pending 2 ept.walk_pending 315,140,605 cycles 2.002333181 seconds time elapsed # Signed-off-by:Kan Liang <kan.liang@linux.intel.com> Acked-by:
Jiri Olsa <jolsa@redhat.com> Tested-by:
-
Kan Liang authored
Some metric groups have metric constraints. A metric group can be scheduled as a group only when some constraints are applied. For example, Page_Walks_Utilization has a metric constraint, "NO_NMI_WATCHDOG". When NMI watchdog is disabled, the metric group can be scheduled as a group. Otherwise, splitting the metric group into standalone metrics. Add a new function, metricgroup__has_constraint(), to check whether all constraints are applied. If not, splitting the metric group into standalone metrics. Currently, only one constraint, "NO_NMI_WATCHDOG", is checked. Print a warning for the metric group with the constraint, when NMI WATCHDOG is enabled. Signed-off-by:
-
Kan Liang authored
The NMI watchdog status is required for metric group constraint examination. Factor out sysctl__nmi_watchdog_enabled() to retrieve the NMI watchdog status. Users may count more than one metric group each time. If so, the NMI watchdog status may be retrieved several times. To reduce the overhead, cache the NMI watchdog status. Replace the NMI watchdog status checking in print_footer() by sysctl__nmi_watchdog_enabled(). Suggested-by:
Andi Kleen <ak@linux.intel.com> Signed-off-by:
-
Kan Liang authored
Factor out metricgroup__add_metric_weak_group() which add metrics into a weak group. The change can improve code readability. Because following patch will introduce a function which add standalone metrics. Signed-off-by:
-
Kan Liang authored
A new field "MetricConstraint" is introduced in JSON event list. Extend jevents to parse the field and save the value in metric_constraint. Signed-off-by:
-
Thomas Richter authored
Add support for new deflate counters: - Counter 247: cycles CPU spent obtaining access to Deflate unit - Counter 252: cycles CPU is using Deflate unit - Counter 264: Increments by one for every DEFLATE CONVERSION CALL instruction executed. - Counter 265: Increments by one for every DEFLATE CONVERSION CALL instruction executed that ended in Condition Codes 0, 1 or 2. Also adjust the some crypto counter description to latest documentation. Signed-off-by:
Thomas Richter <tmricht@linux.ibm.com> Reviewed-by:
Sumanth Korikkar <sumanthk@linux.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Vasily Gorbik <gor@linux.ibm.com> Link: http://lore.kernel.org/lkml/20200310142937.32045-1-tmricht@linux.ibm.comSigned-off-by:
-
Jin Yao authored
It would be nice to print the block percents with colors. This patch supports the 'Sampled Cycles%' and 'Avg Cycles%' printed in colors. For example, perf record -b ... perf report --total-cycles or perf report --total-cycles --stdio percent > 5%, colored in red percent > 0.5%, colored in green percent < 0.5%, default color Signed-off-by:
-
Jin Yao authored
Currently we use a predefined array to set the block info output formats, it's fixed and inflexible. This patch adds two parameters "block_hpps" and "nr_hpps" in block_info__create_report and other static functions, in order to let user decide which columns to report and with specified report ordering. It should be more flexible. Buffers will be allocated to contain the new fmts, of course, we need to release them before perf exits. Signed-off-by:
-
Jin Yao authored
'perf diff' uses block_pair_cmp() to compare two blocks. But block_info__cmp() has the similar functionality and it's a bit more complete. This patch removes block_pair_cmp() and uses __block_info__cmp() instead. __block_info__cmp() is wrapped by block_info__cmp() and it doesn't receives a perf_hpp_fmt parameter. Signed-off-by:
-
Jin Yao authored
Commit 60414418 ("perf block: Cleanup and refactor block info functions") introduces block_info__cmp(), which compares two blocks. But the issues are: 1. It should return the strcmp cmp value only if it's not 0. 2. When symbol names are matched, we need to compare the addresses of blocks further. But it wrongly uses the symbol addresses for comparison. 3. If the syms are both NULL, we can't consider these two blocks are matched. This patch fixes above 3 issues. Fixes: 60414418 ("perf block: Cleanup and refactor block info functions") Signed-off-by:
-
Jiri Olsa authored
To match the error value of the expr__find_other function, so all exported expr functions return the same values: 0 on success, -1 on error. Signed-off-by:
-
Jiri Olsa authored
Now that we have a flex parser we don't need to update the parsed string pointer, so the interface can just be passed the pointer to the expression instead of a pointer to pointer. Signed-off-by:
-
Jiri Olsa authored
We have metrics that define more than 15 variables, like Branch_Misprediction_Cost. Increasing the allowed variables count to 20. As Andy pointed out, we can't go too high in here, because some of the code has O(n^2) complexity (already_seen) and we might want to do some other changes (like using hash tables) before increasing the maximum even more. Signed-off-by:
-
Jiri Olsa authored
Adding expr flex code instead of the manual parser code. So it's easily extensible in upcoming changes. The new flex code is in flex.l object and gets compiled like all the other flexers we use. It's defined as flex reentrant parser. It's used by both expr__parse and expr__find_other interfaces by separating the starting point. There's no intended change of functionality ;-) the test expr is passing. Signed-off-by:
-
Jiri Olsa authored
Add generic expr code into new expr.c object. The expr.c object will be mainly used in following change that will get rid of the manual flex code, Signed-off-by:
-
Kan Liang authored
The perf.data may be generated by a newer version of perf tool, which support new input bits in attr, e.g. new bit for branch_sample_type. The perf.data may be parsed by an older version of perf tool later. The old perf tool may parse the perf.data incorrectly. There is no warning message for this case. Current perf header never check for unknown input bits in attr. When read the event desc from header, check the stored event attr. The reserved bits, sample type, read format and branch sample type will be checked. Signed-off-by:
-
Kan Liang authored
A new branch sample type PERF_SAMPLE_BRANCH_HW_INDEX has been introduced in latest kernel. Enable HW_INDEX by default in LBR call stack mode. If kernel doesn't support the sample type, switching it off. Add HW_INDEX in attr_fprintf as well. User can check whether the branch sample type is set via debug information or header. Committer testing: First collect some samples with LBR callchains, system wide, for a few seconds: # perf record --call-graph lbr -a sleep 5 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.625 MB perf.data (224 samples) ] # Now lets use 'perf evlist -v' to look at the branch_sample_type: # perf evlist -v cycles: size: 120, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|CALLCHAIN|CPU|PERIOD|BRANCH_STACK, read_format: ID, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1, task: 1, precise_ip: 3, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1, branch_sample_type: USER|CALL_STACK|NO_FLAGS|NO_CYCLES|HW_INDEX # So the machine has the kernel feature, and it was correctly added to perf_event_attr.branch_sample_type, for the default 'cycles' event. If we do it in another machine, where the kernel lacks the HW_INDEX feature, we get: # perf record --call-graph lbr -a sleep 2s [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 1.690 MB perf.data (499 samples) ] # perf evlist -v cycles: size: 120, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|CALLCHAIN|CPU|PERIOD|BRANCH_STACK, read_format: ID, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1, task: 1, precise_ip: 3, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1, branch_sample_type: USER|CALL_STACK|NO_FLAGS|NO_CYCLES # No HW_INDEX in attr.branch_sample_type. Signed-off-by: -
Kan Liang authored
The low level index of raw branch records for the most recent branch can be recorded in a sample with PERF_SAMPLE_BRANCH_HW_INDEX branch_sample_type. Extend struct branch_stack to support it. However, if the PERF_SAMPLE_BRANCH_HW_INDEX is not applied, only nr and entries[] will be output by kernel. The pointer of entries[] could be wrong, since the output format is different with new struct branch_stack. Add a variable no_hw_idx in struct perf_sample to indicate whether the hw_idx is output. Add get_branch_entry() to return corresponding pointer of entries[0]. To make dummy branch sample consistent as new branch sample, add hw_idx in struct dummy_branch_stack for cs-etm and intel-pt. Apply the new struct branch_stack for synthetic events as well. Extend test case sample-parsing to support new struct branch_stack. Committer notes: Renamed get_branch_entries() to perf_sample__branch_entries() to have proper namespacing and pave the way for this to be moved to libperf, eventually. Add 'static' to that inline as it is in a header. Add 'hw_idx' to 'struct dummy_branch_stack' in cs-etm.c to fix the build on arm64. Signed-off-by:
-
- 05 Mar, 2020 1 commit
-
-
Arnaldo Carvalho de Melo authored
To get the changes in: bbfd5e4f ("perf/core: Add new branch sample type for HW index of raw branch records") This silences this perf tools build warning: Warning: Kernel ABI header at 'tools/include/uapi/linux/perf_event.h' differs from latest version at 'include/uapi/linux/perf_event.h' diff -u tools/include/uapi/linux/perf_event.h include/uapi/linux/perf_event.h This update is a prerequisite to adding support for the HW index of raw branch records. Acked-by:
-
- 04 Mar, 2020 5 commits
-
-
Steven Rostedt (VMware) authored
If the precision of print_event_time() is zero or greater than the timestamp, it uses a different format. But that format had an extra new line at the end, and caused the output to not look right: cpus=2 sleep-3946 [001]111264306005 : function: inotify_inode_queue_event sleep-3946 [001]111264307158 : function: __fsnotify_parent sleep-3946 [001]111264307637 : function: inotify_dentry_parent_queue_event sleep-3946 [001]111264307989 : function: fsnotify sleep-3946 [001]111264308401 : function: audit_syscall_exit Fixes: 38847db9 ("libtraceevent, perf tools: Changes in tep_print_event_* APIs") Signed-off-by:Steven Rostedt (VMware) <rostedt@goodmis.org> Cc: Jiri Olsa <jolsa@redhat.com> Link: http://lore.kernel.org/lkml/20200303231852.6ab6882f@oasis.local.homeSigned-off-by:
-
Michael Petlan authored
Current libperf man pages mention file counting.c "coming with libperf package", however, the file is missing. Add the file then. Fixes: 81de3bf3 ("libperf: Add man pages") Signed-off-by:
-
Ravi Bangoria authored
The 'nr_jumps' field in 'struct annotation' is not used since it's inception in commit 2402e4a9 ("perf annotate browser: Show 'jumpy' functions"). Get rid of it. Signed-off-by:
Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by:
-
Arnaldo Carvalho de Melo authored
To help in debugging, add this extra message: detect_kbuild_dir: Couldn't find "/lib/modules/5.4.20-200.fc31.x86_64/build/include/generated/autoconf.h", missing kernel-devel package?. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Signed-off-by: -
Jin Yao authored
We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by:
-
