Pull more perf tools updates from Arnaldo Carvalho de Melo:
- Use BPF CO-RE (Compile Once, Run Everywhere) to support old kernels
when using bperf (perf BPF based counters) with cgroups.
- Support HiSilicon PCIe Performance Monitoring Unit (PMU), that
monitors bandwidth, latency, bus utilization and buffer occupancy.
Documented in Documentation/admin-guide/perf/hisi-pcie-pmu.rst.
- User space tasks can migrate between CPUs, so when tracing selected
CPUs, system-wide sideband is still needed, fix it in the setup of
Intel PT on hybrid systems.
- Fix metricgroups title message in 'perf list', it should state that
the metrics groups are to be used with the '-M' option, not '-e'.
- Sync the msr-index.h copy with the kernel sources, adding support for
using "AMD64_TSC_RATIO" in filter expressions in 'perf trace' as well
as decoding it when printing the MSR tracepoint arguments.
- Fix program header size and alignment when generating a JIT ELF in
'perf inject'.
- Add multiple new Intel PT 'perf test' entries, including a jitdump
one.
- Fix the 'perf test' entries for 'perf stat' CSV and JSON output when
running on PowerPC due to an invalid topology number in that arch.
- Fix the 'perf test' for arm_coresight failures on the ARM Juno
system.
- Fix the 'perf test' attr entry for PERF_FORMAT_LOST, adding this
option to the or expression expected in the intercepted
perf_event_open() syscall.
- Add missing condition flags ('hs', 'lo', 'vc', 'vs') for arm64 in the
'perf annotate' asm parser.
- Fix 'perf mem record -C' option processing, it was being chopped up
when preparing the underlying 'perf record -e mem-events' and thus
being ignored, requiring using '-- -C CPUs' as a workaround.
- Improvements and tidy ups for 'perf test' shell infra.
- Fix Intel PT information printing segfault in uClibc, where a NULL
format was being passed to fprintf.
* tag 'perf-tools-for-v6.1-2-2022-10-16' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (23 commits)
tools arch x86: Sync the msr-index.h copy with the kernel sources
perf auxtrace arm64: Add support for parsing HiSilicon PCIe Trace packet
perf auxtrace arm64: Add support for HiSilicon PCIe Tune and Trace device driver
perf auxtrace arm: Refactor event list iteration in auxtrace_record__init()
perf tests stat+json_output: Include sanity check for topology
perf tests stat+csv_output: Include sanity check for topology
perf intel-pt: Fix system_wide dummy event for hybrid
perf intel-pt: Fix segfault in intel_pt_print_info() with uClibc
perf test: Fix attr tests for PERF_FORMAT_LOST
perf test: test_intel_pt.sh: Add 9 tests
perf inject: Fix GEN_ELF_TEXT_OFFSET for jit
perf test: test_intel_pt.sh: Add jitdump test
perf test: test_intel_pt.sh: Tidy some alignment
perf test: test_intel_pt.sh: Print a message when skipping kernel tracing
perf test: test_intel_pt.sh: Tidy some perf record options
perf test: test_intel_pt.sh: Fix return checking again
perf: Skip and warn on unknown format 'configN' attrs
perf list: Fix metricgroups title message
perf mem: Fix -C option behavior for perf mem record
perf annotate: Add missing condition flags for arm64
...
To pick up the changes in:
b8d1d16360 ("x86/apic: Don't disable x2APIC if locked")
ca5b7c0d96 ("perf/x86/amd/lbr: Add LbrExtV2 branch record support")
Addressing these tools/perf build warnings:
diff -u tools/arch/x86/include/asm/msr-index.h arch/x86/include/asm/msr-index.h
Warning: Kernel ABI header at 'tools/arch/x86/include/asm/msr-index.h' differs from latest version at 'arch/x86/include/asm/msr-index.h'
That makes the beautification scripts to pick some new entries:
$ tools/perf/trace/beauty/tracepoints/x86_msr.sh > before
$ cp arch/x86/include/asm/msr-index.h tools/arch/x86/include/asm/msr-index.h
$ tools/perf/trace/beauty/tracepoints/x86_msr.sh > after
$ diff -u before after
--- before 2022-10-14 18:06:34.294561729 -0300
+++ after 2022-10-14 18:06:41.285744044 -0300
@@ -264,6 +264,7 @@
[0xc0000102 - x86_64_specific_MSRs_offset] = "KERNEL_GS_BASE",
[0xc0000103 - x86_64_specific_MSRs_offset] = "TSC_AUX",
[0xc0000104 - x86_64_specific_MSRs_offset] = "AMD64_TSC_RATIO",
+ [0xc000010e - x86_64_specific_MSRs_offset] = "AMD64_LBR_SELECT",
[0xc000010f - x86_64_specific_MSRs_offset] = "AMD_DBG_EXTN_CFG",
[0xc0000300 - x86_64_specific_MSRs_offset] = "AMD64_PERF_CNTR_GLOBAL_STATUS",
[0xc0000301 - x86_64_specific_MSRs_offset] = "AMD64_PERF_CNTR_GLOBAL_CTL",
$
Now one can trace systemwide asking to see backtraces to where that MSR
is being read/written, see this example with a previous update:
# perf trace -e msr:*_msr/max-stack=32/ --filter="msr>=IA32_U_CET && msr<=IA32_INT_SSP_TAB"
^C#
If we use -v (verbose mode) we can see what it does behind the scenes:
# perf trace -v -e msr:*_msr/max-stack=32/ --filter="msr>=IA32_U_CET && msr<=IA32_INT_SSP_TAB"
Using CPUID AuthenticAMD-25-21-0
0x6a0
0x6a8
New filter for msr:read_msr: (msr>=0x6a0 && msr<=0x6a8) && (common_pid != 597499 && common_pid != 3313)
0x6a0
0x6a8
New filter for msr:write_msr: (msr>=0x6a0 && msr<=0x6a8) && (common_pid != 597499 && common_pid != 3313)
mmap size 528384B
^C#
Example with a frequent msr:
# perf trace -v -e msr:*_msr/max-stack=32/ --filter="msr==IA32_SPEC_CTRL" --max-events 2
Using CPUID AuthenticAMD-25-21-0
0x48
New filter for msr:read_msr: (msr==0x48) && (common_pid != 2612129 && common_pid != 3841)
0x48
New filter for msr:write_msr: (msr==0x48) && (common_pid != 2612129 && common_pid != 3841)
mmap size 528384B
Looking at the vmlinux_path (8 entries long)
symsrc__init: build id mismatch for vmlinux.
Using /proc/kcore for kernel data
Using /proc/kallsyms for symbols
0.000 Timer/2525383 msr:write_msr(msr: IA32_SPEC_CTRL, val: 6)
do_trace_write_msr ([kernel.kallsyms])
do_trace_write_msr ([kernel.kallsyms])
__switch_to_xtra ([kernel.kallsyms])
__switch_to ([kernel.kallsyms])
__schedule ([kernel.kallsyms])
schedule ([kernel.kallsyms])
futex_wait_queue_me ([kernel.kallsyms])
futex_wait ([kernel.kallsyms])
do_futex ([kernel.kallsyms])
__x64_sys_futex ([kernel.kallsyms])
do_syscall_64 ([kernel.kallsyms])
entry_SYSCALL_64_after_hwframe ([kernel.kallsyms])
__futex_abstimed_wait_common64 (/usr/lib64/libpthread-2.33.so)
0.030 :0/0 msr:write_msr(msr: IA32_SPEC_CTRL, val: 2)
do_trace_write_msr ([kernel.kallsyms])
do_trace_write_msr ([kernel.kallsyms])
__switch_to_xtra ([kernel.kallsyms])
__switch_to ([kernel.kallsyms])
__schedule ([kernel.kallsyms])
schedule_idle ([kernel.kallsyms])
do_idle ([kernel.kallsyms])
cpu_startup_entry ([kernel.kallsyms])
secondary_startup_64_no_verify ([kernel.kallsyms])
#
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Daniel Sneddon <daniel.sneddon@linux.intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sandipan Das <sandipan.das@amd.com>
Link: https://lore.kernel.org/lkml/Y0nQkz2TUJxwfXJd@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Testcase stat+json_output.sh fails in powerpc:
86: perf stat JSON output linter : FAILED!
The testcase "stat+json_output.sh" verifies perf stat JSON output. The
test covers aggregation modes like per-socket, per-core, per-die, -A
(no_aggr mode) along with few other tests. It counts expected fields for
various commands. For example say -A (i.e, AGGR_NONE mode), expects 7
fields in the output having "CPU" as first field. Same way, for
per-socket, it expects the first field in result to point to socket id.
The testcases compares the result with expected count.
The values for socket, die, core and cpu are fetched from topology
directory:
/sys/devices/system/cpu/cpu*/topology.
For example, socket value is fetched from "physical_package_id" file of
topology directory. (cpu__get_topology_int() in util/cpumap.c)
If a platform fails to fetch the topology information, values will be
set to -1. For example, incase of pSeries platform of powerpc, value for
"physical_package_id" is restricted and not exposed. So, -1 will be
assigned.
Perf code has a checks for valid cpu id in "aggr_printout"
(stat-display.c), which displays the fields. So, in cases where topology
values not exposed, first field of the output displaying will be empty.
This cause the testcase to fail, as it counts number of fields in the
output.
Incase of -A (AGGR_NONE mode,), testcase expects 7 fields in the output,
becos of -1 value obtained from topology files for some, only 6 fields
are printed. Hence a testcase failure reported due to mismatch in number
of fields in the output.
Patch here adds a sanity check in the testcase for topology. Check will
help to skip the test if -1 value found.
Fixes: 0c343af2a2 ("perf test: JSON format checking")
Reported-by: Disha Goel <disgoel@linux.vnet.ibm.com>
Suggested-by: Ian Rogers <irogers@google.com>
Suggested-by: James Clark <james.clark@arm.com>
Signed-off-by: Athira Jajeev <atrajeev@linux.vnet.ibm.com>
Cc: Claire Jensen <cjense@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nageswara R Sastry <rnsastry@linux.ibm.com>
Link: https://lore.kernel.org/r/20221006155149.67205-2-atrajeev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Testcase stat+csv_output.sh fails in powerpc:
84: perf stat CSV output linter: FAILED!
The testcase "stat+csv_output.sh" verifies perf stat CSV output. The
test covers aggregation modes like per-socket, per-core, per-die, -A
(no_aggr mode) along with few other tests. It counts expected fields for
various commands. For example say -A (i.e, AGGR_NONE mode), expects 7
fields in the output having "CPU" as first field. Same way, for
per-socket, it expects the first field in result to point to socket id.
The testcases compares the result with expected count.
The values for socket, die, core and cpu are fetched from topology
directory:
/sys/devices/system/cpu/cpu*/topology.
For example, socket value is fetched from "physical_package_id" file of
topology directory. (cpu__get_topology_int() in util/cpumap.c)
If a platform fails to fetch the topology information, values will be
set to -1. For example, incase of pSeries platform of powerpc, value for
"physical_package_id" is restricted and not exposed. So, -1 will be
assigned.
Perf code has a checks for valid cpu id in "aggr_printout"
(stat-display.c), which displays the fields. So, in cases where topology
values not exposed, first field of the output displaying will be empty.
This cause the testcase to fail, as it counts number of fields in the
output.
Incase of -A (AGGR_NONE mode,), testcase expects 7 fields in the output,
becos of -1 value obtained from topology files for some, only 6 fields
are printed. Hence a testcase failure reported due to mismatch in number
of fields in the output.
Patch here adds a sanity check in the testcase for topology. Check will
help to skip the test if -1 value found.
Fixes: 7473ee56db ("perf test: Add checking for perf stat CSV output.")
Reported-by: Disha Goel <disgoel@linux.vnet.ibm.com>
Suggested-by: Ian Rogers <irogers@google.com>
Suggested-by: James Clark <james.clark@arm.com>
Signed-off-by: Athira Jajeev <atrajeev@linux.vnet.ibm.com>
Cc: Claire Jensen <cjense@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nageswara R Sastry <rnsastry@linux.ibm.com>
Link: https://lore.kernel.org/r/20221006155149.67205-1-atrajeev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add tests:
Test with MTC and TSC disabled
Test with branches disabled
Test with/without CYC
Test recording with sample mode
Test with kernel trace
Test virtual LBR
Test power events
Test with TNT packets disabled
Test with event_trace
These tests mostly check that perf record works with the corresponding
Intel PT config terms, sometimes also checking that certain packets do or
do not appear in the resulting trace as appropriate.
The "Test virtual LBR" is slightly trickier, using a Python script to
check that branch stacks are actually synthesized.
Signed-off-by: Ammy Yi <ammy.yi@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.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/20221014170905.64069-8-adrian.hunter@intel.com
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Pull more MM updates from Andrew Morton:
- fix a race which causes page refcounting errors in ZONE_DEVICE pages
(Alistair Popple)
- fix userfaultfd test harness instability (Peter Xu)
- various other patches in MM, mainly fixes
* tag 'mm-stable-2022-10-13' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (29 commits)
highmem: fix kmap_to_page() for kmap_local_page() addresses
mm/page_alloc: fix incorrect PGFREE and PGALLOC for high-order page
mm/selftest: uffd: explain the write missing fault check
mm/hugetlb: use hugetlb_pte_stable in migration race check
mm/hugetlb: fix race condition of uffd missing/minor handling
zram: always expose rw_page
LoongArch: update local TLB if PTE entry exists
mm: use update_mmu_tlb() on the second thread
kasan: fix array-bounds warnings in tests
hmm-tests: add test for migrate_device_range()
nouveau/dmem: evict device private memory during release
nouveau/dmem: refactor nouveau_dmem_fault_copy_one()
mm/migrate_device.c: add migrate_device_range()
mm/migrate_device.c: refactor migrate_vma and migrate_deivce_coherent_page()
mm/memremap.c: take a pgmap reference on page allocation
mm: free device private pages have zero refcount
mm/memory.c: fix race when faulting a device private page
mm/damon: use damon_sz_region() in appropriate place
mm/damon: move sz_damon_region to damon_sz_region
lib/test_meminit: add checks for the allocation functions
...
If the kernel exposes a new perf_event_attr field in a format attr, perf
will return an error stating the specified PMU can't be found. For
example, a format attr with 'config3:0-63' causes an error as config3 is
unknown to perf. This causes a compatibility issue between a newer
kernel with older perf tool.
Before this change with a kernel adding 'config3' I get:
$ perf record -e arm_spe// -- true
event syntax error: 'arm_spe//'
\___ Cannot find PMU `arm_spe'. Missing kernel support?
Run 'perf list' for a list of valid events
Usage: perf record [<options>] [<command>]
or: perf record [<options>] -- <command> [<options>]
-e, --event <event> event selector. use 'perf list' to list
available events
After this change, I get:
$ perf record -e arm_spe// -- true
WARNING: 'arm_spe_0' format 'inv_event_filter' requires 'perf_event_attr::config3' which is not supported by this version of perf!
[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 0.091 MB perf.data ]
To support unknown configN formats, rework the YACC implementation to
pass any config[0-9]+ format to perf_pmu__new_format() to handle with a
warning.
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220914-arm-perf-tool-spe1-2-v2-v4-1-83c098e6212e@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The -C/--cpu option was maily for report but it also affected record as
it ate the option. So users needed to use "--" after perf mem record to
pass the info to the perf record properly.
Check if this option is set for record, and pass it to the actual perf
record.
Before)
$ sudo perf --debug perf-event-open mem record -C 0 2>&1 | grep -a sys_perf_event_open
...
sys_perf_event_open: pid -1 cpu 0 group_fd -1 flags 0x8 = 4
sys_perf_event_open: pid -1 cpu 1 group_fd -1 flags 0x8 = 5
sys_perf_event_open: pid -1 cpu 2 group_fd -1 flags 0x8 = 6
sys_perf_event_open: pid -1 cpu 3 group_fd -1 flags 0x8 = 7
sys_perf_event_open: pid -1 cpu 0 group_fd -1 flags 0x8 = 8
sys_perf_event_open: pid -1 cpu 1 group_fd -1 flags 0x8 = 9
sys_perf_event_open: pid -1 cpu 2 group_fd -1 flags 0x8 = 10
sys_perf_event_open: pid -1 cpu 3 group_fd -1 flags 0x8 = 11
...
After)
$ sudo perf --debug perf-event-open mem record -C 0 2>&1 | grep -a sys_perf_event_open
...
sys_perf_event_open: pid -1 cpu 0 group_fd -1 flags 0x8 = 4
sys_perf_event_open: pid -1 cpu 0 group_fd -1 flags 0x8 = 5
sys_perf_event_open: pid -1 cpu 0 group_fd -1 flags 0x8 = 6
sys_perf_event_open: pid -1 cpu 0 group_fd -1 flags 0x8 = 7
Reported-by: Ravi Bangoria <ravi.bangoria@amd.com>
Reviewed-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221004200211.1444521-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This test commonly fails on Arm Juno because the instruction interval
is large enough to miss generating any samples for Perf in system-wide
mode.
Fix this by lowering the interval until a comfortable number of Perf
instructions are generated. The test is still quick to run because only
a small amount of trace is gathered.
Before:
sudo ./perf test coresight -vvv
...
Recording trace with system wide mode
Looking at perf.data file for dumping branch samples:
Looking at perf.data file for reporting branch samples:
Looking at perf.data file for instruction samples:
CoreSight system wide testing: FAIL
...
After:
sudo ./perf test coresight -vvv
...
Recording trace with system wide mode
Looking at perf.data file for dumping branch samples:
Looking at perf.data file for reporting branch samples:
Looking at perf.data file for instruction samples:
CoreSight system wide testing: PASS
...
Reviewed-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: James Clark <james.clark@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
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: coresight@lists.linaro.org
Link: https://lore.kernel.org/r/20221005140508.1537277-1-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
