perf test: Add test for events re-armed with PERF_EVENT_IOC_REFRESH
Subsystems (2)
PATCH 1/1 perf test: Add test for events re-armed with PERF_EVENT_IOC_REFRESH
Open a sampling event with sample_period 1, re-arm it with PERF_EVENT_IOC_REFRESH from its SIGIO handler, as self-profilers such as async-profiler do, and check that each of 1000 events delivers a signal. One case uses a software event (page faults), the other the syscalls:sys_enter_getppid tracepoint.
The tracepoint case fails since
commit 18dbcbfabfff ("perf: Fix the POLL_HUP delivery breakage"):
only the first overflow is delivered. It passes on the parent commit. The software case passes on both.
The test was written with an AI coding assistant (Claude) from a standalone reproducer; I reviewed it and ran it in QEMU on current mainline, on that commit and on its parent.
---
Notes:
To run it on any x86_64 box with virtme-ng:
vng --kconfig && make -j$(nproc)
make -C tools/perf NO_LIBTRACEEVENT=1
vng -- tools/perf/perf test -v REFRESH
Results, 2 vCPUs:
b320789d6883 (parent) Tracepoint event: Ok
18dbcbfabfff Tracepoint event: FAILED, 1000 events, 1 overflows
a804c5742bff (master) Tracepoint event: FAILED, 1000 events, 1 overflows
The software event passes on all three.
Leo Yan's series "perf: Revert direct PMU stop for refresh limit" (linux-perf-users, 2026-05-12) reverts 18dbcbfabfff for a different reason; I haven't tested it against this.
#regzbot introduced: 18dbcbfabfff
tools/perf/tests/Build | 1 + tools/perf/tests/builtin-test.c | 1 + tools/perf/tests/event_refresh.c | 136 +++++++++++++++++++++++++++++++ tools/perf/tests/tests.h | 1 + 4 files changed, 139 insertions(+) create mode 100644 tools/perf/tests/event_refresh.c
tools/perf/tests/Build +1 -0
@@ -23,6 +23,7 @@ perf-test-y += hists_output.o perf-test-y += hists_cumulate.o perf-test-y += bp_signal.o perf-test-y += bp_signal_overflow.o +perf-test-y += event_refresh.o perf-test-y += bp_account.o perf-test-y += wp.o perf-test-y += task-exit.o
tools/perf/tests/builtin-test.c +1 -0
@@ -98,6 +98,7 @@ static struct test_suite *generic_tests[] = { &suite__hists_link, &suite__bp_signal, &suite__bp_signal_overflow, + &suite__event_refresh, &suite__bp_accounting, &suite__wp, &suite__task_exit,
tools/perf/tests/event_refresh.c new file +136 -0
136 lines not on this page · raw message
tools/perf/tests/tests.h +1 -0
@@ -124,6 +124,7 @@ DECLARE_SUITE(parse_events);DECLARE_SUITE (hists_link);DECLARE_SUITE (bp_signal);DECLARE_SUITE (bp_signal_overflow); +DECLARE_SUITE (event_refresh);DECLARE_SUITE (bp_accounting);DECLARE_SUITE (wp);DECLARE_SUITE (task_exit);
Signature · 1 line
2.55.0
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] perf test: missing MADV_NOHUGEPAGE in test__event_refresh_sw() fails on THP --
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] perf test: missing MADV_NOHUGEPAGE in test__event_refresh_sw() fails on THP --
commit f4e87f46b616d4e4e30869fe7bd06aad9b7c96f9
Author: Francesco Nigro <[email protected]>
perf test: Add test for events re-armed with PERF_EVENT_IOC_REFRESH
This commit adds a perf test suite that opens a sampling event with sample_period 1, re-arms it via PERF_EVENT_IOC_REFRESH from a SIGIO handler, and verifies that 1000 software page-fault and tracepoint events each deliver an overflow signal.
Quoted 6 lines
tools/perf/tests/event_refresh.c new file +0
-0@@ -0,0 +1,136 @@
[ ... ]
Quoted 22 lines
tools/perf/tests/event_refresh.c new file +22
-0+pr_debug ("%d events, %d overflows\n" ,NR_EVENTS , overflows); +if (overflows ==NR_EVENTS ) + err =TEST_OK ; + +sigaction (SIGIO , &old_sa,NULL ); +out_close: +close (fd); +return err; +} + +static char *pages; + +static void touch_page (int i) +{ + pages[(long )i * page_size] =1 ; +} + +static int test__event_refresh_sw (struct test_suite *test__maybe_unused , +int subtest__maybe_unused ) +{ +struct perf_event_attr attr = { + .type =PERF_TYPE_SOFTWARE ,
tools/perf/tests/event_refresh.c line 100
Quoted 10 lines
tools/perf/tests/event_refresh.c new file +10
-0+ .config = PERF_COUNT_SW_PAGE_FAULTS, + }; +int err; + + pages =mmap (NULL , (long )NR_EVENTS * page_size,PROT_READ |PROT_WRITE , +MAP_PRIVATE |MAP_ANONYMOUS ,-1 ,0 ); +if (pages ==MAP_FAILED ) +return TEST_FAIL ; + + err =count_overflows (&attr, touch_page);
[Severity: Medium]
In tools/perf/tests/event_refresh.c, test__event_refresh_sw() maps NR_EVENTS * page_size bytes (~4 MiB with 4 KiB base pages) as a private anonymous mapping and relies on each touch_page() write at page_size stride to trigger a distinct PERF_COUNT_SW_PAGE_FAULTS event.
On systems with Transparent Huge Pages (THP) or multi-size THP (mTHP) set to "always" (such as when CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y or /sys/kernel/mm/transparent_hugepage/enabled is set to "always"), the first write in touch_page(0) can be satisfied by __handle_mm_fault() -> create_huge_pmd() with a 2 MiB huge page (or a multi-page folio).
That maps up to 512 base pages in a single page fault, so subsequent touch_page() calls within that huge page hit already-present PTEs/PMDs without faulting. As a result, overflows only reaches 489 (or fewer) instead of NR_EVENTS (1000), and count_overflows() returns TEST_FAIL.
Should test__event_refresh_sw() call madvise(pages, (long)NR_EVENTS * page_size, MADV_NOHUGEPAGE) before count_overflows() so that each touch_page() write faults in a single base page?
Quoted 3 lines
tools/perf/tests/event_refresh.c new file +3
-0+munmap (pages, (long )NR_EVENTS * page_size); +return err; +}
[ ... ]
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Quoted 3 lines
On systems with Transparent Huge Pages (THP) or multi-size THP (mTHP) set to "always" [...] As a result, overflows only reaches 489 (or fewer) instead of
NR_EVENTS(1000), andcount_overflows()returnsTEST_FAIL.
Confirmed: on master with transparent_hugepage=always the software case gets 489 of 1000 overflows. With madvise(MADV_NOHUGEPAGE) on the buffer it gets all 1000, and the tracepoint case still gets 1. I'll send a v2 with that change.
Thanks,
Francesco