perf tools: Fix segfault accessing sample_id xyarray
Adrian Hunter
reviewerPATCH 1/1 perf tools: Fix segfault accessing sample_id xyarray
perf_evsel sample_id is an xyarray which can cause a segfault when accessed beyond its size. e.g.
# perf record -e intel_pt// -C 1 sleep 1 Segmentation fault (core dumped)
That is happening because a dummy event is opened to capture text poke events accoss all CPUs, however the mmap logic is allocating according to the number of user_requested_cpus.
In general, perf sometimes uses the evsel cpus to open events, and sometimes the evlist user_requested_cpus. However, it is not necessary to determine which case is which because the opened event file descriptors are also in an xyarray, the size of whch can be used to correctly allocate the size of the sample_id xyarray, because there is one ID per file descriptor. Note, in the affected code path, perf_evsel fd array is subsequently used to get the file descriptor for the mmap, so it makes sense for the xyarrays to be the same size there.
--- tools/lib/perf/evlist.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
tools/lib/perf/evlist.c +1 -2
@@ -577,7 +577,6 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist, {struct perf_evsel *evsel;const struct perf_cpu_map *cpus = evlist->user_requested_cpus; -const struct perf_thread_map *threads = evlist->threads;if (!ops || !ops->get || !ops->mmap)return -EINVAL ;
@@ -589,7 +588,7 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist,perf_evlist__for_each_entry (evlist, evsel) {if ((evsel->attr.read_format &PERF_FORMAT_ID ) && evsel->sample_id ==NULL && -perf_evsel__alloc_id (evsel,perf_cpu_map__nr (cpus), threads->nr) <0 ) +perf_evsel__alloc_id (evsel, evsel->fd->max_x, evsel->fd->max_y) <0 )return -ENOMEM ; }
Signature · 1 line
2.25.1
Quoted 10 lines · Adrian Hunter
perf_evselsample_idis an xyarray which can cause a segfault when accessed beyond its size. e.g.# perf record -e intel_pt// -C 1 sleep 1 Segmentation fault (core dumped)That is happening because a dummy event is opened to capture text poke events accoss all CPUs, however the mmap logic is allocating according to the number of
user_requested_cpus.
Nit: typo on 'accoss'
Quoted 9 lines
In general, perf sometimes uses the evsel cpus to open events, and sometimes the evlist
user_requested_cpus. However, it is not necessary to determine which case is which because the opened event file descriptors are also in an xyarray, the size of whch can be used to correctly allocate the size of thesample_idxyarray, because there is one ID per file descriptor. Note, in the affected code path,perf_evselfd array is subsequently used to get the file descriptor for the mmap, so it makes sense for the xyarrays to be the same size there.
Quoted 4 lines
Fixes 246eba8e9041c4 ("perf tools: Add support for PERF_RECORD_TEXT_POKE")Fixes d1a177595b3a82 ("libperf: Adopt perf_evlist__mmap()/munmap() from tools/perf")Cc [email protected] # 5.5+Signed-off-by Adrian Hunter <[email protected]>
Thanks,
Ian
Quoted 28 lines
--- tools/lib/perf/evlist.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)tools/lib/perf/evlist.c +1
-3@@ -577,7 +577,6 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist, {struct perf_evsel *evsel;const struct perf_cpu_map *cpus = evlist->user_requested_cpus; -const struct perf_thread_map *threads = evlist->threads;if (!ops || !ops->get || !ops->mmap)return -EINVAL ;@@ -589,7 +588,7 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist,perf_evlist__for_each_entry (evlist, evsel) {if ((evsel->attr.read_format &PERF_FORMAT_ID ) && evsel->sample_id ==NULL && -perf_evsel__alloc_id (evsel,perf_cpu_map__nr (cpus), threads->nr) <0 ) +perf_evsel__alloc_id (evsel, evsel->fd->max_x, evsel->fd->max_y) <0 )return -ENOMEM ; } --2.25.1
Em Wed, Apr 13, 2022 at 02:11:41PM -0700, Ian Rogers escreveu:
Quoted 5 lines · Adrian Hunter
Quoted 4 lines
perf_evselsample_idis an xyarray which can cause a segfault when accessed beyond its size. e.g.
Quoted 8 lines
Quoted 6 lines
# perf record -e intel_pt// -C 1 sleep 1 Segmentation fault (core dumped)That is happening because a dummy event is opened to capture text poke events accoss all CPUs, however the mmap logic is allocating according to the number of
user_requested_cpus.Nit: typo on 'accoss'
Fixed
Quoted 9 lines
Quoted 9 lines
In general, perf sometimes uses the evsel cpus to open events, and sometimes the evlist
user_requested_cpus. However, it is not necessary to determine which case is which because the opened event file descriptors are also in an xyarray, the size of whch can be used to correctly allocate the size of thesample_idxyarray, because there is one ID per file descriptor. Note, in the affected code path,perf_evselfd array is subsequently used to get the file descriptor for the mmap, so it makes sense for the xyarrays to be the same size there.
Quoted 6 lines
Quoted 4 lines
Fixes 246eba8e9041c4 ("perf tools: Add support for PERF_RECORD_TEXT_POKE")Fixes d1a177595b3a82 ("libperf: Adopt perf_evlist__mmap()/munmap() from tools/perf")Cc [email protected] # 5.5+Signed-off-by Adrian Hunter <[email protected]>Acked-by Ian Rogers <[email protected]>
Thanks, applied.
- Arnaldo
Quoted 31 lines
Thanks,
IanQuoted 28 lines
--- tools/lib/perf/evlist.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)tools/lib/perf/evlist.c +1
-3@@ -577,7 +577,6 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist, {struct perf_evsel *evsel;const struct perf_cpu_map *cpus = evlist->user_requested_cpus; -const struct perf_thread_map *threads = evlist->threads;if (!ops || !ops->get || !ops->mmap)return -EINVAL ;@@ -589,7 +588,7 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist,perf_evlist__for_each_entry (evlist, evsel) {if ((evsel->attr.read_format &PERF_FORMAT_ID ) && evsel->sample_id ==NULL && -perf_evsel__alloc_id (evsel,perf_cpu_map__nr (cpus), threads->nr) <0 ) +perf_evsel__alloc_id (evsel, evsel->fd->max_x, evsel->fd->max_y) <0 )return -ENOMEM ; } --2.25.1
Signature · 1 line
- Arnaldo