Message ID | 20240521010439.321264-1-irogers@google.com (mailing list archive) |
---|---|
Headers | show |
Series | Use BPF filters for a "perf top -u" workaround | expand |
On Mon, May 20, 2024 at 6:04 PM Ian Rogers <irogers@google.com> wrote: > > Allow uid and gid to be terms in BPF filters by first breaking the > connection between filter terms and PERF_SAMPLE_xx values. Calculate > the uid and gid using the bpf_get_current_uid_gid helper, rather than > from a value in the sample. Allow filters to be passed to perf top, this allows: > > $ perf top -e cycles:P --filter "uid == $(id -u)" > > to work as a "perf top -u" workaround, as "perf top -u" usually fails > due to processes/threads terminating between the /proc scan and the > perf_event_open. > > v2. Allow PERF_SAMPLE_xx to be computed from the PBF_TERM_xx value > using a shift as requested by Namhyung. > > Ian Rogers (3): > perf bpf filter: Give terms their own enum > perf bpf filter: Add uid and gid terms > perf top: Allow filters on events Acked-by: Namhyung Kim <namhyung@kernel.org> Thanks, Namhyung > > tools/perf/Documentation/perf-record.txt | 2 +- > tools/perf/Documentation/perf-top.txt | 4 ++ > tools/perf/builtin-top.c | 9 +++ > tools/perf/util/bpf-filter.c | 33 ++++++---- > tools/perf/util/bpf-filter.h | 5 +- > tools/perf/util/bpf-filter.l | 66 ++++++++++---------- > tools/perf/util/bpf-filter.y | 7 ++- > tools/perf/util/bpf_skel/sample-filter.h | 61 +++++++++++++++++- > tools/perf/util/bpf_skel/sample_filter.bpf.c | 54 +++++++++++----- > 9 files changed, 171 insertions(+), 70 deletions(-) > > -- > 2.45.0.rc1.225.g2a3ae87e7f-goog >
On Wed, May 22, 2024 at 09:38:57PM -0700, Namhyung Kim wrote: > On Mon, May 20, 2024 at 6:04 PM Ian Rogers <irogers@google.com> wrote: > > > > Allow uid and gid to be terms in BPF filters by first breaking the > > connection between filter terms and PERF_SAMPLE_xx values. Calculate > > the uid and gid using the bpf_get_current_uid_gid helper, rather than > > from a value in the sample. Allow filters to be passed to perf top, this allows: > > > > $ perf top -e cycles:P --filter "uid == $(id -u)" > > > > to work as a "perf top -u" workaround, as "perf top -u" usually fails > > due to processes/threads terminating between the /proc scan and the > > perf_event_open. > > > > v2. Allow PERF_SAMPLE_xx to be computed from the PBF_TERM_xx value > > using a shift as requested by Namhyung. > > > > Ian Rogers (3): > > perf bpf filter: Give terms their own enum > > perf bpf filter: Add uid and gid terms > > perf top: Allow filters on events Hmm.. I'm seeing this. $ make build-test ... cd . && make GEN_VMLINUX_H=1 FEATURES_DUMP=/home/namhyung/project/linux/tools/perf/BUILD_TEST_FEATURE_DUMP -j64 O=/tmp/tmp.EeXFOfLPt5 DESTDIR=/tmp/tmp.Y0eiZKvc9D ... CLANG /tmp/tmp.EeXFOfLPt5/util/bpf_skel/.tmp/sample_filter.bpf.o In file included from util/bpf_skel/sample_filter.bpf.c:8: In file included from util/bpf_skel/sample-filter.h:4: /home/namhyung/project/linux/tools/include/uapi/linux/perf_event.h:29:6: error: redefinition of 'perf_type_id' enum perf_type_id { ^ /tmp/tmp.EeXFOfLPt5/util/bpf_skel/.tmp/../vmlinux.h:54086:6: note: previous definition is here enum perf_type_id { ^ In file included from util/bpf_skel/sample_filter.bpf.c:8: In file included from util/bpf_skel/sample-filter.h:4: /home/namhyung/project/linux/tools/include/uapi/linux/perf_event.h:30:2: error: redefinition of enumerator 'PERF_TYPE_HARDWARE' PERF_TYPE_HARDWARE = 0, ^ /tmp/tmp.EeXFOfLPt5/util/bpf_skel/.tmp/../vmlinux.h:54087:2: note: previous definition is here PERF_TYPE_HARDWARE = 0, ^ ... make[3]: *** [Makefile.perf:264: sub-make] Error 2 make[2]: *** [Makefile:70: all] Error 2 make[1]: *** [tests/make:340: make_gen_vmlinux_h_O] Error 1 make: *** [Makefile:103: build-test] Error 2 Thanks, Namhyung
On Fri, May 24, 2024 at 10:58 AM Namhyung Kim <namhyung@kernel.org> wrote: > > On Wed, May 22, 2024 at 09:38:57PM -0700, Namhyung Kim wrote: > > On Mon, May 20, 2024 at 6:04 PM Ian Rogers <irogers@google.com> wrote: > > > > > > Allow uid and gid to be terms in BPF filters by first breaking the > > > connection between filter terms and PERF_SAMPLE_xx values. Calculate > > > the uid and gid using the bpf_get_current_uid_gid helper, rather than > > > from a value in the sample. Allow filters to be passed to perf top, this allows: > > > > > > $ perf top -e cycles:P --filter "uid == $(id -u)" > > > > > > to work as a "perf top -u" workaround, as "perf top -u" usually fails > > > due to processes/threads terminating between the /proc scan and the > > > perf_event_open. > > > > > > v2. Allow PERF_SAMPLE_xx to be computed from the PBF_TERM_xx value > > > using a shift as requested by Namhyung. > > > > > > Ian Rogers (3): > > > perf bpf filter: Give terms their own enum > > > perf bpf filter: Add uid and gid terms > > > perf top: Allow filters on events > > Hmm.. I'm seeing this. > > $ make build-test > ... > cd . && make GEN_VMLINUX_H=1 FEATURES_DUMP=/home/namhyung/project/linux/tools/perf/BUILD_TEST_FEATURE_DUMP -j64 O=/tmp/tmp.EeXFOfLPt5 DESTDIR=/tmp/tmp.Y0eiZKvc9D > ... > CLANG /tmp/tmp.EeXFOfLPt5/util/bpf_skel/.tmp/sample_filter.bpf.o > In file included from util/bpf_skel/sample_filter.bpf.c:8: > In file included from util/bpf_skel/sample-filter.h:4: > /home/namhyung/project/linux/tools/include/uapi/linux/perf_event.h:29:6: error: redefinition of 'perf_type_id' > enum perf_type_id { > ^ > /tmp/tmp.EeXFOfLPt5/util/bpf_skel/.tmp/../vmlinux.h:54086:6: note: previous definition is here > enum perf_type_id { > ^ > In file included from util/bpf_skel/sample_filter.bpf.c:8: > In file included from util/bpf_skel/sample-filter.h:4: > /home/namhyung/project/linux/tools/include/uapi/linux/perf_event.h:30:2: error: redefinition of enumerator 'PERF_TYPE_HARDWARE' > PERF_TYPE_HARDWARE = 0, > ^ > /tmp/tmp.EeXFOfLPt5/util/bpf_skel/.tmp/../vmlinux.h:54087:2: note: previous definition is here > PERF_TYPE_HARDWARE = 0, > ^ > ... > make[3]: *** [Makefile.perf:264: sub-make] Error 2 > make[2]: *** [Makefile:70: all] Error 2 > make[1]: *** [tests/make:340: make_gen_vmlinux_h_O] Error 1 > make: *** [Makefile:103: build-test] Error 2 IIRC this only exists for the sake of asserts. I can move the asserts to placate the generated vmlinux.h. I'll send a v3. Thanks, Ian > Thanks, > Namhyung >