perf/core fixes and improvements
Subsystems (15)
Arnaldo Carvalho de Melo
maintainerHi Ingo/Thomas,
Please consider pulling, the CAP_PERFMON got lots of reviewed-by
and acked-by from the relevant people, its just that it took more time on my side to test everything, working less time than normal in the current situation, delaying this lot more than I expected :-\
Best regards,
- Arnaldo
Test results at the end of this message, as usual.
The following changes since commit 7dc41b9b99cd0037a418ac47e342d56a438df649:
Merge tag 'perf-urgent-for-mingo-5.7-20200403' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent (2020-04-04 10:35:15 +0200)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-core-for-mingo-5.7-20200413
for you to fetch changes up to 80af1ec0f23235271feda1436f1b5c74078a0453:
perf script: Add flamegraph.py script (2020-04-13 11:03:08 -0300)
----------------------------------------------------------------
perf/core fixes and improvements:
kernel + tools/perf:
Alexey Budankov:
- Introduce CAP_PERFMON to kernel and user space.
perf stat:
Jin Yao:
- Fix no metric header if --per-socket and --metric-only set.
perf python:
Arnaldo Carvalho de Melo:
- Check if clang supports -fno-semantic-interposition, fixing the python binding when building with clang.
perf script:
Andreas Gerstmayr:
- Add flamegraph.py script
tools arch x86:
Arnaldo Carvalho de Melo:
- Sync the msr-index.h copy with the kernel sources.
perf bench:
Ian Rogers:
- Add event synthesis benchmark.
BPF:
Jiri Olsa:
- Synthesize bpf_trampoline/dispatcher ksymbol event.
tools api fs:
Stephane Eranian:
- Make xxx__mountpoint() more scalable
----------------------------------------------------------------
Adrian Hunter (1):
perf script: Simplify auxiliary event printing functions
Alexey Budankov (12):
capabilities: Introduce CAP_PERFMON to kernel and user space perf/core: Open access to the core for CAP_PERFMON privileged process perf/core: open access to probes for CAP_PERFMON privileged process perf tools: Support CAP_PERFMON capability
drm/i915/perf: Open access for CAP_PERFMON privileged process trace/bpf_trace: Open access for CAP_PERFMON privileged process powerpc/perf: open access for CAP_PERFMON privileged process parisc/perf: open access for CAP_PERFMON privileged process drivers/perf: Open access for CAP_PERFMON privileged process drivers/oprofile: Open access for CAP_PERFMON privileged process doc/admin-guide: Update perf-security.rst with CAP_PERFMON information doc/admin-guide: update kernel.rst with CAP_PERFMON information
Andreas Gerstmayr (1):
perf script: Add flamegraph.py script
Arnaldo Carvalho de Melo (2):
tools arch x86: Sync the msr-index.h copy with the kernel sources perf python: Check if clang supports -fno-semantic-interposition
Ian Rogers (2):
perf bench: Add event synthesis benchmark perf synthetic-events: save 4kb from 2 stack frames
Jin Yao (1):
perf stat: Fix no metric header if --per-socket and --metric-only set
Jiri Olsa (5):
perf tools: Synthesize bpf_trampoline/dispatcher ksymbol event perf machine: Set ksymbol dso as loaded on arrival perf annotate: Add basic support for bpf_image
perf expr: Add expr_ prefix for parse_ctx and parse_id perf expr: Add expr_scanner_ctx object
Kajol Jain (1):
perf metrictroup: Split the metricgroup__add_metric function
Stephane Eranian (1):
tools api fs: Make xxx__mountpoint() more scalable
Documentation/admin-guide/perf-security.rst | 86 +++++-- Documentation/admin-guide/sysctl/kernel.rst | 16 +- arch/parisc/kernel/perf.c | 2 +- arch/powerpc/perf/imc-pmu.c | 4 +- drivers/gpu/drm/i915/i915_perf.c | 13 +- drivers/oprofile/event_buffer.c | 2 +- drivers/perf/arm_spe_pmu.c | 4 +- include/linux/capability.h | 4 + include/linux/perf_event.h | 6 +- include/uapi/linux/capability.h | 8 +- kernel/events/core.c | 6 +- kernel/trace/bpf_trace.c | 2 +- security/selinux/include/classmap.h | 4 +- tools/arch/x86/include/asm/msr-index.h | 9 + tools/lib/api/fs/fs.c | 17 ++ tools/lib/api/fs/fs.h | 12 + tools/perf/Documentation/perf-bench.txt | 8 + tools/perf/bench/Build | 2 +- tools/perf/bench/bench.h | 2 +- tools/perf/bench/synthesize.c | 101 ++++++++ tools/perf/builtin-bench.c | 6 + tools/perf/builtin-ftrace.c | 5 +- tools/perf/builtin-script.c | 304 +++++------------------- tools/perf/design.txt | 3 +- tools/perf/scripts/python/bin/flamegraph-record | 2 + tools/perf/scripts/python/bin/flamegraph-report | 3 + tools/perf/scripts/python/flamegraph.py | 124 ++++++++++ tools/perf/tests/expr.c | 4 +- tools/perf/util/annotate.c | 20 ++ tools/perf/util/bpf-event.c | 93 ++++++++ tools/perf/util/cap.h | 4 + tools/perf/util/dso.c | 1 + tools/perf/util/dso.h | 1 + tools/perf/util/evsel.c | 10 +- tools/perf/util/expr.c | 16 +- tools/perf/util/expr.h | 16 +- tools/perf/util/expr.l | 10 +- tools/perf/util/expr.y | 6 +- tools/perf/util/machine.c | 12 + tools/perf/util/metricgroup.c | 60 +++-- tools/perf/util/setup.py | 2 + tools/perf/util/stat-shadow.c | 9 +- tools/perf/util/symbol.c | 1 + tools/perf/util/synthetic-events.c | 22 +- tools/perf/util/util.c | 1 + 45 files changed, 679 insertions(+), 364 deletions(-) create mode 100644 tools/perf/bench/synthesize.c create mode 100755 tools/perf/scripts/python/bin/flamegraph-record create mode 100755 tools/perf/scripts/python/bin/flamegraph-report create mode 100755 tools/perf/scripts/python/flamegraph.py
Test results:
The first ones are container based builds of tools/perf with and without libelf support. Where clang is available, it is also used to build perf with/without libelf, and building with LIBCLANGLLVM=1 (built-in clang) with gcc and clang when clang and its devel libraries are installed.
The objtool and samples/bpf/ builds are disabled now that I'm switching from using the sources in a local volume to fetching them from a http server to build it inside the container, to make it easier to build in a container cluster. Those will come back later.
Several are cross builds, the ones with -x-ARCH and the android one, and those may not have all the features built, due to lack of multi-arch devel packages, available and being used so far on just a few, like debian:experimental-x-{arm64,mipsel}.
The 'perf test' one will perform a variety of tests exercising tools/perf/util/, tools/lib/{bpf,traceevent,etc}, as well as run perf commands with a variety of command line event specifications to then intercept the sys_perf_event syscall to check that the perf_event_attr fields are set up as expected, among a variety of other unit tests.
Then there is the 'make -C tools/perf build-test' ones, that build tools/perf/ with a variety of feature sets, exercising the build with an incomplete set of features as well as with a complete one. It is planned to have it run on each of the containers mentioned above, using some container orchestration infrastructure. Get in contact if interested in helping having this in place.
Ubuntu 19.10 is failing when linking against libllvm, which isn't the default, needs to be investigated, haven't tested with CC=gcc, but should be the same problem:
+ make ARCH= CROSS_COMPILE= EXTRA_CFLAGS= LIBCLANGLLVM=1 -C /git/linux/tools/perf O=/tmp/build/perf CC=clang
.../usr/bin/ld: /usr/lib/llvm-9/lib/libclangAnalysis.a(ExprMutationAnalyzer.cpp.o): in function `clang::ast_matchers::internal::matcher_ignoringImpCasts0Matcher::matches(clang::Expr const&, clang::ast_matchers::internal::ASTMatchFinder*, clang::ast_matchers::internal::BoundNodesTreeBuilder*) const':
(.text._ZNK5clang12ast_matchers8internal32matcher_ignoringImpCasts0Matcher7matchesERKNS_4ExprEPNS1_14ASTMatchFinderEPNS1_21BoundNodesTreeBuilderE[_ZNK5clang12ast_matchers8internal32matcher_ignoringImpCasts0Matcher7matchesERKNS_4ExprEPNS1_14ASTMatchFinderEPNS1_21BoundNodesTreeBuilderE]+0x43): undefined reference to `clang::ast_matchers::internal::DynTypedMatcher::matches(clang::ast_type_traits::DynTypedNode const&, clang::ast_matchers::internal::ASTMatchFinder*, clang::ast_matchers::internal::BoundNodesTreeBuilder*) const' /usr/bin/ld: /usr/lib/llvm-9/lib/libclangAnalysis.a(ExprMutationAnalyzer.cpp.o): in function `clang::ast_matchers::internal::matcher_hasLoopVariable0Matcher::matches(clang::CXXForRangeStmt const&, clang::ast_matchers::internal::ASTMatchFinder*, clang::ast_matchers::internal::BoundNodesTreeBuilder*) const':
(.text._ZNK5clang12ast_matchers8internal31matcher_hasLoopVariable0Matcher7matchesERKNS_15CXXForRangeStmtEPNS1_14ASTMatchFinderEPNS1_21BoundNodesTreeBuilderE[_ZNK5clang12ast_matchers8internal31matcher_hasLoopVariable0Matcher7matchesERKNS_15CXXForRangeStmtEPNS1_14ASTMatchFinderEPNS1_21BoundNodesTreeBuilderE]+0x48): undefined reference to `clang::ast_matchers::internal::DynTypedMatcher::matches(clang::ast_type_traits::DynTypedNode const&, clang::ast_matchers::internal::ASTMatchFinder*, clang::ast_matchers::internal::BoundNodesTreeBuilder*) const'
...
It builds ok with the default set of options.
$ export PERF_TARBALL=http://192.168.122.1/perf/perf-5.6.0.tar.xz $ dm 1 alpine:3.4 : Ok gcc (Alpine 5.3.0) 5.3.0, clang version 3.8.0 (tags/RELEASE_380/final) 2 alpine:3.5 : Ok gcc (Alpine 6.2.1) 6.2.1 20160822, clang version 3.8.1 (tags/RELEASE_381/final) 3 alpine:3.6 : Ok gcc (Alpine 6.3.0) 6.3.0, clang version 4.0.0 (tags/RELEASE_400/final) 4 alpine:3.7 : Ok gcc (Alpine 6.4.0) 6.4.0, Alpine clang version 5.0.0 (tags/RELEASE_500/final) (based on LLVM 5.0.0) 5 alpine:3.8 : Ok gcc (Alpine 6.4.0) 6.4.0, Alpine clang version 5.0.1 (tags/RELEASE_501/final) (based on LLVM 5.0.1) 6 alpine:3.9 : Ok gcc (Alpine 8.3.0) 8.3.0, Alpine clang version 5.0.1 (tags/RELEASE_502/final) (based on LLVM 5.0.1) 7 alpine:3.10 : Ok gcc (Alpine 8.3.0) 8.3.0, Alpine clang version 8.0.0 (tags/RELEASE_800/final) (based on LLVM 8.0.0) 8 alpine:3.11 : Ok gcc (Alpine 9.2.0) 9.2.0, Alpine clang version 9.0.0 (https://git.alpinelinux.org/aports f7f0d2c2b8bcd6a5843401a9a702029556492689) (based on LLVM 9.0.0) 9 alpine:edge : Ok gcc (Alpine 9.2.0) 9.2.0, Alpine clang version 9.0.1 (git://git.alpinelinux.org/aports 6c34b9a10bcdcdac04a11569c50b61fb50c4ea6e) (based on LLVM 9.0.1) 10 alt:p8 : Ok x86_64-alt-linux-gcc (GCC) 5.3.1 20151207 (ALT p8 5.3.1-alt3.M80P.1), clang version 3.8.0 (tags/RELEASE_380/final) 11 alt:p9 : Ok x86_64-alt-linux-gcc (GCC) 8.3.1 20190507 (ALT p9 8.3.1-alt5), clang version 7.0.1 12 alt:sisyphus : Ok x86_64-alt-linux-gcc (GCC) 9.2.1 20200123 (ALT Sisyphus 9.2.1-alt3), clang version 9.0.1 13 amazonlinux:1 : Ok gcc (GCC) 7.2.1 20170915 (Red Hat 7.2.1-2), clang version 3.6.2 (tags/RELEASE_362/final) 14 amazonlinux:2 : Ok gcc (GCC) 7.3.1 20180712 (Red Hat 7.3.1-6), clang version 7.0.1 (Amazon Linux 2 7.0.1-1.amzn2.0.2) 15 android-ndk:r12b-arm : Ok arm-linux-androideabi-gcc (GCC) 4.9.x 20150123 (prerelease) 16 android-ndk:r15c-arm : Ok arm-linux-androideabi-gcc (GCC) 4.9.x 20150123 (prerelease) 17 centos:5 : Ok gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-55) 18 centos:6 : Ok gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23) 19 centos:7 : Ok gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-39) 20 centos:8 : Ok gcc (GCC) 8.3.1 20190507 (Red Hat 8.3.1-4), clang version 8.0.1 (Red Hat 8.0.1-1.module_el8.1.0+215+a01033fb) 21 clearlinux:latest : Ok gcc (Clear Linux OS for Intel Architecture) 9.3.1 20200325 releases/gcc-9.3.0-55-gdff885cdc0, clang version 9.0.1 22 debian:8 : Ok gcc (Debian 4.9.2-10+deb8u2) 4.9.2, Debian clang version 3.5.0-10 (tags/RELEASE_350/final) (based on LLVM 3.5.0) 23 debian:9 : Ok gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516, clang version 3.8.1-24 (tags/RELEASE_381/final) 24 debian:10 : Ok gcc (Debian 8.3.0-6) 8.3.0, clang version 7.0.1-8 (tags/RELEASE_701/final) 25 debian:experimental : Ok gcc (Debian 9.3.0-7) 9.3.0 26 debian:experimental-x-mips : Ok mips-linux-gnu-gcc (Debian 9.2.1-24) 9.2.1 20200117 27 fedora:20 : Ok gcc (GCC) 4.8.3 20140911 (Red Hat 4.8.3-7) 28 fedora:22 : Ok gcc (GCC) 5.3.1 20160406 (Red Hat 5.3.1-6), clang version 3.5.0 (tags/RELEASE_350/final) 29 fedora:23 : Ok gcc (GCC) 5.3.1 20160406 (Red Hat 5.3.1-6), clang version 3.7.0 (tags/RELEASE_370/final) 30 fedora:24 : Ok gcc (GCC) 6.3.1 20161221 (Red Hat 6.3.1-1), clang version 3.8.1 (tags/RELEASE_381/final) 31 fedora:24-x-ARC-uClibc : Ok arc-linux-gcc (ARCompact ISA Linux uClibc toolchain 2017.09-rc2) 7.1.1 20170710 32 fedora:25 : Ok gcc (GCC) 6.4.1 20170727 (Red Hat 6.4.1-1), clang version 3.9.1 (tags/RELEASE_391/final) 33 fedora:26 : Ok gcc (GCC) 7.3.1 20180130 (Red Hat 7.3.1-2), clang version 4.0.1 (tags/RELEASE_401/final) 34 fedora:27 : Ok gcc (GCC) 7.3.1 20180712 (Red Hat 7.3.1-6), clang version 5.0.2 (tags/RELEASE_502/final) 35 fedora:28 : Ok gcc (GCC) 8.3.1 20190223 (Red Hat 8.3.1-2), clang version 6.0.1 (tags/RELEASE_601/final) 36 fedora:29 : Ok gcc (GCC) 8.3.1 20190223 (Red Hat 8.3.1-2), clang version 7.0.1 (Fedora 7.0.1-6.fc29) 37 fedora:30 : Ok gcc (GCC) 9.2.1 20190827 (Red Hat 9.2.1-1), clang version 8.0.0 (Fedora 8.0.0-3.fc30) 38 fedora:30-x-ARC-glibc : Ok arc-linux-gcc (ARC HS GNU/Linux glibc toolchain 2019.03-rc1) 8.3.1 20190225 39 fedora:30-x-ARC-uClibc : Ok arc-linux-gcc (ARCv2 ISA Linux uClibc toolchain 2019.03-rc1) 8.3.1 20190225 40 fedora:31 : Ok gcc (GCC) 9.2.1 20190827 (Red Hat 9.2.1-1), clang version 9.0.1 (Fedora 9.0.1-2.fc31) 41 fedora:32 : Ok gcc (GCC) 10.0.1 20200216 (Red Hat 10.0.1-0.8), clang version 10.0.0 (Fedora 10.0.0-0.1.rc2.fc32) 42 fedora:33 : Ok gcc (GCC) 10.0.1 20200311 (Red Hat 10.0.1-0.9), clang version 10.0.0 (Fedora 10.0.0-0.9.rc5.fc33) 43 fedora:rawhide : Ok gcc (GCC) 10.0.1 20200216 (Red Hat 10.0.1-0.8), clang version 10.0.0 (Fedora 10.0.0-0.5.rc3.fc33) 44 gentoo-stage3-amd64:latest : Ok gcc (Gentoo 9.2.0-r2 p3) 9.2.0 45 mageia:5 : Ok gcc (GCC) 4.9.2, clang version 3.5.2 (tags/RELEASE_352/final) 46 mageia:6 : Ok gcc (Mageia 5.5.0-1.mga6) 5.5.0, clang version 3.9.1 (tags/RELEASE_391/final) 47 mageia:7 : Ok gcc (Mageia 8.4.0-1.mga7) 8.4.0, clang version 8.0.0 (Mageia 8.0.0-1.mga7) 48 manjaro:latest : Ok gcc (Arch Linux 9.3.0-1) 9.3.0, clang version 9.0.1 49 openmandriva:cooker : Ok gcc (GCC) 10.0.0 20200301 (OpenMandriva), clang version 10.0.0 50 opensuse:15.0 : Ok gcc (SUSE Linux) 7.4.1 20190905 [gcc-7-branch revision 275407], clang version 5.0.1 (tags/RELEASE_501/final 312548) 51 opensuse:15.1 : Ok gcc (SUSE Linux) 7.5.0, clang version 7.0.1 (tags/RELEASE_701/final 349238) 52 opensuse:42.3 : Ok gcc (SUSE Linux) 4.8.5, clang version 3.8.0 (tags/RELEASE_380/final 262553) 53 opensuse:tumbleweed : Ok gcc (SUSE Linux) 9.2.1 20200128 [revision 83f65674e78d97d27537361de1a9d74067ff228d], clang version 9.0.1 54 oraclelinux:6 : Ok gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23.0.1) 55 oraclelinux:7 : Ok gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-39.0.3) 56 oraclelinux:8 : Ok gcc (GCC) 8.3.1 20190507 (Red Hat 8.3.1-4.5.0.5), clang version 8.0.1 (Red Hat 8.0.1-1.0.1.module+el8.1.0+5428+345cee14) 57 ubuntu:12.04 : Ok gcc (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3, Ubuntu clang version 3.0-6ubuntu3 (tags/RELEASE_30/final) (based on LLVM 3.0) 58 ubuntu:14.04 : Ok gcc (Ubuntu 4.8.4-2ubuntu1~14.04.4) 4.8.4 59 ubuntu:16.04 : Ok gcc (Ubuntu 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609, clang version 3.8.0-2ubuntu4 (tags/RELEASE_380/final) 60 ubuntu:16.04-x-arm : Ok arm-linux-gnueabihf-gcc (Ubuntu/Linaro 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609 61 ubuntu:16.04-x-arm64 : Ok aarch64-linux-gnu-gcc (Ubuntu/Linaro 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609 62 ubuntu:16.04-x-powerpc : Ok powerpc-linux-gnu-gcc (Ubuntu 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609 63 ubuntu:16.04-x-powerpc64 : Ok powerpc64-linux-gnu-gcc (Ubuntu/IBM 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609 64 ubuntu:16.04-x-powerpc64el : Ok powerpc64le-linux-gnu-gcc (Ubuntu/IBM 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609 65 ubuntu:16.04-x-s390 : Ok s390x-linux-gnu-gcc (Ubuntu 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609 66 ubuntu:18.04 : Ok gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0, clang version 6.0.0-1ubuntu2 (tags/RELEASE_600/final) 67 ubuntu:18.04-x-arm : Ok arm-linux-gnueabihf-gcc (Ubuntu/Linaro 7.5.0-3ubuntu1~18.04) 7.5.0 68 ubuntu:18.04-x-arm64 : Ok aarch64-linux-gnu-gcc (Ubuntu/Linaro 7.5.0-3ubuntu1~18.04) 7.5.0 69 ubuntu:18.04-x-m68k : Ok m68k-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 70 ubuntu:18.04-x-powerpc : Ok powerpc-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 71 ubuntu:18.04-x-powerpc64 : Ok powerpc64-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 72 ubuntu:18.04-x-powerpc64el : Ok powerpc64le-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 73 ubuntu:18.04-x-riscv64 : Ok riscv64-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 74 ubuntu:18.04-x-s390 : Ok s390x-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 75 ubuntu:18.04-x-sh4 : Ok sh4-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 76 ubuntu:18.04-x-sparc64 : Ok sparc64-linux-gnu-gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 77 ubuntu:19.04 : Ok gcc (Ubuntu 8.3.0-6ubuntu1) 8.3.0, clang version 8.0.0-3 (tags/RELEASE_800/final) 78 ubuntu:19.04-x-alpha : Ok alpha-linux-gnu-gcc (Ubuntu 8.3.0-6ubuntu1) 8.3.0 79 ubuntu:19.04-x-arm64 : Ok aarch64-linux-gnu-gcc (Ubuntu/Linaro 8.3.0-6ubuntu1) 8.3.0 80 ubuntu:19.04-x-hppa : Ok hppa-linux-gnu-gcc (Ubuntu 8.3.0-6ubuntu1) 8.3.0 81 ubuntu:19.10 : FAIL gcc (Ubuntu 9.2.1-9ubuntu2) 9.2.1 20191008, clang version 9.0.0-2 (tags/RELEASE_900/final) 82 ubuntu:20.04 : Ok gcc (Ubuntu 9.3.0-8ubuntu1) 9.3.0, clang version 10.0.0-1ubuntu1 $
# uname -a
Linux seventh 5.5.16-100.fc30.x86_64 #1 SMP Wed Apr 8 17:09:38 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
# git log --oneline -1
80af1ec0f232 perf script: Add flamegraph.py script
# perf version --build-options
perf version 5.6.g80af1ec0f232
dwarf: [ on ] # HAVE_DWARF_SUPPORT
dwarf_getlocations: [ on ] # HAVE_DWARF_GETLOCATIONS_SUPPORT
glibc: [ on ] # HAVE_GLIBC_SUPPORT
gtk2: [ on ] # HAVE_GTK2_SUPPORT
syscall_table: [ on ] # HAVE_SYSCALL_TABLE_SUPPORT
libbfd: [ on ] # HAVE_LIBBFD_SUPPORT
libelf: [ on ] # HAVE_LIBELF_SUPPORT
libnuma: [ on ] # HAVE_LIBNUMA_SUPPORT
numa_num_possible_cpus: [ on ] # HAVE_LIBNUMA_SUPPORT
libperl: [ on ] # HAVE_LIBPERL_SUPPORT
libpython: [ on ] # HAVE_LIBPYTHON_SUPPORT
libslang: [ on ] # HAVE_SLANG_SUPPORT
libcrypto: [ on ] # HAVE_LIBCRYPTO_SUPPORT
libunwind: [ on ] # HAVE_LIBUNWIND_SUPPORT
libdw-dwarf-unwind: [ on ] # HAVE_DWARF_SUPPORT
zlib: [ on ] # HAVE_ZLIB_SUPPORT
lzma: [ on ] # HAVE_LZMA_SUPPORT
get_cpuid: [ on ] # HAVE_AUXTRACE_SUPPORT
bpf: [ on ] # HAVE_LIBBPF_SUPPORT
aio: [ on ] # HAVE_AIO_SUPPORT
zstd: [ on ] # HAVE_ZSTD_SUPPORT
# perf test
1: vmlinux symtab matches kallsyms : Ok
2: Detect openat syscall event : Ok
3: Detect openat syscall event on all cpus : Ok
4: Read samples using the mmap interface : Ok
5: Test data source output : Ok
6: Parse event definition strings : Ok
7: Simple expression parser : Ok
8: PERF_RECORD_* events & perf_sample fields : Ok
9: Parse perf pmu format : Ok
10: PMU events : Ok
11: DSO data read : Ok
12: DSO data cache : Ok
13: DSO data reopen : Ok
14: Roundtrip evsel->name : Ok
15: Parse sched tracepoints fields : Ok
16: syscalls:sys_enter_openat event fields : Ok
17: Setup struct perf_event_attr : Ok
18: Match and link multiple hists : Ok
19: 'import perf' in python : Ok
20: Breakpoint overflow signal handler : Ok
21: Breakpoint overflow sampling : Ok
22: Breakpoint accounting : Ok
23: Watchpoint :
23.1: Read Only Watchpoint : Skip
23.2: Write Only Watchpoint : Ok
23.3: Read / Write Watchpoint : Ok
23.4: Modify Watchpoint : Ok
24: Number of exit events of a simple workload : Ok
25: Software clock events period values : Ok
26: Object code reading : Ok
27: Sample parsing : Ok
28: Use a dummy software event to keep tracking : Ok
29: Parse with no sample_id_all bit set : Ok
30: Filter hist entries : Ok
31: Lookup mmap thread : Ok
32: Share thread maps : Ok
33: Sort output of hist entries : Ok
34: Cumulate child hist entries : Ok
35: Track with sched_switch : Ok
36: Filter fds with revents mask in a fdarray : Ok
37: Add fd to a fdarray, making it autogrow : Ok
38: kmod_path__parse : Ok
39: Thread map : Ok
40: LLVM search and compile :
40.1: Basic BPF llvm compile : Ok
40.2: kbuild searching : Ok
40.3: Compile source for BPF prologue generation : Ok
40.4: Compile source for BPF relocation : Ok
41: Session topology : Ok
42: BPF filter :
42.1: Basic BPF filtering : Ok
42.2: BPF pinning : Ok
42.3: BPF prologue generation : Ok
42.4: BPF relocation checker : Ok
43: Synthesize thread map : Ok
44: Remove thread map : Ok
45: Synthesize cpu map : Ok
46: Synthesize stat config : Ok
47: Synthesize stat : Ok
48: Synthesize stat round : Ok
49: Synthesize attr update : Ok
50: Event times : Ok
51: Read backward ring buffer : Ok
52: Print cpu map : Ok
53: Merge cpu map : Ok
54: Probe SDT events : Ok
55: is_printable_array : Ok
56: Print bitmap : Ok
57: perf hooks : Ok
58: builtin clang support : Skip (not compiled in)
59: unit_number__scnprintf : Ok
60: mem2node : Ok
61: time utils : Ok
62: Test jit_write_elf : Ok
63: maps__merge_in : Ok
64: x86 rdpmc : Ok
65: Convert perf time to TSC : Ok
66: DWARF unwind : Ok
67: x86 instruction decoder - new instructions : Ok
68: Intel PT packet decoder : Ok
69: x86 bp modify : Ok
70: Use vfs_getname probe to get syscall args filenames : Ok
71: Check open filename arg using perf trace + vfs_getname: Ok
72: Zstd perf.data compression/decompression : Ok
73: probe libc's inet_pton & backtrace it with ping : Ok
74: Add vfs_getname probe to get syscall args filenames : Ok
# $ git log --oneline -1 ; make -C tools/perf build-test
80af1ec0f232 (HEAD -> perf/core, quaco/perf/core, five/perf/core) perf script: Add flamegraph.py script
make: Entering directory '/home/acme/git/perf/tools/perf'
- tarpkg: ./tests/perf-targz-src-pkg .
make_pure_O: make
make_debug_O: make DEBUG=1
make_no_libpython_O: make NO_LIBPYTHON=1
make_util_map_o_O: make util/map.o
make_no_libbpf_O: make NO_LIBBPF=1
make_with_babeltrace_O: make LIBBABELTRACE=1
make_no_libperl_O: make NO_LIBPERL=1
make_static_O: make LDFLAGS=-static NO_PERF_READ_VDSO32=1 NO_PERF_READ_VDSOX32=1 NO_JVMTI=1
make_util_pmu_bison_o_O: make util/pmu-bison.o
make_perf_o_O: make perf.o
make_no_libaudit_O: make NO_LIBAUDIT=1
make_install_O: make install
make_cscope_O: make cscope
make_no_backtrace_O: make NO_BACKTRACE=1
make_install_prefix_slash_O: make install prefix=/tmp/krava/
make_minimal_O: make NO_LIBPERL=1 NO_LIBPYTHON=1 NO_NEWT=1 NO_GTK2=1 NO_DEMANGLE=1 NO_LIBELF=1 NO_LIBUNWIND=1 NO_BACKTRACE=1 NO_LIBNUMA=1 NO_LIBAUDIT=1 NO_LIBBIONIC=1 NO_LIBDW_DWARF_UNWIND=1 NO_AUXTRACE=1 NO_LIBBPF=1 NO_LIBCRYPTO=1 NO_SDT=1 NO_JVMTI=1 NO_LIBZSTD=1 NO_LIBCAP=1
make_help_O: make help
make_no_libnuma_O: make NO_LIBNUMA=1
make_doc_O: make doc
make_clean_all_O: make clean all
make_install_prefix_O: make install prefix=/tmp/krava
make_no_scripts_O: make NO_LIBPYTHON=1 NO_LIBPERL=1
make_no_auxtrace_O: make NO_AUXTRACE=1
make_no_demangle_O: make NO_DEMANGLE=1
make_no_libelf_O: make NO_LIBELF=1
make_tags_O: make tags
make_no_slang_O: make NO_SLANG=1
make_no_ui_O: make NO_NEWT=1 NO_SLANG=1 NO_GTK2=1
make_no_libunwind_O: make NO_LIBUNWIND=1
make_install_bin_O: make install-bin
make_no_newt_O: make NO_NEWT=1
make_no_libbionic_O: make NO_LIBBIONIC=1
make_with_clangllvm_O: make LIBCLANGLLVM=1
make_no_libdw_dwarf_unwind_O: make NO_LIBDW_DWARF_UNWIND=1
make_no_gtk2_O: make NO_GTK2=1
OK
make: Leaving directory '/home/acme/git/perf/tools/perf'
$Arnaldo Carvalho de Melo
maintainerPATCH 1/26 perf tools: Synthesize bpf_trampoline/dispatcher ksymbol event
From: Jiri Olsa <[email protected]>
Synthesize bpf images (trampolines/dispatchers) on start, as ksymbol events from /proc/kallsyms. Having this perf can recognize samples from those images and perf report and top shows them correctly.
The rest of the ksymbol handling is already in place from for the bpf programs monitoring, so only the initial state was needed.
perf report output:
# Overhead Command Shared Object Symbol
12.37% test_progs [kernel.vmlinux] [k] entry_SYSCALL_64
11.80% test_progs [kernel.vmlinux] [k] syscall_return_via_sysret
9.63% test_progs bpf_prog_bcf7977d3b93787c_prog2 [k] bpf_prog_bcf7977d3b93787c_prog2
6.90% test_progs bpf_trampoline_24456 [k] bpf_trampoline_24456
6.36% test_progs [kernel.vmlinux] [k] memcpy_ermsCommitter notes:
Use scnprintf() instead of strncpy() to overcome this on fedora:32, rawhide and OpenMandriva Cooker:
CC /tmp/build/perf/util/bpf-event.o
In file included from /usr/include/string.h:495,
from /git/linux/tools/lib/bpf/libbpf_common.h:12,
from /git/linux/tools/lib/bpf/bpf.h:31,
from util/bpf-event.c:4:
In function 'strncpy',
inlined from 'process_bpf_image' at util/bpf-event.c:323:2,
inlined from 'kallsyms_process_symbol' at util/bpf-event.c:358:9:
/usr/include/bits/string_fortified.h:106:10: error: '__builtin_strncpy' specified bound 256 equals destination size [-Werror=stringop-truncation]
106 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~cc1: all warnings being treated as errors
--- tools/perf/util/bpf-event.c | 93 +++++++++++++++++++++++++++++++++++++ 1 file changed, 93 insertions(+)
tools/perf/util/bpf-event.c +93 -0
@@ -6,6 +6,9 @@#include <bpf/libbpf.h> #include <linux/btf.h> #include <linux/err.h> +#include <linux/string.h> +#include <internal/lib.h> +#include <symbol/kallsyms.h> #include "bpf-event.h" #include "debug.h" #include "dso.h"
@@ -290,11 +293,82 @@ static int perf_event__synthesize_one_bpf_prog(struct perf_session *session,return err ?-1 :0 ; } +struct kallsyms_parse { +union perf_event *event; +perf_event__handler_t process; +struct machine *machine; +struct perf_tool *tool; +}; + +static int +process_bpf_image (char *name,u64 addr,struct kallsyms_parse *data) +{ +struct machine *machine = data->machine; +union perf_event *event = data->event; +struct perf_record_ksymbol *ksymbol; +int len; + + ksymbol = &event->ksymbol; + + *ksymbol = (struct perf_record_ksymbol ) { + .header = { + .type =PERF_RECORD_KSYMBOL , + .size = offsetof(struct perf_record_ksymbol , name), + }, + .addr = addr, + .len = page_size, + .ksym_type =PERF_RECORD_KSYMBOL_TYPE_BPF , + .flags =0 , + }; + + len =scnprintf (ksymbol->name,KSYM_NAME_LEN ,"%s" , name); + ksymbol->header.size +=PERF_ALIGN (len +1 ,sizeof (u64)); +memset ((void *) event + event->header.size,0 , machine->id_hdr_size); + event->header.size += machine->id_hdr_size; + +return perf_tool__process_synth_event (data->tool, event, machine, + data->process); +} + +static int +kallsyms_process_symbol (void *data,const char *_name, +char type__maybe_unused ,u64 start) +{ +char disp[KSYM_NAME_LEN ]; +char *module, *name; +unsigned long id; +int err =0 ; + + module =strchr (_name,'\t' ); +if (!module) +return 0 ; + +/* We are going after [bpf] module ... */ +if (strcmp (module +1 ,"[bpf]" )) +return 0 ; + + name =memdup (_name, (module - _name) +1 ); +if (!name) +return -ENOMEM ; + + name[module - _name] =0 ; + +/* .. and only for trampolines and dispatchers */ +if ((sscanf (name,"bpf_trampoline_%lu" , &id) ==1 ) || + (sscanf (name,"bpf_dispatcher_%s" , disp) ==1 )) + err =process_bpf_image (name, start, data); + +free (name); +return err; +} +int perf_event__synthesize_bpf_events (struct perf_session *session,perf_event__handler_t process,struct machine *machine,struct record_opts *opts) { +const char *kallsyms_filename ="/proc/kallsyms" ; +struct kallsyms_parse arg;union perf_event *event;__u32 id =0 ;int err;
@@ -303,6 +377,8 @@ int perf_event__synthesize_bpf_events(struct perf_session *session, event =malloc (sizeof (event->bpf) +KSYM_NAME_LEN + machine->id_hdr_size);if (!event)return -1 ; + +/* Synthesize all the bpf programs in system. */ while (true) { err =bpf_prog_get_next_id (id, &id);if (err) {
@@ -335,6 +411,23 @@ int perf_event__synthesize_bpf_events(struct perf_session *session,break ; } } + +/* Synthesize all the bpf images - trampolines/dispatchers. */ +if (symbol_conf.kallsyms_name !=NULL ) + kallsyms_filename = symbol_conf.kallsyms_name; + + arg = (struct kallsyms_parse ) { + .event = event, + .process = process, + .machine = machine, + .tool = session->tool, + }; + +if (kallsyms__parse (kallsyms_filename, &arg, kallsyms_process_symbol)) { +pr_err ("%s: failed to synthesize bpf images: %s\n" , + __func__,strerror (errno)); + } +free (event);return err; }
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 2/26 perf machine: Set ksymbol dso as loaded on arrival
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
There's no special load action for ksymbol data on map__load/dso__load action, where the kernel is getting loaded. It only gets confused with kernel kallsyms/vmlinux load for bpf object, which fails and could mess up with the map.
Disabling any further load of the map for ksymbol related dso/map.
--- tools/perf/util/machine.c | 1 + 1 file changed, 1 insertion(+)
tools/perf/util/machine.c +1 -0
@@ -759,6 +759,7 @@ static int machine__process_ksymbol_register(struct machine *machine, map->start = event->ksymbol.addr; map->end = map->start + event->ksymbol.len;maps__insert (&machine->kmaps, map); +dso__set_loaded (dso); } sym = symbol__new(map->map_ip(map, map->start),
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 3/26 perf annotate: Add basic support for bpf_image
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Add the DSO_BINARY_TYPE__BPF_IMAGE dso binary type to recognize BPF images that carry trampoline or dispatcher.
Upcoming patches will add support to read the image data, store it within the BPF feature in perf.data and display it for annotation purposes.
Currently we only display following message:
# ./perf annotate bpf_trampoline_24456 --stdio
Percent | Source code & Disassembly of . for cycles (504 ...
--------------------------------------------------------------- ...
: to be implemented--- tools/perf/util/annotate.c | 20 ++++++++++++++++++++ tools/perf/util/dso.c | 1 + tools/perf/util/dso.h | 1 + tools/perf/util/machine.c | 11 +++++++++++ tools/perf/util/symbol.c | 1 + 5 files changed, 34 insertions(+)
tools/perf/util/annotate.c +20 -0
@@ -1821,6 +1821,24 @@ static int symbol__disassemble_bpf(struct symbol *sym __maybe_unused, }#endif // defined(HAVE_LIBBFD_SUPPORT) && defined(HAVE_LIBBPF_SUPPORT) +static int +symbol__disassemble_bpf_image (struct symbol *sym, +struct annotate_args *args) +{ +struct annotation *notes =symbol__annotation (sym); +struct disasm_line *dl; + + args->offset =-1 ; + args->line =strdup ("to be implemented" ); + args->line_nr =0 ; + dl =disasm_line__new (args); +if (dl) +annotation_line__add (&dl->al, ¬es->src->source); + +free (args->line); +return 0 ; +} +/* * Possibly create a new version of line with tabs expanded. Returns the * existing or new line, storage is updated if a new line is allocated. If
@@ -1920,6 +1938,8 @@ static int symbol__disassemble(struct symbol *sym, struct annotate_args *args)if (dso->binary_type ==DSO_BINARY_TYPE__BPF_PROG_INFO ) {return symbol__disassemble_bpf (sym, args); + }else if (dso->binary_type ==DSO_BINARY_TYPE__BPF_IMAGE ) { +return symbol__disassemble_bpf_image (sym, args); }else if (dso__is_kcore (dso)) { kce.kcore_filename = symfs_filename; kce.addr =map__rip_2objdump (map, sym->start);
tools/perf/util/dso.c +1 -0
@@ -191,6 +191,7 @@ int dso__read_binary_type_filename(const struct dso *dso,case DSO_BINARY_TYPE__GUEST_KALLSYMS :case DSO_BINARY_TYPE__JAVA_JIT :case DSO_BINARY_TYPE__BPF_PROG_INFO : +case DSO_BINARY_TYPE__BPF_IMAGE :case DSO_BINARY_TYPE__NOT_FOUND : ret =-1 ;break ;
tools/perf/util/dso.h +1 -0
@@ -40,6 +40,7 @@ enum dso_binary_type {DSO_BINARY_TYPE__GUEST_KCORE ,DSO_BINARY_TYPE__OPENEMBEDDED_DEBUGINFO ,DSO_BINARY_TYPE__BPF_PROG_INFO , +DSO_BINARY_TYPE__BPF_IMAGE , DSO_BINARY_TYPE__NOT_FOUND, };
tools/perf/util/machine.c +11 -0
@@ -736,6 +736,12 @@ int machine__process_switch_event(struct machine *machine __maybe_unused,return 0 ; } +static int is_bpf_image (const char *name) +{ +return strncmp (name,"bpf_trampoline_" ,sizeof ("bpf_trampoline_" ) -1 ) || +strncmp (name,"bpf_dispatcher_" ,sizeof ("bpf_dispatcher_" ) -1 ); +} +static int machine__process_ksymbol_register (struct machine *machine,union perf_event *event,struct perf_sample *sample__maybe_unused )
@@ -760,6 +766,11 @@ static int machine__process_ksymbol_register(struct machine *machine, map->end = map->start + event->ksymbol.len;maps__insert (&machine->kmaps, map);dso__set_loaded (dso); + +if (is_bpf_image (event->ksymbol.name)) { + dso->binary_type =DSO_BINARY_TYPE__BPF_IMAGE ; +dso__set_long_name (dso,"" , false); + } } sym = symbol__new(map->map_ip(map, map->start),
tools/perf/util/symbol.c +1 -0
@@ -1544,6 +1544,7 @@ static bool dso__is_compatible_symtab_type(struct dso *dso, bool kmod,return true;case DSO_BINARY_TYPE__BPF_PROG_INFO : +case DSO_BINARY_TYPE__BPF_IMAGE :case DSO_BINARY_TYPE__NOT_FOUND :default :return false;
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 4/26 tools arch x86: Sync the msr-index.h copy with the kernel sources
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
To pick up the changes in:
6650cdd9a8cc ("x86/split_lock: Enable split lock detection by kernel")
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' diff -u tools/arch/x86/include/asm/msr-index.h arch/x86/include/asm/msr-index.h
Which causes these changes in tooling:
$ 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 2020-04-01 12:11:14.789344795 -0300 +++ after 2020-04-01 12:11:56.907798879 -0300 @@ -10,6 +10,7 @@ [0x00000029] = "KNC_EVNTSEL1", [0x0000002a] = "IA32_EBL_CR_POWERON", [0x0000002c] = "EBC_FREQUENCY_ID", + [0x00000033] = "TEST_CTRL", [0x00000034] = "SMI_COUNT", [0x0000003a] = "IA32_FEAT_CTL", [0x0000003b] = "IA32_TSC_ADJUST", @@ -27,6 +28,7 @@ [0x000000c2] = "IA32_PERFCTR1", [0x000000cd] = "FSB_FREQ", [0x000000ce] = "PLATFORM_INFO", + [0x000000cf] = "IA32_CORE_CAPS", [0x000000e2] = "PKG_CST_CONFIG_CONTROL", [0x000000e7] = "IA32_MPERF", [0x000000e8] = "IA32_APERF", $
$ make -C tools/perf O=/tmp/build/perf install-bin
<SNIP>
CC /tmp/build/perf/trace/beauty/tracepoints/x86_msr.o
LD /tmp/build/perf/trace/beauty/tracepoints/perf-in.o
LD /tmp/build/perf/trace/beauty/perf-in.o
LD /tmp/build/perf/perf-in.o
LINK /tmp/build/perf/perf
<SNIP>Now one can do:
perf trace -e msr:* --filter=msr==IA32_CORE_CAPS
or:
perf trace -e msr:* --filter='msr==IA32_CORE_CAPS || msr==TEST_CTRL'
And see only those MSRs being accessed via:
# perf trace -v -e msr:* --filter='msr==IA32_CORE_CAPS || msr==TEST_CTRL' New filter for msr:read_msr: (msr==0xcf || msr==0x33) && (common_pid != 8263 && common_pid != 23250) New filter for msr:write_msr: (msr==0xcf || msr==0x33) && (common_pid != 8263 && common_pid != 23250) New filter for msr:rdpmc: (msr==0xcf || msr==0x33) && (common_pid != 8263 && common_pid != 23250)
--- tools/arch/x86/include/asm/msr-index.h | 9 +++++++++ 1 file changed, 9 insertions(+)
tools/arch/x86/include/asm/msr-index.h +9 -0
@@ -41,6 +41,10 @@/* Intel MSRs. Some also available on other CPUs */ +#define MSR_TEST_CTRL 0x00000033 +#define MSR_TEST_CTRL_SPLIT_LOCK_DETECT_BIT 29 +#define MSR_TEST_CTRL_SPLIT_LOCK_DETECT BIT(MSR_TEST_CTRL_SPLIT_LOCK_DETECT_BIT) +#define MSR_IA32_SPEC_CTRL 0x00000048/* Speculation Control */ #define SPEC_CTRL_IBRS BIT(0)/* Indirect Branch Restricted Speculation */ #define SPEC_CTRL_STIBP_SHIFT 1/* Single Thread Indirect Branch Predictor (STIBP) bit */
@@ -70,6 +74,11 @@*/ #define MSR_IA32_UMWAIT_CONTROL_TIME_MASK (~0x03U) +/* Abbreviated from Intel SDM name IA32_CORE_CAPABILITIES */ +#define MSR_IA32_CORE_CAPS 0x000000cf +#define MSR_IA32_CORE_CAPS_SPLIT_LOCK_DETECT_BIT 5 +#define MSR_IA32_CORE_CAPS_SPLIT_LOCK_DETECT BIT(MSR_IA32_CORE_CAPS_SPLIT_LOCK_DETECT_BIT) +#define MSR_PKG_CST_CONFIG_CONTROL 0x000000e2#define NHM_C3_AUTO_DEMOTE (1UL << 25)#define NHM_C1_AUTO_DEMOTE (1UL << 26)
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 5/26 capabilities: Introduce CAP_PERFMON to kernel and user space
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <alexey.budankov-VuQAYsv1563Yd54FQh9/[email protected]>
Introduce the CAP_PERFMON capability designed to secure system performance monitoring and observability operations so that CAP_PERFMON can assist CAP_SYS_ADMIN capability in its governing role for performance monitoring and observability subsystems.
CAP_PERFMON hardens system security and integrity during performance monitoring and observability operations by decreasing attack surface that is available to a CAP_SYS_ADMIN privileged process [2]. Providing the access to system performance monitoring and observability operations under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes the operation more secure.
Thus, CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e:
2.2.2.39 principle of least privilege: A security design principle that
states that a process or program be granted only those privileges
(e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
CAP_PERFMON meets the demand to secure system performance monitoring and observability operations for adoption in security sensitive, restricted, multiuser production environments (e.g. HPC clusters, cloud and virtual compute environments), where root or CAP_SYS_ADMIN credentials are not available to mass users of a system, and securely unblocks applicability and scalability of system performance monitoring and observability operations beyond root and CAP_SYS_ADMIN use cases.
CAP_PERFMON takes over CAP_SYS_ADMIN credentials related to system performance monitoring and observability operations and balances amount of CAP_SYS_ADMIN credentials following the recommendations in the capabilities man page [1] for CAP_SYS_ADMIN: "Note: this capability is overloaded; see Notes to kernel developers, below." For backward compatibility reasons access to system performance monitoring and observability subsystems of the kernel remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN capability usage for secure system performance monitoring and observability operations is discouraged with respect to the designed CAP_PERFMON capability.
Although the software running under CAP_PERFMON can not ensure avoidance of related hardware issues, the software can still mitigate these issues following the official hardware issues mitigation procedure [2]. The bugs in the software itself can be fixed following the standard kernel development process [3] to maintain and harden security of system performance monitoring and observability operations.
[1] http://man7.org/linux/man-pages/man7/capabilities.7.html
[2] https://www.kernel.org/doc/html/latest/process/embargoed-hardware-issues.html
[3] https://www.kernel.org/doc/html/latest/admin-guide/security-bugs.html
--- include/linux/capability.h | 4 ++++ include/uapi/linux/capability.h | 8 +++++++- security/selinux/include/classmap.h | 4 ++-- 3 files changed, 13 insertions(+), 3 deletions(-)
include/linux/capability.h +4 -0
@@ -251,6 +251,10 @@ extern bool privileged_wrt_inode_uidgid(struct user_namespace *ns, const structextern bool capable_wrt_inode_uidgid (const struct inode *inode,int cap);extern bool file_ns_capable (const struct file *file,struct user_namespace *ns,int cap);extern bool ptracer_capable (struct task_struct *tsk,struct user_namespace *ns); +static inline bool perfmon_capable (void ) +{ +return capable (CAP_PERFMON ) ||capable (CAP_SYS_ADMIN ); +}/* audit system wants to get cap info from files as well */ extern int get_vfs_caps_from_disk (const struct dentry *dentry,struct cpu_vfs_cap_data *cpu_caps);
include/uapi/linux/capability.h +7 -1
@@ -367,8 +367,14 @@ struct vfs_ns_cap_data {#define CAP_AUDIT_READ 37 +/* +* Allow system performance and observability privileged operations +* using perf_events, i915_perf and other kernel subsystems +*/ + +#define CAP_PERFMON 38 -#define CAP_LAST_CAP CAP_AUDIT_READ +#define CAP_LAST_CAP CAP_PERFMON#define cap_valid (x) ((x) >= 0 && (x) <= CAP_LAST_CAP)
security/selinux/include/classmap.h +2 -2
@@ -27,9 +27,9 @@ "audit_control", "setfcap"#define COMMON_CAP2_PERMS "mac_override", "mac_admin", "syslog", \ - "wake_alarm", "block_suspend", "audit_read" + "wake_alarm", "block_suspend", "audit_read", "perfmon" -#if CAP_LAST_CAP >CAP_AUDIT_READ +#if CAP_LAST_CAP >CAP_PERFMON #error New capability defined, please update COMMON_CAP2_PERMS.#endif
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 6/26 perf/core: Open access to the core for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to monitoring of kernel code, CPUs, tracepoints and namespaces data for a CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons the access to perf_events subsystem remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure perf_events monitoring is discouraged with respect to CAP_PERFMON capability.
--- include/linux/perf_event.h | 6 +++--- kernel/events/core.c | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-)
include/linux/perf_event.h +3 -3
@@ -1305,7 +1305,7 @@ static inline int perf_is_paranoid(void)static inline int perf_allow_kernel (struct perf_event_attr *attr) { -if (sysctl_perf_event_paranoid >1 && !capable (CAP_SYS_ADMIN )) +if (sysctl_perf_event_paranoid >1 && !perfmon_capable ())return -EACCES ;return security_perf_event_open (attr,PERF_SECURITY_KERNEL );
@@ -1313,7 +1313,7 @@ static inline int perf_allow_kernel(struct perf_event_attr *attr)static inline int perf_allow_cpu (struct perf_event_attr *attr) { -if (sysctl_perf_event_paranoid >0 && !capable (CAP_SYS_ADMIN )) +if (sysctl_perf_event_paranoid >0 && !perfmon_capable ())return -EACCES ;return security_perf_event_open (attr,PERF_SECURITY_CPU );
@@ -1321,7 +1321,7 @@ static inline int perf_allow_cpu(struct perf_event_attr *attr)static inline int perf_allow_tracepoint (struct perf_event_attr *attr) { -if (sysctl_perf_event_paranoid >-1 && !capable (CAP_SYS_ADMIN )) +if (sysctl_perf_event_paranoid >-1 && !perfmon_capable ())return -EPERM ;return security_perf_event_open (attr,PERF_SECURITY_TRACEPOINT );
kernel/events/core.c +1 -1
@@ -11486,7 +11486,7 @@ SYSCALL_DEFINE5(perf_event_open, }if (attr.namespaces) { -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EACCES ; }
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 7/26 perf/core: open access to probes for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to monitoring via kprobes and uprobes and eBPF tracing for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
perf kprobes and uprobes are used by ftrace and eBPF. perf probe uses ftrace to define new kprobe events, and those events are treated as tracepoint events. eBPF defines new probes via perf_event_open interface and then the probes are used in eBPF tracing.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to perf_events subsystem remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure perf_events monitoring is discouraged with respect to CAP_PERFMON capability.
--- kernel/events/core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/events/core.c +2 -2
@@ -9377,7 +9377,7 @@ static int perf_kprobe_event_init(struct perf_event *event)if (event->attr.type != perf_kprobe.type)return -ENOENT ; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EACCES ;/*
@@ -9437,7 +9437,7 @@ static int perf_uprobe_event_init(struct perf_event *event)if (event->attr.type != perf_uprobe.type)return -ENOENT ; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EACCES ;/*
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 8/26 perf tools: Support CAP_PERFMON capability
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Extend error messages to mention CAP_PERFMON capability as an option to substitute CAP_SYS_ADMIN capability for secure system performance monitoring and observability operations. Make perf_event_paranoid_check() and __cmd_ftrace() to be aware of CAP_PERFMON capability.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to perf_events subsystem remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure perf_events monitoring is discouraged with respect to CAP_PERFMON capability.
Committer testing:
Using a libcap with this patch:
diff --git a/libcap/include/uapi/linux/capability.h b/libcap/include/uapi/linux/capability.h
index 78b2fd4c8a95..89b5b0279b60 100644
--- a/libcap/include/uapi/linux/capability.h
+++ b/libcap/include/uapi/linux/capability.h
@@ -366,8 +366,9 @@ struct vfs_ns_cap_data {#define CAP_AUDIT_READ 37
+#define CAP_PERFMON 38
-#define CAP_LAST_CAP CAP_AUDIT_READ
+#define CAP_LAST_CAP CAP_PERFMON
#define cap_valid(x) ((x) >= 0 && (x) <= CAP_LAST_CAP)
Note that using '38' in place of 'cap_perfmon' works to some degree with an old libcap, its only when cap_get_flag() is called that libcap performs an error check based on the maximum value known for capabilities that it will fail.
This makes determining the default of perf_event_attr.exclude_kernel to fail, as it can't determine if CAP_PERFMON is in place.
Using 'perf top -e cycles' avoids the default check and sets perf_event_attr.exclude_kernel to 1.
As root, with a libcap supporting CAP_PERFMON:
# groupadd perf_users # adduser perf -g perf_users # mkdir ~perf/bin # cp ~acme/bin/perf ~perf/bin/ # chgrp perf_users ~perf/bin/perf # setcap "cap_perfmon,cap_sys_ptrace,cap_syslog=ep" ~perf/bin/perf # getcap ~perf/bin/perf /home/perf/bin/perf = cap_sys_ptrace,cap_syslog,cap_perfmon+ep # ls -la ~perf/bin/perf -rwxr-xr-x. 1 root perf_users 16968552 Apr 9 13:10 /home/perf/bin/perf
As the 'perf' user in the 'perf_users' group:
$ perf top -a --stdio Error: Failed to mmap with 1 (Operation not permitted) $
Either add the cap_ipc_lock capability to the perf binary or reduce the ring buffer size to some smaller value:
$ perf top -m10 -a --stdio rounding mmap pages size to 64K (16 pages) Error: Failed to mmap with 1 (Operation not permitted) $ perf top -m4 -a --stdio Error: Failed to mmap with 1 (Operation not permitted) $ perf top -m2 -a --stdio PerfTop: 762 irqs/sec kernel:49.7% exact: 100.0% lost: 0/0 drop: 0/0 [4000Hz cycles], (all, 4 CPUs) ------------------------------------------------------------------------------------------------------
9.83% perf [.] __symbols__insert
8.58% perf [.] rb_next
5.91% [kernel] [k] module_get_kallsym
5.66% [kernel] [k] kallsyms_expand_symbol.constprop.0
3.98% libc-2.29.so [.] __GI_____strtoull_l_internal
3.66% perf [.] rb_insert_color
2.34% [kernel] [k] vsnprintf
2.30% [kernel] [k] string_nocheck
2.16% libc-2.29.so [.] _IO_getdelim
2.15% [kernel] [k] number
2.13% [kernel] [k] format_decode
1.58% libc-2.29.so [.] _IO_feof
1.52% libc-2.29.so [.] __strcmp_avx2
1.50% perf [.] rb_set_parent_color
1.47% libc-2.29.so [.] __libc_calloc
1.24% [kernel] [k] do_syscall_64
1.17% [kernel] [k] __x86_indirect_thunk_rax $ perf record -a sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.552 MB perf.data (74 samples) ]
$ perf evlist
cycles
$ perf evlist -v
cycles: size: 120, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|CPU|PERIOD, 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
$ perf report | head -20
# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 74 of event 'cycles'
# Event count (approx.): 15694834
#
# Overhead Command Shared Object Symbol
# ........ ............... .......................... ......................................
#
19.62% perf [kernel.vmlinux] [k] strnlen_user
13.88% swapper [kernel.vmlinux] [k] intel_idle
13.83% ksoftirqd/0 [kernel.vmlinux] [k] pfifo_fast_dequeue
13.51% swapper [kernel.vmlinux] [k] kmem_cache_free
6.31% gnome-shell [kernel.vmlinux] [k] kmem_cache_free
5.66% kworker/u8:3+ix [kernel.vmlinux] [k] delay_tsc
4.42% perf [kernel.vmlinux] [k] __set_cpus_allowed_ptr
3.45% kworker/2:1-eve [kernel.vmlinux] [k] shmem_truncate_range
2.29% gnome-shell libgobject-2.0.so.0.6000.7 [.] g_closure_ref
$--- tools/perf/builtin-ftrace.c | 5 +++-- tools/perf/design.txt | 3 ++- tools/perf/util/cap.h | 4 ++++ tools/perf/util/evsel.c | 10 +++++----- tools/perf/util/util.c | 1 + 5 files changed, 15 insertions(+), 8 deletions(-)
tools/perf/builtin-ftrace.c +3 -2
@@ -284,10 +284,11 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace, int argc, const char **argv) .events = POLLIN, }; - if (!perf_cap__capable(CAP_SYS_ADMIN)) { + if (!(perf_cap__capable(CAP_PERFMON) || + perf_cap__capable(CAP_SYS_ADMIN))) { pr_err("ftrace only works for %s!\n", #ifdef HAVE_LIBCAP_SUPPORT - "users with the SYS_ADMIN capability" + "users with the CAP_PERFMON or CAP_SYS_ADMIN capability" #else "root" #endif
tools/perf/design.txt +2 -1
@@ -258,7 +258,8 @@ gets schedule to. Per task counters can be created by any user, for their own tasks. A 'pid == -1' and 'cpu == x' counter is a per CPU counter that counts -all events on CPU-x. Per CPU counters need CAP_SYS_ADMIN privilege. +all events on CPU-x. Per CPU counters need CAP_PERFMON or CAP_SYS_ADMIN +privilege. The 'flags' parameter is currently unused and must be zero.
tools/perf/util/cap.h +4 -0
@@ -29,4 +29,8 @@ static inline bool perf_cap__capable(int cap __maybe_unused)#define CAP_SYSLOG 34#endif +#ifndef CAP_PERFMON +#define CAP_PERFMON 38 +#endif +#endif /* __PERF_CAP_H */
tools/perf/util/evsel.c +5 -5
@@ -2523,14 +2523,14 @@ int perf_evsel__open_strerror(struct evsel *evsel, struct target *target,"You may not have permission to collect %sstats.\n\n" "Consider tweaking /proc/sys/kernel/perf_event_paranoid,\n" "which controls use of the performance events system by\n" -"unprivileged users (without CAP_SYS_ADMIN).\n\n" +"unprivileged users (without CAP_PERFMON or CAP_SYS_ADMIN).\n\n" "The current value is %d:\n\n" " -1: Allow use of (almost) all events by all users\n" " Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK\n" -">= 0: Disallow ftrace function tracepoint by users without CAP_SYS_ADMIN\n" -" Disallow raw tracepoint access by users without CAP_SYS_ADMIN\n" -">= 1: Disallow CPU event access by users without CAP_SYS_ADMIN\n" -">= 2: Disallow kernel profiling by users without CAP_SYS_ADMIN\n\n" +">= 0: Disallow ftrace function tracepoint by users without CAP_PERFMON or CAP_SYS_ADMIN\n" +" Disallow raw tracepoint access by users without CAP_SYS_PERFMON or CAP_SYS_ADMIN\n" +">= 1: Disallow CPU event access by users without CAP_PERFMON or CAP_SYS_ADMIN\n" +">= 2: Disallow kernel profiling by users without CAP_PERFMON or CAP_SYS_ADMIN\n\n" "To make this setting permanent, edit /etc/sysctl.conf too, e.g.:\n\n" " kernel.perf_event_paranoid = -1\n" , target->system_wide ? "system-wide " : "",
tools/perf/util/util.c +1 -0
@@ -290,6 +290,7 @@ int perf_event_paranoid(void)bool perf_event_paranoid_check (int max_level) {return perf_cap__capable (CAP_SYS_ADMIN ) || +perf_cap__capable (CAP_PERFMON ) ||perf_event_paranoid () <= max_level; }
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 9/26 drm/i915/perf: Open access for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to i915_perf monitoring for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to i915_events subsystem remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure i915_events monitoring is discouraged with respect to CAP_PERFMON capability.
--- drivers/gpu/drm/i915/i915_perf.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/i915_perf.c +6 -7
@@ -3402,10 +3402,10 @@ i915_perf_open_ioctl_locked(struct i915_perf *perf,/* Similar to perf's kernel.perf_paranoid_cpu sysctl option * we check a dev.i915.perf_stream_paranoid sysctl option * to determine if it's ok to access system wide OA counters -* without CAP_SYS_ADMIN privileges. +* without CAP_PERFMON or CAP_SYS_ADMIN privileges. */ if (privileged_op && - i915_perf_stream_paranoid && !capable (CAP_SYS_ADMIN )) { + i915_perf_stream_paranoid && !perfmon_capable ()) {DRM_DEBUG ("Insufficient privileges to open i915 perf stream\n" ); ret = -EACCES ; goto err_ctx;
@@ -3598,9 +3598,8 @@ static int read_properties_unlocked(struct i915_perf *perf, }else oa_freq_hz =0 ; -if (oa_freq_hz > i915_oa_max_sample_rate && - !capable (CAP_SYS_ADMIN )) { -DRM_DEBUG ("OA exponent would exceed the max sampling frequency (sysctl dev.i915.oa_max_sample_rate) %uHz without root privileges\n" , +if (oa_freq_hz > i915_oa_max_sample_rate && !perfmon_capable ()) { +DRM_DEBUG ("OA exponent would exceed the max sampling frequency (sysctl dev.i915.oa_max_sample_rate) %uHz without CAP_PERFMON or CAP_SYS_ADMIN privileges\n" , i915_oa_max_sample_rate);return -EACCES ; }
@@ -4021,7 +4020,7 @@ int i915_perf_add_config_ioctl(struct drm_device *dev, void *data,return -EINVAL ; } -if (i915_perf_stream_paranoid && !capable (CAP_SYS_ADMIN )) { +if (i915_perf_stream_paranoid && !perfmon_capable ()) {DRM_DEBUG ("Insufficient privileges to add i915 OA config\n" );return -EACCES ; }
@@ -4168,7 +4167,7 @@ int i915_perf_remove_config_ioctl(struct drm_device *dev, void *data,return -ENOTSUPP ; } -if (i915_perf_stream_paranoid && !capable (CAP_SYS_ADMIN )) { +if (i915_perf_stream_paranoid && !perfmon_capable ()) {DRM_DEBUG ("Insufficient privileges to remove i915 OA config\n" );return -EACCES ; }
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 10/26 trace/bpf_trace: Open access for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to bpf_trace monitoring for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to bpf_trace monitoring remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure bpf_trace monitoring is discouraged with respect to CAP_PERFMON capability.
--- kernel/trace/bpf_trace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/trace/bpf_trace.c +1 -1
@@ -1416,7 +1416,7 @@ int perf_event_query_prog_array(struct perf_event *event, void __user *info)u32 *ids, prog_cnt, ids_len;int ret; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EPERM ;if (event->attr.type !=PERF_TYPE_TRACEPOINT )return -EINVAL ;
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 11/26 powerpc/perf: open access for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to monitoring for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to the monitoring remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure monitoring is discouraged with respect to CAP_PERFMON capability.
--- arch/powerpc/perf/imc-pmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/powerpc/perf/imc-pmu.c +2 -2
@@ -898,7 +898,7 @@ static int thread_imc_event_init(struct perf_event *event)if (event->attr.type != event->pmu->type)return -ENOENT ; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EACCES ;/* Sampling not supported */
@@ -1307,7 +1307,7 @@ static int trace_imc_event_init(struct perf_event *event)if (event->attr.type != event->pmu->type)return -ENOENT ; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EACCES ;/* Return if this is a couting event */
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 12/26 parisc/perf: open access for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to monitoring for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to the monitoring remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure monitoring is discouraged with respect to CAP_PERFMON capability.
--- arch/parisc/kernel/perf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/parisc/kernel/perf.c +1 -1
@@ -300,7 +300,7 @@ static ssize_t perf_write(struct file *file, const char __user *buf,else return -EFAULT ; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EACCES ;if (count !=sizeof (uint32_t ))
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 13/26 drivers/perf: Open access for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to monitoring for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to the monitoring remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure monitoring is discouraged with respect to CAP_PERFMON capability.
--- drivers/perf/arm_spe_pmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/perf/arm_spe_pmu.c +2 -2
@@ -274,7 +274,7 @@ static u64 arm_spe_event_to_pmscr(struct perf_event *event)if (!attr->exclude_kernel) reg |=BIT (SYS_PMSCR_EL1_E1SPE_SHIFT ); -if (IS_ENABLED (CONFIG_PID_IN_CONTEXTIDR ) &&capable (CAP_SYS_ADMIN )) +if (IS_ENABLED (CONFIG_PID_IN_CONTEXTIDR ) &&perfmon_capable ()) reg |=BIT (SYS_PMSCR_EL1_CX_SHIFT );return reg;
@@ -700,7 +700,7 @@ static int arm_spe_pmu_event_init(struct perf_event *event)return -EOPNOTSUPP ; reg =arm_spe_event_to_pmscr (event); -if (!capable (CAP_SYS_ADMIN ) && +if (!perfmon_capable () && (reg & (BIT (SYS_PMSCR_EL1_PA_SHIFT ) |BIT (SYS_PMSCR_EL1_CX_SHIFT ) |BIT (SYS_PMSCR_EL1_PCT_SHIFT ))))
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 14/26 drivers/oprofile: Open access for CAP_PERFMON privileged process
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Open access to monitoring for CAP_PERFMON privileged process. Providing the access under CAP_PERFMON capability singly, without the rest of CAP_SYS_ADMIN credentials, excludes chances to misuse the credentials and makes operation more secure.
CAP_PERFMON implements the principle of least privilege for performance monitoring and observability operations (POSIX IEEE 1003.1e 2.2.2.39 principle of least privilege: A security design principle that states that a process or program be granted only those privileges (e.g., capabilities) necessary to accomplish its legitimate function, and only for the time that such privileges are actually required)
For backward compatibility reasons access to the monitoring remains open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure monitoring is discouraged with respect to CAP_PERFMON capability.
--- drivers/oprofile/event_buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/oprofile/event_buffer.c +1 -1
@@ -113,7 +113,7 @@ static int event_buffer_open(struct inode *inode, struct file *file) {int err = -EPERM ; -if (!capable (CAP_SYS_ADMIN )) +if (!perfmon_capable ())return -EPERM ;if (test_and_set_bit_lock (0 , &buffer_opened))
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 15/26 doc/admin-guide: Update perf-security.rst with CAP_PERFMON information
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Update perf-security.rst documentation file with the information related to usage of CAP_PERFMON capability to secure performance monitoring and observability operations in system.
Committer notes:
While testing 'perf top' under cap_perfmon I noticed that it needs some more capability and Alexey pointed out cap_ipc_lock, as needed by this kernel chunk:
kernel/events/core.c: 6101
if ((locked > lock_limit) && perf_is_paranoid() &&
!capable(CAP_IPC_LOCK)) {
ret = -EPERM;
goto unlock;
}So I added it to the documentation, and also mentioned that if the libcap version doesn't yet supports 'cap_perfmon', its numeric value can be used instead, i.e. if:
# setcap "cap_perfmon,cap_ipc_lock,cap_sys_ptrace,cap_syslog=ep" perf
Fails, try:
# setcap "38,cap_ipc_lock,cap_sys_ptrace,cap_syslog=ep" perf
I also added a paragraph stating that using an unpatched libcap will fail the check for CAP_PERFMON, as it checks the cap number against a maximum to see if it is valid, which makes it use as the default the 'cycles:u' event, even tho a cap_perfmon capable perf binary can get kernel samples, to workaround that just use, e.g.:
# perf top -e cycles # perf record -e cycles
And it will sample kernel and user modes.
--- Documentation/admin-guide/perf-security.rst | 86 +++++++++++++++------ 1 file changed, 61 insertions(+), 25 deletions(-)
Documentation/admin-guide/perf-security.rst +61 -25
@@ -1,6 +1,6 @@ .. _perf_security: -Perf Events and tool security +Perf events and tool security ============================= Overview
@@ -42,11 +42,11 @@ categories: Data that belong to the fourth category can potentially contain sensitive process data. If PMUs in some monitoring modes capture values of execution context registers or data from process memory then access -to such monitoring capabilities requires to be ordered and secured -properly. So, perf_events/Perf performance monitoring is the subject for -security access control management [5]_ . +to such monitoring modes requires to be ordered and secured properly. +So, perf_events performance monitoring and observability operations are +the subject for security access control management [5]_ . -perf_events/Perf access control +perf_events access control ------------------------------- To perform security checks, the Linux implementation splits processes
@@ -66,11 +66,25 @@ into distinct units, known as capabilities [6]_ , which can be independently enabled and disabled on per-thread basis for processes and files of unprivileged users. -Unprivileged processes with enabled CAP_SYS_ADMIN capability are treated +Unprivileged processes with enabled CAP_PERFMON capability are treated as privileged processes with respect to perf_events performance -monitoring and bypass *scope* permissions checks in the kernel. - -Unprivileged processes using perf_events system call API is also subject +monitoring and observability operations, thus, bypass *scope* permissions +checks in the kernel. CAP_PERFMON implements the principle of least +privilege [13]_ (POSIX 1003.1e: 2.2.2.39) for performance monitoring and +observability operations in the kernel and provides a secure approach to +perfomance monitoring and observability in the system. + +For backward compatibility reasons the access to perf_events monitoring and +observability operations is also open for CAP_SYS_ADMIN privileged +processes but CAP_SYS_ADMIN usage for secure monitoring and observability +use cases is discouraged with respect to the CAP_PERFMON capability. +If system audit records [14]_ for a process using perf_events system call +API contain denial records of acquiring both CAP_PERFMON and CAP_SYS_ADMIN +capabilities then providing the process with CAP_PERFMON capability singly +is recommended as the preferred secure approach to resolve double access +denial logging related to usage of performance monitoring and observability. + +Unprivileged processes using perf_events system call are also subject for PTRACE_MODE_READ_REALCREDS ptrace access mode check [7]_ , whose outcome determines whether monitoring is permitted. So unprivileged processes provided with CAP_SYS_PTRACE capability are effectively
@@ -82,14 +96,14 @@ performance analysis of monitored processes or a system. For example, CAP_SYSLOG capability permits reading kernel space memory addresses from /proc/kallsyms file. -perf_events/Perf privileged users +Privileged Perf users groups --------------------------------- Mechanisms of capabilities, privileged capability-dumb files [6]_ and -file system ACLs [10]_ can be used to create a dedicated group of -perf_events/Perf privileged users who are permitted to execute -performance monitoring without scope limits. The following steps can be -taken to create such a group of privileged Perf users. +file system ACLs [10]_ can be used to create dedicated groups of +privileged Perf users who are permitted to execute performance monitoring +and observability without scope limits. The following steps can be +taken to create such groups of privileged Perf users. 1. Create perf_users group of privileged Perf users, assign perf_users group to Perf tool executable and limit access to the executable for
@@ -108,30 +122,51 @@ taken to create such a group of privileged Perf users. -rwxr-x--- 2 root perf_users 11M Oct 19 15:12 perf 2. Assign the required capabilities to the Perf tool executable file and - enable members of perf_users group with performance monitoring + enable members of perf_users group with monitoring and observability privileges [6]_ : :: - # setcap "cap_sys_admin,cap_sys_ptrace,cap_syslog=ep" perf - # setcap -v "cap_sys_admin,cap_sys_ptrace,cap_syslog=ep" perf + # setcap "cap_perfmon,cap_sys_ptrace,cap_syslog=ep" perf + # setcap -v "cap_perfmon,cap_sys_ptrace,cap_syslog=ep" perf perf: OK # getcap perf - perf = cap_sys_ptrace,cap_sys_admin,cap_syslog+ep + perf = cap_sys_ptrace,cap_syslog,cap_perfmon+ep + +If the libcap installed doesn't yet support "cap_perfmon", use "38" instead, +i.e.: + +:: + + # setcap "38,cap_ipc_lock,cap_sys_ptrace,cap_syslog=ep" perf + +Note that you may need to have 'cap_ipc_lock' in the mix for tools such as +'perf top', alternatively use 'perf top -m N', to reduce the memory that +it uses for the perf ring buffer, see the memory allocation section below. + +Using a libcap without support for CAP_PERFMON will make cap_get_flag(caps, 38, +CAP_EFFECTIVE, &val) fail, which will lead the default event to be 'cycles:u', +so as a workaround explicitly ask for the 'cycles' event, i.e.: + +:: + + # perf top -e cycles + +To get kernel and user samples with a perf binary with just CAP_PERFMON. As a result, members of perf_users group are capable of conducting -performance monitoring by using functionality of the configured Perf -tool executable that, when executes, passes perf_events subsystem scope -checks. +performance monitoring and observability by using functionality of the +configured Perf tool executable that, when executes, passes perf_events +subsystem scope checks. This specific access control management is only available to superuser or root running processes with CAP_SETPCAP, CAP_SETFCAP [6]_ capabilities. -perf_events/Perf unprivileged users +Unprivileged users ----------------------------------- -perf_events/Perf *scope* and *access* control for unprivileged processes +perf_events *scope* and *access* control for unprivileged processes is governed by perf_event_paranoid [2]_ setting: -1:
@@ -166,7 +201,7 @@ is governed by perf_event_paranoid [2]_ setting: perf_event_mlock_kb locking limit is imposed but ignored for unprivileged processes with CAP_IPC_LOCK capability. -perf_events/Perf resource control +Resource control --------------------------------- Open file descriptors
@@ -227,4 +262,5 @@ Bibliography .. [10] `<http://man7.org/linux/man-pages/man5/acl.5.html>`_ .. [11] `<http://man7.org/linux/man-pages/man2/getrlimit.2.html>`_ .. [12] `<http://man7.org/linux/man-pages/man5/limits.conf.5.html>`_ - +.. [13] `<https://sites.google.com/site/fullycapable>`_ +.. [14] `<http://man7.org/linux/man-pages/man8/auditd.8.html>`_
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 16/26 doc/admin-guide: update kernel.rst with CAP_PERFMON information
In reply to Arnaldo Carvalho de MeloFrom: Alexey Budankov <[email protected]>
Update the kernel.rst documentation file with the information related to usage of CAP_PERFMON capability to secure performance monitoring and observability operations in system.
--- Documentation/admin-guide/sysctl/kernel.rst | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-)
Documentation/admin-guide/sysctl/kernel.rst +11 -5
@@ -709,7 +709,13 @@ perf_event_paranoid =================== Controls use of the performance events system by unprivileged -users (without CAP_SYS_ADMIN). The default value is 2. +users (without CAP_PERFMON). The default value is 2. + +For backward compatibility reasons access to system performance +monitoring and observability remains open for CAP_SYS_ADMIN +privileged processes but CAP_SYS_ADMIN usage for secure system +performance monitoring and observability operations is discouraged +with respect to CAP_PERFMON use cases. === ================================================================== -1 Allow use of (almost) all events by all users.
@@ -718,13 +724,13 @@ users (without CAP_SYS_ADMIN). The default value is 2. ``CAP_IPC_LOCK``. >=0 Disallow ftrace function tracepoint by users without - ``CAP_SYS_ADMIN``. + ``CAP_PERFMON``. - Disallow raw tracepoint access by users without ``CAP_SYS_ADMIN``. + Disallow raw tracepoint access by users without ``CAP_PERFMON``. ->=1 Disallow CPU event access by users without ``CAP_SYS_ADMIN``. +>=1 Disallow CPU event access by users without ``CAP_PERFMON``. ->=2 Disallow kernel profiling by users without ``CAP_SYS_ADMIN``. +>=2 Disallow kernel profiling by users without ``CAP_PERFMON``. === ==================================================================
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 17/26 perf python: Check if clang supports -fno-semantic-interposition
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
The set of C compiler options used by distros to build python bindings may include options that are unknown to clang, we check for a variety of such options, add -fno-semantic-interposition to that mix:
This fixes the build on, among others, Manjaro Linux:
GEN /tmp/build/perf/python/perf.so
clang-9: error: unknown argument: '-fno-semantic-interposition'
error: command 'clang' failed with exit status 1
make: Leaving directory '/git/perf/tools/perf'
[perfbuilder@602aed1c266d ~]$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-pc-linux-gnu/9.3.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /build/gcc/src/gcc/configure --prefix=/usr --libdir=/usr/lib --libexecdir=/usr/lib --mandir=/usr/share/man --infodir=/usr/share/info --with-pkgversion='Arch Linux 9.3.0-1' --with-bugurl=https://bugs.archlinux.org/ --enable-languages=c,c++,ada,fortran,go,lto,objc,obj-c++,d --enable-shared --enable-threads=posix --with-system-zlib --with-isl --enable-__cxa_atexit --disable-libunwind-exceptions --enable-clocale=gnu --disable-libstdcxx-pch --disable-libssp --enable-gnu-unique-object --enable-linker-build-id --enable-lto --enable-plugin --enable-install-libiberty --with-linker-hash-style=gnu --enable-gnu-indirect-function --enable-multilib --disable-werror --enable-checking=release --enable-default-pie --enable-default-ssp --enable-cet=auto gdc_include_dir=/usr/include/dlang/
gdc
Thread model: posix
gcc version 9.3.0 (Arch Linux 9.3.0-1)
[perfbuilder@602aed1c266d ~]$
--- tools/perf/util/setup.py | 2 ++ 1 file changed, 2 insertions(+)
tools/perf/util/setup.py +2 -0
@@ -21,6 +21,8 @@ if cc_is_clang: vars[var] = sub("-fstack-clash-protection", "", vars[var]) if not clang_has_option("-fstack-protector-strong"): vars[var] = sub("-fstack-protector-strong", "", vars[var]) + if not clang_has_option("-fno-semantic-interposition"): + vars[var] = sub("-fno-semantic-interposition", "", vars[var]) from distutils.core import setup, Extension
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 18/26 perf script: Simplify auxiliary event printing functions
In reply to Arnaldo Carvalho de MeloFrom: Adrian Hunter <[email protected]>
This simplifies the print functions for the following perf script options:
--show-task-events
--show-namespace-events
--show-cgroup-events
--show-mmap-events
--show-switch-events
--show-lost-events
--show-bpf-events
Example:
# perf record --switch-events -a -e cycles -c 10000 sleep 1 Before: # perf script --show-task-events --show-namespace-events --show-cgroup-events --show-mmap-events --show-switch-events --show-lost-events --show-bpf-events > out-before.txt After: # perf script --show-task-events --show-namespace-events --show-cgroup-events --show-mmap-events --show-switch-events --show-lost-events --show-bpf-events > out-after.txt # diff -s out-before.txt out-after.txt Files out-before.txt and out-after.tx are identical
--- tools/perf/builtin-script.c | 304 ++++++++---------------------------- 1 file changed, 66 insertions(+), 238 deletions(-)
tools/perf/builtin-script.c +66 -238
@@ -2040,7 +2040,7 @@ static int cleanup_scripting(void)static bool filter_cpu (struct perf_sample *sample) { -if (cpu_list) +if (cpu_list && sample->cpu != (u32)-1 )return !test_bit (sample->cpu, cpu_bitmap);return false; }
@@ -2138,41 +2138,59 @@ static int process_attr(struct perf_tool *tool, union perf_event *event,return err; } -static int process_comm_event (struct perf_tool *tool, -union perf_event *event, -struct perf_sample *sample, -struct machine *machine) +static int print_event_with_time (struct perf_tool *tool, +union perf_event *event, +struct perf_sample *sample, +struct machine *machine, +pid_t pid,pid_t tid,u64 timestamp) { -struct thread *thread; struct perf_script *script = container_of(tool, struct perf_script, tool);struct perf_session *session = script->session;struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); -int ret =-1 ; +struct thread *thread =NULL ; - thread =machine__findnew_thread (machine, event->comm.pid, event->comm.tid); -if (thread ==NULL ) { -pr_debug ("problem processing COMM event, skipping it.\n" ); -return -1 ; +if (evsel && !evsel->core.attr.sample_id_all) { + sample->cpu =0 ; + sample->time = timestamp; + sample->pid = pid; + sample->tid = tid; } -if (perf_event__process_comm (tool, event, sample, machine) <0 ) - goto out; +if (filter_cpu (sample)) +return 0 ; -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time =0 ; - sample->tid = event->comm.tid; - sample->pid = event->comm.pid; - } -if (!filter_cpu (sample)) { +if (tid !=-1 ) + thread =machine__findnew_thread (machine, pid, tid); + +if (thread && evsel) {perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_COMM , stdout); -perf_event__fprintf (event, stdout); + event->header.type, stdout); } - ret =0 ; -out: + +perf_event__fprintf (event, stdout); +thread__put (thread); -return ret; + +return 0 ; +} + +static int print_event (struct perf_tool *tool,union perf_event *event, +struct perf_sample *sample,struct machine *machine, +pid_t pid,pid_t tid) +{ +return print_event_with_time (tool, event, sample, machine, pid, tid,0 ); +} + +static int process_comm_event (struct perf_tool *tool, +union perf_event *event, +struct perf_sample *sample, +struct machine *machine) +{ +if (perf_event__process_comm (tool, event, sample, machine) <0 ) +return -1 ; + +return print_event (tool, event, sample, machine, event->comm.pid, + event->comm.tid); } static int process_namespaces_event(struct perf_tool *tool,
@@ -2180,37 +2198,11 @@ static int process_namespaces_event(struct perf_tool *tool, struct perf_sample *sample, struct machine *machine) { - struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); -int ret =-1 ; - - thread =machine__findnew_thread (machine, event->namespaces.pid, - event->namespaces.tid); -if (thread ==NULL ) { -pr_debug ("problem processing NAMESPACES event, skipping it.\n" ); -return -1 ; - } - if (perf_event__process_namespaces(tool, event, sample, machine) < 0) - goto out; + return -1; -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time =0 ; - sample->tid = event->namespaces.tid; - sample->pid = event->namespaces.pid; - } -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_NAMESPACES , stdout); -perf_event__fprintf (event, stdout); - } - ret =0 ; -out: -thread__put (thread); -return ret; + return print_event(tool, event, sample, machine, event->namespaces.pid, + event->namespaces.tid); } static int process_cgroup_event(struct perf_tool *tool,
@@ -2218,34 +2210,11 @@ static int process_cgroup_event(struct perf_tool *tool, struct perf_sample *sample, struct machine *machine) { - struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); -int ret =-1 ; - - thread =machine__findnew_thread (machine, sample->pid, sample->tid); -if (thread ==NULL ) { -pr_debug ("problem processing CGROUP event, skipping it.\n" ); -return -1 ; - } - if (perf_event__process_cgroup(tool, event, sample, machine) < 0) - goto out; + return -1; -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time =0 ; - } -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_CGROUP , stdout); -perf_event__fprintf (event, stdout); - } - ret =0 ; -out: -thread__put (thread); -return ret; + return print_event(tool, event, sample, machine, sample->pid, + sample->tid); } static int process_fork_event(struct perf_tool *tool,
@@ -2253,69 +2222,24 @@ static int process_fork_event(struct perf_tool *tool, struct perf_sample *sample, struct machine *machine) { - struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); - if (perf_event__process_fork(tool, event, sample, machine) < 0) return -1; - thread =machine__findnew_thread (machine, event->fork.pid, event->fork.tid); -if (thread ==NULL ) { -pr_debug ("problem processing FORK event, skipping it.\n" ); -return -1 ; - } - -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time = event->fork.time; - sample->tid = event->fork.tid; - sample->pid = event->fork.pid; - } -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_FORK , stdout); -perf_event__fprintf (event, stdout); - } -thread__put (thread); - -return 0 ; + return print_event_with_time(tool, event, sample, machine, + event->fork.pid, event->fork.tid, + event->fork.time); }static int process_exit_event (struct perf_tool *tool,union perf_event *event,struct perf_sample *sample,struct machine *machine) { -int err =0 ; -struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); - - thread =machine__findnew_thread (machine, event->fork.pid, event->fork.tid); -if (thread ==NULL ) { -pr_debug ("problem processing EXIT event, skipping it.\n" ); +/* Print before 'exit' deletes anything */ +if (print_event_with_time (tool, event, sample, machine, event->fork.pid, + event->fork.tid, event->fork.time))return -1 ; - } - -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time =0 ; - sample->tid = event->fork.tid; - sample->pid = event->fork.pid; - } -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_EXIT , stdout); -perf_event__fprintf (event, stdout); - } -if (perf_event__process_exit (tool, event, sample, machine) <0 ) - err =-1 ; - -thread__put (thread); -return err; +return perf_event__process_exit (tool, event, sample, machine); } static int process_mmap_event(struct perf_tool *tool,
@@ -2323,33 +2247,11 @@ static int process_mmap_event(struct perf_tool *tool, struct perf_sample *sample, struct machine *machine) { - struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); - if (perf_event__process_mmap(tool, event, sample, machine) < 0) return -1; - thread =machine__findnew_thread (machine, event->mmap.pid, event->mmap.tid); -if (thread ==NULL ) { -pr_debug ("problem processing MMAP event, skipping it.\n" ); -return -1 ; - } - -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time =0 ; - sample->tid = event->mmap.tid; - sample->pid = event->mmap.pid; - } -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_MMAP , stdout); -perf_event__fprintf (event, stdout); - } -thread__put (thread); -return 0 ; + return print_event(tool, event, sample, machine, event->mmap.pid, + event->mmap.tid); } static int process_mmap2_event(struct perf_tool *tool,
@@ -2357,33 +2259,11 @@ static int process_mmap2_event(struct perf_tool *tool, struct perf_sample *sample, struct machine *machine) { - struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); - if (perf_event__process_mmap2(tool, event, sample, machine) < 0) return -1; - thread =machine__findnew_thread (machine, event->mmap2.pid, event->mmap2.tid); -if (thread ==NULL ) { -pr_debug ("problem processing MMAP2 event, skipping it.\n" ); -return -1 ; - } - -if (!evsel->core.attr.sample_id_all) { - sample->cpu =0 ; - sample->time =0 ; - sample->tid = event->mmap2.tid; - sample->pid = event->mmap2.pid; - } -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_MMAP2 , stdout); -perf_event__fprintf (event, stdout); - } -thread__put (thread); -return 0 ; + return print_event(tool, event, sample, machine, event->mmap2.pid, + event->mmap2.tid); } static int process_switch_event(struct perf_tool *tool,
@@ -2391,10 +2271,7 @@ static int process_switch_event(struct perf_tool *tool,struct perf_sample *sample, struct machine *machine) { - struct thread *thread; struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id);if (perf_event__process_switch (tool, event, sample, machine) <0 )return -1 ;
@@ -2405,20 +2282,8 @@ static int process_switch_event(struct perf_tool *tool,if (!script->show_switch_events)return 0 ; - thread =machine__findnew_thread (machine, sample->pid, - sample->tid); -if (thread ==NULL ) { -pr_debug ("problem processing SWITCH event, skipping it.\n" ); -return -1 ; - } - -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_SWITCH , stdout); -perf_event__fprintf (event, stdout); - } -thread__put (thread); -return 0 ; +return print_event (tool, event, sample, machine, sample->pid, + sample->tid); } static int
@@ -2427,23 +2292,8 @@ process_lost_event(struct perf_tool *tool,struct perf_sample *sample, struct machine *machine) { - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); -struct thread *thread; - - thread =machine__findnew_thread (machine, sample->pid, - sample->tid); -if (thread ==NULL ) -return -1 ; - -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, -PERF_RECORD_LOST , stdout); -perf_event__fprintf (event, stdout); - } -thread__put (thread); -return 0 ; + return print_event(tool, event, sample, machine, sample->pid, + sample->tid); } static int
@@ -2462,33 +2312,11 @@ process_bpf_events(struct perf_tool *tool __maybe_unused, struct perf_sample *sample, struct machine *machine) { - struct thread *thread; - struct perf_script *script = container_of(tool, struct perf_script, tool); -struct perf_session *session = script->session; -struct evsel *evsel =perf_evlist__id2evsel (session->evlist, sample->id); - if (machine__process_ksymbol(machine, event, sample) < 0) return -1; -if (!evsel->core.attr.sample_id_all) { -perf_event__fprintf (event, stdout); -return 0 ; - } - - thread =machine__findnew_thread (machine, sample->pid, sample->tid); -if (thread ==NULL ) { -pr_debug ("problem processing MMAP event, skipping it.\n" ); -return -1 ; - } - -if (!filter_cpu (sample)) { -perf_sample__fprintf_start (sample, thread, evsel, - event->header.type, stdout); -perf_event__fprintf (event, stdout); - } - -thread__put (thread); -return 0 ; + return print_event(tool, event, sample, machine, sample->pid, + sample->tid); }static void sig_handler (int sig__maybe_unused )
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 19/26 perf bench: Add event synthesis benchmark
In reply to Arnaldo Carvalho de MeloFrom: Ian Rogers <[email protected]>
Event synthesis may occur at the start or end (tail) of a perf command. In system-wide mode it can scan every process in /proc, which may add seconds of latency before event recording. Add a new benchmark that times how long event synthesis takes with and without data synthesis.
An example execution looks like:
$ perf bench internals synthesize # Running 'internals/synthesize' benchmark: Average synthesis took: 168.253800 usec Average data synthesis took: 208.104700 usec
--- tools/perf/Documentation/perf-bench.txt | 8 ++ tools/perf/bench/Build | 2 +- tools/perf/bench/bench.h | 2 +- tools/perf/bench/synthesize.c | 101 ++++++++++++++++++++++++ tools/perf/builtin-bench.c | 6 ++ 5 files changed, 117 insertions(+), 2 deletions(-) create mode 100644 tools/perf/bench/synthesize.c
tools/perf/Documentation/perf-bench.txt +8 -0
@@ -61,6 +61,9 @@ SUBSYSTEM 'epoll':: Eventpoll (epoll) stressing benchmarks. +'internals':: + Benchmark internal perf functionality. + 'all':: All benchmark subsystems.
@@ -214,6 +217,11 @@ Suite for evaluating concurrent epoll_wait calls. *ctl*:: Suite for evaluating multiple epoll_ctl calls. +SUITES FOR 'internals' +~~~~~~~~~~~~~~~~~~~~~~ +*synthesize*:: +Suite for evaluating perf's event synthesis performance. + SEE ALSO -------- linkperf:perf[1]
tools/perf/bench/Build +1 -1
@@ -6,9 +6,9 @@ perf-y += futex-wake.o perf-y += futex-wake-parallel.o perf-y += futex-requeue.o perf-y += futex-lock-pi.o - perf-y += epoll-wait.o perf-y += epoll-ctl.o +perf-y += synthesize.o perf-$(CONFIG_X86_64) += mem-memcpy-x86-64-lib.o perf-$(CONFIG_X86_64) += mem-memcpy-x86-64-asm.o
tools/perf/bench/bench.h +1 -1
@@ -41,9 +41,9 @@ int bench_futex_wake_parallel(int argc, const char **argv);int bench_futex_requeue (int argc,const char **argv);/* pi futexes */ int bench_futex_lock_pi (int argc,const char **argv); -int bench_epoll_wait (int argc,const char **argv);int bench_epoll_ctl (int argc,const char **argv); +int bench_synthesize (int argc,const char **argv);#define BENCH_FORMAT_DEFAULT_STR "default"#define BENCH_FORMAT_DEFAULT 0
tools/perf/bench/synthesize.c new file +101 -0
101 lines not on this page · raw message
tools/perf/builtin-bench.c +6 -0
@@ -76,6 +76,11 @@ static struct bench epoll_benchmarks[] = { };#endif // HAVE_EVENTFD +static struct bench internals_benchmarks[] = { + {"synthesize" ,"Benchmark perf event synthesis" , bench_synthesize }, + {NULL ,NULL ,NULL } +}; +struct collection {const char *name;const char *summary;
@@ -92,6 +97,7 @@ static struct collection collections[] = { #ifdef HAVE_EVENTFD {"epoll", "Epoll stressing benchmarks", epoll_benchmarks }, #endif + { "internals", "Perf-internals benchmarks", internals_benchmarks }, { "all", "All benchmarks", NULL }, { NULL, NULL, NULL } };
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 20/26 tools api fs: Make xxx__mountpoint() more scalable
In reply to Arnaldo Carvalho de MeloFrom: Stephane Eranian <[email protected]>
The xxx_mountpoint() interface provided by fs.c finds mount points for common pseudo filesystems. The first time xxx_mountpoint() is invoked, it scans the mount table (/proc/mounts) looking for a match. If found, it is cached. The price to scan /proc/mounts is paid once if the mount is found.
When the mount point is not found, subsequent calls to xxx_mountpoint() scan /proc/mounts over and over again. There is no caching.
This causes a scaling issue in perf record with hugeltbfs__mountpoint(). The function is called for each process found in synthesize__mmap_events(). If the machine has thousands of processes and if the /proc/mounts has many entries this could cause major overhead in perf record. We have observed multi-second slowdowns on some configurations.
As an example on a laptop:
Before:
$ sudo umount /dev/hugepages $ strace -e trace=openat -o /tmp/tt perf record -a ls $ fgrep mounts /tmp/tt 285
After:
$ sudo umount /dev/hugepages $ strace -e trace=openat -o /tmp/tt perf record -a ls $ fgrep mounts /tmp/tt 1
One could argue that the non-caching in case the moint point is not found is intentional. That way subsequent calls may discover a moint point if the sysadmin mounts the filesystem. But the same argument could be made against caching the mount point. It could be unmounted causing errors. It all depends on the intent of the interface. This patch assumes it is expected to scan /proc/mounts once. The patch documents the caching behavior in the fs.h header file.
An alternative would be to just fix perf record. But it would solve the problem with hugetlbs__mountpoint() but there could be similar issues (possibly down the line) with other xxx_mountpoint() calls in perf or other tools.
--- tools/lib/api/fs/fs.c | 17 +++++++++++++++++ tools/lib/api/fs/fs.h | 12 ++++++++++++ 2 files changed, 29 insertions(+)
tools/lib/api/fs/fs.c +17 -0
@@ -90,6 +90,7 @@ struct fs {const char *const *mounts;char path[PATH_MAX ];bool found; +bool checked;long magic; };
@@ -111,31 +112,37 @@ static struct fs fs__entries[] = { .name = "sysfs", .mounts = sysfs__fs_known_mountpoints, .magic = SYSFS_MAGIC, + .checked = false, }, [FS__PROCFS] = { .name = "proc", .mounts = procfs__known_mountpoints, .magic = PROC_SUPER_MAGIC, + .checked = false, }, [FS__DEBUGFS] = { .name = "debugfs", .mounts = debugfs__known_mountpoints, .magic = DEBUGFS_MAGIC, + .checked = false, }, [FS__TRACEFS] = { .name = "tracefs", .mounts = tracefs__known_mountpoints, .magic = TRACEFS_MAGIC, + .checked = false, }, [FS__HUGETLBFS] = { .name = "hugetlbfs", .mounts = hugetlbfs__known_mountpoints, .magic = HUGETLBFS_MAGIC, + .checked = false, }, [FS__BPF_FS] = { .name = "bpf", .mounts = bpf_fs__known_mountpoints, .magic = BPF_FS_MAGIC, + .checked = false, }, };
@@ -158,6 +165,7 @@ static bool fs__read_mounts(struct fs *fs) }fclose (fp); + fs->checked = true;return fs->found = found; }
@@ -220,6 +228,7 @@ static bool fs__env_override(struct fs *fs)return false; fs->found = true; + fs->checked = true;strncpy (fs->path, override_path,sizeof (fs->path) -1 ); fs->path[sizeof (fs->path) -1 ] ='\0' ;return true;
@@ -246,6 +255,14 @@ static const char *fs__mountpoint(int idx)if (fs->found)return (const char *)fs->path; +/* the mount point was already checked for the mount point +* but and did not exist, so return NULL to avoid scanning again. +* This makes the found and not found paths cost equivalent +* in case of multiple calls. +*/ +if (fs->checked) +return NULL ; +return fs__get_mountpoint (fs); }
tools/lib/api/fs/fs.h +12 -0
@@ -18,6 +18,18 @@ const char *name##__mount(void); \ bool name##__configured(void); \ +/* +* The xxxx__mountpoint() entry points find the first match mount point for each +* filesystems listed below, where xxxx is the filesystem type. +* +* The interface is as follows: +* +* - If a mount point is found on first call, it is cached and used for all +* subsequent calls. +* +* - If a mount point is not found, NULL is returned on first call and all +* subsequent calls. +*/ FS(sysfs) FS(procfs) FS(debugfs)
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 21/26 perf synthetic-events: save 4kb from 2 stack frames
In reply to Arnaldo Carvalho de MeloFrom: Ian Rogers <[email protected]>
Reuse an existing char buffer to avoid two PATH_MAX sized char buffers.
Reduces stack frame sizes by 4kb.
perf_event__synthesize_mmap_events before 'sub $0x45b8,%rsp' after 'sub $0x35b8,%rsp'.
perf_event__get_comm_ids before 'sub $0x2028,%rsp' after 'sub $0x1028,%rsp'.
The performance impact of this change is negligible.
--- tools/perf/util/synthetic-events.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-)
tools/perf/util/synthetic-events.c +10 -12
@@ -71,7 +71,6 @@ int perf_tool__process_synth_event(struct perf_tool *tool,static int perf_event__get_comm_ids (pid_t pid,char *comm,size_t len,pid_t *tgid,pid_t *ppid) { -char filename[PATH_MAX ];char bf[4096 ];int fd;size_t size =0 ;
@@ -81,11 +80,11 @@ static int perf_event__get_comm_ids(pid_t pid, char *comm, size_t len, *tgid =-1 ; *ppid =-1 ; -snprintf (filename,sizeof (filename),"/proc/%d/status" , pid); +snprintf (bf,sizeof (bf),"/proc/%d/status" , pid); - fd =open (filename,O_RDONLY ); + fd =open (bf,O_RDONLY );if (fd <0 ) { -pr_debug ("couldn't open %s\n" , filename); +pr_debug ("couldn't open %s\n" , bf);return -1 ; }
@@ -281,9 +280,9 @@ int perf_event__synthesize_mmap_events(struct perf_tool *tool, struct machine *machine, bool mmap_data) { - char filename[PATH_MAX];FILE *fp;unsigned long long t; +char bf[BUFSIZ ];bool truncation = false;unsigned long long timeout = proc_map_timeout *1000000ULL ;int rc =0 ;
@@ -293,15 +292,15 @@ int perf_event__synthesize_mmap_events(struct perf_tool *tool,if (machine__is_default_guest (machine))return 0 ; -snprintf (filename,sizeof (filename),"%s/proc/%d/task/%d/maps" , - machine->root_dir, pid, pid); +snprintf (bf,sizeof (bf),"%s/proc/%d/task/%d/maps" , + machine->root_dir, pid, pid); - fp =fopen (filename,"r" ); + fp =fopen (bf,"r" );if (fp ==NULL ) {/* * We raced with a task exiting - just return: */ -pr_debug ("couldn't open %s\n" , filename); +pr_debug ("couldn't open %s\n" , bf);return -1 ; }
@@ -309,7 +308,6 @@ int perf_event__synthesize_mmap_events(struct perf_tool *tool, t =rdclock ();while (1 ) { -char bf[BUFSIZ ];char prot[5 ];char execname[PATH_MAX ];char anonstr[] ="//anon" ;
@@ -321,10 +319,10 @@ int perf_event__synthesize_mmap_events(struct perf_tool *tool,break ;if ((rdclock () - t) > timeout) { -pr_warning ("Reading %s time out. " +pr_warning ("Reading %s/proc/%d/task/%d/maps time out. " "You may want to increase " "the time limit by --proc-map-timeout\n" , - filename); + machine->root_dir, pid, pid); truncation = true; goto out; }
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 22/26 perf stat: Fix no metric header if --per-socket and --metric-only set
In reply to Arnaldo Carvalho de MeloFrom: Jin Yao <[email protected]>
We received a report that was no metric header displayed if --per-socket and --metric-only were both set.
It's hard for script to parse the perf-stat output. This patch fixes this issue.
Before:
root@kbl-ppc:~# perf stat -a -M CPI --metric-only --per-socket ^C Performance counter stats for 'system wide':
S0 8 2.6
2.215270071 seconds time elapsed
root@kbl-ppc:~# perf stat -a -M CPI --metric-only --per-socket -I1000
# time socket cpus
1.000411692 S0 8 2.2
2.001547952 S0 8 3.4
3.002446511 S0 8 3.4
4.003346157 S0 8 4.0
5.004245736 S0 8 0.3After:
root@kbl-ppc:~# perf stat -a -M CPI --metric-only --per-socket ^C Performance counter stats for 'system wide':
CPI
S0 8 2.1
1.813579830 seconds time elapsed
root@kbl-ppc:~# perf stat -a -M CPI --metric-only --per-socket -I1000
# time socket cpus CPI
1.000415122 S0 8 3.2
2.001630051 S0 8 2.9
3.002612278 S0 8 4.3
4.003523594 S0 8 3.0
5.004504256 S0 8 3.7--- tools/perf/util/stat-shadow.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
tools/perf/util/stat-shadow.c +5 -2
@@ -803,8 +803,11 @@ static void generic_metric(struct perf_stat_config *config, out->force_header ? (metric_name ? metric_name : name) : "", 0); } - }else -print_metric (config, ctxp,NULL ,NULL ,"" ,0 ); + }else { +print_metric (config, ctxp,NULL ,NULL , + out->force_header ? + (metric_name ? metric_name : name) :"" ,0 ); + }for (i =1 ; i < pctx.num_ids; i++)zfree (&pctx.ids[i].name);
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 23/26 perf expr: Add expr_ prefix for parse_ctx and parse_id
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Adding expr_ prefix for parse_ctx and parse_id, to straighten out the expr* namespace.
There's no functional change.
--- tools/perf/tests/expr.c | 4 ++-- tools/perf/util/expr.c | 10 +++++----- tools/perf/util/expr.h | 12 ++++++------ tools/perf/util/expr.y | 6 +++--- tools/perf/util/stat-shadow.c | 2 +- 5 files changed, 17 insertions(+), 17 deletions(-)
tools/perf/tests/expr.c +2 -2
@@ -6,7 +6,7 @@#include <string.h> #include <linux/zalloc.h> -static int test (struct parse_ctx *ctx,const char *e,double val2) +static int test (struct expr_parse_ctx *ctx,const char *e,double val2) {double val;
@@ -22,7 +22,7 @@ int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused)const char **other;double val;int i, ret; -struct parse_ctx ctx; +struct expr_parse_ctx ctx;int num_other;expr__ctx_init (&ctx);
tools/perf/util/expr.c +5 -5
@@ -11,7 +11,7 @@ extern int expr_debug;#endif /* Caller must make sure id is allocated */ -void expr__add_id (struct parse_ctx *ctx,const char *name,double val) +void expr__add_id (struct expr_parse_ctx *ctx,const char *name,double val) {int idx;
@@ -21,13 +21,13 @@ void expr__add_id(struct parse_ctx *ctx, const char *name, double val) ctx->ids[idx].val = val; } -void expr__ctx_init (struct parse_ctx *ctx) +void expr__ctx_init (struct expr_parse_ctx *ctx) { ctx->num_ids =0 ; }static int -__expr__parse (double *val,struct parse_ctx *ctx,const char *expr, +__expr__parse (double *val,struct expr_parse_ctx *ctx,const char *expr,int start) {YY_BUFFER_STATE buffer;
@@ -52,7 +52,7 @@ __expr__parse(double *val, struct parse_ctx *ctx, const char *expr,return ret; } -int expr__parse (double *final_val,struct parse_ctx *ctx,const char *expr) +int expr__parse (double *final_val,struct expr_parse_ctx *ctx,const char *expr) {return __expr__parse (final_val, ctx, expr,EXPR_PARSE ) ?-1 :0 ; }
@@ -75,7 +75,7 @@ int expr__find_other(const char *expr, const char *one, const char ***other, int *num_other) {int err, i =0 , j =0 ; -struct parse_ctx ctx; +struct expr_parse_ctx ctx;expr__ctx_init (&ctx); err =__expr__parse (NULL , &ctx, expr,EXPR_OTHER );
tools/perf/util/expr.h +6 -6
@@ -5,19 +5,19 @@#define EXPR_MAX_OTHER 20#define MAX_PARSE_ID EXPR_MAX_OTHER -struct parse_id { +struct expr_parse_id {const char *name;double val; }; -struct parse_ctx { +struct expr_parse_ctx {int num_ids; -struct parse_id ids[MAX_PARSE_ID ]; +struct expr_parse_id ids[MAX_PARSE_ID ]; }; -void expr__ctx_init (struct parse_ctx *ctx); -void expr__add_id (struct parse_ctx *ctx,const char *id,double val); -int expr__parse (double *final_val,struct parse_ctx *ctx,const char *expr); +void expr__ctx_init (struct expr_parse_ctx *ctx); +void expr__add_id (struct expr_parse_ctx *ctx,const char *id,double val); +int expr__parse (double *final_val,struct expr_parse_ctx *ctx,const char *expr);int expr__find_other (const char *expr,const char *one,const char ***other,int *num_other);
tools/perf/util/expr.y +3 -3
@@ -15,7 +15,7 @@ %define api.pure full %parse-param { double *final_val } -%parse-param { struct parse_ctx *ctx } +%parse-param { struct expr_parse_ctx *ctx } %parse-param {void *scanner} %lex-param {void* scanner}
@@ -39,14 +39,14 @@ %{ static void expr_error(double *final_val __maybe_unused, - struct parse_ctx *ctx __maybe_unused, + struct expr_parse_ctx *ctx __maybe_unused, void *scanner, const char *s) { pr_debug("%s\n", s); } -static int lookup_id(struct parse_ctx *ctx, char *id, double *val) +static int lookup_id(struct expr_parse_ctx *ctx, char *id, double *val) { int i;
tools/perf/util/stat-shadow.c +1 -1
@@ -729,7 +729,7 @@ static void generic_metric(struct perf_stat_config *config, struct runtime_stat *st) {print_metric_t print_metric = out->print_metric; -struct parse_ctx pctx; +struct expr_parse_ctx pctx;double ratio, scale;int i;void *ctxp = out->ctx;
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 24/26 perf expr: Add expr_scanner_ctx object
In reply to Arnaldo Carvalho de MeloFrom: Jiri Olsa <[email protected]>
Add the expr_scanner_ctx object to hold user data for the expr scanner. Currently it holds only start_token, Kajol Jain will use it to hold 24x7 runtime param.
--- tools/perf/util/expr.c | 6 ++++-- tools/perf/util/expr.h | 4 ++++ tools/perf/util/expr.l | 10 +++++----- 3 files changed, 13 insertions(+), 7 deletions(-)
tools/perf/util/expr.c +4 -2
@@ -3,7 +3,6 @@#include <assert.h> #include "expr.h" #include "expr-bison.h" -#define YY_EXTRA_TYPE int#include "expr-flex.h" #ifdef PARSER_DEBUG
@@ -30,11 +29,14 @@ static int __expr__parse(double *val, struct expr_parse_ctx *ctx, const char *expr, int start) { +struct expr_scanner_ctx scanner_ctx = { + .start_token = start, + };YY_BUFFER_STATE buffer;void *scanner;int ret; - ret =expr_lex_init_extra (start, &scanner); + ret =expr_lex_init_extra (&scanner_ctx, &scanner);if (ret)return ret;
tools/perf/util/expr.h +4 -0
@@ -15,6 +15,10 @@ struct expr_parse_ctx {struct expr_parse_id ids[MAX_PARSE_ID ]; }; +struct expr_scanner_ctx { +int start_token; +}; +void expr__ctx_init (struct expr_parse_ctx *ctx);void expr__add_id (struct expr_parse_ctx *ctx,const char *id,double val);int expr__parse (double *final_val,struct expr_parse_ctx *ctx,const char *expr);
tools/perf/util/expr.l +5 -5
@@ -76,13 +76,13 @@ sym [0-9a-zA-Z_\.:@]+ symbol {spec}*{sym}*{spec}*{sym}* %% - { - int start_token; + struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner); - start_token = expr_get_extra(yyscanner); + { + int start_token = sctx->start_token; - if (start_token) { - expr_set_extra(NULL, yyscanner); + if (sctx->start_token) { + sctx->start_token = 0; return start_token; } }
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 25/26 perf metrictroup: Split the metricgroup__add_metric function
In reply to Arnaldo Carvalho de MeloFrom: Kajol Jain <[email protected]>
This patch refactors metricgroup__add_metric function where some part of it move to function metricgroup__add_metric_param. No logic change.
--- tools/perf/util/metricgroup.c | 60 ++++++++++++++++++++--------------- 1 file changed, 35 insertions(+), 25 deletions(-)
tools/perf/util/metricgroup.c +35 -25
@@ -485,6 +485,39 @@ static bool metricgroup__has_constraint(struct pmu_event *pe)return false; } +static int __metricgroup__add_metric (struct strbuf *events, +struct list_head *group_list,struct pmu_event *pe) +{ + +const char **ids; +int idnum; +struct egroup *eg; + +if (expr__find_other (pe->metric_expr,NULL , &ids, &idnum) <0 ) +return -EINVAL ; + +if (events->len >0 ) +strbuf_addf (events,"," ); + +if (metricgroup__has_constraint (pe)) +metricgroup__add_metric_non_group (events, ids, idnum); +else +metricgroup__add_metric_weak_group (events, ids, idnum); + + eg =malloc (sizeof (*eg)); +if (!eg) +return -ENOMEM ; + + eg->ids = ids; + eg->idnum = idnum; + eg->metric_name = pe->metric_name; + eg->metric_expr = pe->metric_expr; + eg->metric_unit = pe->unit; +list_add_tail (&eg->nd, group_list); + +return 0 ; +} +static int metricgroup__add_metric (const char *metric,struct strbuf *events,struct list_head *group_list) {
@@ -504,35 +537,12 @@ static int metricgroup__add_metric(const char *metric, struct strbuf *events,continue ;if (match_metric (pe->metric_group, metric) ||match_metric (pe->metric_name, metric)) { -const char **ids; -int idnum; -struct egroup *eg;pr_debug ("metric expr %s for %s\n" , pe->metric_expr, pe->metric_name); -if (expr__find_other (pe->metric_expr, -NULL , &ids, &idnum) <0 ) -continue ; -if (events->len >0 ) -strbuf_addf (events,"," ); - -if (metricgroup__has_constraint (pe)) -metricgroup__add_metric_non_group (events, ids, idnum); -else -metricgroup__add_metric_weak_group (events, ids, idnum); - - eg =malloc (sizeof (struct egroup )); -if (!eg) { - ret = -ENOMEM ; + ret =__metricgroup__add_metric (events, group_list, pe); +if (ret == -ENOMEM )break ; - } - eg->ids = ids; - eg->idnum = idnum; - eg->metric_name = pe->metric_name; - eg->metric_expr = pe->metric_expr; - eg->metric_unit = pe->unit; -list_add_tail (&eg->nd, group_list); - ret =0 ; } }return ret;
Signature · 1 line
2.21.1
Arnaldo Carvalho de Melo
maintainerPATCH 26/26 perf script: Add flamegraph.py script
In reply to Arnaldo Carvalho de MeloFrom: Andreas Gerstmayr <[email protected]>
This script works in tandem with d3-flame-graph to generate flame graphs from perf. It supports two output formats: JSON and HTML (the default). The HTML format will look for a standalone d3-flame-graph template file in /usr/share/d3-flame-graph/d3-flamegraph-base.html and fill in the collected stacks.
Usage:
perf record -a -g -F 99 sleep 60
perf script report flamegraph
Combined:
perf script flamegraph -a -F 99 sleep 60
Committer testing:
Tested both with "PYTHON=python3" and with the default, that uses python2-devel:
Complete set of instructions:
$ mkdir /tmp/build/perf $ make PYTHON=python3 -C tools/perf O=/tmp/build/perf install-bin $ export PATH=~/bin:$PATH $ perf record -a -g -F 99 sleep 60 $ perf script report flamegraph
Now go and open the generated flamegraph.html file in a browser.
At first this required building with PYTHON=python3, but after I reported this Andreas was kind enough to send a patch making it work with both python and python3.
--- .../perf/scripts/python/bin/flamegraph-record | 2 + .../perf/scripts/python/bin/flamegraph-report | 3 + tools/perf/scripts/python/flamegraph.py | 124 ++++++++++++++++++ 3 files changed, 129 insertions(+) create mode 100755 tools/perf/scripts/python/bin/flamegraph-record create mode 100755 tools/perf/scripts/python/bin/flamegraph-report create mode 100755 tools/perf/scripts/python/flamegraph.py
tools/perf/scripts/python/bin/flamegraph-record new file +2 -0
@@ -0,0 +1,2 @@ +#!/usr/bin/sh +perf record -g "$@"
tools/perf/scripts/python/bin/flamegraph-report new file +3 -0
@@ -0,0 +1,3 @@ +#!/usr/bin/sh +# description: create flame graphs +perf script -s "$PERF_EXEC_PATH"/scripts/python/flamegraph.py -- "$@"
tools/perf/scripts/python/flamegraph.py new file +124 -0
124 lines not on this page · raw message
Signature · 1 line
2.21.1