- 04 Jul, 2016 2 commits
-
-
Hemant Kumar authored
This patch serves the initial support to identify and list SDT events in binaries. When programs containing SDT markers are compiled, gcc with the help of assembler directives identifies them and places them in the section ".note.stapsdt". To find these markers from the binaries, one needs to traverse through this section and parse the relevant details like the name, type and location of the marker. Also, the original location could be skewed due to the effect of prelinking. If that is the case, the locations need to be adjusted. The functions in this patch open a given ELF, find out the SDT section, parse the relevant details, adjust the location (if necessary) and populate them in a list. A typical note entry in ".note.stapsdt" section is as follows : |--nhdr.n_namesz--| ------------------------------------ | nhdr | "stapsdt" | ----- |----------------------------------| | | <location> <base_address> | | | <semaphore> | nhdr.n_descsize | "provider_name" "note_name" | | | <args> | ----- |----------------------------------| | nhdr | "stapsdt" | |... The above shows an excerpt from the section ".note.stapsdt". 'nhdr' is a structure which has the note name size (n_namesz), note description size (n_desc_sz) and note type (n_type). So, in order to parse the note note info, we need nhdr to tell us where to start from. As can be seen from <sys/sdt.h>, the name of the SDT notes given is "stapsdt". But this is not the identifier of the note. After that, we go to description of the note to find out its location, the address of the ".stapsdt.base" section and the semaphore address. Then, we find the provider name and the SDT marker name and then follow the arguments. Signed-off-by:Hemant Kumar <hemant@linux.vnet.ibm.com> Reviewed-by:
Masami Hiramatsu <mhiramat@kernel.org> Acked-by:
Namhyung Kim <namhyung@kernel.org> Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com> Cc: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/146736022628.27797.1201368329092908163.stgit@devboxSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Arnaldo Carvalho de Melo authored
That appeared after 0.140, and will be used in the SDT code, so, to avoid bisection break on older systems, add a feature detection and provide a stub with a pr_debug() to keep it building. Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com> Cc: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Hemant Kumar <hemant@linux.vnet.ibm.com> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-80y0eldgweorqnwha9rvfxjr@git.kernel.orgSigned-off-by:
-
- 01 Jul, 2016 4 commits
-
-
Masami Hiramatsu authored
'perf probe --del' removes caches when '--cache' is given. Note that the delete pattern is not the same as for normal events. If you cached probes with event name, --del "eventname" works as expected. However, if you skipped it, the cached probes doesn't have actual event name. In that case --del "probe-desc" is required (wildcard is acceptable). For example a cache entry has the probe-desc "vfs_read $params", you can remove it with --del 'vfs_read*'. ----- # perf probe --cache --list /[kernel.kallsyms] (1466a0a250b5d0070c6d0f03c5fed30b237970a1): vfs_read $params /usr/lib64/libc-2.17.so (c31ffe7942bfd77b2fca8f9bd5709d387a86d3bc): getaddrinfo $params # perf probe --cache --del vfs_read\* Removed cached event: probe:vfs_read # perf probe --cache --list /[kernel.kallsyms] (1466a0a250b5d0070c6d0f03c5fed30b237970a1): /usr/lib64/libc-2.17.so (c31ffe7942bfd77b2fca8f9bd5709d387a86d3bc): getaddrinfo $params ----- Signed-off-by:
Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Tested-by:
-
Masami Hiramatsu authored
perf probe --list shows all cached probes when --cache is given. Each caches are shown with on which binary that probed. E.g.: ----- # perf probe --cache vfs_read \$params # perf probe --cache -x /lib64/libc-2.17.so getaddrinfo \$params # perf probe --cache --list [kernel.kallsyms] (1466a0a250b5d0070c6d0f03c5fed30b237970a1): vfs_read $params /usr/lib64/libc-2.17.so (c31ffe7942bfd77b2fca8f9bd5709d387a86d3bc): getaddrinfo $params ----- Note that $params requires debuginfo. Signed-off-by:
-
Masami Hiramatsu authored
Before analyzing debuginfo, try to find a corresponding entry from probe cache always. This does not depend on --cache, the --cache enables to store/update cache, but looking up the cache is always enabled. Signed-off-by:
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo-20160630' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo: New features: - Allow running 'perf test' entries in the same process, not forking to test each testcase, useful for debugging (Jiri Olsa) - Show number of samples in the stdio annotate header (Peter Zijlstra) Documentation changes: - Add documentation for perf.data on disk format (Andi Kleen) Build fixes: - Fix 'perf trace' build on old systems wrt missing SCHED_RESET_ON_FORK and eventfd.h (Arnaldo Carvalho de Melo) Infrastructure changes: - Utility function to fetch arch from evsel/evlist (Ravi Bangoria) Trivial changes: - Fix spelling mistake: "missmatch" -> "mismatch" in libbpf (Colin Ian King) Signed-off-by:
Ingo Molnar <mingo@kernel.org>
-
- 30 Jun, 2016 9 commits
-
-
Jiri Olsa authored
Display cpu map in standard list form. (perf report -D output on perf stat data). before: 0x590 [0x18]: PERF_RECORD_CPU_MAP nr: 4 cpus: 0, 1, 2, 3 after: 0x590 [0x18]: PERF_RECORD_CPU_MAP: 0-3 Adding automated testcase. Signed-off-by:
Jiri Olsa <jolsa@kernel.org> Tested-by:
-
Jiri Olsa authored
Adding -F/--dont-fork option to bypass forking for each test. It's useful for debugging test. Signed-off-by:
Nilay Vaish <nilayvaish@gmail.com> Cc: David Ahern <dsahern@gmail.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1467113345-12669-1-git-send-email-jolsa@kernel.orgSigned-off-by:
-
Jiri Olsa authored
I hit a bug when running test suite without forking each test (-F option): $ perf test -Fv ... 34: Test thread map : --- start --- FAILED tests/thread-map.c:24 wrong comm ---- end ---- Test thread map: FAILED! The reason was the process name wasn't 'perf' as expected by the test, because other tests set the name as well. Setting it explicitly now. Signed-off-by:
-
Jiri Olsa authored
I hit a bug when running test suite without forking each test (-F option): $ perf test -F dso 8: Test dso data read : Ok 9: Test dso data cache : FAILED! 10: Test dso data reopen : FAILED! The reason the session file limit is set just once for perf process so we need to reset it for each test, otherwise wrong limit is taken into account. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Old systems such as RHEL5 lack this file, and what we need is already under ifdefs, so just ditch this #include. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-dzbjfllw6znuoy37skwnwa4r@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
RHEL5 for instance doesn't have this one, help it. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-3adewnii78zi110eovfciopy@git.kernel.orgSigned-off-by:
-
Peter Zijlstra (Intel) authored
Staring at annotations of large functions is useless if there's only a few samples in them. Report the number of samples in the header to make this easier to determine. Committer note: The change amounts to: - Percent | Source code & Disassembly of perf-vdso.so for cycles:u ------------------------------------------------------------------ + Percent | Source code & Disassembly of perf-vdso.so for cycles:u (3278 samples) +-------------------------------------------------------------------------------- Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Jiri Olsa <jolsa@kernel.org> Link: http://lkml.kernel.org/r/20160630082955.GA30921@twins.programming.kicks-ass.net [ split from a larger patch ] Signed-off-by:
-
Peter Zijlstra (Intel) authored
No need to use strlen, etc to figure that out, just use the return from printf(), it will tell how wide the following line needs to be. Signed-off-by:
-
Ravi Bangoria authored
Add Utility function to fetch arch using evsel. (evsel->env->arch) Signed-off-by:
Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Anton Blanchard <anton@ozlabs.org> Cc: Daniel Axtens <dja@axtens.net> Cc: David Laight <David.Laight@ACULAB.COM> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com> Link: http://lkml.kernel.org/r/1467267262-4589-2-git-send-email-ravi.bangoria@linux.vnet.ibm.comSigned-off-by:
-
- 29 Jun, 2016 3 commits
-
-
Colin Ian King authored
Trivial fix to spelling mistake Signed-off-by:
Colin King <colin.king@canonical.com> Acked-by:
Wang Nan <wangnan0@huawei.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: He Kuang <hekuang@huawei.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1467116617-8318-1-git-send-email-colin.king@canonical.comSigned-off-by:
-
Andi Kleen authored
Add some documentation for the on disk format of perf.data. This is not documenting the actual perf events -- which are documented in perf_event.h -- but just the additional headers that perf record adds around them when writing the data to disk. Signed-off-by:
Andi Kleen <ak@linux.intel.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Link: http://lkml.kernel.org/r/1466800885-12974-1-git-send-email-andi@firstfloor.orgSigned-off-by:
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo-20160628' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo: New features: - Generate comm, fork and exit events when converting perf.data files to CTF (Wang Nan) Infrastructure changes: - Add libbabeltrace to build-test (Wang Nan) - 'perf record' prep work to support multiple evlists (Wang Nan) - Remove unused hist_entry__annotate function (Ravi Bangoria) - Add more toolchain triplets (Ravi Bangoria) - Update message for slang devel packages on Ubuntu (Neeraj Badlani) - Generalize handling of 'ret' instructions in the annotate TUI (Naveen N. Rao) - Use proper dso name for is_regular_file, fixing device file handling (Jiri Olsa) Build Fixes: - Add missing config.h include, fixing the build with libbabeltrace (Jiri Olsa) Signed-off-by:
-
- 28 Jun, 2016 13 commits
-
-
Wang Nan authored
If 'all' is selected, convert fork and exit events to output CTF stream. Signed-off-by:
-
Wang Nan authored
After this patch, 'perf data convert' convert comm events to output CTF stream. Result: # perf record -a sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.378 MB perf.data (73 samples) ] # perf data convert --to-ctf ./out.ctf [ perf data convert: Converted 'perf.data' into CTF data './out.ctf' ] [ perf data convert: Converted and wrote 0.003 MB (73 samples) ] # babeltrace --clock-seconds ./out.ctf/ [10627.402515791] (+?.?????????) cycles:ppp: { cpu_id = 0 }, { perf_ip = 0xFFFFFFFF81065AF4, perf_tid = 0, perf_pid = 0, perf_period = 1 } [10627.402518972] (+0.000003181) cycles:ppp: { cpu_id = 0 }, { perf_ip = 0xFFFFFFFF81065AF4, perf_tid = 0, perf_pid = 0, perf_period = 1 } ... // only sample event is converted # perf data convert --all --to-ctf ./out.ctf [ perf data convert: Converted 'perf.data' into CTF data './out.ctf' ] [ perf data convert: Converted and wrote 0.023 MB (73 samples, 384 non-samples) ] # babeltrace --clock-seconds ./out.ctf/ [ 0.000000000] (+?.?????????) perf_comm: { cpu_id = 0 }, { pid = 1, tid = 1, comm = "init" } [ 0.000000000] (+0.000000000) perf_comm: { cpu_id = 0 }, { pid = 2, tid = 2, comm = "kthreadd" } [ 0.000000000] (+0.000000000) perf_comm: { cpu_id = 0 }, { pid = 3, tid = 3, comm = "ksoftirqd/0" } ... // comm events are converted [10627.402515791] (+10627.402515791) cycles:ppp: { cpu_id = 0 }, { perf_ip = 0xFFFFFFFF81065AF4, perf_tid = 0, perf_pid = 0, perf_period = 1 } [10627.402518972] (+0.000003181) cycles:ppp: { cpu_id = 0 }, { perf_ip = 0xFFFFFFFF81065AF4, perf_tid = 0, perf_pid = 0, perf_period = 1 } ... // samples are also converted Signed-off-by: -
Wang Nan authored
If 'all' is selected, convert comm event to output CTF stream. setup_non_sample_events() is called if non_sample is selected. It creates a comm_class for comm event. Use macros to generate and process_comm_event and add_comm_event. These macros can be reused for other non-sample events. Signed-off-by:
-
Wang Nan authored
Following commits are going to allow 'perf data convert' to collect not only samples, but also non-sample events like comm and fork. In this patch we count non-sample events using c.non_sample_count, and prepare to print number of both type of events like: # ~/perf data convert --all --to-ctf ./out.ctf [ perf data convert: Converted 'perf.data' into CTF data './out.ctf' ] [ perf data convert: Converted and wrote 0.846 MB (6508 samples, 686 non-samples) ] Signed-off-by:
-
Wang Nan authored
If 'all' option is selected, 'perf data convert' should convert not only samples, but non-sample events such as comm and fork. Add this option in perf_data_convert_opts. Following commits will add cmdline option to select it. Signed-off-by:
-
Wang Nan authored
Following commits will add new option to 'perf data convert'. All options should be grouped into a structure and passed to low level converter (currently there's only one converter). Introduce data-convert.h and define 'struct perf_data_convert_opts' in it. Pass 'force' through opts. Signed-off-by:
-
Wang Nan authored
There are many value_set_##x helper for integer, but only for integer. This patch adds value_set_string() helper to help following commits create string fields. Signed-off-by:
-
Jiri Olsa authored
Marc reported use of uninitialized memory: > In commit "40356721 perf symbols: Do not read symbols/data from > device files" a check to uninitialzied memory was added. This leads to > the following valgrind output: > > ==24515== Syscall param stat(file_name) points to uninitialised byte(s) > ==24515== at 0x75B26D5: _xstat (in /lib/x86_64-linux-gnu/libc-2.22.so) > ==24515== by 0x4E548D: stat (stat.h:454) > ==24515== by 0x4E548D: is_regular_file (util.c:687) > ==24515== by 0x4A5BEE: dso__load (symbol.c:1435) > ==24515== by 0x4BB1AE: map__load (map.c:289) > ==24515== by 0x4BB1AE: map__find_symbol (map.c:333) > ==24515== by 0x4835B3: thread__find_addr_location (event.c:1300) > ==24515== by 0x4B5342: add_callchain_ip (machine.c:1652) > ==24515== by 0x4B5342: thread__resolve_callchain_sample (machine.c:1906) > ==24515== by 0x4B9E7D: thread__resolve_callchain (machine.c:1958) > ==24515== by 0x441B3E: process_event (builtin-script.c:795) > ==24515== by 0x441B3E: process_sample_event (builtin-script.c:920) > ==24515== by 0x4BEE29: perf_evlist__deliver_sample (session.c:1192) > ==24515== by 0x4BEE29: machines__deliver_event (session.c:1229) > ==24515== by 0x4BF770: perf_session__deliver_event (session.c:1286) > ==24515== by 0x4BF770: ordered_events__deliver_event (session.c:114) > ==24515== by 0x4C1D17: __ordered_events__flush (ordered-events.c:207) > ==24515== by 0x4C1D17: ordered_events__flush.part.3 (ordered-events.c:274) > ==24515== by 0x4BF44C: perf_session__process_user_event (session.c:1325) > ==24515== by 0x4BF44C: perf_session__process_event (session.c:1451) > ==24515== Address 0x807c6a0 is 0 bytes inside a block of size 4,096 alloc'd > ==24515== at 0x4C29C0F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so) > ==24515== by 0x4A5BCB: dso__load (symbol.c:1421) > ==24515== by 0x4BB1AE: map__load (map.c:289) > ==24515== by 0x4BB1AE: map__find_symbol (map.c:333) > ==24515== by 0x4835B3: thread__find_addr_location (event.c:1300) > ==24515== by 0x4B5342: add_callchain_ip (machine.c:1652) > ==24515== by 0x4B5342: thread__resolve_callchain_sample (machine.c:1906) > ==24515== by 0x4B9E7D: thread__resolve_callchain (machine.c:1958) > ==24515== by 0x441B3E: process_event (builtin-script.c:795) > ==24515== by 0x441B3E: process_sample_event (builtin-script.c:920) > ==24515== by 0x4BEE29: perf_evlist__deliver_sample (session.c:1192) > ==24515== by 0x4BEE29: machines__deliver_event (session.c:1229) > ==24515== by 0x4BF770: perf_session__deliver_event (session.c:1286) > ==24515== by 0x4BF770: ordered_events__deliver_event (session.c:114) > ==24515== by 0x4C1D17: __ordered_events__flush (ordered-events.c:207) > ==24515== by 0x4C1D17: ordered_events__flush.part.3 (ordered-events.c:274) > ==24515== by 0x4BF44C: perf_session__process_user_event (session.c:1325) > ==24515== by 0x4BF44C: perf_session__process_event (session.c:1451) > ==24515== by 0x4C0EAC: __perf_session__process_events (session.c:1804) > ==24515== by 0x4C0EAC: perf_session__process_events (session.c:1858) The reason was a typo that passed global 'name' variable as the is_regular_file argument instead dso->long_name. Reported-by:
Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by:
-
Wang Nan authored
Following commits introduce new evlists to record. This patch adjusts record__pick_pc() and introduces perf_evlist__pick_pc() to read control page from one specific evlist. record__pick_pc() will be improved to search control page from multiple evlists. Signed-off-by:
-
Wang Nan authored
Following commits introduce new evlists to record. This patch adjusts record__mmap_read_all() and record__mmap_read(): converting original record__mmap_read_all() to record__mmap_read_evlist(), read from one evlist; makes record__mmap_read() reading from specific evlist. record__mmap_read_all() will be improved to read from multiple evlists. Signed-off-by:
-
Wang Nan authored
Following commits introduce multiple evlists to record. This patch extracts perf_evlist__mmap_ex() processing to a new function, creates record__mmap() and record__mmap_evlist() to wrap perf_evlist__mmap_ex() and its error processing. They will be improvemented to create mmap for all evlists. Signed-off-by:
-
Jiri Olsa authored
Otherwise some compiler might scream: $ make LIBBABELTRACE_DIR=/opt/libbabeltrace/ LIBBABELTRACE=1 BUILD: Doing 'make -j4' parallel build CC util/data-convert-bt.o util/data-convert-bt.c: In function ‘convert__config’: util/data-convert-bt.c:1299:19: error: implicit declaration of function ‘perf_config_u64’ [-Werror=implicit-function-declaration] c->queue_size = perf_config_u64(var, value); ... Signed-off-by: -
Wang Nan authored
'make build-test' doesn't test LIBBABELTRACE=1. It misses a building failure caused by commit 41840d21 ("perf config: Move config declarations from util/cache.h to util/config.h"), breaks bisect. Add LIBBABELTRACE=1 to build-test. Signed-off-by:
-
- 27 Jun, 2016 9 commits
-
-
Ravi Bangoria authored
Add few more triplets based on Fedora and Ubuntu binutils (cross tools). Before applying patch on x86: ( Install binutils-powerpc64-linux-gnu.x86_64 ) $ perf report -i perf.data.powerpc --vmlinux vmlinux.powerpc \ --objdump powerpc64-linux-gnu-objdump After applying patch on x86: $ perf report -i perf.data.powerpc --vmlinux vmlinux.powerpc I.e. it will find the right objdump from the environment data recorded in the perf.data file + these triplets. Signed-off-by: -
Naveen N. Rao authored
Introduce helper to detect 'ret' instructions and use the same in the TUI. A helper is needed since some architectures such as powerpc have more than one return instruction. Signed-off-by:
Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Anton Blanchard <anton@ozlabs.org> Cc: Daniel Axtens <dja@axtens.net> Cc: Michael Ellerman <mpe@ellerman.id.au> Link: http://lkml.kernel.org/r/1466769240-12376-5-git-send-email-ravi.bangoria@linux.vnet.ibm.comSigned-off-by:
-
Neeraj Badlani authored
In case of missing library (libslang), give hint to install library (libslang2-dev), since libslang-dev is not provided by Ubuntu. Signed-off-by:
Neeraj Badlani <neerajbadlani@gmail.com> Link: http://lkml.kernel.org/r/1467035997-9100-1-git-send-email-neerajbadlani@gmail.com [ removed excessive 'or' usage ] Signed-off-by:
-
Ravi Bangoria authored
hist_entry__annotate looks part of API but I don't find any caller of this function. Removing it. Signed-off-by:
-
Peter Zijlstra authored
The whole rdmsr()/wrmsr() for lbr_from got a little unweildy with the sign extension quirk, provide a few simple wrappers to clean things up. Signed-off-by:
-
David Carrillo-Cisneros authored
Add quirk for context switch to save/restore the value of MSR_LAST_BRANCH_FROM_x when LBR is enabled and there is potential for kernel addresses to be in the lbr_from register. To test this patch, use a perf tool and kernel with the patch next in this series. That patch removes the work around that masked the hw bug: $ ./lbr_perf record --call-graph lbr -e cycles:k sleep 1 where lbr_perf is the patched perf tool, that allows to specify :k on lbr mode. The above command will trigger a #GPF : WARNING: CPU: 28 PID: 14096 at arch/x86/mm/extable.c:65 ex_handler_wrmsr_unsafe+0x70/0x80 unchecked MSR access error: WRMSR to 0x681 (tried to write 0x1fffffff81010794) ... Call Trace: [<ffffffff8167af49>] dump_stack+0x4d/0x63 [<ffffffff810b9b15>] __warn+0xe5/0x100 [<ffffffff810b9be9>] warn_slowpath_fmt+0x49/0x50 [<ffffffff810abb40>] ex_handler_wrmsr_unsafe+0x70/0x80 [<ffffffff810abc42>] fixup_exception+0x42/0x50 [<ffffffff81079d1a>] do_general_protection+0x8a/0x160 [<ffffffff81684ec2>] general_protection+0x22/0x30 [<ffffffff810101b9>] ? intel_pmu_lbr_sched_task+0xc9/0x380 [<ffffffff81009d7c>] intel_pmu_sched_task+0x3c/0x60 [<ffffffff81003a2b>] x86_pmu_sched_task+0x1b/0x20 [<ffffffff81192a5b>] perf_pmu_sched_task+0x6b/0xb0 [<ffffffff8119746d>] __perf_event_task_sched_in+0x7d/0x150 [<ffffffff810dd9dc>] finish_task_switch+0x15c/0x200 [<ffffffff8167f894>] __schedule+0x274/0x6cc [<ffffffff8167fdd9>] schedule+0x39/0x90 [<ffffffff81675398>] exit_to_usermode_loop+0x39/0x89 [<ffffffff810028ce>] prepare_exit_to_usermode+0x2e/0x30 [<ffffffff81683c1b>] retint_user+0x8/0x10 Signed-off-by:
David Carrillo-Cisneros <davidcc@google.com> Signed-off-by:
Stephane Eranian <eranian@google.com> Reviewed-by:
-
David Carrillo-Cisneros authored
Replace spaces by tabs in LBR_FROM_* constants to align with newly defined constant. Use BIT_ULL. Signed-off-by:
-
David Carrillo-Cisneros authored
Intel's SDM states that bits 61:62 in MSR_LAST_BRANCH_FROM_x are the TSX flags for formats with LBR_TSX flags (i.e. LBR_FORMAT_EIP_EFLAGS2). However, when the CPU has TSX support deactivated, bits 61:62 actually behave as follows: - For wrmsr(), bits 61:62 are considered part of the sign extension. - When capturing branches, the LBR hw will always clear bits 61:62. regardless of the sign extension. Therefore, if: 1) LBR has TSX format. 2) CPU has no TSX support enabled. ... then any value passed to wrmsr() must be sign extended to 63 bits and any value from rdmsr() must be converted to have a sign extension of 61 bits, ignoring the values at TSX flags. This bug was masked by the work-around to the Intel's CPU bug: BJ94. "LBR May Contain Incorrect Information When Using FREEZE_LBRS_ON_PMI" in Document Number: 324643-037US. The aforementioned work-around uses hw flags to filter out all kernel branches, limiting LBR callstack to user level execution only. Since user addresses are not sign extended, they do not trigger the wrmsr() bug in MSR_LAST_BRANCH_FROM_x when saved/restored at context switch. To verify the hw bug: $ perf record -b -e cycles sleep 1 $ rdmsr -p 0 0x680 0x1fffffffb0b9b0cc $ wrmsr -p 0 0x680 0x1fffffffb0b9b0cc write(): Input/output error The quirk for LBR_FROM_ MSRs is required before calls to wrmsrl() and after rdmsrl(). This patch introduces it for wrmsrl()'s done for testing LBR support. Future patch in series adds the quirk for context switch, that would be required if LBR callstack is to be enabled for ring 0. Signed-off-by: -
David Carrillo-Cisneros authored
The following commit: 338b522c ("perf/x86/intel: Protect LBR and extra_regs against KVM lying") added an additional test to LBR support detection that is performed after printing the LBR support statement to dmesg. Move the LBR support output after the very last test, to make sure we print the true status of LBR support. Signed-off-by:
-
