- 05 Jul, 2012 4 commits
-
-
Arnaldo Carvalho de Melo authored
Merge tag 'libtraceevent-core-for-acme' of git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf into perf/core libtraceevent/core improvements * Remaining backport of trace-cmd's libparseevent Signed-off-by:
Namhyung Kim <namhyung@kernel.org> Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Jiri Olsa authored
It'll be convenient in upcoming patch to access hw event symbols strings via enum perf_hw_id indexes. In order not to duplicate the data, creating two separate arrays: event_symbols_hw for enum perf_hw_id events event_symbols_sw for enum perf_sw_ids events Changing the current event list code to follow the change. Signed-off-by:
Jiri Olsa <jolsa@redhat.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1341352848-11833-7-git-send-email-jolsa@redhat.comSigned-off-by:
-
Jiri Olsa authored
Spliting PE_VALUE_SYM token to PE_VALUE_SYM_HW and PE_VALUE_SYM_SW tokens to separate hardware and software symbols. This will be useful in upcomming patch where we want to be able to parse out only hardware events. Signed-off-by:
-
Jiri Olsa authored
The flex generator prints out each input character that is ignored by lex rules. Since the alias processing, we can have '\n' characters on input. We need to assign empty rule to it, so it's not printed out. Signed-off-by:
-
- 04 Jul, 2012 17 commits
-
-
Jiri Olsa authored
Use ARRAY_SIZE instead of defining the sizes separately for each test arrays. Signed-off-by:
-
Namhyung Kim authored
The if branch is completely unnecessary since 'realloc' can handle NULL pointers for the first parameter. This patch is just an adoption of Ulrich Drepper's recent patch on perf tools. Signed-off-by:
Namhyung Kim <namhyung.kim@lge.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ulrich Drepper <drepper@gmail.com> Link: http://lkml.kernel.org/r/1335230984-7613-1-git-send-email-namhyung.kim@lge.comSigned-off-by:
Steven Rostedt <rostedt@goodmis.org> Signed-off-by:
-
Peter Huewe authored
In the current code we assign vsize=8 and then fall through to the default and assign vsize=1. -> probably the break is missing here, otherwise we can remove the case. Signed-off-by:
Peter Huewe <peterhuewe@gmx.de> Signed-off-by:
-
Namhyung Kim authored
The arg_to_str() can fail so we should handle that case properly. Signed-off-by:
-
Namhyung Kim authored
The malloc can fail so the return value should be checked. For now, just use malloc_or_die(). Signed-off-by:
-
Namhyung Kim authored
Update and add missing argument descriptions and fix some typo on function comments. Signed-off-by:
-
Namhyung Kim authored
When memory allocation for the field name is failed, do not goto event_failed since we added the event already. Signed-off-by:
-
Namhyung Kim authored
It seems PEVENT_FUNC_ARG_STRING missed passing the allocated string to the args array. Fix it. Signed-off-by:
-
Namhyung Kim authored
The realloc can fail so that we should handle it properly. Signed-off-by:
-
Namhyung Kim authored
There were some places didn't check return value of the strdup and had unneeded/duplicated checks. Fix it. Signed-off-by:
-
Namhyung Kim authored
The __read_token() function has some duplicated code to handle internal buffer overflow. Factor them out to new extend_token(). According to the man pages of realloc/free(3), they can handle NULL pointer input so that it can be ended up to compact the code. Also handle error path correctly. Signed-off-by:
-
Namhyung Kim authored
The printk_cmp function should use printk_map instead of func_map. Also rename the variables for consistency. Signed-off-by:
-
Wolfgang Mauerer authored
On 32 bit systems, a conversion of the trace_printk format string "%lu" -> "%llu" is intended (similar for %lx etc.) when a trace was taken on a machine with 64 bit long integers. However, the current code computes the bogus transformation "%lu" -> "%u". Fix this. Besides that, the transformation is only required on systems that don't use 64 bits for long integers natively. Signed-off-by:
Wolfgang Mauerer <wolfgang.mauerer@siemens.com> Signed-off-by:
-
Steven Rostedt authored
When %pM is used, the arg value must be a 6 byte character that will be printed as a 6 byte MAC address. But the code does a break over the main code which updates the current processing arg to point to the next arg. If there are other print arguments after a %pM, they will be off by one. The next arg will still be processing the %pM arg. Reported-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
Steven Rostedt authored
The RT kernel added a migrate disable counter in all events. Add support to show this in the latency format. Signed-off-by:
-
Steven Rostedt authored
The arg notation of '*' in bprintks is not handled by the parser. Implement it so that they show up properly in the output and do not kill the tracer from reporting events. Reported-by:
-
Steven Rostedt authored
As a pointer can be converted into a function name, let the filters work with the function name as well as with the pointer number. If the comparison expects a string, then convert numbers into functions, but only when the number is the same size as a long. Signed-off-by:
-
- 02 Jul, 2012 1 commit
-
-
Hitoshi Mitake authored
As Namhyung Kim pointed, there are confused namings and descriptions of words "cycle" and "clock" in mem-memset.c and mem-memcpy.c. With the option "-c" (or "--clock", now renamed as "--cycle"), mem subsystem measures cost of memset() and memcpy() with cpu-cycles event. But current mem subsystem source code contains lots of confused variable namings and descriptions with "clock" (e.g. the variable use_clock). This is a very bad style because there is another software event named "cpu-clock". This patch replaces wrong usage of "clock" to "cycle". v2: modified Documentation/perf-bench.txt for the descriptions of --cycle option Signed-off-by:
Hitoshi Mitake <h.mitake@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1341236777-18457-1-git-send-email-h.mitake@gmail.comSigned-off-by:
-
- 29 Jun, 2012 13 commits
-
-
Jiri Olsa authored
Adding round_up and round_down macros. They will be used in upcoming patches. Signed-off-by:
-
Jiri Olsa authored
Adding '.note' section name to be check when looking for notes section. The '.note' name is used by kernel VDSO. Signed-off-by:
-
Robert Richter authored
The internal pmu list was never used. With each perf_pmu__find() call the pmu structure was created new by parsing sysfs. Beside this it caused memory leaks. We now keep all pmus by adding them to the list. Also, pmu_lookup() should return pmus that do not expose the format specifier in sysfs. We need a valid internal pmu list in a later patch to iterate over all pmus that exist in the system. Signed-off-by:
Robert Richter <robert.richter@amd.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1339706321-8802-3-git-send-email-jolsa@redhat.comSigned-off-by:
-
Jiri Olsa authored
Adding hw breakpoint events hook in the perf_evsel__name function, to display event names properly all over the perf tools. Updated hw breakpoints events tests. Signed-off-by:
-
Jiri Olsa authored
Fixing the hw breakpoint's type modifier parsing to allow all possible combinations of 'rwx' characters. Adding automated tests to the parsing test suite. Reported-by:
Jovi Zhang <bookjovi@gmail.com> Original-patch-by:
-
Namhyung Kim authored
Because the only caller of the alloc_event() (pevent_parse_event) checks return value properly, it can be changed to use plain malloc. Signed-off-by:
-
Namhyung Kim authored
Since the __print_hex() function is used in print fmt now, add corresponding parser routines. This makes the output of perf script on the kvm_emulate_insn event not to fail any more. before: kvm_emulate_insn: [FAILED TO PARSE] rip=3238197797 ... after: kvm_emulate_insn: 0:c102fa25:89 10 (prot32) Signed-off-by: -
Namhyung Kim authored
Use local variable 'field' to reduce typing. It is needed by later patch not to exceed 80 column. Signed-off-by:
-
Namhyung Kim authored
When libtraceevent parses format fields, it assumes that array of 1 byte is string but it's not always true. The kvm_emulate_insn contains 15 u8 array of insn that contains (binary) instructions. Thus when it's printed, it'll have broken output like below: kvm_emulate_insn: [FAILED TO PARSE] rip=3238197797 csbase=0 len=2 \ insn=<89>P^]<B4>& flags=5 failed=0 With this patch: kvm_emulate_insn: [FAILED TO PARSE] rip=3238197797 csbase=0 len=2 \ insn=ARRAY[89, 10, 5d, c3, 8d, b4, 26, 00, 00, 00, 00, 55, 89, e5, 3e] flags=5 failed=0 Suggested-by:
-
Namhyung Kim authored
Ingo reported that libtraceevent doesn't clean out dependency (.d) files and it can cause a build error when the libgcc package upgraded: comet:~/tip/tools/perf> make -j SUBDIR ../lib/traceevent/ make[1]: *** No rule to make target `/usr/lib/gcc/x86_64-redhat-linux/4.7.0/include/stddef.h', needed by `event-parse.o'. Stop. make: *** [../lib/traceevent//libtraceevent.a] Error 2 So this patch makes the .d files depends on the source and header files also, so that it can be re-generated as needed. NOTE: This code is copied from the GNU make manual page (4.14 Generating Prerequisites Automatically). Reported-by:Ingo Molnar <mingo@kernel.org> Signed-off-by:
-
Namhyung Kim authored
As we use a macro trick to sync each error codes with its description string, teach [ce]tags to process them properly. This patch modifies the libtraceevent's Makefile not a kernel one. Suggested-by:
-
Konstantin Stepanyuk authored
Dependency files were not cleaned up. Add missing space to fix the issue. Signed-off-by:
Konstantin Stepanyuk <konstantin.stepanyuk@gmail.com> Acked-by:
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core perf/core improvements and fixes: * Improve 'perf bench' docs, by Namhyung Kim * Fix build when O= is not used, from David Ahern * Fix cross compilation build, from Namhyung Kim * Fix pipe mode when callchains are used, from David Ahern * Follow .gnu_debuglink section to find separate symbols, from Pierre-Loup A. Griffais * Fix 'perf test' raw events entries, from Jiri Olsa * Use the events description in the perf.data file, not the sysfs ones. Signed-off-by:
-
- 27 Jun, 2012 5 commits
-
-
David Ahern authored
References to OUTPUT should not be followed by a '/'. When a build output directory is not specified for this case you get: gcc -o builtin-annotate.o -c ... -I/util ... which is wrong. Signed-off-by:
David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tim Chen <tim.c.chen@linux.intel.com> Link: http://lkml.kernel.org/r/1339463612-30937-1-git-send-email-dsahern@gmail.comSigned-off-by:
-
David Ahern authored
The pipeline: perf record -a -g -o - sleep 5 |perf inject -v -b | perf report -g -i - generates the warning: Selected -g but no callchain data. Did you call 'perf record' without -g? The problem is that the header data is not written to the pipe, so the sample_type has not been available when perf_report__setup_sample_type is called. For pipe mode, record dumps the sample type as part of the synthesized events stream -- perf_event__synthesize_attrs(). Handle this be detecting pipe mode and not doing early sanity checks on sample_type. Signed-off-by:
Tim Chen <tim.c.chen@linux.intel.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tim Chen <tim.c.chen@linux.intel.com> Link: http://lkml.kernel.org/r/1339444121-26236-1-git-send-email-dsahern@gmail.comSigned-off-by:
-
Namhyung Kim authored
The current perf-bench documentation has a couple of typos and even lacks entire description of mem subsystem. Fix it. Reported-by:
Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1340172486-17805-1-git-send-email-namhyung@kernel.orgSigned-off-by:
-
Pierre-Loup A. Griffais authored
The .gnu_debuglink section is specified to contain the filename of the debug info file, as well as a CRC that can be used to validate it. This doesn't currently use the checksum and relies on the usual build-id matching for validation. This provides more context: http://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.htmlSigned-off-by:
Pierre-Loup A. Griffais <pgriffais@nvidia.com> Reported-by:
Mike Sartain <mikesart@valvesoftware.com> Tested-by:
-
Arnaldo Carvalho de Melo authored
The pevent thing is per perf.data file, so I made it stop being static and become a perf_session member, so tools processing perf.data files use perf_session and _there_ we read the trace events description into session->pevent and then change everywhere to stop using that single global pevent variable and use the per session one. Note that it _doesn't_ fall backs to trace__event_id, as we're not interested at all in what is present in the /sys/kernel/debug/tracing/events in the workstation doing the analysis, just in what is in the perf.data file. This patch also introduces perf_session__set_tracepoints_handlers that is the perf perf.data/session way to associate handlers to tracepoint events by resolving their IDs using the events descriptions stored in a perf.data file. Make 'perf sched' use it. Reported-by:
Dmitry Antipov <dmitry.antipov@linaro.org> Tested-by:
-
