- 14 Oct, 2013 14 commits
-
-
Ingo Molnar authored
Namhyung Kim noticed that the volatile-register-var feature check is superfluous: > The gcc manpage says this warning is enabled by -Wall, and we add -Wall > to CFLAGS before doing feature checks. So all gcc versions that support > -Wvolatile-register-var enables it by default without this check and > older gcc versions will always fail the feature check. Remove it - this will further speed up feature checks. Reported-by:
Namhyung Kim <namhyung@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Signed-off-by:
Ingo Molnar <mingo@kernel.org> Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Ingo Molnar authored
Ulrich Drepper and Namhyung Kim reported that the libelf logic in config/Makefile is duplicated in part. Remove the duplication, and also remove the now unused FLAGS_LIBELF variable. Reported-by:
Ulrich Drepper <drepper@gmail.com> Reported-by:
-
Ingo Molnar authored
Namhyung Kim reported these duplicate DPACKAGE definitions: test-libbfd: $(BUILD) -DPACKAGE='perf' -DPACKAGE=perf -lbfd -ldl Fix all affected places and use Namhyung's suggestion that the definition should look like a normal C string: -DPACKAGE='"perf"'. Reported-by: -
Ingo Molnar authored
Arnaldo reported that 'make DEBUG=1' does not work anymore. The reason is that 'Makefile' only passes it through to 'Makefile.perf' via the environment, but 'Makefile.perf' checks that it's a command line option: ifeq ("$(origin DEBUG)", "command line") PERF_DEBUG = $(DEBUG) endif So pass it through properly, and also clean up DEBUG parameter handling while at it and fix a couple of annoyances: - DEBUG=0 used to be interpreted as 'debugging on'. Turn it into 'debugging off' instead. - Same was the case for 'DEBUG=' - turn that into debug-off as well. - Pass in just a clean, sanitized 'DEBUG' value and get rid of the intermediate, unnecessary PERF_DEBUG variable. Reported-by: -
Ingo Molnar authored
Arnaldo reported that non-existent build directories were not recognized properly. The reason is readlink failure causing 'O' to become empty. Solve it by passing through the 'O' variable unmodified if readlink fails. Reported-by:
Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/20131009150023.GA10167@gmail.comSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Add a function to find a symbol using an ip that might be on a different map. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Namhyung Kim <namhyung@gmail.com> 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/1381747424-3557-2-git-send-email-adrian.hunter@intel.comSigned-off-by:
Adrian Hunter <adrian.hunter@intel.com> Signed-off-by:
-
Adrian Hunter authored
The objdump tool fails to annotate module symbols when looking at kcore. Workaround this by extracting object code from kcore and putting it in a temporary file for objdump to use instead. The temporary file is created to look like kcore but contains only the function being disassembled. Signed-off-by:
-
Adrian Hunter authored
Before using kcore we need to check that modules are in memory at the same addresses that they were when data was recorded. This is done because, while we could remap symbols to different addresses, the object code linkages would still be different which would provide an erroneous view of the object code. Signed-off-by:
-
Jiri Olsa authored
We fail build with NO_DEMANGLE with missing -lbfd externals error. The reason is that we now use bfd code in srcline object: perf tools: Implement addr2line directly using libbfd So we need to check/add -lbfd always now. Signed-off-by:Jiri Olsa <jolsa@redhat.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
-
Felipe Pena authored
Fix for a memory leak on test_file() function in dso-data.c. Signed-off-by:
Felipe Pena <felipensp@gmail.com> Acked-by:
-
David Ahern authored
trace-event-parse.c:parse_proc_kallsyms() Old GCC (4.4.2) does not see through the code flow of get_srcline() and gets confused about the status of 'file' and 'line': CC /tmp/build/perf/util/srcline.o cc1: warnings being treated as errors util/srcline.c: In function ¿get_srcline¿: util/srcline.c:226: error: ¿file¿ may be used uninitialized in this function util/srcline.c:227: error: ¿line¿ may be used uninitialized in this function make[1]: *** [/tmp/build/perf/util/srcline.o] Error 1 make: *** [install] Error 2 make: Leaving directory `/home/acme/git/linux/tools/perf' [acme@fedora12 linux]$ Help out GCC by initializing 'file' and 'line'. Signed-off-by:David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@gmail.com> Link: http://lkml.kernel.org/n/tip-h8k7h49z3cndqgjdftkmm9f8@git.kernel.orgSigned-off-by:
-
David Ahern authored
When enabled dumps a summary of all syscalls by task with the usual statistics -- min, max, average and relative stddev. For example, make - 26341 : 3344 [ 17.4% ] 0.000 ms read : 52 0.000 4.802 0.644 30.08 write : 20 0.004 0.036 0.010 21.72 open : 24 0.003 0.046 0.014 23.68 close : 64 0.002 0.055 0.008 22.53 stat : 2714 0.002 0.222 0.004 4.47 fstat : 18 0.001 0.041 0.006 46.26 mmap : 30 0.003 0.009 0.006 5.71 mprotect : 8 0.006 0.039 0.016 32.16 munmap : 12 0.007 0.077 0.020 38.25 brk : 48 0.002 0.014 0.004 10.18 rt_sigaction : 18 0.002 0.002 0.002 2.11 rt_sigprocmask : 60 0.002 0.128 0.010 32.88 access : 2 0.006 0.006 0.006 0.00 pipe : 12 0.004 0.048 0.013 35.98 vfork : 34 0.448 0.980 0.692 3.04 execve : 20 0.000 0.387 0.046 56.66 wait4 : 34 0.017 9923.287 593.221 68.45 fcntl : 8 0.001 0.041 0.013 48.79 getdents : 48 0.002 0.079 0.013 19.62 getcwd : 2 0.005 0.005 0.005 0.00 chdir : 2 0.070 0.070 0.070 0.00 getrlimit : 2 0.045 0.045 0.045 0.00 arch_prctl : 2 0.002 0.002 0.002 0.00 setrlimit : 2 0.002 0.002 0.002 0.00 openat : 94 0.003 0.005 0.003 2.11 Signed-off-by: -
David Ahern authored
Similar to other findnew based methods if the requested object is not found, add it to the list. v2: followed format of other findnew methods per acme's request Signed-off-by:
-
Ramkumar Ramachandra authored
Currently, execution of 'perf trace' reports the following cryptic message to the user: $ perf trace Couldn't read the raw_syscalls tracepoints information! Typically this happens because the user does not have permissions to read the debugfs filesystem. Also handle the case when the kernel was not compiled with debugfs support or when it isn't mounted. Now, the tool prints detailed error messages: $ perf trace Error: Unable to find debugfs Hint: Was your kernel was compiled with debugfs support? Hint: Is the debugfs filesystem mounted? Hint: Try 'sudo mount -t debugfs nodev /sys/kernel/debug' $ perf trace Error: No permissions to read /sys/kernel/debug//tracing/events/raw_syscalls Hint: Try 'sudo mount -o remount,mode=755 /sys/kernel/debug/' Signed-off-by:
Ramkumar Ramachandra <artagnon@gmail.com> Cc: David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Link: http://lkml.kernel.org/r/1380863851-14460-1-git-send-email-artagnon@gmail.com [ Added ready to use commands to fix the issues as extra hints, use the current debugfs mount point when reporting permission error, use strerror_r instead of the deprecated sys_errlist, as reported by David Ahern ] Signed-off-by:
-
- 11 Oct, 2013 26 commits
-
-
Ramkumar Ramachandra authored
While at it, update the synopsis to show both forms. Signed-off-by:
-
Ingo Molnar authored
'make install' used to show all the install lines, which is way too verbose to be really informative to the user. Implement summary output instead: comet:~/tip/tools/perf> make install BUILD: Doing 'make -j12' parallel build SUBDIR Documentation INSTALL Documentation-man INSTALL binaries INSTALL libexec INSTALL perf-archive INSTALL perl-scripts INSTALL python-scripts INSTALL bash_completion-script INSTALL tests 'make install V=1' will still show the old, detailed output. Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: -
Ingo Molnar authored
Before: CC util/pmu.o CC util/parse-events.o PERF_VERSION = 3.12.rc4.g1b30c CC util/parse-events-flex.o GEN perf-archive After: CC util/pmu.o CC util/parse-events.o PERF_VERSION = 3.12.rc4.g1b30c CC util/parse-events-flex.o GEN perf-archive Signed-off-by:
-
Ingo Molnar authored
The various build lines from libtraceevent and perf mix up during a parallel build and produce unaligned output like: CC builtin-buildid-list.o CC builtin-buildid-cache.o CC builtin-list.o CC FPIC trace-seq.o CC builtin-record.o CC FPIC parse-filter.o CC builtin-report.o CC builtin-stat.o CC FPIC parse-utils.o CC FPIC kbuffer-parse.o CC builtin-timechart.o CC builtin-top.o CC builtin-script.o BUILD STATIC LIB libtraceevent.a CC builtin-probe.o CC builtin-kmem.o CC builtin-lock.o To solve this, harmonize all the build message alignments to be similar to the kernel's kbuild output: prefixed by two spaces and 11-char wide. After the patch the output looks pretty tidy, even if output lines get mixed up: CC builtin-annotate.o FLAGS: * new build flags or cross compiler CC builtin-bench.o AR liblk.a CC bench/sched-messaging.o CC FPIC event-parse.o CC bench/sched-pipe.o CC FPIC trace-seq.o CC bench/mem-memcpy.o CC bench/mem-memset.o CC FPIC parse-filter.o CC builtin-diff.o CC builtin-evlist.o CC builtin-help.o Signed-off-by: -
Ingo Molnar authored
'make clean' used to show all the rm lines, which isn't really informative in any way and spams the console. Implement summary output: comet:~/tip/tools/perf> make clean CLEAN libtraceevent CLEAN liblk CLEAN config CLEAN core-objs CLEAN core-progs CLEAN core-gen CLEAN Documentation CLEAN python 'make clean V=1' will still show the old, detailed output. Signed-off-by:
-
Ingo Molnar authored
Fix the duplicate util/util printout Arnaldo reported: $ make V=1 O=/tmp/build/perf -C tools/perf/ util/srcline.o ... # Redirected target util/srcline.o => /tmp/build/perf/util/util/srcline.o Reported-by: -
Arnaldo Carvalho de Melo authored
[root@zoo linux]# trace -e ioctl | grep -v "cmd: 0x" | head -10 0.386 ( 0.001 ms): trace/1602 ioctl(fd: 1<pipe:[127057]>, cmd: TCGETS, arg: 0x7fff59fcb4d0 ) = -1 ENOTTY Inappropriate ioctl for device 1459.368 ( 0.002 ms): inotify_reader/10352 ioctl(fd: 18<anon_inode:inotify>, cmd: FIONREAD, arg: 0x7fb835228bcc ) = 0 1463.586 ( 0.002 ms): inotify_reader/10352 ioctl(fd: 18<anon_inode:inotify>, cmd: FIONREAD, arg: 0x7fb835228bcc ) = 0 1463.611 ( 0.002 ms): inotify_reader/10352 ioctl(fd: 18<anon_inode:inotify>, cmd: FIONREAD, arg: 0x7fb835228bcc ) = 0 3740.526 ( 0.002 ms): awk/1612 ioctl(fd: 1<pipe:[128265]>, cmd: TCGETS, arg: 0x7fff4d166b90 ) = -1 ENOTTY Inappropriate ioctl for device 3740.704 ( 0.001 ms): awk/1612 ioctl(fd: 3</proc/meminfo>, cmd: TCGETS, arg: 0x7fff4d1669a0 ) = -1 ENOTTY Inappropriate ioctl for device 3742.550 ( 0.002 ms): ps/1614 ioctl(fd: 1<pipe:[128266]>, cmd: TIOCGWINSZ, arg: 0x7fff591762b0 ) = -1 ENOTTY Inappropriate ioctl for device 3742.555 ( 0.003 ms): ps/1614 ioctl(fd: 2<socket:[19550]>, cmd: TIOCGWINSZ, arg: 0x7fff591762b0 ) = -1 ENOTTY Inappropriate ioctl for device 3742.558 ( 0.002 ms): ps/1614 ioctl(cmd: TIOCGWINSZ, arg: 0x7fff591762b0 ) = -1 ENOTTY Inappropriate ioctl for device 3742.572 ( 0.002 ms): ps/1614 ioctl(fd: 1<pipe:[128266]>, cmd: TCGETS, arg: 0x7fff59176220 ) = -1 ENOTTY Inappropriate ioctl for device [root@zoo linux]# Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-afajwap3mr60dfl4qpdl1pxn@git.kernel.orgSigned-off-by: -
Arnaldo Carvalho de Melo authored
Right now when an index passed to that method has no string associated it'll print the index as a decimal number, prepare it so that we can use it to print it in hex as well, for ioctls, for instance. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-nsvy06sqj64qvnkmzvwxsx2v@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
So that the index passed doesn't have to start at zero, being decremented from an offset specified when declaring the strarray before being used as the real array index. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-k1ce6uqyt4qar9edrj3mevod@git.kernel.orgSigned-off-by:
-
Adrian Hunter authored
Make a separate function to parse /proc/modules so that it can be reused. Signed-off-by:
-
David Ahern authored
Allows commands to leverage intlist infrastructure for opaque structures. For example an upcoming perf-trace change will use this as a means of tracking syscalls statistics by task. Signed-off-by:
-
David Ahern authored
Use the new machine method that loops over threads to dump summary data. Signed-off-by:
-
David Ahern authored
Loop over all threads within a machine - including threads moved to the dead threads list -- and invoked a function. This allows commands to run some specific function on each thread (eg., dump statistics) yet hides how the threads are maintained within the machine. Signed-off-by:
-
David Ahern authored
The record option is a convience alias to include the -e raw_syscalls:* argument to perf-record. All other options are passed to perf-record's handler. Resulting data file can be analyzed by perf-trace -i. Signed-off-by:
-
David Ahern authored
Task comm's are getting lost when processing events from a file. The problem is that the trace struct used by the live processing has its host machine and the perf-session used for file based processing has its host machine. Fix by having both references point to the same machine. Before: 0.030 ( 0.001 ms): :27743/27743 brk( ... 0.057 ( 0.004 ms): :27743/27743 mmap(len: 4096, prot: READ|WRITE, flags: ... 0.075 ( 0.006 ms): :27743/27743 access(filename: 0x7f3809fbce00, mode: R ... 0.091 ( 0.005 ms): :27743/27743 open(filename: 0x7f3809fba14c, flags: CLOEXEC ... ... After: 0.030 ( 0.001 ms): make/27743 brk( ... 0.057 ( 0.004 ms): make/27743 mmap(len: 4096, prot: READ|WRITE, flags: ... 0.075 ( 0.006 ms): make/27743 access(filename: 0x7f3809fbce00, mode: R ... 0.091 ( 0.005 ms): make/27743 open(filename: 0x7f3809fba14c, flags: CLOEXEC ... ... Signed-off-by: -
David Ahern authored
Ingo pointed out that the task-clock counter should have the units explicitly stated since it is not a counter. Before: perf stat -a -- sleep 1 Performance counter stats for 'sleep 1': 16186.874834 task-clock # 16.154 CPUs utilized ... After: perf stat -a -- sleep 1 Performance counter stats for 'system wide': 16146.402138 task-clock (msec) # 16.125 CPUs utilized ... Reported-by: -
David Ahern authored
The "perf stat" command can do system wide counters or one or more cpus. For these options do not require a workload to be specified. v2: use perf_target__none per Namhyung's comment. Signed-off-by:
-
David Ahern authored
The "perf stat" tool displays the command run in its summary output which is misleading when using a cpu list or system wide collection. Before: perf stat -a -- sleep 1 Performance counter stats for 'sleep 1': 16152.670249 task-clock # 16.132 CPUs utilized 417 context-switches # 0.002 M/sec 7 cpu-migrations # 0.030 K/sec ... After: perf stat -a -- sleep 1 Performance counter stats for 'system wide': 16206.931120 task-clock # 16.144 CPUs utilized 395 context-switches # 0.002 M/sec 5 cpu-migrations # 0.030 K/sec ... or perf stat -C1 -- sleep 1 Performance counter stats for 'CPU(s) 1': 1001.669257 task-clock # 1.000 CPUs utilized 4,264 context-switches # 0.004 M/sec 3 cpu-migrations # 0.003 K/sec ... Signed-off-by: -
Jiri Olsa authored
The perf_evlist__mmap_read used 'union perf_event' as a placeholder for event crossing the mmap boundary. This is ok for sample shorter than ~PATH_MAX. However we could grow up to the maximum sample size which is 16 bits max. I hit this overflow issue when using 'perf top -G dwarf' which produces sample with the size around 8192 bytes. We could configure any valid sample size here using: '-G dwarf,size'. Using array with sample max size instead for the event placeholder. Also adding another safe check for the dynamic size of the user stack. TODO: The 'struct perf_mmap' is quite big now, maybe we could use some lazy allocation for event_copy size. Signed-off-by:
-
Petr Holasek authored
Patch adds more subtle handling of -C and -N parameters in parse_{cpu,node}_setup_list() functions when there isn't enough NUMA nodes or CPUs present. Instead of assertion and terminating benchmark, partial test is skipped with error message and perf will continue to the next one. Fixed problem can be easily reproduced on machine with only one NUMA node: # Running numa/mem benchmark... # Running main, "perf bench numa mem -a" ... # Running RAM-bw-remote, "perf bench numa mem -p 1 -t 1 -P 1024 -C 0 -M 1 -s perf: bench/numa.c:622: parse_setup_node_list: Assertion `!(bind_node_0 < 0 || bind_node_0 >= g->p.nr_nodes)' failed. Aborted Signed-off-by:Petr Holasek <pholasek@redhat.com> Acked-by:
Petr Benas <pbenas@redhat.com> Signed-off-by:
-
Ramkumar Ramachandra authored
The default output file produced by the 'perf timechart' tool is called output.svg, add it to .gitignore. Signed-off-by:
-
Ramkumar Ramachandra authored
When only the instructions event is requested: $ perf stat -e instructions git s M builtin-stat.c Performance counter stats for 'git s': 917,453,420 instructions # 0.00 insns per cycle 0.213002926 seconds time elapsed The 0.00 insns per cycle comment in the output is totally bogus and misleading. It happens because update_shadow_stats() doesn't touch runtime_cycles_stats when only the instructions event is requested. So, omit printing the bogus data altogether. Signed-off-by: -
Ramkumar Ramachandra authored
When only the cycles event is requested: $ perf stat -e cycles dd if=/dev/zero of=/dev/null count=1000000 1000000+0 records in 1000000+0 records out 512000000 bytes (512 MB) copied, 0.26123 s, 2.0 GB/s Performance counter stats for 'dd if=/dev/zero of=/dev/null count=1000000': 911,626,453 cycles # 0.000 GHz 0.262113350 seconds time elapsed The 0.000 GHz comment in the output is totally bogus and misleading. It happens because update_shadow_stats() doesn't touch runtime_nsecs_stats; it is only written when a requested counter matches a SW_TASK_CLOCK. In our case, since we have only requested HW_CPU_CYCLES, runtime_nsecs_stats is unavailable. So, omit printing the comment altogether. Signed-off-by: -
Jiri Olsa authored
Moving start conditions to start of the flex file so it's clear what the INITIAL condition rules are. Plus adding default rule for INITIAL condition. This prevents default space to be printed for events like: $ ./perf stat -e "cycles " kill 2>/dev/null $ ^^^^^^^^ Signed-off-by: -
Jiri Olsa authored
If we build perf with NO_LIBPYTHON=1 NO_LIBPERL=1 the '-ldl' is not added to libs build fails if we have gtk2 code in, because it depends on it. Signed-off-by:
-
Adrian Hunter authored
To avoid buffer overruns. Signed-off-by:
-
