When multiple events were used in 'perf record', allow the user to
choose which one is wanted before showing the per event histograms.
Annotations will be performed on the chosen event.
Allow going back and forth from event to event quickly using just the
arrow keys and enter.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: William Cohen <wcohen@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
By creating an perf_evlist out of the attributes in the perf.data file
header, so that we can use evlists and evsels when reading recorded
sessions in addition to when we record sessions.
More work is needed to allow tools to allow the user to select which
events are wanted when browsing sessions, be it just one or a subset of
them, aggregated or showed at the same time but with different
indications on the UI to allow seeing workloads thru different views at
the same time.
But the overall goal/trend is to more uniformly use evsels and evlists.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The previous situation was to receive an fd from where to read the event
ID.
Spin off a routine for when we have the ID handy, not having to read it
from some fd.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To support multiple events we need to do these calcs per 'struct hists'
instance, and it turns out we already do that at:
__hists__add_entry
hists__inc_nr_entries
hists__calc_col_len
for all the unfiltered hist_entry instances we stash in the rb tree, so
trow away the dead code.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
TAB/UNTAB were not hotkeys, so didn't exit hists__browse back to
hists__tui_browse_tree, allowing just the first event to be browsed.
Reported-by: William Cohen <wcohen@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: William Cohen <wcohen@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
PyVarObject_HEAD_INIT is undefined in python 2.5, resulting
in a build crash:
util/python.c:81: attention : déclaration implicite de la fonction « «PyVarObject_HEAD_INIT» »
util/python.c:82: erreur: request for member «tp_name» in something not a structure or union
util/python.c:117: erreur: request for member «tp_name» in something not a structure or union
util/python.c:146: erreur: request for member «tp_name» in something not a structure or union
util/python.c:177: erreur: request for member «tp_name» in something not a structure or union
util/python.c:290: erreur: request for member «tp_name» in something not a structure or union
util/python.c:359: erreur: request for member «tp_name» in something not a structure or union
util/python.c:532: erreur: request for member «tp_name» in something not a structure or union
util/python.c:761: erreur: request for member «tp_name» in something not a structure or union
error: command 'gcc' failed with exit status 1
make: *** [python/perf.so] Erreur 1
We can fix that by defining PyVarObject_HEAD_INIT as a wrapper on
PyObject_HEAD_INIT, thanks to a trick found on biopython:
d4eaf57946
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
<ctype.h> is included first without _GNU_SOURCE, so it ends up
including <string.h> without declaring strndup(). And further
<string.h> declarations, even with _GNU_SOURCE defined, are
of course without effect.
Therefore:
util/strfilter.c: Dans la fonction «strfilter_node__new» :
util/strfilter.c:134: attention : déclaration implicite de la fonction « «strndup» »
util/strfilter.c:134: attention : incompatible implicit declaration of built-in function «strndup»
make: *** [util/strfilter.o] Erreur 1
Just don't include ctype.h as it doesn't appear to be necessary
anyway.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
We currently set the filters after we mmap the events, this is a
race that let undesired events record themselves in the buffer before
we had the time to set the filters.
So set the filters before they can be recorded. That also librarizes
the filters setting so that filtering can be done more easily
from other tools than perf record later.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Using ui__warning, that will, in --tui, show a window with the message,
waiting for the user to press Ok.
Also run exit_browser() to let newt do its final cleaning of the screen.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
By taking the ui__lock so that no other screen updates take place while
waiting for the user.
That was happening when handling an invalid --vmlinux parameter in 'perf
top --tui', with the screen refresh routine repainting the screen and
removing the warning window.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fixing a SEGV. An empty list could happen when not being able to resolve
symbols, for instance when --vmlinux invalid-file is used.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The ec5761e cset introduced the symfs feature with a bug for loading vmlinux
files that ended up causing this failure:
[root@emilia v2.6.38-rc5+]# strace -e trace=open perf top --vmlinux ./vmlinux 2>&1 | tail -3
open("/./vmlinux", O_RDONLY) = -1 ENOENT (No such file or directory)
./vmlinux with build id b9266bf40e98dadb5d43a2f3e95d3c5d4aff46dc not found, continuing without symbols
The ./vmlinux file can't be used
[root@emilia v2.6.38-rc5+]#
Remove the extra slash, just like is done in the DSO__ORIG_DSO handling in
dso__load() and other parts of the ec5761e cset.
Reported-by: Ingo Molnar <mingo@elte.hu>
Cc: David Ahern <daahern@cisco.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Currently numcpus is determined in pid_put_sample which is only
called on sched_switch/sched_wakeup sample processing.
On a machine with a lot cpus I often saw the last cpu missing.
Check for (max) numcpus on every event happening and in the
beginning. -> fixes the issue for me.
Signed-off-by: Thomas Renninger <trenn@suse.de>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: lenb@kernel.org
LKML-Reference: <1298842606-55712-6-git-send-email-trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This fix is needed for eye of gnome and firefox svg viewers.
Only Inkscape can handle the broken case.
Compare with the other svg_legenda_box declarations, looks
like a typo slipped in at this place.
Signed-off-by: Thomas Renninger <trenn@suse.de>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: lenb@kernel.org
LKML-Reference: <1298842606-55712-5-git-send-email-trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Needed because we were only showing the title in ui_browser__show,
not in ui_browser__run, and in the run loop we may be calling other
browsers that would then change the title, when we go back to the
previous browser, we need to redraw the title.
We could have done this as the Newt help line, with pop, etc, but I
don't think its worth, doing it explicitely, when needed (some browsers
may not use the title area at all) seems enough/more flexible.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The left key was exiting 'perf top --tui' when it really shouldn't, it
was too easy to leave the live annotation window and then press one too
many <- and get out of the tool altogether.
Do just like the report TUI does, ignore the left key for exit and also
ask the user when pressing ESC if that is really what is wanted.
Reported-by: Mike Galbraith <efault@gmx.de>
Suggested-by: Ingo Molnar <mingo@elte.hu>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In both --tui and --stdio, in 'annotate', 'top', 'report' when trying to
annotate a kernel symbol having just access to a kallsyms file, that
doesn't have the DWARF info needed for annotation.
Suggested-by: Ingo Molnar <mingo@elte.hu>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This is not a problem when we're not at the bottom of the active symbols
list, so was not noticed, but at the end of the screen it falls apart.
Fix it by adjusting the ui_browser indexes according to the new number
of entries in the rb_tree and by seeking from the start of the rb_tree
to find the new symbol at the top of the screen.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Now one has just to press the right key, 'a' or Enter on the main 'perf
top --tui' screen to live annotate the symbol under the cursor.
The annotate window starts centered on the hottest line (the one with
most samples so far) then TAB and shift+TAB can be used to go to the
prev/next hot line.
Pressing 'H' at any point will center again the screen on the hottest
line.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
While fixing an error propagating problem in f809b25 I added two
redundant checks.
I did that because I didn't expect the checks to be on the while and for
loop condition expression, where they are tested before we run the loop,
where the 'ret' variable is set.
So remove it from there and leave it just after it is actually set,
eliminating unneded tests.
Reported-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.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>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The kernel refuses mmapping an event with the inherit flag set for
something that is systemwide (cpu == -1), and the evsel layer got this
reversed at some point, fix it.
The symtom was that the --pid and --tid parameters for 'perf record' and
'perf top' returned with -EINVAL, like:
# /tmp/build-perf/perf record -v -fo/tmp/perf.data -p 1042
Warning: ... trying to fall back to cpu-clock-ticks
Fatal: failed to mmap with 22 (Invalid argument)
Reported-by: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
There are two hunks in this patch that stops probe processing as soon as one
error is found, breaking out of loops, the other fix an error propagation that
should return a negative error number but instead was returning the result of
"ret < 0", which is 1 and thus made several error checks fail because they test
agains < 0.
The problem could be triggered by asking for a variable that was optimized out,
fact that should stop the whole probe processing but instead was segfaulting
while installing broken probes:
[root@emilia ~]# probe perf_mmap:55 user_lock_limit
Failed to find the location of user_lock_limit at this address.
Perhaps, it has been optimized out.
Failed to find 'user_lock_limit' in this function.
Add new events:
probe:perf_mmap (on perf_mmap:55 with user_lock_limit)
probe:perf_mmap_1 (on perf_mmap:55 with user_lock_limit)
Segmentation fault (core dumped)
[root@emilia ~]# perf probe -l
probe:perf_mmap (on perf_mmap:55@git/linux/kernel/perf_event.c with user_lock_limit)
probe:perf_mmap_1 (on perf_mmap:55@git/linux/kernel/perf_event.c with user_lock_limit)
[root@emilia ~]#
After the fix:
[root@emilia ~]# probe perf_mmap:55 user_lock_limit
Failed to find the location of user_lock_limit at this address.
Perhaps, it has been optimized out.
Failed to find 'user_lock_limit' in this function.
Error: Failed to add events. (-2)
[root@emilia ~]#
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.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>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This commit squashes several commits that remove:
unnecessary uname calls
`sh -c'
BUILT_INS and QUIET_BUILT_IN
They have no effect, and the `fixup-builtins' and `check-builtins.sh'
scripts don't even exist.
RUNTIME_PREFIX
It's currently never anything but unset, and it's apparently
only meaningful when Microsoft Windows is the operating system
(according to the source for git).
TEST_PROGRAMS
EXTRA_PROGRAMS
unused SHELL_PATH_SQ portions
unused test for V=2
useless exports
Only when `V' is undefined (that is, only when the value of `V'
is empty) is `export V' performed, which just has the effect of
placing the empty-valued variable `V' in the environment.
The only other script to make use of `V' is `Documentation/Makefile',
which only checks whether `V' is undefined (that is, whether the value
of `V' is empty); hence, the `export V' has no effect whatsoever.
Similarly, `export QUIET_GEN' is useless because it will only have
a non-empty value when `V' has an empty-value, and when `V' has
an empty-value, `QUIET_GEN' is always explicitly set in every
script in which it is used.
`DESTDIR' is only ever defined by the user via the environment
or the command line, both of which are automatically exported
to sub-make processes. Furthermore, no non-make sub-scripts
make use of `DESTDIR' as an environment variable.
No other scripts use `perfexec_instdir'.
unused QUIET_SUBDIR{0,1}
TAR and RPMBUILD
PTHREAD_LIBS
Maintainer's dist rules and commands
distclean target
Test suite coverage testing
PRINT_DIR and NO_SUBDIR
`configure' target
NO_CURL
@@PERF_VERSION@@ substitution
Without the sed command, all of the rule's commands can be reduced
to a single line that copies a file and sets the permissions properly
in the process.
`make test' echo line
template_instdir
PERF-BUILD-OPTIONS
double-colon rules
The use of double-colon rules seems misguided or vestigial git.
Essentially hard-coded $(SCRIPTS) expansion
Signed-off-by: Michael Witten <mfwitten@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This commit squashes several commits that remove:
NO_C99_FORMAT
CURLDIR and EXPATDIR
NO_DEFLATE_BOUND
CC_LD_DYNPATH and NO_R_TO_GCC_LINKER
NO_PERL_MAKEMAKER
INTERNAL_QSORT
NO_EXTERNAL_GREP
NO_PERL
SCRIPT_PERL
PERL_PATH_SQ
Signed-off-by: Michael Witten <mfwitten@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
While it makes sense that this tool could be used on
other platforms at least to parse data, there doesn't
appear to be any real support for such usage.
This commit squashes several commits that remove:
SNPRINTF_RETURNS_BOGUS
FREAD_READS_DIRECTORIES
NO_D_{INO,TYPE}_IN_DIRENT
NO_STRCASESTR
NO_MEMMEM
NO_STRTOUMAX and NO_STRTOULL
NO_SETENV
NO_UNSETENV
NO_MKDTEMP
NEEDS_LIBICONV
NEEDS_SOCKET
NO_MMAP
NO_PTHREADS
NO_PREAD
NO_TRUSTABLE_FILEMODE
NO_IPV6 and NO_SOCKADDR_STORAGE
NO_ICONV and OLD_ICONV
NO_NSEC, USE_NSEC, and USE_ST_TIMESPEC
NO_ST_BLOCKS_IN_STRUCT_STAT
NO_FINK and NO_DARWIN_PORTS
NO_SYS_SELECT_H
NO_HSTRERROR
DIR_HAS_BSD_GROUP_SEMANTICS and FORCE_DIR_SET_GID
NEEDS_NSL, NO_UINTMAX_T, NO_INET_{N,P}TON
COMPAT_{CFLAGS,OBJS}
Executable extension `X'
Signed-off-by: Michael Witten <mfwitten@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
[root@emilia ~]# perf report --stdio
The perf.data file has no samples!
[root@emilia ~]#
The TUI shows a popup warning message with the same message.
Reported-by: Ingo Molnar <mingo@elte.hu>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
While testing the --filter option I noticed that we were writing lots of
unneeded stuff to the perf.data header when the filter ioctl fails, so
move the atexit(atexit_header) call to after we create the counters
successfully.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch changes the way perf stat prints event names at the end of a
run. Until now, it was trying to reconstruct the event name from its
encoding. The problem is that it would only print generic events without
their modifiers (u, k, pp).
This patch saves the event name as passed by the user in the evsel
struct and uses it to print the final event name.
This would also work in case perf is linked with a library (such as
libpfm4) which provides full PMU event tables.
$ perf stat -e cycles:u,cycles:k date
Wed Feb 16 14:58:52 CET 2011
Performance counter stats for 'date':
568600 cycles:u
2779715 cycles:k
0.001908182 seconds time elapsed
Cc: Arun Sharma <arun@sharma-home.net>
Cc: David S. Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@gmail.com>
LPU-Reference: <4d5bdc64.98a1df0a.7aa3.06c2@mx.google.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
[ committer note: Fixed a merge problem with 023695d "Add cgroup support" ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The 023695d cset added a new file, util/cgroup.c, that is referenced from
util/evsel.c, so it needs to be present in util/setup.py so that the python
shared object binding works, fixing this:
[root@emilia linux]# export PYTHONPATH=~acme/git/build/perf/python/
[root@emilia linux]# ./tools/perf/python/twatch.py
Traceback (most recent call last):
File "./tools/perf/python/twatch.py", line 16, in <module>
import perf
ImportError: /home/acme/git/build/perf/python/perf.so: undefined symbol: close_cgroup
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Show filename which contains a target function with the function name on
"--lines" mode, because perf-probe just shows the first function even if
there are many same-name functions.
Originally adopted by Franck Bui-Huu's patch which shows file name
instead of function name. I've just modified it to show both of function
name and file name, because of completeness of output.
E.g.)
$ perf probe -L t_show
<t_show@/home/mhiramat/ksrc/linux-2.6-tip/kernel/trace/ftrace.c:0>
0 static int t_show(struct seq_file *m, void *v)
1 {
2 struct ftrace_iterator *iter = m->private;
...
$ perf probe -L t_show@trace/trace.c
<t_show@/home/mhiramat/ksrc/linux-2.6-tip/kernel/trace/trace.c:0>
0 static int t_show(struct seq_file *m, void *v)
1 {
struct tracer *t = v;
...
Original-patch-by: Franck Bui-Huu <fbuihuu@gmail.com>
Cc: 2nddept-manager@sdl.hitachi.co.jp
Cc: Franck Bui-Huu <fbuihuu@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110210090816.1809.43426.stgit@ltc236.sdl.hitachi.co.jp>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The perf makefile is nicely complete except for
a) an uninstall option
b) a 'make help' description
This patch implements b)
it also comments out other non-working makefile targets
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Not just for the percentage number, to see the hot lines more easily.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Just like done on symbol__inc_addr_samples to catch misparsed offsets
from objdump.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The ui operations so far were used by just one thread, but 'perf top
--tui' now has two threads updating the screen, so we need to use a
mutex to avoid garbling the screen.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch adds the ability to filter monitoring based on container groups
(cgroups) for both perf stat and perf record. It is possible to monitor
multiple cgroup in parallel. There is one cgroup per event. The cgroups to
monitor are passed via a new -G option followed by a comma separated list of
cgroup names.
The cgroup filesystem has to be mounted. Given a cgroup name, the perf tool
finds the corresponding directory in the cgroup filesystem and opens it. It
then passes that file descriptor to the kernel.
Example:
$ perf stat -B -a -e cycles:u,cycles:u,cycles:u -G test1,,test2 -- sleep 1
Performance counter stats for 'sleep 1':
2,368,667,414 cycles test1
2,369,661,459 cycles
<not counted> cycles test2
1.001856890 seconds time elapsed
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4d590290.825bdf0a.7d0a.4890@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We only allocate it when in TUI mode. In --stdio mode unconditionally
initializing this area leads to memory corruption.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fixups due to rename of event_t routines from event__ to perf_event__
done in perf/core.
Conflicts:
tools/perf/builtin-record.c
tools/perf/builtin-top.c
tools/perf/util/event.c
tools/perf/util/event.h
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Jeff Moyer reported these messages:
Warning: ... trying to fall back to cpu-clock-ticks
couldn't open /proc/-1/status
couldn't open /proc/-1/maps
[ls output]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.008 MB perf.data (~363 samples) ]
That lead me and David Ahern to see that something was fishy on the thread
synthesizing routines, at least for the case where the workload is started
from 'perf record', as -1 is the default for target_tid in 'perf record --tid'
parameter, so somehow we were trying to synthesize the PERF_RECORD_MMAP and
PERF_RECORD_COMM events for the thread -1, a bug.
So I investigated this and noticed that when we introduced support for
recording a process and its threads using --pid some bugs were introduced and
that the way to fix it was to instead of passing the target_tid to the event
synthesizing routines we should better pass the thread_map that has the list of
threads for a --pid or just the single thread for a --tid.
Checked in the following ways:
On a 8-way machine run cyclictest:
[root@emilia ~]# perf record cyclictest -a -t -n -p99 -i100 -d50
policy: fifo: loadavg: 0.00 0.13 0.31 2/139 28798
T: 0 (28791) P:99 I:100 C: 25072 Min: 4 Act: 5 Avg: 6 Max: 122
T: 1 (28792) P:98 I:150 C: 16715 Min: 4 Act: 6 Avg: 5 Max: 27
T: 2 (28793) P:97 I:200 C: 12534 Min: 4 Act: 5 Avg: 4 Max: 8
T: 3 (28794) P:96 I:250 C: 10028 Min: 4 Act: 5 Avg: 5 Max: 96
T: 4 (28795) P:95 I:300 C: 8357 Min: 5 Act: 6 Avg: 5 Max: 12
T: 5 (28796) P:94 I:350 C: 7163 Min: 5 Act: 6 Avg: 5 Max: 12
T: 6 (28797) P:93 I:400 C: 6267 Min: 4 Act: 5 Avg: 5 Max: 9
T: 7 (28798) P:92 I:450 C: 5571 Min: 4 Act: 5 Avg: 5 Max: 9
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.108 MB perf.data (~4719 samples) ]
[root@emilia ~]#
This will create one extra thread per CPU:
[root@emilia ~]# tuna -t cyclictest -CP
thread ctxt_switches
pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
28825 OTHER 0 0xff 2169 671 cyclictest
28832 FIFO 93 6 52338 1 cyclictest
28833 FIFO 92 7 46524 1 cyclictest
28826 FIFO 99 0 209360 1 cyclictest
28827 FIFO 98 1 139577 1 cyclictest
28828 FIFO 97 2 104686 0 cyclictest
28829 FIFO 96 3 83751 1 cyclictest
28830 FIFO 95 4 69794 1 cyclictest
28831 FIFO 94 5 59825 1 cyclictest
[root@emilia ~]#
So we should expect only samples for the above 9 threads when using the
--dump-raw-trace|-D perf report switch to look at the column with the tid:
[root@emilia ~]# perf report -D | grep RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort | uniq -c
629 28825
110 28826
491 28827
308 28828
198 28829
621 28830
225 28831
203 28832
89 28833
[root@emilia ~]#
So for workloads started by 'perf record' seems to work, now for existing workloads,
just run cyclictest first, without 'perf record':
[root@emilia ~]# tuna -t cyclictest -CP
thread ctxt_switches
pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
28859 OTHER 0 0xff 594 200 cyclictest
28864 FIFO 95 4 16587 1 cyclictest
28865 FIFO 94 5 14219 1 cyclictest
28866 FIFO 93 6 12443 0 cyclictest
28867 FIFO 92 7 11062 1 cyclictest
28860 FIFO 99 0 49779 1 cyclictest
28861 FIFO 98 1 33190 1 cyclictest
28862 FIFO 97 2 24895 1 cyclictest
28863 FIFO 96 3 19918 1 cyclictest
[root@emilia ~]#
and then later did:
[root@emilia ~]# perf record --pid 28859 sleep 3
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.027 MB perf.data (~1195 samples) ]
[root@emilia ~]#
To collect 3 seconds worth of samples for pid 28859 and its children:
[root@emilia ~]# perf report -D | grep RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort | uniq -c
15 28859
33 28860
19 28861
13 28862
13 28863
10 28864
11 28865
9 28866
255 28867
[root@emilia ~]#
Works, last thing is to check if looking at just one of those threads also works:
[root@emilia ~]# perf record --tid 28866 sleep 3
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.006 MB perf.data (~242 samples) ]
[root@emilia ~]# perf report -D | grep RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort | uniq -c
3 28866
[root@emilia ~]#
Works too.
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jeff Moyer <jmoyer@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>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The live annotation done in 'perf top' needs to limit the context before
lines that aren't filtered out by the min percent filter, if we don't do
that, the screen in a tty often is not enough for showing what is
interesting: lines with hits and a few source code lines before it.
Reported-by: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Since we'll need it when implementing the live annotate TUI browser.
This also simplifies things a bit by having the list head for the source
code to be in the dynamicly allocated part of struct annotation, that
way we don't have to pass it around, it can be found from the struct
symbol that is passed everywhere.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The checks for not using a max_lines parameter were b0rked, problem
introduced in 3653246.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
