When trying to map a bunch of instruction addresses to their respective
threads, I kept getting a lot of bogus entries [I forget the exact
reason as I patched my code months ago].
Looking through ip__resolve_ams, I noticed the check for
if (al.sym)
and realized, most times I have an al.map definition but sometimes an
al.sym is undefined. In the cases where al.sym is undefined, the loop
keeps going even though a valid al.map exists.
Modify this check to use the more reliable al.map. This fixed my bogus
entries.
Signed-off-by: Don Zickus <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Stephane Eranian <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
When compiling perf tool code with gcc 4.4.7 I'm getting
following error:
CC util/session.o
cc1: warnings being treated as errors
util/session.c: In function ‘perf_session_deliver_event’:
tools/perf/util/include/linux/bitops.h:109: error: dereferencing pointer ‘p’ does break strict-aliasing rules
tools/perf/util/include/linux/bitops.h:101: error: dereferencing pointer ‘p’ does break strict-aliasing rules
util/session.c:697: note: initialized from here
tools/perf/util/include/linux/bitops.h:101: note: initialized from here
make[1]: *** [util/session.o] Error 1
make: *** [util/session.o] Error 2
The aliased types here are u64 and unsigned long pointers, which is safe
for the find_first_bit processing.
This error shows up for me only for gcc 4.4 on 32bit x86, even for
-Wstrict-aliasing=3, while newer gcc are quiet and scream here for
-Wstrict-aliasing={2,1}. Looks like newer gcc changed the rules for
strict alias warnings.
The gcc documentation offers workaround for valid aliasing by using
__may_alias__ attribute:
http://gcc.gnu.org/onlinedocs/gcc-4.4.0/gcc/Type-Attributes.html
Using this workaround for the find_first_bit function.
Signed-off-by: Jiri Olsa <[email protected]>
Cc: Corey Ashford <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Stephane reported that perf report and annotate failed to process data
using lots of (> 500) shared libraries. It was because of the limit on
number of open files (ulimit -n).
Currently when perf loads a DSO, it'll look for normal and dynamic
symbol tables. And if it fails to find out both tables, it'll iterate
all of possible symtab types. But many of them are useless since they
have no additional information and the problem is that it's not closing
those files even though they're not used. Fix it.
Reported-by: Stephane Eranian <[email protected]>
Signed-off-by: Namhyung Kim <[email protected]>
Cc: Cody P Schafer <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
The TUI of perf report and top support annotation, but stdio and GTK
don't. So it should be checked before calling hist_entry__inc_addr_
samples() to avoid wasting resources that will never be used.
perf annotate need it regardless of UI and sort keys, so the check
of whether to allocate resources should be on the tools that have
annotate as an option in the TUI, 'report' and 'top', not on the
function called by all of them.
It caused perf annotate on ppc64 to produce zero output, since the
buckets were not being allocated.
Reported-by: Anton Blanchard <[email protected]>
Signed-off-by: Namhyung Kim <[email protected]>
Cc: Anton Blanchard <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Renamed (report,top)__needs_annotate() to ui__has_annotation() ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
"perf list" listing of hardware events doesn't work on older ARM devices.
The change enabling event detection:
commit b41f1cec91
Author: Namhyung Kim <[email protected]>
Date: Tue Aug 27 11:41:53 2013 +0900
perf list: Skip unsupported events
uses the following code in tools/perf/util/parse-events.c:
struct perf_event_attr attr = {
.type = type,
.config = config,
.disabled = 1,
.exclude_kernel = 1,
};
On ARM machines pre-dating the Cortex-A15 this doesn't work, as these
machines don't support .exclude_kernel. So starting with 3.12 "perf
list" does not report any hardware events at all on older machines (seen
on Rasp-Pi, Pandaboard, Beagleboard, etc).
This version of the patch makes changes suggested by Namhyung Kim to
check for EACCESS and retry (instead of just dropping the
exclude_kernel) so we can properly handle machines where
/proc/sys/kernel/perf_event_paranoid is set to 2.
Reported-by: Chad Paradis <[email protected]>
Signed-off-by: Vince Weaver <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Chad Paradis <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1312301536150.28814@vincent-weaver-1.um.maine.edu
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
We removed event types from data file in following commits:
6065210 perf tools: Remove event types framework completely
44b3c57 perf tools: Remove event types from perf data file
We no longer need this information, because we can get it directly from
tracepoints.
But we still need to handle PERF_RECORD_HEADER_EVENT_TYPE event for the
sake of old perf data files created in pipe mode like:
$ perf.3.4 record -o - foo >perf.data
$ perf.312 report -i - < perf.data
Reported-by: Stephane Eranian <[email protected]>
Signed-off-by: Jiri Olsa <[email protected]>
Cc: Corey Ashford <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Fix perf-probe not to add offset value twice to uprobe probe address
when post processing.
The tevs[i].point.address struct member is the address of symbol+offset,
but current perf-probe adjusts the point.address by adding the offset.
As a result, the probe address becomes symbol+offset+offset. This may
cause unexpected code corruption. Urgent fix is needed.
Without this fix:
---
# ./perf probe -x ./perf dso__load_vmlinux+4
# ./perf probe -l
probe_perf:dso__load_vmlinux (on 0x000000000006d2b8)
# nm ./perf.orig | grep dso__load_vmlinux\$
000000000046d0a0 T dso__load_vmlinux
---
You can see the given offset is 3 but the actual probed address is
dso__load_vmlinux+8.
With this fix:
---
# ./perf probe -x ./perf dso__load_vmlinux+4
# ./perf probe -l
probe_perf:dso__load_vmlinux (on 0x000000000006d2b4)
---
Now the problem is fixed.
Note: This bug is introduced by
commit fb7345bbf7
Signed-off-by: Masami Hiramatsu <[email protected]>
Cc: "David A. Long" <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Srikar Dronamraju <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/20140205051858.6519.27314.stgit@kbuild-fedora.yrl.intra.hitachi.co.jp
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Commit 71ae8aac ("lib: introduce arch optimized hash library") added an
include to <linux/hash.h> for setting up an architecture specific fast
hash.
Since perf includes directly the non-uapi kernel header, it cannot find
<asm/hash.h> on non-x86 and thus prevents perf to be compiled on every
architecture other than x86.
The problem is the inclusion of <asm/hash.h> in hash.h that results in
the following error originating from util/evlist.c:
fatal error: asm/hash.h: No such file or directory
This commit simply adds an empty <asm/hash.h> stub/file to fix the
compile issue on non-x86 architectures.
As perf does not use any of these new functions, it fixes the
compilation and therefore seems to be the most appropriate solution to
go with.
Signed-off-by: Francesco Fusco <[email protected]>
Link: http://lkml.kernel.org/r/2cf8143aad65a6aa6fe30325ef8a65847141afa2.1390829373.git.ffusco@redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
In map_groups__find_symbol() map->map_ip is used without ensuring the
map is loaded. Then the address passed to map->map_ip isn't mapped at
the first time.
E.g. below code always fails to get a symbol at the first call;
addr = /* Somewhere in the kernel text */
symbol_conf.try_vmlinux_path = true;
symbol__init();
host_machine = machine__new_host();
sym = machine__find_kernel_function(host_machine,
addr, NULL, NULL);
/* Note that machine__find_kernel_function calls
map_groups__find_symbol */
This ensures it by calling map__load before using it in
map_groups__find_symbol().
Signed-off-by: Masami Hiramatsu <[email protected]>
Cc: "David A. Long" <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Srikar Dronamraju <[email protected]>
Cc: "Steven Rostedt (Red Hat)" <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/20140123022950.7206.17357.stgit@kbuild-fedora.yrl.intra.hitachi.co.jp
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
The plugindir_SQ definition contains $(prefix) which is not needed as
the $(libdir) definition already contains prefix in it. This leads to
the path including an extra prefix in it, e.g. /usr/usr/lib64.
The -DPLUGIN_DIR defintion includes DESTDIR. This is incorrect, as it
sets the plugin search path to include the value of DESTDIR. DESTDIR is
a mechanism to install in a non-standard location such as a chroot or an
RPM build root. In the RPM case, this leads to the search path being
incorrect after the resulting RPM is installed (or in some cases an RPM
build failure).
Remove both of these unnecessary inclusions.
Signed-off-by: Josh Boyer <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Steven Rostedt <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Gaurav reported that perf cannot profile JIT program if it executes the
code on heap. This was because current map__new() only handle JIT on
anon mappings - extends it to handle no_dso (heap, stack) case too.
This patch assumes JIT profiling only provides dynamic function symbols
so check the mapping type to distinguish the case. It'd provide no
symbols for data mapping - if we need to support symbols on data
mappings later it should be changed.
Reported-by: Gaurav Jain <[email protected]>
Signed-off-by: Namhyung Kim <[email protected]>
Tested-by: Gaurav Jain <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Gaurav Jain <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
This patch fixes a memory corruption problem with the xyarray when the
evsel fds get closed at the end of the run_perf_stat() call.
It could be triggered with:
# perf stat -a -e power/energy-cores/ ls
When cpumask are used by events (.e.g, RAPL or uncores) then the evsel
fds are allocated based on the actual number of CPUs monitored. That
number can be smaller than the total number of CPUs on the system.
The problem arises at the end by perf stat closes the fds twice. When
fds are closed, their entry in the xyarray are set to -1.
The first close() on the evsel is made from __run_perf_stat() and it
uses the actual number of CPUS for the event which is how the xyarray
was allocated for.
The second is from perf_evlist_close() but that one is on the total
number of CPUs in the system, so it assume the xyarray was allocated to
cover it. However it was not, and some writes corrupt memory.
The fix is in perf_evlist_close() is to first try with the evsel->cpus
if present, if not use the evlist->cpus. That fixes the problem.
Signed-off-by: Stephane Eranian <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
This patch fixes a problem with the handling of the newly introduced
optional event unit. The following cmdline caused a segfault:
$ perf stat -e cpu/event-0x3c/ ls
This patch fixes the problem with the default setting for alias->unit
which was eventually causing the segfault.
Signed-off-by: Stephane Eranian <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
When a new callchain child branch matches an existing one in the rbtree,
the comparison of its first entry is performed twice:
1) From append_chain_children() on branch lookup
2) If 1) reports a match, append_chain() then compares all entries of
the new branch against the matching node in the rbtree, and this
comparison includes the first entry of the new branch again.
Lets shortcut this by performing the whole comparison only from
append_chain() which then returns the result of the comparison between
the first entry of the new branch and the iterating node in the rbtree.
If the first entry matches, the lookup on the current level of siblings
stops and propagates to the children of the matching nodes.
This results in less comparisons performed by the CPU.
Signed-off-by: Frederic Weisbecker <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
The comm overriding API ignores memory allocation failures by silently
keeping the previous and out of date comm.
As a result, the user may get buggy events without ever being notified
about the problem and its source.
Lets start to fix this by propagating the error from the API. Not all
callers may be doing proper error handling on comm set yet but this is
the first step toward it.
Signed-off-by: Frederic Weisbecker <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Currently the feature-checks Makefile does not inherit $(CC), and calls
cc rather than $(CROSS_COMPILE)gcc. Thus the feature checks invoke the
native toolchain rather than the cross toolchain, and can identify
features as available when they are not. This can break the build.
Additionally the native pkg-config is always called as opposed to
$(CROSS_COMPILE)pkg-config, so the wrong flags and paths may be passed
to the cross compiler.
This patch passes CROSS_COMPILE down to the feature-checks Makefile, and
forces its use. Additionally pkg-config is replaced with
$(CROSS_COMPILE)pkg-config via a new $(PKG_CONFIG) variable. This patch
has been build tested on x86_64 and arm.
Signed-off-by: Mark Rutland <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Will Deacon <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
It failed to build perf on my ubuntu 10.04 box (gcc 4.4.3):
CC util/strlist.o
cc1: warnings being treated as errors
util/strlist.c: In function ‘str_node__delete’:
util/strlist.c:42: error: dereferencing type-punned pointer will break strict-aliasing rules
util/strlist.c:42: error: dereferencing type-punned pointer will break strict-aliasing rules
CC util/strfilter.o
make: *** [util/strlist.o] Error 1
CC util/srcline.o
cc1: warnings being treated as errors
util/srcline.c: In function ‘addr2line_init’:
util/srcline.c:132: error: dereferencing type-punned pointer will break strict-aliasing rules
util/srcline.c:132: error: dereferencing type-punned pointer will break strict-aliasing rules
util/srcline.c: In function ‘addr2line_cleanup’:
util/srcline.c:143: error: dereferencing type-punned pointer will break strict-aliasing rules
util/srcline.c:143: error: dereferencing type-punned pointer will break strict-aliasing rules
make: *** [util/srcline.o] Error 1
It seems it only allows to remove 'const' qualifier.
Signed-off-by: Namhyung Kim <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Don Zickus