linux.git/tools/perf, branch v4.2-rc5

perf tools: Really allow to specify custom CC, AR or LD

2015-07-15T14:57:28+00:00

Commit 5ef7bbb09f7b ("perf tools: Allow to specify custom linker
command") was meant to enable usage non $(CROSS_COMPILE)ld linker during
perf building.

But implementation didn't take into account the fact that LD is a
pre-defined variable in GNU Make. I.e. it is always defined.

Which means there's no point to check "LD ?= ..." because it will never
succeed.

And so LD will be either that explicitly passed to make like this:

 ------->8-------
 make LD=path_to_my_ld ...
 ------->8-------
 or default value, which is host's "ld".

Latter leads to failure of cross-linkage because instead of cross linker
"$(CROSS_COMPILE)ld" host's "ld" is used.

Fortunately there's a way to do correct substitution of $(CROSS_COMPILE)ld
with user defined LD on command-line.

As a reference was used implementation in "tools/lib/traceevent/Makefile".

Build tested for x86_64 and ARC.

Thanks Jiri for this hint.

Signed-off-by: Alexey Brodkin 
Fixes: 5ef7bbb09f7b ("perf tools: Allow to specify custom linker command")
Cc: Aaro Koskinen 
Cc: Jiri Olsa 
Cc: Paul Mackerras 
Cc: Peter Zijlstra 
Cc: Vineet Gupta 
Cc: Vineet Gupta 
Cc: linux-arch@vger.kernel.org
Link: http://lkml.kernel.org/r/1436864720-26316-1-git-send-email-abrodkin@synopsys.com
Signed-off-by: Arnaldo Carvalho de Melo

perf auxtrace: Fix misplaced check for HAVE_SYNC_COMPARE_AND_SWAP_SUPPORT

2015-07-15T14:57:28+00:00

Move the checking for HAVE_SYNC_COMPARE_AND_SWAP_SUPPORT for AUX area mmaps
until after checking if such mmaps are used anyway.

Reported-by: Alexey Brodkin 
Signed-off-by: Adrian Hunter 
Tested-by: Alexey Brodkin 
Cc: David Ahern 
Cc: Frederic Weisbecker 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Stephane Eranian 
Cc: Vineet Gupta 
Cc: linux-arch@vger.kernel.org
Link: http://lkml.kernel.org/r/55A5023C.7020907@intel.com
Signed-off-by: Arnaldo Carvalho de Melo

perf hists browser: Take the --comm, --dsos, etc filters into account

2015-07-13T19:06:09+00:00

At some point:

  commit 2c86c7ca7606
  Author: Namhyung Kim 
  Date:   Mon Mar 17 18:18:54 2014 -0300

    perf report: Merge al->filtered with hist_entry->filtered

We stopped dropping samples for things filtered via the --comms, --dsos,
--symbols, etc, i.e. things marked as filtered in the symbol resolution
routines (thread__find_addr_map(), perf_event__preprocess_sample(),
etc).

But then, in:

  commit 268397cb2a47
  Author: Namhyung Kim 
  Date:   Tue Apr 22 14:49:31 2014 +0900

    perf top/tui: Update nr_entries properly after a filter is applied

We don't take into account entries that were filtered in
perf_event__preprocess_sample() and friends, which leads to
inconsistency in the browser seek routines, that expects the number of
hist_entry->filtered entries to match what it thinks is the number of
unfiltered, browsable entries.

So, for instance, when we do:

  perf top --symbols ___non_existent_symbol___

the hist_browser__nr_entries() routine thinks there are no filters in
place, uses the hists->nr_entries but all entries are filtered, leading
to a segfault.

Tested with:

   perf top --symbols malloc,free --percentage=relative

Freezing, by pressing 'f', at any time and doing the math on the
percentages ends up with 100%, ditto for:

   perf top --dsos libpthread-2.20.so,libxul.so --percentage=relative

Both were segfaulting, all fixed now.

More work needed to do away with checking if filters are in place, we
should just use the nr_non_filtered_samples counter, no need to
conditionally use it or hists.nr_filter, as what the browser does is
just show unfiltered stuff. An audit of how it is being accounted is
needed, this is the minimal fix.

Reported-by: Michael Petlan 
Fixes: 268397cb2a47 ("perf top/tui: Update nr_entries properly after a filter is applied")
Cc: Adrian Hunter 
Cc: Borislav Petkov 
Cc: David Ahern 
Cc: Frederic Weisbecker 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Stephane Eranian 
Link: http://lkml.kernel.org/n/tip-6w01d5q97qk0d64kuojme5in@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

perf symbols: Store if there is a filter in place

2015-07-13T11:21:57+00:00

When setting yup the symbols library we setup several filter lists,
for dsos, comms, symbols, etc, and there is code that, if there are
filters, do certain operations, like recalculate the number of non
filtered histogram entries in the top/report TUI.

But they were considering just the "Zoom" filters, when they need to
take into account as well the above mentioned filters (perf top --comms,
--dsos, etc).

So store in symbol_conf.has_filter true if any of those filters is in
place.

Cc: Adrian Hunter 
Cc: Borislav Petkov 
Cc: David Ahern 
Cc: Frederic Weisbecker 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Stephane Eranian 
Link: http://lkml.kernel.org/n/tip-f5edfmhq69vfvs1kmikq1wep@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

tools: Copy lib/hweight.c from the kernel sources

2015-07-09T19:29:56+00:00

Instead of accessing it directly, as it uses EXPORT_SYMBOL, that has
no meaning in tools/perf and because we removed the stubs for it, i.e.
we removed the tools/include/linux/export.h file.

This fixes the build for the detached tarball sources cases and removes
one more source of entanglement with the kernel sources.

Cc: Adrian Hunter 
Cc: Borislav Petkov 
Cc: David Ahern 
Cc: Frederic Weisbecker 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Stephane Eranian 
Link: http://lkml.kernel.org/n/tip-oyqx541o7apa2cskjhcxi6nx@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

perf tools: Fix the detached tarball wrt rbtree copy

2015-07-09T19:23:57+00:00

The python binding build process was still looking at the kernel
rbtree.c file, so, when doing a in-tree build it would work, but when
creating a tarball using tools/perf/MANIFEST as the contents list and
then trying to build the resulting detached sources, it failed.

Fix it by removing one level of indirection from rbtree.c in the
tools/perf/util/python-ext-sources file.

Cc: Adrian Hunter 
Cc: Borislav Petkov 
Cc: David Ahern 
Cc: Frederic Weisbecker 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Stephane Eranian 
Link: http://lkml.kernel.org/n/tip-8u83c2k5guyhxdlkaaqis8k4@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

perf thread_map: Fix the sizeof() calculation for map entries

2015-07-09T15:28:53+00:00

When we started adding extra stuff per array entry, growing the size of
those entries to more than sizeof(pid_t), we had to convert those sizeof
operations to the more robust sizeof(map->map[0]) idiom, that is future
proof, i.e. if/when we add more stuff to those entries, that expression
will produce the new per-entry size.

And besides that, we need to zero out those extra fields, that sometimes
may not get filled, like when we couldn't care less about the comms,
since we don't need those, but since we will try freeing it at
thread_map__delete(), we better fix it.

That is why a thread_map__realloc() was provided.

But that method wasn't used in thread_map__new_by_uid(), fix it.

Reported-by: Ingo Molnar 
Fixes: 792402fd5c0a ("perf thrad_map: Add comm string into array")
Fixes: 9d7e8c3a96e5 ("perf tools: Add thread_map__(alloc|realloc) helpers")
Cc: Adrian Hunter 
Cc: Borislav Petkov 
Cc: David Ahern 
Cc: Frederic Weisbecker 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Stephane Eranian 
Link: http://lkml.kernel.org/n/tip-6a0swlm6m8lnu3wpjv284hkb@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

perf stat: Fix shadow declaration of close

2015-07-08T13:04:58+00:00

Vinson reported shadow declaration of close introduced
by the following commit:

  106a94a0f8c2 perf stat: Introduce read_counters function

Using close_counters name instead.

Reported-by: Vinson Lee 
Signed-off-by: Jiri Olsa 
Cc: Jiri Olsa 
Cc: Peter Zijlstra 
Fixes: 106a94a0f8c2 ("perf stat: Introduce read_counters function")
Link: http://lkml.kernel.org/r/20150708111731.GA3512@krava.redhat.com
Signed-off-by: Arnaldo Carvalho de Melo

perf tools: Fix lockup using 32-bit compat vdso

2015-07-07T14:05:08+00:00

The __machine__findnew_compat() function is called only from
__machine__findnew_vdso_compat() which is called only from
machine__findnew_vdso() which already holds machine->dsos.lock, so
remove locking from __machine__findnew_compat().

This manifests itself tracing 32-bit programs with a 64-bit perf.

Signed-off-by: Adrian Hunter 
Cc: David Ahern 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Link: http://lkml.kernel.org/r/1436267618-20521-1-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo

Merge branch 'perf/rbtree_copy' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent

2015-07-06T07:24:41+00:00

Pull rbtree build fix from Arnaldo Carvalho de Melo.

Signed-off-by: Ingo Molnar