openwrt/staging/blogic.git
13 years agoperf hist: Better displaying of unresolved DSOs and symbols
Ian Munsie [Mon, 6 Dec 2010 02:37:04 +0000 (13:37 +1100)]
perf hist: Better displaying of unresolved DSOs and symbols

In the event that a DSO has not been identified, just print out [unknown]
instead of the instruction pointer as we previously were doing, which is pretty
meaningless for a shared object (at least to the users perspective).

The IP we print out is fairly meaningless in general anyway - it's just one
(the first) of the many addresses that were lumped together as unidentified,
and could span many shared objects and symbols. In reality if we see this
[unknown] output then the report -D output is going to be more useful anyway as
we can see all the different address that it represents.

If we are printing the symbols we are still going to see this IP in that column
anyway since they shouldn't resolve either.

This patch also changes the symbol address printouts so that they print out 0x
before the address, are left aligned, and changes the %L format string (which
relies on a glibc bug) to %ll.

Before:
    74.11%    :3259               4a6c  [k]     4a6c
After:
    74.11%    :3259  [unknown]          [k] 0x4a6c

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291603026-11785-2-git-send-email-imunsie@au1.ibm.com>
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events
Arnaldo Carvalho de Melo [Thu, 2 Dec 2010 12:25:28 +0000 (10:25 -0200)]
perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events

So that we can use -T == --timestamp, asking for PERF_SAMPLE_TIME:

  $ perf record -aT
  $ perf report -D | grep PERF_RECORD_
  <SNIP>
   3   5951915425 0x47530 [0x58]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff8138c1a2 period: 215979 cpu:3
   3   5952026879 0x47588 [0x90]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff810cb480 period: 215979 cpu:3
   3   5952059959 0x47618 [0x38]: PERF_RECORD_FORK(6853:6853):(16811:16811)
   3   5952138878 0x47650 [0x78]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff811bac35 period: 431478 cpu:3
   3   5952375068 0x476c8 [0x30]: PERF_RECORD_COMM: find:6853
   3   5952395923 0x476f8 [0x50]: PERF_RECORD_MMAP 6853/6853: [0x400000(0x25000) @ 0]: /usr/bin/find
   3   5952413756 0x47748 [0xa0]: PERF_RECORD_SAMPLE(IP, 1): 6853/6853: 0xffffffff810d080f period: 859332 cpu:3
   3   5952419837 0x477e8 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44600000(0x21d000) @ 0]: /lib64/ld-2.5.so
   3   5952437929 0x47840 [0x48]: PERF_RECORD_MMAP 6853/6853: [0x7fff7e1c9000(0x1000) @ 0x7fff7e1c9000]: [vdso]
   3   5952570127 0x47888 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f46200000(0x218000) @ 0]: /lib64/libselinux.so.1
   3   5952623637 0x478e0 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44a00000(0x356000) @ 0]: /lib64/libc-2.5.so
   3   5952675720 0x47938 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44e00000(0x204000) @ 0]: /lib64/libdl-2.5.so
   3   5952710080 0x47990 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f45a00000(0x246000) @ 0]: /lib64/libsepol.so.1
   3   5952847802 0x479e8 [0x58]: PERF_RECORD_SAMPLE(IP, 1): 6853/6853: 0xffffffff813897f0 period: 1142536 cpu:3
  <SNIP>

First column is the cpu and the second the timestamp.

That way we can investigate problems in the event stream.

If the new perf binary is run on an older kernel, it will disable this feature
automatically.

Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-5-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf session: Parse sample earlier
Arnaldo Carvalho de Melo [Thu, 2 Dec 2010 16:10:21 +0000 (14:10 -0200)]
perf session: Parse sample earlier

At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.

This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.

Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.

There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.

Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf events: Make sample_type identity fields available in all PERF_RECORD_ events
Arnaldo Carvalho de Melo [Sun, 5 Dec 2010 01:02:20 +0000 (23:02 -0200)]
perf events: Make sample_type identity fields available in all PERF_RECORD_ events

If perf_event_attr.sample_id_all is set it will add the PERF_SAMPLE_ identity
info:

TID, TIME, ID, CPU, STREAM_ID

As a trailer, so that older perf tools can process new files, just ignoring the
extra payload.

With this its possible to do further analysis on problems in the event stream,
like detecting reordering of MMAP and FORK events, etc.

V2: Fixup header size in comm, mmap and task processing, as we have to take into
account different sample_types for each matching event, noticed by Thomas Gleixner.

Thomas also noticed a problem in v2 where if we didn't had space in the buffer we
wouldn't restore the header size.

Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf events: Separate the routines handling the PERF_SAMPLE_ identity fields
Arnaldo Carvalho de Melo [Fri, 3 Dec 2010 18:36:35 +0000 (16:36 -0200)]
perf events: Separate the routines handling the PERF_SAMPLE_ identity fields

Those will be made available in sample like events like MMAP, EXEC, etc in a
followup patch. So precalculate the extra id header space and have a separate
routine to fill them up.

V2: Thomas noticed that the id header needs to be precalculated at
inherit_events too:

LKML-Reference: <alpine.LFD.2.00.1012031245220.2653@localhost6.localdomain6>

Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1291318772-30880-2-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf events: Fix event inherit fallout of precalculated headers
Thomas Gleixner [Fri, 3 Dec 2010 18:24:32 +0000 (16:24 -0200)]
perf events: Fix event inherit fallout of precalculated headers

The precalculated header size is not updated when an event is inherited. That
results in bogus sample entries for all child events. Bug introduced in c320c7b.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <alpine.LFD.2.00.1012031245220.2653@localhost6.localdomain6>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoMerge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux...
Ingo Molnar [Thu, 2 Dec 2010 10:20:11 +0000 (11:20 +0100)]
Merge branch 'perf/core' of git://git./linux/kernel/git/acme/linux-2.6 into perf/core

14 years agoperf stat: Add csv-style output
Stephane Eranian [Wed, 1 Dec 2010 16:49:05 +0000 (18:49 +0200)]
perf stat: Add csv-style output

This patch adds an option (-x/--field-separator) to print counts using a
CSV-style output. The user can pass a custom separator. This makes it very easy
to import counts directly into your favorite spreadsheet without having to
write scripts.

Example:
$ perf stat --field-separator=,  -a -- sleep 1
4009.961740,task-clock-msecs
13,context-switches
2,CPU-migrations
189,page-faults
9596385684,cycles
3493659441,instructions
872897069,branches
41562,branch-misses
22424,cache-references
1289,cache-misses

Works also in non-aggregated mode:

$ perf stat -x ,  -a -A -- sleep 1
CPU0,1002.526168,task-clock-msecs
CPU1,1002.528365,task-clock-msecs
CPU2,1002.523360,task-clock-msecs
CPU3,1002.519878,task-clock-msecs
CPU0,1,context-switches
CPU1,5,context-switches
CPU2,5,context-switches
CPU3,6,context-switches
CPU0,0,CPU-migrations
CPU1,1,CPU-migrations
CPU2,0,CPU-migrations
CPU3,1,CPU-migrations
CPU0,2,page-faults
CPU1,6,page-faults
CPU2,9,page-faults
CPU3,174,page-faults
CPU0,2399439771,cycles
CPU1,2380369063,cycles
CPU2,2399142710,cycles
CPU3,2373161192,cycles
CPU0,872900618,instructions
CPU1,873030960,instructions
CPU2,872714525,instructions
CPU3,874460580,instructions
CPU0,221556839,branches
CPU1,218134342,branches
CPU2,218161730,branches
CPU3,218284093,branches
CPU0,18556,branch-misses
CPU1,1449,branch-misses
CPU2,3447,branch-misses
CPU3,12714,branch-misses
CPU0,8330,cache-references
CPU1,313844,cache-references
CPU2,47993728,cache-references
CPU3,826481,cache-references
CPU0,272,cache-misses
CPU1,5360,cache-misses
CPU2,1342193,cache-misses
CPU3,13992,cache-misses

This second version adds the ability to name a separator and uses
field-separator as the long option to be consistent with perf report.

Commiter note: Since we enabled --big-num by default in 201e0b0 and -x can't be
used with it, we need to notice if the user explicitely enabled or disabled -B,
add code to disable big_num if the user didn't explicitely set --big_num when
-x is used.

Cc: David S. Miller <davem@davemloft.net>
Cc: Frederik Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: paulus@samba.org
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
LKML-Reference: <4cf68aa7.0fedd80a.5294.1203@mx.google.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf stat: Use --big-num format by default
Arnaldo Carvalho de Melo [Wed, 1 Dec 2010 19:53:27 +0000 (17:53 -0200)]
perf stat: Use --big-num format by default

[acme@mica linux]$ perf stat ls > /dev/null

 Performance counter stats for 'ls':

           1.512532  task-clock-msecs         #      0.801 CPUs
                  2  context-switches         #      0.001 M/sec
                  0  CPU-migrations           #      0.000 M/sec
                241  page-faults              #      0.159 M/sec
          2,973,331  cycles                   #   1965.797 M/sec
          1,460,802  instructions             #      0.491 IPC
            314,642  branches                 #    208.023 M/sec
             18,475  branch-misses            #      5.872 %
      <not counted>  cache-references
      <not counted>  cache-misses

        0.001887676  seconds time elapsed

To get the previous behaviour just use --no-big-num:

[acme@mica linux]$ perf stat --no-big-num ls > /dev/null

 Performance counter stats for 'ls':

           1.468014  task-clock-msecs         #      0.795 CPUs
                  1  context-switches         #      0.001 M/sec
                  0  CPU-migrations           #      0.000 M/sec
                241  page-faults              #      0.164 M/sec
            2900254  cycles                   #   1975.631 M/sec
            1437991  instructions             #      0.496 IPC
             310905  branches                 #    211.786 M/sec
              17912  branch-misses            #      5.761 %
      <not counted>  cache-references
      <not counted>  cache-misses

        0.001845435  seconds time elapsed

[acme@mica linux]$

Suggested-by: Ingo Molnar <mingo@elte.hu>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf stat: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:19 +0000 (19:57 -0600)]
perf stat: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-12-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf test: Fix spelling mistake in documentation
Shawn Bohrer [Wed, 1 Dec 2010 01:57:20 +0000 (19:57 -0600)]
perf test: Fix spelling mistake in documentation

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-13-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf trace: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:22 +0000 (19:57 -0600)]
perf trace: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-15-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf top: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:21 +0000 (19:57 -0600)]
perf top: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-14-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf sched: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:18 +0000 (19:57 -0600)]
perf sched: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-11-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf report: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:17 +0000 (19:57 -0600)]
perf report: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-10-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf record: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:16 +0000 (19:57 -0600)]
perf record: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-9-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf probe: Fix spelling mistake in documentation
Shawn Bohrer [Wed, 1 Dec 2010 01:57:15 +0000 (19:57 -0600)]
perf probe: Fix spelling mistake in documentation

Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-8-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf lock: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:14 +0000 (19:57 -0600)]
perf lock: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-7-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf kvm: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:13 +0000 (19:57 -0600)]
perf kvm: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-6-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf diff: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:12 +0000 (19:57 -0600)]
perf diff: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-5-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf diff: Fix displacement and modules options short flag
Shawn Bohrer [Wed, 1 Dec 2010 01:57:11 +0000 (19:57 -0600)]
perf diff: Fix displacement and modules options short flag

The --displacement and --modules options to perf diff both use -m as a
short flag.  Change --displacement to use -M since other perf commands
use -m, --modules.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-4-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf buildid-list: Document missing options
Shawn Bohrer [Wed, 1 Dec 2010 01:57:10 +0000 (19:57 -0600)]
perf buildid-list: Document missing options

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-3-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf annotate: Document missing options.
Shawn Bohrer [Wed, 1 Dec 2010 01:57:09 +0000 (19:57 -0600)]
perf annotate: Document missing options.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1291168642-11402-2-git-send-email-shawn.bohrer@gmail.com>
Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoMerge branch 'perf/rename' into perf/core
Ingo Molnar [Wed, 1 Dec 2010 08:18:12 +0000 (09:18 +0100)]
Merge branch 'perf/rename' into perf/core

Merge reason: This is an older commit under testing that was not pushed yet - merge it.

Also fix up the merge in command-list.txt.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Tom Zanussi <tzanussi@gmail.com>
14 years agoperf tools: fix event parsing of comma-separated tracepoint events
Corey Ashford [Tue, 30 Nov 2010 22:27:01 +0000 (14:27 -0800)]
perf tools: fix event parsing of comma-separated tracepoint events

There are number of issues that prevent the use of multiple tracepoint events
being specified in a -e/--event switch, separated by commas.

For example, perf stat -e irq:irq_handler_entry,irq:irq_handler_exit ...  fails
because the tracepoint event parsing code doesn't recognize the comma separator
properly.

This patch corrects those issues.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Julia Lawall <julia@diku.dk>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reported-by: Michael Ellerman <michaele@au1.ibm.com>
LKML-Reference: <1291156021-17711-1-git-send-email-cjashfor@linux.vnet.ibm.com>
Signed-off-by: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf packaging: add memcpy to perf MANIFEST
Don Zickus [Tue, 30 Nov 2010 22:12:13 +0000 (17:12 -0500)]
perf packaging: add memcpy to perf MANIFEST

There seems to be a new dependency on arch/*/lib/memcpy*.S when compiling
the perf tool.  Make sure that file is included in the MANIFEST when
creating the tarball.

Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <1291155133-3499-2-git-send-email-dzickus@redhat.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf debug: Simplify trace_event
Arnaldo Carvalho de Melo [Tue, 30 Nov 2010 19:48:53 +0000 (17:48 -0200)]
perf debug: Simplify trace_event

No need to check that many times if debug_trace is on.

Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Allocate chunks of sample objects
Thomas Gleixner [Tue, 30 Nov 2010 17:49:55 +0000 (17:49 +0000)]
perf session: Allocate chunks of sample objects

The ordered sample code allocates singular reference objects struct
sample_queue which have 48byte size on 64bit and 20 bytes on 32bit. That's
silly. Allocate ~64k sized chunks and hand them out.

Performance gain: ~ 15%

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.398713983@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Cache sample objects
Thomas Gleixner [Tue, 30 Nov 2010 17:49:53 +0000 (17:49 +0000)]
perf session: Cache sample objects

When the sample queue is flushed we free the sample reference objects. Though
we need to malloc new objects when we process further. Stop the malloc/free
orgy and cache the already allocated object for resuage. Only allocate when
the cache is empty.

Performance gain: ~ 10%

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.338488630@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Keep file mmaped instead of malloc/memcpy
Thomas Gleixner [Tue, 30 Nov 2010 17:49:49 +0000 (17:49 +0000)]
perf session: Keep file mmaped instead of malloc/memcpy

Profiling perf with perf revealed that a large part of the processing time is
spent in malloc/memcpy/free in the sample ordering code. That code copies the
data from the mmap into malloc'ed memory. That's silly. We can keep the mmap
and just store the pointer in the queuing data structure. For 64 bit this is
not a problem as we map the whole file anyway. On 32bit we keep 8 maps around
and unmap the oldest before mmaping the next chunk of the file.

Performance gain: 2.95s -> 1.23s (Faktor 2.4)

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.278787719@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Use sensible mmap size
Thomas Gleixner [Tue, 30 Nov 2010 17:49:46 +0000 (17:49 +0000)]
perf session: Use sensible mmap size

On 64bit we can map the whole file in one go, on 32bit we can at least map
32MB and not map/unmap tiny chunks of the file.

Base the progress bar on 1/16 of the data size.

Preparatory patch to get rid of the malloc/memcpy/free of trace data.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.213687773@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Simplify termination checks
Thomas Gleixner [Tue, 30 Nov 2010 17:49:44 +0000 (17:49 +0000)]
perf session: Simplify termination checks

No need to check twice.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.152886642@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Move ui_progress_update in __perf_session__process_events()
Thomas Gleixner [Tue, 30 Nov 2010 17:49:41 +0000 (17:49 +0000)]
perf session: Move ui_progress_update in __perf_session__process_events()

The progress bar is changed when the file offset changes. This happens only
when the next mmap is done. No need to call ui_progress_update() for every
event.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.094836523@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Cleanup __perf_session__process_events()
Thomas Gleixner [Tue, 30 Nov 2010 17:49:38 +0000 (17:49 +0000)]
perf session: Cleanup __perf_session__process_events()

Replace the pseudo C++ self argument with session and give the mmap related
variables a sensible name. shift is a complete misnomer - it took me several
rounds of cursing to figure out that it's not a shift value.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163820.029687218@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Use appropriate pointer type instead of silly typecasting
Thomas Gleixner [Tue, 30 Nov 2010 17:49:35 +0000 (17:49 +0000)]
perf session: Use appropriate pointer type instead of silly typecasting

There is no reason to use a struct sample_event pointer in struct sample_queue
and type cast it when flushing the queue.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163819.969462809@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf session: Fix list sort algorithm
Thomas Gleixner [Tue, 30 Nov 2010 17:49:33 +0000 (17:49 +0000)]
perf session: Fix list sort algorithm

The homebrewn sort algorithm fails to sort in time order. One of the problem
spots is that it fails to deal with equal timestamps correctly.

My first gut reaction was to replace the fancy list with an rbtree, but the
performance is 3 times worse.

Rewrite it so it works.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20101130163819.908482530@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf events: Precalculate the header space for PERF_SAMPLE_ fields
Arnaldo Carvalho de Melo [Wed, 20 Oct 2010 14:50:11 +0000 (12:50 -0200)]
perf events: Precalculate the header space for PERF_SAMPLE_ fields

PERF_SAMPLE_{CALLCHAIN,RAW} have variable lenghts per sample, but the others
can be precalculated, reducing a bit the per sample cost.

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf tools: Fix lost and unknown events handling
Arnaldo Carvalho de Melo [Sat, 27 Nov 2010 04:41:01 +0000 (02:41 -0200)]
perf tools: Fix lost and unknown events handling

Fix it by explaining what can be happening and giving the number of processed
and lost events.

Also holler if unknown events were found, that can be due to processing a
perf.data file collected using a newer tool where newer events got added on
reporting using an older perf tool, that or a bug, so ask for a report to be
made.

Works on both --tui and --stdio.

Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf trace: Handle DT_UNKNOWN on filesystems that don't support d_type
Shawn Bohrer [Sun, 21 Nov 2010 16:09:39 +0000 (10:09 -0600)]
perf trace: Handle DT_UNKNOWN on filesystems that don't support d_type

Some filesystems like xfs and reiserfs will return DT_UNKNOWN for the
d_type.  Handle this case by calling stat() to determine the type.

Cc: Andreas Schwab <schwab@linux-m68k.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290355779-3276-1-git-send-email-sbohrer@rgmadvisors.com>
Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf symbols: Correct final kernel map guesses
Ian Munsie [Thu, 25 Nov 2010 04:12:53 +0000 (15:12 +1100)]
perf symbols: Correct final kernel map guesses

If a 32bit userspace perf is running on a 64bit kernel, the end of the final
map in the kernel would incorrectly be set to 2^32-1 rather than 2^64-1.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290658375-10342-1-git-send-email-imunsie@au1.ibm.com>
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf events: Default to using event__process_lost
Arnaldo Carvalho de Melo [Fri, 26 Nov 2010 20:31:54 +0000 (18:31 -0200)]
perf events: Default to using event__process_lost

Tool developers have to fill in a 'perf_event_ops' method table to
specify how to handle each event, so far the ones that were not
explicitely especified would get a stub that would just discard the
event.

Change that so that tool developers can get the lost event details and
the total number of such events at the end of 'perf report -D' output.

Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
CC: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf record: Add option to disable collecting build-ids
Arnaldo Carvalho de Melo [Fri, 26 Nov 2010 21:39:15 +0000 (19:39 -0200)]
perf record: Add option to disable collecting build-ids

Collecting build-ids for long running sessions may take a long time
because it needs to traverse the whole just collected perf.data stream
of events, marking the DSOs that had hits and then looking for the
.note.gnu.build-id ELF section.

For things like the 'trace' tool that records and right away consumes
the data on systems where its unlikely that the DSOs being monitored
will change while 'trace' runs, it is desirable to remove build id
collection, so add a -B/--no-buildid option to perf record to allow such
use case.

Longer term we'll avoid all this if we, at DSO load time, in the kernel,
take advantage of this slow code path to collect the build-id and stash
it somewhere, so that we can insert it in the PERF_RECORD_MMAP event.

Reported-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf, x86: P4 PMU - describe config format
Cyrill Gorcunov [Fri, 26 Nov 2010 11:32:09 +0000 (14:32 +0300)]
perf, x86: P4 PMU - describe config format

Add description of .config in a sake of RAW events.
At least this should bring some light to those who
will be reading this code.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Stephane Eranian <eranian@google.com>
Cc: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf, arch: Cleanup perf-pmu init vs lockup-detector
Peter Zijlstra [Thu, 25 Nov 2010 17:38:29 +0000 (18:38 +0100)]
perf, arch: Cleanup perf-pmu init vs lockup-detector

The perf hardware pmu got initialized at various points in the boot,
some before early_initcall() some after (notably arch_initcall).

The problem is that the NMI lockup detector is ran from early_initcall()
and expects the hardware pmu to be present.

Sanitize this by moving all architecture hardware pmu implementations to
initialize at early_initcall() and move the lockup detector to an explicit
initcall right after that.

Cc: paulus <paulus@samba.org>
Cc: davem <davem@davemloft.net>
Cc: Michael Cree <mcree@orcon.net.nz>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290707759.2145.119.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agox86: Set cpu masks before calling CPU_STARTING notifiers
Andi Kleen [Thu, 18 Nov 2010 10:47:31 +0000 (11:47 +0100)]
x86: Set cpu masks before calling CPU_STARTING notifiers

When booting up a CPU set the various topology masks before
calling the CPU_STARTING notifier. This way the notifier
can actually use the masks.

This is needed for a perf change.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290077254-12165-2-git-send-email-andi@firstfloor.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf: Ignore non-sampling overflows
Peter Zijlstra [Wed, 24 Nov 2010 17:55:29 +0000 (18:55 +0100)]
perf: Ignore non-sampling overflows

Some arch implementations call perf_event_overflow() by 'accident',
ignore this.

Reported-by: Francis Moreau <francis.moro@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf: Don't bother to init the hrtimer for no SW sampling counters
Franck Bui-Huu [Tue, 23 Nov 2010 15:21:45 +0000 (16:21 +0100)]
perf: Don't bother to init the hrtimer for no SW sampling counters

Signed-off-by: Franck Bui-Huu <fbuihuu@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290525705-6265-3-git-send-email-fbuihuu@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf: Limit event refresh to sampling event
Franck Bui-Huu [Tue, 23 Nov 2010 15:21:44 +0000 (16:21 +0100)]
perf: Limit event refresh to sampling event

Signed-off-by: Franck Bui-Huu <fbuihuu@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290525705-6265-2-git-send-email-fbuihuu@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf: Introduce is_sampling_event()
Franck Bui-Huu [Tue, 23 Nov 2010 15:21:43 +0000 (16:21 +0100)]
perf: Introduce is_sampling_event()

and use it when appropriate.

Signed-off-by: Franck Bui-Huu <fbuihuu@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290525705-6265-1-git-send-email-fbuihuu@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoscripts/tags.sh: Add magic for trace-events
Peter Zijlstra [Wed, 24 Nov 2010 09:43:55 +0000 (10:43 +0100)]
scripts/tags.sh: Add magic for trace-events

Make tags find the trace-event definitions

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: WANG Cong <xiyou.wangcong@gmail.com>
LKML-Reference: <1290591835.2072.438.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoMerge branch 'perf/urgent' into perf/core
Ingo Molnar [Fri, 26 Nov 2010 14:07:02 +0000 (15:07 +0100)]
Merge branch 'perf/urgent' into perf/core

Conflicts:
arch/x86/kernel/apic/hw_nmi.c

Merge reason: Resolve conflict, queue up dependent patch.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoMerge commit 'v2.6.37-rc3' into perf/core
Ingo Molnar [Fri, 26 Nov 2010 14:04:42 +0000 (15:04 +0100)]
Merge commit 'v2.6.37-rc3' into perf/core

Merge reason: Pick up latest fixes.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf: Fix the software context switch counter
Peter Zijlstra [Fri, 26 Nov 2010 12:49:04 +0000 (13:49 +0100)]
perf: Fix the software context switch counter

Stephane noticed that because the perf_sw_event() call is inside the
perf_event_task_sched_out() call it won't get called unless we
have a per-task counter.

Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf, x86: Fixup Kconfig deps
Peter Zijlstra [Tue, 16 Nov 2010 20:49:01 +0000 (21:49 +0100)]
perf, x86: Fixup Kconfig deps

This leads to a Kconfig dep inversion, x86 selects PERF_EVENT (due to
a hw_breakpoint dep) but doesn't unconditionally provide
HAVE_PERF_EVENT.

(This can cause build failures on M386/M486 kernel .config's.)

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101117222055.982965150@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agox86, perf, nmi: Disable perf if counters are not accessible
Don Zickus [Mon, 22 Nov 2010 21:55:23 +0000 (16:55 -0500)]
x86, perf, nmi: Disable perf if counters are not accessible

In a kvm virt guests, the perf counters are not emulated.  Instead they
return zero on a rdmsrl. The perf nmi handler uses the fact that crossing
a zero means the counter overflowed (for those counters that do not have
specific interrupt bits). Therefore on kvm guests, perf will swallow all
NMIs thinking the counters overflowed.

This causes problems for subsystems like kgdb which needs NMIs to do its
magic. This problem was discovered by running kgdb tests.

The solution is to write garbage into a perf counter during the
initialization and hopefully reading back the same number.  On kvm
guests, the value will be read back as zero and we disable perf as
a result.

Reported-by: Jason Wessel <jason.wessel@windriver.com>
Patch-inspired-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1290462923-30734-1-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf: Fix inherit vs. context rotation bug
Thomas Gleixner [Wed, 24 Nov 2010 09:05:55 +0000 (10:05 +0100)]
perf: Fix inherit vs. context rotation bug

It was found that sometimes children of tasks with inherited events had
one extra event. Eventually it turned out to be due to the list rotation
no being exclusive with the list iteration in the inheritance code.

Cure this by temporarily disabling the rotation while we inherit the events.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf bench: Add feature that measures the performance of the arch/x86/lib/memcpy_64...
Hitoshi Mitake [Thu, 25 Nov 2010 07:04:53 +0000 (16:04 +0900)]
perf bench: Add feature that measures the performance of the arch/x86/lib/memcpy_64.S memcpy routines via 'perf bench mem'

This patch ports arch/x86/lib/memcpy_64.S to perf bench mem
memcpy for benchmarking memcpy() in userland with tricky and
dirty way.

util/include/asm/cpufeature.h, util/include/asm/dwarf2.h, and
util/include/linux/linkage.h are mostly dummy files with small
wrappers, so that we are able to include memcpy_64.S
unmodified.

Signed-off-by: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: h.mitake@gmail.com
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: Ma Ling <ling.ma@intel.com>
Cc: Zhao Yakui <yakui.zhao@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Andi Kleen <andi@firstfloor.org>
LKML-Reference: <1290668693-27068-2-git-send-email-mitake@dcl.info.waseda.ac.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf bench: Print both of prefaulted and no prefaulted results by default
Hitoshi Mitake [Thu, 25 Nov 2010 07:04:52 +0000 (16:04 +0900)]
perf bench: Print both of prefaulted and no prefaulted results by default

After applying this patch, perf bench mem memcpy prints
both of prefualted and without prefaulted score of memcpy().

New options --no-prefault and --only-prefault are added
to print single result, mainly for scripting usage.

Usage example:

 | mitake@X201i:~/linux/.../tools/perf% ./perf bench mem memcpy -l 500MB
 | # Running mem/memcpy benchmark...
 | # Copying 500MB Bytes ...
 |
 |      634.969014 MB/Sec
 |        4.828062 GB/Sec (with prefault)
 | mitake@X201i:~/linux/.../tools/perf% ./perf bench mem memcpy -l 500MB --only-prefault
 | # Running mem/memcpy benchmark...
 | # Copying 500MB Bytes ...
 |
 |        4.705192 GB/Sec (with prefault)
 | mitake@X201i:~/linux/.../tools/perf% ./perf bench mem memcpy -l 500MB --no-prefault
 | # Running mem/memcpy benchmark...
 | # Copying 500MB Bytes ...
 |
 |      642.725568 MB/Sec

Signed-off-by: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: h.mitake@gmail.com
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: Ma Ling <ling.ma@intel.com>
Cc: Zhao Yakui <yakui.zhao@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Andi Kleen <andi@firstfloor.org>
LKML-Reference: <1290668693-27068-1-git-send-email-mitake@dcl.info.waseda.ac.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoperf symbols: Remove incorrect open-coded container_of()
Rabin Vincent [Tue, 23 Nov 2010 16:38:18 +0000 (22:08 +0530)]
perf symbols: Remove incorrect open-coded container_of()

At least on ARM, padding is inserted between rb_node and sym in struct
symbol_name_rb_node, causing "((void *)sym) - sizeof(struct rb_node)" to
point inside rb_node rather than to the symbol_name_rb_node.  Fix this
by converting the code to use container_of().

Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ming Lei <tom.leiming@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <20101123163106.GA25677@debian>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf record: Handle restrictive permissions in /proc/{kallsyms,modules}
Arnaldo Carvalho de Melo [Mon, 22 Nov 2010 16:01:55 +0000 (14:01 -0200)]
perf record: Handle restrictive permissions in /proc/{kallsyms,modules}

The 59365d1 commit, even being reverted by 33e0d57, showed a non robust
behavior in 'perf record': it really should just warn the user that some
functionality will not be available.

The new behavior then becomes:

[acme@felicio linux]$ ls -la /proc/{kallsyms,modules}
-r-------- 1 root root 0 Nov 22 12:19 /proc/kallsyms
-r-------- 1 root root 0 Nov 22 12:19 /proc/modules
[acme@felicio linux]$ perf record ls -R > /dev/null
Couldn't record kernel reference relocation symbol
Symbol resolution may be skewed if relocation was used (e.g. kexec).
Check /proc/kallsyms permission or run as root.
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.004 MB perf.data (~161 samples) ]
[acme@felicio linux]$ perf report --stdio
[kernel.kallsyms] with build id 77b05e00e64e4de1c9347d83879779b540d69f00 not found, continuing without symbols
# Events: 98  cycles
#
# Overhead  Command    Shared Object                Symbol
# ........  .......  ...............  ....................
#
    48.26%       ls  [kernel]         [k] ffffffff8102b92b
    22.49%       ls  libc-2.12.90.so  [.] __strlen_sse2
     8.35%       ls  libc-2.12.90.so  [.] __GI___strcoll_l
     8.17%       ls  ls               [.]            11580
     3.35%       ls  libc-2.12.90.so  [.] _IO_new_file_xsputn
     3.33%       ls  libc-2.12.90.so  [.] _int_malloc
     1.88%       ls  libc-2.12.90.so  [.] _int_free
     0.84%       ls  libc-2.12.90.so  [.] malloc_consolidate
     0.84%       ls  libc-2.12.90.so  [.] __readdir64
     0.83%       ls  ls               [.] strlen@plt
     0.83%       ls  libc-2.12.90.so  [.] __GI_fwrite_unlocked
     0.83%       ls  libc-2.12.90.so  [.] __memcpy_sse2

#
# (For a higher level overview, try: perf report --sort comm,dso)
#
[acme@felicio linux]$

It still has the build-ids for DSOs in the maps with hits:

[acme@felicio linux]$ perf buildid-list
77b05e00e64e4de1c9347d83879779b540d69f00 [kernel.kallsyms]
09c4a431a4a8b648fcfc2c2bdda70f56050ddff1 /bin/ls
af75ea9ad951d25e0f038901a11b3846dccb29a4 /lib64/libc-2.12.90.so
[acme@felicio linux]$

That can be used in another machine to resolve kernel symbols.

Cc: Eugene Teo <eugeneteo@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jesper Juhl <jj@chaosbits.net>
Cc: Marcus Meissner <meissner@suse.de>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoLinux 2.6.37-rc3
Linus Torvalds [Sun, 21 Nov 2010 23:18:56 +0000 (15:18 -0800)]
Linux 2.6.37-rc3

14 years agoperf stat: Change and clean up sys_perf_event_open error handling
Corey Ashford [Sat, 20 Nov 2010 01:37:24 +0000 (17:37 -0800)]
perf stat: Change and clean up sys_perf_event_open error handling

This patch makes several changes to "perf stat":

- "perf stat" will no longer go ahead and run the application when one or
more of the specified events could not be opened.
- Use error() and die() instead of pr_err() so that the output is more
consistent with "perf top" and "perf record".
- Handle permission errors in a more robust way, and in a similar way to
"perf record" and "perf top".

In addition, the sys_perf_event_open() error handling of "perf top" and "perf
record" is made more consistent and adds the following phrase when an event
doesn't open (with something ther than an access or permission error):

"/bin/dmesg may provide additional information."

This is added because kernel code doesn't have a good way of expressing
detailed errors to user space, so its only avenue is to use printk's.  However,
many users may not think of looking at dmesg to find out why an event is being
rejected.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <fweisbec@gmail.com>
Cc: Ian Munsie <ianmunsi@au1.ibm.com>
Cc: Michael Ellerman <michaele@au1.ibm.com>
LKML-Reference: <1290217044-26293-1-git-send-email-cjashfor@linux.vnet.ibm.com>
Signed-off-by: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Linus Torvalds [Sat, 20 Nov 2010 03:46:45 +0000 (19:46 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/tytso/ext4

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: Add EXT4_IOC_TRIM ioctl to handle batched discard
  fs: Do not dispatch FITRIM through separate super_operation
  ext4: ext4_fill_super shouldn't return 0 on corruption
  jbd2: fix /proc/fs/jbd2/<dev> when using an external journal
  ext4: missing unlock in ext4_clear_request_list()
  ext4: fix setting random pages PageUptodate

14 years agoext4: Add EXT4_IOC_TRIM ioctl to handle batched discard
Lukas Czerner [Sat, 20 Nov 2010 02:47:07 +0000 (21:47 -0500)]
ext4: Add EXT4_IOC_TRIM ioctl to handle batched discard

Filesystem independent ioctl was rejected as not common enough to be in
core vfs ioctl. Since we still need to access to this functionality this
commit adds ext4 specific ioctl EXT4_IOC_TRIM to dispatch
ext4_trim_fs().

It takes fstrim_range structure as an argument. fstrim_range is definec in
the include/linux/fs.h and its definition is as follows.

struct fstrim_range {
__u64 start;
__u64 len;
__u64 minlen;
}

start - first Byte to trim
len - number of Bytes to trim from start
minlen - minimum extent length to trim, free extents shorter than this
  number of Bytes will be ignored. This will be rounded up to fs
  block size.

After the FITRIM is done, the number of actually discarded Bytes is stored
in fstrim_range.len to give the user better insight on how much storage
space has been really released for wear-leveling.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
14 years agofs: Do not dispatch FITRIM through separate super_operation
Lukas Czerner [Sat, 20 Nov 2010 02:18:35 +0000 (21:18 -0500)]
fs: Do not dispatch FITRIM through separate super_operation

There was concern that FITRIM ioctl is not common enough to be included
in core vfs ioctl, as Christoph Hellwig pointed out there's no real point
in dispatching this out to a separate vector instead of just through
->ioctl.

So this commit removes ioctl_fstrim() from vfs ioctl and trim_fs
from super_operation structure.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
14 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Fri, 19 Nov 2010 23:32:22 +0000 (15:32 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  ceph: fix readdir EOVERFLOW on 32-bit archs
  ceph: fix frag offset for non-leftmost frags
  ceph: fix dangling pointer
  ceph: explicitly specify page alignment in network messages
  ceph: make page alignment explicit in osd interface
  ceph: fix comment, remove extraneous args
  ceph: fix update of ctime from MDS
  ceph: fix version check on racing inode updates
  ceph: fix uid/gid on resent mds requests
  ceph: fix rdcache_gen usage and invalidate
  ceph: re-request max_size if cap auth changes
  ceph: only let auth caps update max_size
  ceph: fix open for write on clustered mds
  ceph: fix bad pointer dereference in ceph_fill_trace
  ceph: fix small seq message skipping
  Revert "ceph: update issue_seq on cap grant"

14 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Fri, 19 Nov 2010 23:25:59 +0000 (15:25 -0800)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (31 commits)
  net: fix kernel-doc for sk_filter_rcu_release
  be2net: Fix to avoid firmware update when interface is not open.
  netfilter: fix IP_VS dependencies
  net: irda: irttp: sync error paths of data- and udata-requests
  ipv6: Expose reachable and retrans timer values as msecs
  ipv6: Expose IFLA_PROTINFO timer values in msecs instead of jiffies
  3c59x: fix build failure on !CONFIG_PCI
  ipg.c: remove id [SUNDANCE, 0x1021]
  net: caif: spi: fix potential NULL dereference
  ath9k_htc: Avoid setting QoS control for non-QoS frames
  net: zero kobject in rx_queue_release
  net: Fix duplicate volatile warning.
  MAINTAINERS: Add stmmac maintainer
  bonding: fix a race in IGMP handling
  cfg80211: fix can_beacon_sec_chan, reenable HT40
  gianfar: fix signedness issue
  net: bnx2x: fix error value sign
  8139cp: fix checksum broken
  r8169: fix checksum broken
  rds: Integer overflow in RDS cmsg handling
  ...

14 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Fri, 19 Nov 2010 19:59:49 +0000 (11:59 -0800)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  sata_via: apply magic FIFO fix to vt6420 too

14 years agoRevert "kernel: make /proc/kallsyms mode 400 to reduce ease of attacking"
Linus Torvalds [Fri, 19 Nov 2010 19:54:40 +0000 (11:54 -0800)]
Revert "kernel: make /proc/kallsyms mode 400 to reduce ease of attacking"

This reverts commit 59365d136d205cc20fe666ca7f89b1c5001b0d5a.

It turns out that this can break certain existing user land setups.
Quoth Sarah Sharp:

 "On Wednesday, I updated my branch to commit 460781b from linus' tree,
  and my box would not boot.  klogd segfaulted, which stalled the whole
  system.

  At first I thought it actually hung the box, but it continued booting
  after 5 minutes, and I was able to log in.  It dropped back to the
  text console instead of the graphical bootup display for that period
  of time.  dmesg surprisingly still works.  I've bisected the problem
  down to this commit (commit 59365d136d205cc20fe666ca7f89b1c5001b0d5a)

  The box is running klogd 1.5.5ubuntu3 (from Jaunty).  Yes, I know
  that's old.  I read the bit in the commit about changing the
  permissions of kallsyms after boot, but if I can't boot that doesn't
  help."

So let's just keep the old default, and encourage distributions to do
the "chmod -r /proc/kallsyms" in their bootup scripts.  This is not
worth a kernel option to change default behavior, since it's so easily
done in user space.

Reported-and-bisected-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Cc: Marcus Meissner <meissner@suse.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Eugene Teo <eugeneteo@kernel.org>
Cc: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoperf tools: Change my maintainer address
Arnaldo Carvalho de Melo [Fri, 19 Nov 2010 18:46:26 +0000 (16:46 -0200)]
perf tools: Change my maintainer address

Also remove old snail mail address from CREDITS, moved years ago.

LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoperf tools: Remove hardcoded include paths for elfutils
Robert Morell [Tue, 16 Nov 2010 22:16:33 +0000 (14:16 -0800)]
perf tools: Remove hardcoded include paths for elfutils

This change removes the use of hardcoded absolute "/usr/include/elfutils" paths
from the perf build.  The problem with hardcoded paths is that it prevents them
from being overridden by $prefix or by -I in CFLAGS (e.g., for cross-compiling
purposes).

Instead, just include the "elfutils/" subdirectory as a relative path when
files are needed from that directory.

Tested by building perf:
- Cross-compiled for ARM on x86_64
- Built natively on x86_64
- Built on x86_64 with /usr/include/elfutils moved to another location
  and manually included in CFLAGS

Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
LKML-Reference: <1289945793-31441-1-git-send-email-rmorell@nvidia.com>
Signed-off-by: Robert Morell <rmorell@nvidia.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 19 Nov 2010 18:31:04 +0000 (10:31 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: fix typo in keycode validation supporting large scancodes
  Input: aiptek - tighten up permissions on sysfs attributes
  Input: sysrq - pass along lone Alt + SysRq

14 years agoMerge branch 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ickle...
Linus Torvalds [Fri, 19 Nov 2010 18:28:13 +0000 (10:28 -0800)]
Merge branch 'drm-intel-fixes' of git://git./linux/kernel/git/ickle/drm-intel

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ickle/drm-intel:
  drm/i915: Disable FBC on Ironlake to save 1W
  drm/i915: Take advantage of auto-polling CRT hotplug detection on PCH hardware
  drm/i915/crt: Introduce struct intel_crt
  drm/i915: Do not hold mutex when faulting in user addresses
  drm: radeon: fix error value sign
  drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx
  drm/i915: Retire any pending operations on the old scanout when switching
  drm/i915: Fix I2C adapter registration

14 years agoMerge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Fri, 19 Nov 2010 18:27:57 +0000 (10:27 -0800)]
Merge branch 'drm-fixes' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6: (40 commits)
  drm/radeon/kms: i2c s/sprintf/snprintf/g for safety
  drm/radeon/kms: fix i2c pad masks on rs4xx
  drm/ttm: Fix up a theoretical deadlock
  drm/radeon/kms: fix tiling info on evergreen
  drm/radeon/kms: fix alignment when allocating buffers
  drm/vmwgfx: Fix up an error path during bo creation
  drm/radeon/kms: register an i2c adapter name for the dp aux bus
  drm/radeon/kms/atom: add proper external encoders support
  drm/radeon/kms/atom: cleanup and unify DVO handling
  drm/radeon/kms: properly power up/down the eDP panel as needed (v4)
  drm/radeon/kms/atom: set sane defaults in atombios_get_encoder_mode()
  drm/radeon/kms: turn the backlight off explicitly for dpms
  drm/radeon/kms: fix typo in r600 cs checker
  drm: radeon: fix error value sign
  drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx
  nouveau: Acknowledge HPD irq in handler, not bottom half
  drm/nouveau: Fix a few confusions between "chipset" and "card_type".
  drm/nouveau: don't expose backlight control when available through ACPI
  drm/nouveau/pm: improve memtiming mappings
  drm/nouveau: Make PCIE GART size depend on the available RAMIN space.
  ...

14 years agoperf stat: Add no-aggregation mode to -a
Stephane Eranian [Tue, 16 Nov 2010 09:05:01 +0000 (11:05 +0200)]
perf stat: Add no-aggregation mode to -a

This patch adds a new -A option to perf stat. If specified then perf stat does
not aggregate counts across all monitored CPUs in system-wide mode, i.e., when
using -a. This option is not supported in per-thread mode.

Being able to get a per-cpu breakdown is useful to detect imbalances between
CPUs when running a uniform workload than spans all monitored CPUs.

The second version corrects the missing cpumap[] support, so that it works when
the -C option is used.

The third version fixes a missing cpumap[] in print_counter() and removes a
stray patch in builtin-trace.c.

Examples on a 4-way system:

# perf stat -a   -e cycles,instructions -- sleep 1
 Performance counter stats for 'sleep 1':
         9592808135  cycles
         3490380006  instructions             #      0.364 IPC
        1.001584632  seconds time elapsed

# perf stat -a -A -e cycles,instructions -- sleep 1
 Performance counter stats for 'sleep 1':
CPU0            2398163767  cycles
CPU1            2398180817  cycles
CPU2            2398217115  cycles
CPU3            2398247483  cycles
CPU0             872282046  instructions             #      0.364 IPC
CPU1             873481776  instructions             #      0.364 IPC
CPU2             872638127  instructions             #      0.364 IPC
CPU3             872437789  instructions             #      0.364 IPC
        1.001556052  seconds time elapsed

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>
LKML-Reference: <4ce257b5.1e07e30a.7b6b.3aa9@mx.google.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
14 years agosata_via: apply magic FIFO fix to vt6420 too
Tejun Heo [Fri, 19 Nov 2010 14:29:19 +0000 (15:29 +0100)]
sata_via: apply magic FIFO fix to vt6420 too

vt6420 has the same FIFO overflow problem as vt6421 when combined with
certain devices.  This patch applies the magic fix to vt6420 too.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Martin Qvist <q@maq.dk>
Reported-by: Peter Zijlstra <peterz@infradead.org>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: stable@kernel.org
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
14 years agonet: fix kernel-doc for sk_filter_rcu_release
Randy Dunlap [Thu, 18 Nov 2010 13:02:37 +0000 (13:02 +0000)]
net: fix kernel-doc for sk_filter_rcu_release

Fix kernel-doc warning for sk_filter_rcu_release():

Warning(net/core/filter.c:586): missing initial short description on line:
 *  sk_filter_rcu_release: Release a socket filter by rcu_head

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Fix to avoid firmware update when interface is not open.
Sarveshwar Bandi [Thu, 18 Nov 2010 23:44:45 +0000 (23:44 +0000)]
be2net: Fix to avoid firmware update when interface is not open.

Since interrupts are enabled only when open is called on the interface,
Attempting a firmware update operation when interface is down could lead to
partial success or failure of operation. This fix fails the request if
netif_running is false.

Signed-off-by: Sarveshwar Bandi <Sarveshwar.Bandi@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoext4: ext4_fill_super shouldn't return 0 on corruption
Darrick J. Wong [Fri, 19 Nov 2010 14:56:44 +0000 (09:56 -0500)]
ext4: ext4_fill_super shouldn't return 0 on corruption

At the start of ext4_fill_super, ret is set to -EINVAL, and any failure path
out of that function returns ret.  However, the generic_check_addressable
clause sets ret = 0 (if it passes), which means that a subsequent failure (e.g.
a group checksum error) returns 0 even though the mount should fail.  This
causes vfs_kern_mount in turn to think that the mount succeeded, leading to an
oops.

A simple fix is to avoid using ret for the generic_check_addressable check,
which was last changed in commit 30ca22c70e3ef0a96ff84de69cd7e8561b416cb2.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
14 years agodrm/i915: Disable FBC on Ironlake to save 1W
Alex Shi [Fri, 19 Nov 2010 09:33:55 +0000 (09:33 +0000)]
drm/i915: Disable FBC on Ironlake to save 1W

Frame buffer compression is broken on Ironlake due to buggy hardware.
Currently it is disabled through chicken bits, but it still consumes
over 1W more than if we simply never attempt to enable the FBC code
paths.

Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
14 years agodrm/i915: Take advantage of auto-polling CRT hotplug detection on PCH hardware
Keith Packard [Tue, 16 Nov 2010 08:03:53 +0000 (16:03 +0800)]
drm/i915: Take advantage of auto-polling CRT hotplug detection on PCH hardware

Both IBX and CPT have an automatic hotplug detection mode which appears to work reliably enough
that we can dispense with the manual force hotplug trigger stuff. This means that
hotplug detection is as simple as reading the current hotplug register values.

The first time the hotplug detection is activated, the code synchronously waits for a hotplug
sequence in case the hardware hasn't bothered to do a detection cycle since being initialized.

Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
14 years agodrm/i915/crt: Introduce struct intel_crt
Chris Wilson [Tue, 16 Nov 2010 10:58:37 +0000 (10:58 +0000)]
drm/i915/crt: Introduce struct intel_crt

We will use this structure in future patches to store CRT specific
information on the encoder.

Split out and tweaked from a patch by Keith Packard.

Signed-off-by: Keith Packard <keithp@kithp.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
14 years agodrm/i915: Do not hold mutex when faulting in user addresses
Chris Wilson [Wed, 17 Nov 2010 09:10:42 +0000 (09:10 +0000)]
drm/i915: Do not hold mutex when faulting in user addresses

Linus Torvalds found that it was rather trivial to trigger a system
freeze:

  In fact, with lockdep, I don't even need to do the sysrq-d thing: it
  shows the bug as it happens. It's the X server taking the same lock
  recursively.

  Here's the problem:

    =============================================
    [ INFO: possible recursive locking detected ]
    2.6.37-rc2-00012-gbdbd01a #7
    ---------------------------------------------
    Xorg/2816 is trying to acquire lock:
     (&dev->struct_mutex){+.+.+.}, at: [<ffffffff812c626c>] i915_gem_fault+0x50/0x17e

    but task is already holding lock:
     (&dev->struct_mutex){+.+.+.}, at: [<ffffffff812c403b>] i915_mutex_lock_interruptible+0x28/0x4a

    other info that might help us debug this:
    2 locks held by Xorg/2816:
     #0:  (&dev->struct_mutex){+.+.+.}, at: [<ffffffff812c403b>] i915_mutex_lock_interruptible+0x28/0x4a
     #1:  (&mm->mmap_sem){++++++}, at: [<ffffffff81022d4f>] page_fault+0x156/0x37b

This recursion was introduced by rearranging the locking to avoid the
double locking on the fast path (4f27b5d and fbd5a26d) and the
introduction of the prefault to encourage the fast paths (b5e4f2b). In
order to undo the problem, we rearrange the code to perform the access
validation upfront, attempt to prefault and then fight for control of the
mutex.  the best case scenario where the mutex is uncontended the
prefaulting is not wasted.

Reported-and-tested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
14 years agodrm/radeon/kms: i2c s/sprintf/snprintf/g for safety
Alex Deucher [Thu, 18 Nov 2010 16:37:18 +0000 (11:37 -0500)]
drm/radeon/kms: i2c s/sprintf/snprintf/g for safety

As per advice from Jean Delvare.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
14 years agodrm/radeon/kms: fix i2c pad masks on rs4xx
Alex Deucher [Thu, 18 Nov 2010 22:18:08 +0000 (17:18 -0500)]
drm/radeon/kms: fix i2c pad masks on rs4xx

These got lost in the last i2c cleanup.  Fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=23222

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
14 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Thu, 18 Nov 2010 23:01:43 +0000 (15:01 -0800)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  [libata] remove SCSI host lock and serial number usage from ata_scsi_queuecmd

14 years agoMAINTAINERS: update documentation entries
Randy Dunlap [Thu, 18 Nov 2010 20:27:37 +0000 (12:27 -0800)]
MAINTAINERS: update documentation entries

Update kernel-doc and Documentation maintainers info.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation/development-process: more staging info
Andres Salomon [Thu, 18 Nov 2010 20:27:36 +0000 (12:27 -0800)]
Documentation/development-process: more staging info

Document things that I would've liked to have known when submitting a driver
to gregkh for staging.

Signed-off-by: Andres Salomon <dilinger@queued.net>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation/development-process: use -next trees instead of staging
Andres Salomon [Thu, 18 Nov 2010 20:27:35 +0000 (12:27 -0800)]
Documentation/development-process: use -next trees instead of staging

This is confusing, as we have "staging" trees for drivers/staging.  Call
them -next trees.

Signed-off-by: Andres Salomon <dilinger@queued.net>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation: change email address for Hans Koch
Hans J. Koch [Thu, 18 Nov 2010 20:27:34 +0000 (12:27 -0800)]
Documentation: change email address for Hans Koch

My old mail address doesn't exist anymore. This changes all occurrences
to my new address.

Signed-off-by: Hans J. Koch <hjk@hansjkoch.de>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation/gpio.txt: explain poll/select usage
Bernhard Walle [Thu, 18 Nov 2010 20:27:33 +0000 (12:27 -0800)]
Documentation/gpio.txt: explain poll/select usage

Add a bit more information how to use poll(2) on GPIO value files
correctly. For me it was not clear that I need to poll(2) for
POLLPRI|POLLERR or select(2) for exceptfds.

Signed-off-by: Bernhard Walle <walle@corscience.de>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation: make configfs example code simpler, clearer
Dan Carpenter [Thu, 18 Nov 2010 20:27:32 +0000 (12:27 -0800)]
Documentation: make configfs example code simpler, clearer

If "p" is NULL then it will cause an oops when we pass it to
simple_strtoul().  In this case "p" can not be NULL so I removed the
check.  I also changed the check a little to make it more explicit that
we are testing whether p points to the NUL char.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokernel-doc: escape xml for structs
Randy Dunlap [Thu, 18 Nov 2010 20:27:31 +0000 (12:27 -0800)]
kernel-doc: escape xml for structs

scripts/kernel-doc was leaving unescaped '<', '>', and '&' in
generated xml output for structs.  This causes xml parser errors.
Convert these characters to "&lt;", "&gt;", and "&amp;" as needed
to prevent errors.

Most of the conversion was already done; complete it just before
output.

Documentation/DocBook/device-drivers.xml:41883: parser error : StartTag: invalid element name
#define INPUT_KEYMAP_BY_INDEX (1 << 0)

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agonetfilter: fix IP_VS dependencies
Patrick McHardy [Thu, 18 Nov 2010 08:20:57 +0000 (08:20 +0000)]
netfilter: fix IP_VS dependencies

When NF_CONNTRACK is enabled, IP_VS uses conntrack symbols.
Therefore IP_VS can't be linked statically when conntrack
is built modular.

Reported-by: Justin P. Mattock <justinmattock@gmail.com>
Tested-by: Justin P. Mattock <justinmattock@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: irda: irttp: sync error paths of data- and udata-requests
Wolfram Sang [Tue, 16 Nov 2010 09:40:02 +0000 (09:40 +0000)]
net: irda: irttp: sync error paths of data- and udata-requests

irttp_data_request() returns meaningful errorcodes, while irttp_udata_request()
just returns -1 in similar situations. Sync the two and the loglevels of the
accompanying output.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: Samuel Ortiz <sameo@linux.intel.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoipv6: Expose reachable and retrans timer values as msecs
Thomas Graf [Wed, 17 Nov 2010 04:12:02 +0000 (04:12 +0000)]
ipv6: Expose reachable and retrans timer values as msecs

Expose reachable and retrans timer values in msecs instead of jiffies.
Both timer values are already exposed as msecs in the neighbour table
netlink interface.

The creation timestamp format with increased precision is kept but
cleaned up.

Signed-off-by: Thomas Graf <tgraf@infradead.org>
Cc: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Thu, 18 Nov 2010 19:56:09 +0000 (11:56 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

14 years agoFix build failure at cx25821-video driver
Youquan Song [Thu, 18 Nov 2010 23:28:20 +0000 (18:28 -0500)]
Fix build failure at cx25821-video driver

Kernel build fail for cx25821-video has depends on smp_lock.h header
file, but the dependency is removed in recent commit 451a3c24b013.

Signed-off-by: Youquan Song <youquan.song@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoMerge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic...
Ingo Molnar [Thu, 18 Nov 2010 19:07:12 +0000 (20:07 +0100)]
Merge branch 'perf/core' of git://git./linux/kernel/git/frederic/random-tracing into perf/core

14 years agoipv6: Expose IFLA_PROTINFO timer values in msecs instead of jiffies
Thomas Graf [Wed, 17 Nov 2010 01:44:24 +0000 (01:44 +0000)]
ipv6: Expose IFLA_PROTINFO timer values in msecs instead of jiffies

IFLA_PROTINFO exposes timer related per device settings in jiffies.
Change it to expose these values in msecs like the sysctl interface
does.

I did not find any users of IFLA_PROTINFO which rely on any of these
values and even if there are, they are likely already broken because
there is no way for them to reliably convert such a value to another
time format.

Signed-off-by: Thomas Graf <tgraf@infradead.org>
Cc: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>