openwrt/staging/blogic.git
13 years agoperf, x86: Fix event scheduler for constraints with overlapping counters
Robert Richter [Fri, 18 Nov 2011 11:35:22 +0000 (12:35 +0100)]
perf, x86: Fix event scheduler for constraints with overlapping counters

The current x86 event scheduler fails to resolve scheduling problems
of certain combinations of events and constraints. This happens if the
counter mask of such an event is not a subset of any other counter
mask of a constraint with an equal or higher weight, e.g. constraints
of the AMD family 15h pmu:

                        counter mask    weight

 amd_f15_PMC30          0x09            2  <--- overlapping counters
 amd_f15_PMC20          0x07            3
 amd_f15_PMC53          0x38            3

The scheduler does not find then an existing solution. Here is an
example:

 event code     counter         failure         possible solution

 0x02E          PMC[3,0]        0               3
 0x043          PMC[2:0]        1               0
 0x045          PMC[2:0]        2               1
 0x046          PMC[2:0]        FAIL            2

The event scheduler may not select the correct counter in the first
cycle because it needs to know which subsequent events will be
scheduled. It may fail to schedule the events then.

To solve this, we now save the scheduler state of events with
overlapping counter counstraints.  If we fail to schedule the events
we rollback to those states and try to use another free counter.

Constraints with overlapping counters are marked with a new introduced
overlap flag. We set the overlap flag for such constraints to give the
scheduler a hint which events to select for counter rescheduling. The
EVENT_CONSTRAINT_OVERLAP() macro can be used for this.

Care must be taken as the rescheduling algorithm is O(n!) which will
increase scheduling cycles for an over-commited system dramatically.
The number of such EVENT_CONSTRAINT_OVERLAP() macros and its counter
masks must be kept at a minimum. Thus, the current stack is limited to
2 states to limit the number of loops the algorithm takes in the worst
case.

On systems with no overlapping-counter constraints, this
implementation does not increase the loop count compared to the
previous algorithm.

V2:
* Renamed redo -> overlap.
* Reimplementation using perf scheduling helper functions.

V3:
* Added WARN_ON_ONCE() if out of save states.
* Changed function interface of perf_sched_restore_state() to use bool
  as return value.

Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1321616122-1533-3-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf, x86: Implement event scheduler helper functions
Robert Richter [Fri, 18 Nov 2011 11:35:21 +0000 (12:35 +0100)]
perf, x86: Implement event scheduler helper functions

This patch introduces x86 perf scheduler code helper functions. We
need this to later add more complex functionality to support
overlapping counter constraints (next patch).

The algorithm is modified so that the range of weight values is now
generated from the constraints. There shouldn't be other functional
changes.

With the helper functions the scheduler is controlled. There are
functions to initialize, traverse the event list, find unused counters
etc. The scheduler keeps its own state.

V3:
* Added macro for_each_set_bit_cont().
* Changed functions interfaces of perf_sched_find_counter() and
  perf_sched_next_event() to use bool as return value.
* Added some comments to make code better understandable.

V4:
* Fix broken event assignment if weight of the first event is not
  wmin (perf_sched_init()).

Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1321616122-1533-2-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf: Avoid a useless pmu_disable() in the perf-tick
Peter Zijlstra [Wed, 16 Nov 2011 13:38:16 +0000 (14:38 +0100)]
perf: Avoid a useless pmu_disable() in the perf-tick

Gleb writes:

 > Currently pmu is disabled and re-enabled on each timer interrupt even
 > when no rotation or frequency adjustment is needed. On Intel CPU this
 > results in two writes into PERF_GLOBAL_CTRL MSR per tick. On bare metal
 > it does not cause significant slowdown, but when running perf in a virtual
 > machine it leads to 20% slowdown on my machine.

Cure this by keeping a perf_event_context::nr_freq counter that counts the
number of active events that require frequency adjustments and use this in a
similar fashion to the already existing nr_events != nr_active test in
perf_rotate_context().

By being able to exclude both rotation and frequency adjustments a-priory for
the common case we can avoid the otherwise superfluous PMU disable.

Suggested-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-515yhoatehd3gza7we9fapaa@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge branch 'perf/urgent' into perf/core
Ingo Molnar [Tue, 6 Dec 2011 05:42:35 +0000 (06:42 +0100)]
Merge branch 'perf/urgent' into perf/core

Merge reason: Add these cherry-picked commits so that future changes
              on perf/core don't conflict.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoftrace: Fix hash record accounting bug
Steven Rostedt [Sat, 5 Nov 2011 00:32:39 +0000 (20:32 -0400)]
ftrace: Fix hash record accounting bug

If the set_ftrace_filter is cleared by writing just whitespace to
it, then the filter hash refcounts will be decremented but not
updated. This causes two bugs:

1) No functions will be enabled for tracing when they all should be

2) If the users clears the set_ftrace_filter twice, it will crash ftrace:

------------[ cut here ]------------
WARNING: at /home/rostedt/work/git/linux-trace.git/kernel/trace/ftrace.c:1384 __ftrace_hash_rec_update.part.27+0x157/0x1a7()
Modules linked in:
Pid: 2330, comm: bash Not tainted 3.1.0-test+ #32
Call Trace:
 [<ffffffff81051828>] warn_slowpath_common+0x83/0x9b
 [<ffffffff8105185a>] warn_slowpath_null+0x1a/0x1c
 [<ffffffff810ba362>] __ftrace_hash_rec_update.part.27+0x157/0x1a7
 [<ffffffff810ba6e8>] ? ftrace_regex_release+0xa7/0x10f
 [<ffffffff8111bdfe>] ? kfree+0xe5/0x115
 [<ffffffff810ba51e>] ftrace_hash_move+0x2e/0x151
 [<ffffffff810ba6fb>] ftrace_regex_release+0xba/0x10f
 [<ffffffff8112e49a>] fput+0xfd/0x1c2
 [<ffffffff8112b54c>] filp_close+0x6d/0x78
 [<ffffffff8113a92d>] sys_dup3+0x197/0x1c1
 [<ffffffff8113a9a6>] sys_dup2+0x4f/0x54
 [<ffffffff8150cac2>] system_call_fastpath+0x16/0x1b
---[ end trace 77a3a7ee73794a02 ]---

Link: http://lkml.kernel.org/r/20111101141420.GA4918@debian
Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agoperf: Fix parsing of __print_flags() in TP_printk()
Steven Rostedt [Fri, 4 Nov 2011 20:32:25 +0000 (16:32 -0400)]
perf: Fix parsing of __print_flags() in TP_printk()

A update is made to the sched:sched_switch event that adds some
logic to the first parameter of the __print_flags() that shows the
state of tasks. This change cause perf to fail parsing the flags.

A simple fix is needed to have the parser be able to process ops
within the argument.

Cc: stable@vger.kernel.org
Reported-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agojump_label: jump_label_inc may return before the code is patched
Gleb Natapov [Tue, 18 Oct 2011 17:55:51 +0000 (19:55 +0200)]
jump_label: jump_label_inc may return before the code is patched

If cpu A calls jump_label_inc() just after atomic_add_return() is
called by cpu B, atomic_inc_not_zero() will return value greater then
zero and jump_label_inc() will return to a caller before jump_label_update()
finishes its job on cpu B.

Link: http://lkml.kernel.org/r/20111018175551.GH17571@redhat.com
Cc: stable@vger.kernel.org
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Jason Baron <jbaron@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agoftrace: Remove force undef config value left for testing
Steven Rostedt [Fri, 4 Nov 2011 14:45:23 +0000 (10:45 -0400)]
ftrace: Remove force undef config value left for testing

A forced undef of a config value was used for testing and was
accidently left in during the final commit. This causes x86 to
run slower than needed while running function tracing as well
as causes the function graph selftest to fail when DYNMAIC_FTRACE
is not set. This is because the code in MCOUNT expects the ftrace
code to be processed with the config value set that happened to
be forced not set.

The forced config option was left in by:
    commit 6331c28c962561aee59e5a493b7556a4bb585957
    ftrace: Fix dynamic selftest failure on some archs

Link: http://lkml.kernel.org/r/20111102150255.GA6973@debian
Cc: stable@vger.kernel.org
Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: Restore system filter behavior
Li Zefan [Tue, 1 Nov 2011 01:09:35 +0000 (09:09 +0800)]
tracing: Restore system filter behavior

Though not all events have field 'prev_pid', it was allowed to do this:

  # echo 'prev_pid == 100' > events/sched/filter

but commit 75b8e98263fdb0bfbdeba60d4db463259f1fe8a2 (tracing/filter: Swap
entire filter of events) broke it without any reason.

Link: http://lkml.kernel.org/r/4EAF46CF.8040408@cn.fujitsu.com
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: fix event_subsystem ref counting
Ilya Dryomov [Mon, 31 Oct 2011 09:07:42 +0000 (11:07 +0200)]
tracing: fix event_subsystem ref counting

Fix a bug introduced by e9dbfae5, which prevents event_subsystem from
ever being released.

Ref_count was added to keep track of subsystem users, not for counting
events.  Subsystem is created with ref_count = 1, so there is no need to
increment it for every event, we have nr_events for that.  Fix this by
touching ref_count only when we actually have a new user -
subsystem_open().

Cc: stable@vger.kernel.org
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Link: http://lkml.kernel.org/r/1320052062-7846-1-git-send-email-idryomov@gmail.com
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agox86/tools: Add decoded instruction dump mode
Masami Hiramatsu [Mon, 5 Dec 2011 12:06:03 +0000 (21:06 +0900)]
x86/tools: Add decoded instruction dump mode

Add instruction dump mode to insn_sanity tool for
checking decoder really decoded instructions.

This mode is enabled when passing double -v (-vv) to
insn_sanity. It is useful for who wants to check whether
the decoder can decode some instructions correctly.
e.g.
 $ echo 0f 73 10 11 | ./insn_sanity -y -vv -i -
 Instruction = {
        .prefixes = {
                .value = 0, bytes[] = {0, 0, 0, 0},
                .got = 1, .nbytes = 0},
        .rex_prefix = {
                .value = 0, bytes[] = {0, 0, 0, 0},
                .got = 1, .nbytes = 0},
        .vex_prefix = {
                .value = 0, bytes[] = {0, 0, 0, 0},
                .got = 1, .nbytes = 0},
        .opcode = {
                .value = 29455, bytes[] = {f, 73, 0, 0},
                .got = 1, .nbytes = 2},
        .modrm = {
                .value = 16, bytes[] = {10, 0, 0, 0},
                .got = 1, .nbytes = 1},
        .sib = {
                .value = 0, bytes[] = {0, 0, 0, 0},
                .got = 1, .nbytes = 0},
        .displacement = {
                .value = 0, bytes[] = {0, 0, 0, 0},
                .got = 1, .nbytes = 0},
        .immediate1 = {
                .value = 17, bytes[] = {11, 0, 0, 0},
                .got = 1, .nbytes = 1},
        .immediate2 = {
                .value = 0, bytes[] = {0, 0, 0, 0},
                .got = 0, .nbytes = 0},
        .attr = 44800, .opnd_bytes = 4, .addr_bytes = 8,
        .length = 4, .x86_64 = 1, .kaddr = 0x7fff0f7d9430}
 Success: decoded and checked 1 given instructions with 0 errors (seed:0x0)

Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20111205120603.15475.91192.stgit@cloud
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: Update instruction decoder to support new AVX formats
Masami Hiramatsu [Mon, 5 Dec 2011 12:05:57 +0000 (21:05 +0900)]
x86: Update instruction decoder to support new AVX formats

Since new Intel software developers manual introduces
new format for AVX instruction set (including AVX2),
it is important to update x86-opcode-map.txt to fit
those changes.

Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20111205120557.15475.13236.stgit@cloud
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86/tools: Fix insn_sanity message outputs
Masami Hiramatsu [Mon, 5 Dec 2011 12:05:50 +0000 (21:05 +0900)]
x86/tools: Fix insn_sanity message outputs

Fix x86 instruction decoder test to dump all error messages
to stderr and others to stdout.

Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20111205120550.15475.70149.stgit@cloud
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86/tools: Fix instruction decoder message output
Masami Hiramatsu [Mon, 5 Dec 2011 12:05:45 +0000 (21:05 +0900)]
x86/tools: Fix instruction decoder message output

Fix instruction decoder test (insn_sanity), so that
it doesn't show both info and error messages twice on
same instruction. (In that case, show only error message)

Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20111205120545.15475.7928.stgit@cloud
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: Fix instruction decoder to handle grouped AVX instructions
Masami Hiramatsu [Mon, 5 Dec 2011 12:05:39 +0000 (21:05 +0900)]
x86: Fix instruction decoder to handle grouped AVX instructions

For reducing memory usage of attribute table, x86 instruction
decoder puts "Group" attribute only on "no-last-prefix"
attribute table (same as vex_p == 0 case).

Thus, the decoder should look no-last-prefix table first, and
then only if it is not a group, move on to "with-last-prefix"
table (vex_p != 0).

However, current implementation, inat_get_avx_attribute()
looks with-last-prefix directly. So, when decoding
a grouped AVX instruction, the decoder fails to find correct
group because there is no "Group" attribute on the table.
This ends up with the mis-decoding of instructions, as Ingo
reported in http://thread.gmane.org/gmane.linux.kernel/1214103

This patch fixes it to check no-last-prefix table first
even if that is an AVX instruction, and get an attribute from
"with last-prefix" table only if that is not a group.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20111205120539.15475.91428.stgit@cloud
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86/tools: Fix Makefile to build all test tools
Masami Hiramatsu [Mon, 5 Dec 2011 12:05:33 +0000 (21:05 +0900)]
x86/tools: Fix Makefile to build all test tools

Fix arch/x86/tools/Makefile to compile both test tools
correctly. This bug leads build error.

Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20111205120533.15475.62047.stgit@cloud
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge branch 'tip/perf/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/roste...
Ingo Molnar [Mon, 5 Dec 2011 13:34:00 +0000 (14:34 +0100)]
Merge branch 'tip/perf/urgent' of git://git./linux/kernel/git/rostedt/linux-trace into perf/urgent

13 years agoMerge branch 'perf/urgent' of git://github.com/acmel/linux into perf/urgent
Ingo Molnar [Mon, 5 Dec 2011 09:32:39 +0000 (10:32 +0100)]
Merge branch 'perf/urgent' of git://github.com/acmel/linux into perf/urgent

13 years agoperf: Fix loss of notification with multi-event
Peter Zijlstra [Sat, 26 Nov 2011 01:47:31 +0000 (02:47 +0100)]
perf: Fix loss of notification with multi-event

When you do:
        $ perf record -e cycles,cycles,cycles noploop 10

You expect about 10,000 samples for each event, i.e., 10s at
1000samples/sec. However, this is not what's happening. You
get much fewer samples, maybe 3700 samples/event:

$ perf report -D | tail -15
Aggregated stats:
           TOTAL events:      10998
            MMAP events:         66
            COMM events:          2
          SAMPLE events:      10930
cycles stats:
           TOTAL events:       3644
          SAMPLE events:       3644
cycles stats:
           TOTAL events:       3642
          SAMPLE events:       3642
cycles stats:
           TOTAL events:       3644
          SAMPLE events:       3644

On a Intel Nehalem or even AMD64, there are 4 counters capable
of measuring cycles, so there is plenty of space to measure those
events without multiplexing (even with the NMI watchdog active).
And even with multiplexing, we'd expect roughly the same number
of samples per event.

The root of the problem was that when the event that caused the buffer
to become full was not the first event passed on the cmdline, the user
notification would get lost. The notification was sent to the file
descriptor of the overflowed event but the perf tool was not polling
on it.  The perf tool aggregates all samples into a single buffer,
i.e., the buffer of the first event. Consequently, it assumes
notifications for any event will come via that descriptor.

The seemingly straight forward solution of moving the waitq into the
ringbuffer object doesn't work because of life-time issues. One could
perf_event_set_output() on a fd that you're also blocking on and cause
the old rb object to be freed while its waitq would still be
referenced by the blocked thread -> FAIL.

Therefore link all events to the ringbuffer and broadcast the wakeup
from the ringbuffer object to all possible events that could be waited
upon. This is rather ugly, and we're open to better solutions but it
works for now.

Reported-by: Stephane Eranian <eranian@google.com>
Finished-by: Stephane Eranian <eranian@google.com>
Reviewed-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20111126014731.GA7030@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf, x86: Force IBS LVT offset assignment for family 10h
Robert Richter [Tue, 8 Nov 2011 18:20:44 +0000 (19:20 +0100)]
perf, x86: Force IBS LVT offset assignment for family 10h

On AMD family 10h we see firmware bug messages like the following:

 [Firmware Bug]: cpu 6, try to use APIC500 (LVT offset 0) for vector 0x10400, but the register is already in use for vector 0xf9 on another cpu
 [Firmware Bug]: cpu 6, IBS interrupt offset 0 not available (MSRC001103A=0x0000000000000100)
 [Firmware Bug]: using offset 1 for IBS interrupts
 [Firmware Bug]: workaround enabled for IBS LVT offset
 perf: AMD IBS detected (0x00000007)

We always see this, since the offsets are not assigned by the BIOS for
this family. Force LVT offset assignment in this case. If the OS
assignment fails, fallback to BIOS settings and try to setup this.

The fallback to BIOS settings weakens the family check since
force_ibs_eilvt_setup() may fail e.g. in case of virtual machines.
But setup may still succeed if BIOS offsets are correct.

Other families don't have a workaround implemented that assigns LVT
offsets. It's ok, to drop calling force_ibs_eilvt_setup() for that
families.

With the patch the [Firmware Bug] messages vanish. We see now:

 IBS: LVT offset 1 assigned
 perf: AMD IBS detected (0x00000007)

Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20111109162225.GO12451@erda.amd.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf, x86: Disable PEBS on SandyBridge chips
Peter Zijlstra [Tue, 15 Nov 2011 09:51:15 +0000 (10:51 +0100)]
perf, x86: Disable PEBS on SandyBridge chips

Cc: Stephane Eranian <eranian@google.com>
Cc: stable@kernel.org
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf test: Soft errors shouldn't stop the "Validate PERF_RECORD_" test
Arnaldo Carvalho de Melo [Fri, 2 Dec 2011 15:53:04 +0000 (13:53 -0200)]
perf test: Soft errors shouldn't stop the "Validate PERF_RECORD_" test

For errors that don't preclude checking for further errors, aka "soft"
errors, just  continue testing for other errors.

Better coverage in verbose mode.

Suggested-by: David Ahern <dsahern@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-jafcokbj26m845dsgm2hx6az@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf test: Validate PERF_RECORD_ events and perf_sample fields
Arnaldo Carvalho de Melo [Fri, 2 Dec 2011 13:13:50 +0000 (11:13 -0200)]
perf test: Validate PERF_RECORD_ events and perf_sample fields

This new test will validate these new routines extracted from 'perf
record':

 - perf_evlist__config_attrs
 - perf_evlist__prepare_workload
 - perf_evlist__start_workload

In addition to several other perf_evlist methods.

It consists of starting a simple workload, setting up just one event to
monitor ("cycles") requesting that several PERF_SAMPLE_ fields be
present in all events.

It then will check that the expected PERF_RECORD_ events are produced
and will sanity check all its fields.

Some checks performed:

. PERF_SAMPLE_TIME monotonically increases.

. PERF_SAMPLE_CPU is the one requested with sched_setaffinity

. PERF_SAMPLE_TID and PERF_SAMPLE_PID matches the one we forked
  in perf_evlist__prepare_workload and that is stored in
  evlist->workload.pid

. For the events where these fields are also present in its
  pre-sample_id_all fields (e.g. event->mmap.pid), that they are what
  is expected too.

. That we get a bunch of mmaps:

  PATH/libcSUFFIX
  PATH/ldSUFFIX
  [vdso]
  PATH/sleep

Example:

  [root@emilia ~]# taskset -c 3,4 perf test -v1 perf_sample
   6: Validate PERF_RECORD_* events & perf_sample fields:
  --- start ---
  7159480799825 3 PERF_RECORD_SAMPLE
  7159480805584 3 PERF_RECORD_SAMPLE
  7159480807814 3 PERF_RECORD_SAMPLE
  7159480810430 3 PERF_RECORD_SAMPLE
  7159480861511 3 PERF_RECORD_MMAP 8086/8086: [0x7fffffffd000(0x2000) @ 0x7fffffffd000]: //anon
  7159481052516 3 PERF_RECORD_COMM: sleep:8086
  7159481070188 3 PERF_RECORD_MMAP 8086/8086: [0x400000(0x6000) @ 0]: /bin/sleep
  7159481077104 3 PERF_RECORD_MMAP 8086/8086: [0x3d06400000(0x221000) @ 0]: /lib64/ld-2.12.so
  7159481092912 3 PERF_RECORD_MMAP 8086/8086: [0x7fff1adff000(0x1000) @ 0x7fff1adff000]: [vdso]
  7159481196779 3 PERF_RECORD_MMAP 8086/8086: [0x3d06800000(0x37f000) @ 0]: /lib64/libc-2.12.so
  7160481558435 3 PERF_RECORD_EXIT(8086:8086):(8086:8086)
  ---- end ----
  Validate PERF_RECORD_* events & perf_sample fields: Ok
  [root@emilia ~]#

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-svag18v2z4idas0dyz3umjpq@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf event: Introduce perf_event__fprintf
Arnaldo Carvalho de Melo [Fri, 2 Dec 2011 13:06:37 +0000 (11:06 -0200)]
perf event: Introduce perf_event__fprintf

So that tools like 'perf test' can print the events when in verbose
mode, for instance.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xnovdqfi25nc48gy6604k7yp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agotrace_events_filter: Use rcu_assign_pointer() when setting ftrace_event_call->filter
Tejun Heo [Wed, 23 Nov 2011 16:49:49 +0000 (08:49 -0800)]
trace_events_filter: Use rcu_assign_pointer() when setting ftrace_event_call->filter

ftrace_event_call->filter is sched RCU protected but didn't use
rcu_assign_pointer().  Use it.

TODO: Add proper __rcu annotation to call->filter and all its users.

-v2: Use RCU_INIT_POINTER() for %NULL clearing as suggested by Eric.

Link: http://lkml.kernel.org/r/20111123164949.GA29639@google.com
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: stable@kernel.org # (2.6.39+)
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agoperf test: Allow running just a subset of the available tests
Arnaldo Carvalho de Melo [Tue, 29 Nov 2011 14:52:07 +0000 (12:52 -0200)]
perf test: Allow running just a subset of the available tests

To obtain a list of available tests:

[root@emilia linux]# perf test list
 1: vmlinux symtab matches kallsyms
 2: detect open syscall event
 3: detect open syscall event on all cpus
 4: read samples using the mmap interface
 5: parse events tests
[root@emilia linux]#

To list just a subset:

[root@emilia linux]# perf test list syscall
 2: detect open syscall event
 3: detect open syscall event on all cpus
[root@emilia linux]#

To run a subset:

[root@emilia linux]# perf test detect
 2: detect open syscall event: Ok
 3: detect open syscall event on all cpus: Ok
[root@emilia linux]#

Specific tests can be chosen by number:

[root@emilia linux]# perf test 1 3 parse
 1: vmlinux symtab matches kallsyms: Ok
 3: detect open syscall event on all cpus: Ok
 5: parse events tests: Ok
[root@emilia linux]#

Now to write more tests!

Suggested-by: Peter Zijlstra <peterz@infradead.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-nqec2145qfxdgimux28aw7v8@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evlist: Always do automatic allocation of pollfd and mmap structures
Arnaldo Carvalho de Melo [Tue, 29 Nov 2011 10:05:52 +0000 (08:05 -0200)]
perf evlist: Always do automatic allocation of pollfd and mmap structures

At first tools were required to do that, but while writing the python
bindings to simplify the API I made them auto-allocate when needed.

This just makes record, stat and top use that auto allocation,
simplifying them a bit.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-iokhcvkzzijr3keioubx8hlq@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Save some loops using perf_evlist__id2evsel
Arnaldo Carvalho de Melo [Mon, 28 Nov 2011 19:57:40 +0000 (17:57 -0200)]
perf tools: Save some loops using perf_evlist__id2evsel

Since we already ask for PERF_SAMPLE_ID and use it to quickly find the
associated evsel, add handler func + data to struct perf_evsel to avoid
using chains of if(strcmp(event_name)) and also to avoid all the linear
list searches via trace_event_find.

To demonstrate the technique convert 'perf sched' to it:

 # perf sched record sleep 5m

And then:

 Performance counter stats for '/tmp/oldperf sched lat':

        646.929438 task-clock                #    0.999 CPUs utilized
                 9 context-switches          #    0.000 M/sec
                 0 CPU-migrations            #    0.000 M/sec
            20,901 page-faults               #    0.032 M/sec
     1,290,144,450 cycles                    #    1.994 GHz
   <not supported> stalled-cycles-frontend
   <not supported> stalled-cycles-backend
     1,606,158,439 instructions              #    1.24  insns per cycle
       339,088,395 branches                  #  524.151 M/sec
         4,550,735 branch-misses             #    1.34% of all branches

       0.647524759 seconds time elapsed

Versus:

 Performance counter stats for 'perf sched lat':

        473.564691 task-clock                #    0.999 CPUs utilized
                 9 context-switches          #    0.000 M/sec
                 0 CPU-migrations            #    0.000 M/sec
            20,903 page-faults               #    0.044 M/sec
       944,367,984 cycles                    #    1.994 GHz
   <not supported> stalled-cycles-frontend
   <not supported> stalled-cycles-backend
     1,442,385,571 instructions              #    1.53  insns per cycle
       308,383,106 branches                  #  651.195 M/sec
         4,481,784 branch-misses             #    1.45% of all branches

       0.474215751 seconds time elapsed

[root@emilia ~]#

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-1kbzpl74lwi6lavpqke2u2p3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf script: Add comm filtering option
David Ahern [Mon, 21 Nov 2011 17:02:52 +0000 (10:02 -0700)]
perf script: Add comm filtering option

Allows collecting events system wide and then pulling out events for a
specific task name(s). e.g,

    perf script -c gnome-shell,gnome-terminal

Applies on top of:
    https://lkml.org/lkml/2011/11/13/74

v2->v3
- update Documentation

v1->v2
- use comm_list from symbol_conf

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1321894972-24246-1-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: make -C consistent across commands (for cpu list arg)
David Ahern [Sun, 13 Nov 2011 18:30:08 +0000 (11:30 -0700)]
perf tools: make -C consistent across commands (for cpu list arg)

Currently the meaning of -C varies by perf command: for perf-top,
perf-stat, perf-record it means cpu list. For perf-report it means comm
list. Then perf-annotate, perf-report and perf-script use -c for cpu
list.

Fix annotate, report and script to use -C for cpu list to be consistent
with top, stat and record. This means report needs to use -c for comm
list which does introduce a backward compatibility change.

v1 -> v2
- update perf-script.txt too

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1321209008-7004-1-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf top: Stop using globals for tool state
Arnaldo Carvalho de Melo [Mon, 28 Nov 2011 11:37:05 +0000 (09:37 -0200)]
perf top: Stop using globals for tool state

Use its 'perf_tool' base class instead.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-i33q40wwvk2zna8fd36ex6sm@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Rename perf_event_ops to perf_tool
Arnaldo Carvalho de Melo [Mon, 28 Nov 2011 10:30:20 +0000 (08:30 -0200)]
perf tools: Rename perf_event_ops to perf_tool

To better reflect that it became the base class for all tools, that must
be in each tool struct and where common stuff will be put.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-qgpc4msetqlwr8y2k7537cxe@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Resolve machine earlier and pass it to perf_event_ops
Arnaldo Carvalho de Melo [Mon, 28 Nov 2011 09:56:39 +0000 (07:56 -0200)]
perf tools: Resolve machine earlier and pass it to perf_event_ops

Reducing the exposure of perf_session further, so that we can use the
classes in cases where no perf.data file is created.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-stua66dcscsezzrcdugvbmvd@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Pass tool context in the the perf_event_ops functions
Arnaldo Carvalho de Melo [Fri, 25 Nov 2011 10:19:45 +0000 (08:19 -0200)]
perf tools: Pass tool context in the the perf_event_ops functions

So that we don't need to have that many globals.

Next steps will remove the 'session' pointer, that in most cases is
not needed.

Then we can rename perf_event_ops to 'perf_tool' that better describes
this class hierarchy.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-wp4djox7x6w1i2bab1pt4xxp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf annotate: Group options in a struct
Arnaldo Carvalho de Melo [Thu, 17 Nov 2011 14:33:21 +0000 (12:33 -0200)]
perf annotate: Group options in a struct

Paving the way to remove these globals when we change the perf_event_ops
to receive as a first parameter a pointer to a perf_event_ops that will
then provide access to perf_annotate via container_of.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xduzibqrdg3h5cttmk6p5wwc@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf report: Group options in a struct
Arnaldo Carvalho de Melo [Thu, 17 Nov 2011 14:19:04 +0000 (12:19 -0200)]
perf report: Group options in a struct

Paving the way to remove these globals when we change the perf_event_ops
to receive as a first parameter a pointer to a perf_event_ops that will
then provide access to perf_report via container_of.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-2eh2vi2nb5z3tg1lvoxv09xu@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Use evsel->attr.sample_type instead of session->sample_type
Arnaldo Carvalho de Melo [Wed, 16 Nov 2011 19:02:54 +0000 (17:02 -0200)]
perf tools: Use evsel->attr.sample_type instead of session->sample_type

Eventually session->sample_type will go away as we want to support
multiple sample types per session, so use it from the evsel which is a
step in that direction.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-0vwdpjcwbjezw459lw5n3ew1@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf session: Remove superfluous callchain_cursor member
Arnaldo Carvalho de Melo [Sat, 12 Nov 2011 01:10:26 +0000 (23:10 -0200)]
perf session: Remove superfluous callchain_cursor member

Since we have it in evsel->hists.callchain_cursor, remove it from
perf_session.

One more step in disentangling several places from requiring a
perf_session pointer.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-rxr5dj3di7ckyfmnz0naku1z@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf event: perf_event_ops->attr() manipulates only an evlist
Arnaldo Carvalho de Melo [Sat, 12 Nov 2011 00:45:41 +0000 (22:45 -0200)]
perf event: perf_event_ops->attr() manipulates only an evlist

Removing another case where a perf_session is required when processing
events.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-ug1wtjbnva4bxwknflkkrlrh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evlist: Introduce id_hdr_size method out of perf_session
Arnaldo Carvalho de Melo [Sat, 12 Nov 2011 00:28:50 +0000 (22:28 -0200)]
perf evlist: Introduce id_hdr_size method out of perf_session

We will need this when not using perf_session in cases like 'perf top'
and strace where no perf.data file is created nor consumed.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-za923wjc41q5xot5vrhuhj3j@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf symbols: Add nr_events to symbol_conf
Arnaldo Carvalho de Melo [Sat, 12 Nov 2011 00:17:32 +0000 (22:17 -0200)]
perf symbols: Add nr_events to symbol_conf

Since symbol__alloc_hists need it, to avoid passing it around in many
functions have it in the symbol_conf struct.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-cwv8ysvpywzjq4v3xtbd4zwv@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf ui progress: Fix divide by zero
Arnaldo Carvalho de Melo [Sat, 12 Nov 2011 00:08:07 +0000 (22:08 -0200)]
perf ui progress: Fix divide by zero

Happens in a perf.data file where one of the events had no samples.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-j7st3oyiotvfxqde2nc41kxb@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf record: Move 'group' to perf_event_ops
Arnaldo Carvalho de Melo [Fri, 11 Nov 2011 17:12:56 +0000 (15:12 -0200)]
perf record: Move 'group' to perf_event_ops

Will be used in other tools to share the command line parsing code.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-8x0yr77r6lrd2t699s499m8n@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf session: Move threads to struct machine
Arnaldo Carvalho de Melo [Wed, 9 Nov 2011 15:24:25 +0000 (13:24 -0200)]
perf session: Move threads to struct machine

The 'machine' abstraction was introduced with 'perf kvm' where we could
have samples for the host and multiple guests, but at the time we ended
up keeping the list of all machines threads all in
session->host_machine.

Move the threads rb_tree to struct machine to separate the namespaces.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-mdg7sm6j3va09vtgj49gbsrp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf record: Move mmap_pages to perf_record_opts
Arnaldo Carvalho de Melo [Wed, 9 Nov 2011 11:16:26 +0000 (09:16 -0200)]
perf record: Move mmap_pages to perf_record_opts

Tools being developed will need this to allow the user to override this
value.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-zydc1yhxfm0z35fuy95bsn1l@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evlist: Handle default value for 'pages' on mmap method
Arnaldo Carvalho de Melo [Wed, 9 Nov 2011 11:10:47 +0000 (09:10 -0200)]
perf evlist: Handle default value for 'pages' on mmap method

Every tool that calls this and allows the user to override the value
needs this logic.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-lwscxpg57xfzahz5dmdfp9uz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evlist: Introduce {prepare,start}_workload refactored from 'perf record'
Arnaldo Carvalho de Melo [Wed, 9 Nov 2011 10:47:15 +0000 (08:47 -0200)]
perf evlist: Introduce {prepare,start}_workload refactored from 'perf record'

So that we can easily start a workload in other tools.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-zdsksd4aphu0nltg2lpwsw3x@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Introduce config attr method
Arnaldo Carvalho de Melo [Tue, 8 Nov 2011 16:41:57 +0000 (14:41 -0200)]
perf evsel: Introduce config attr method

Out of the code in 'perf record', so that we can share option parsing,
etc. Eventually will be used by 'perf top', but first 'trace' will use
it.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-hzjqsgnte1esk90ytq0ap98v@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evlist: Introduce add_tracepoints method
Arnaldo Carvalho de Melo [Sat, 5 Nov 2011 10:41:51 +0000 (08:41 -0200)]
perf evlist: Introduce add_tracepoints method

Convenient way of asking for tracepoint events to be added to an
existing evlist.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-0ylj4wrg54791u0baqb9swbb@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evlist: Introduce perf_evlist__add_attrs
Arnaldo Carvalho de Melo [Fri, 4 Nov 2011 11:10:59 +0000 (09:10 -0200)]
perf evlist: Introduce perf_evlist__add_attrs

Replacing the open coded equivalents in 'perf stat'.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-1btwadnf2tds2g07hsccsdse@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Simplify debugfs mountpoint handling code
Arnaldo Carvalho de Melo [Wed, 16 Nov 2011 16:03:07 +0000 (14:03 -0200)]
perf tools: Simplify debugfs mountpoint handling code

We don't need to have two PATH_MAX char sized arrays holding it, just
one in util/debugfs.c will do.

Also rename debugfs_path to tracing_events_path, as it is not the path
to debugfs, that is debugfs_mountpoint. Both are now accessible.

This will allow accessing this code in the perf python binding without
having to drag in perf.c and util/parse-events.c.

The defaults for these variables are the canonical "/sys/kernel/debug"
and "/sys/kernel/debug/tracing/events/", removing the need for simple
tools to call debugfs_mount(NULL).

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-ug9jvtjrsqbluuhqqxpvg30f@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Eliminate duplicate code and use PATH_MAX consistently
Arnaldo Carvalho de Melo [Wed, 16 Nov 2011 14:55:59 +0000 (12:55 -0200)]
perf tools: Eliminate duplicate code and use PATH_MAX consistently

No need for multiple definitions for STR() and die(), also use SuSv2's
PATH_MAX instead of adding MAX_PATH.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-qpujjkw7u0bf0tr4wt55cr9y@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf session: Fix crash with invalid CPU list
David Ahern [Sun, 13 Nov 2011 17:45:27 +0000 (10:45 -0700)]
perf session: Fix crash with invalid CPU list

commit 5d67be9 added the option to specify a range of CPUs of interest,
but does not catch an invalid CPU list:

$ perf script -c foo
Segmentation fault (core dumped)

Cc: Anton Blanchard <anton@samba.org>
Link: http://lkml.kernel.org/r/1321206327-5881-1-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf python: Fix undefined symbol problem
Arnaldo Carvalho de Melo [Fri, 4 Nov 2011 10:16:58 +0000 (08:16 -0200)]
perf python: Fix undefined symbol problem

Recently we made perf_evsel__init call hists__init, which broke the perf
python binding:

[root@emilia linux]# ./tools/perf/python/twatch.py
Traceback (most recent call last):
  File "./tools/perf/python/twatch.py", line 16, in <module>
    import perf
ImportError: /home/acme/git/build/perf/python/perf.so: undefined symbol: hists__init

Fix it by moving the hists__init function to its only caller, evsel.c.

This way we avoid dragging in other parts of tools/perf/util/ to the
perf python binding.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-5nffmdt5mu6ozxgj54oi4qon@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoMerge branch 'core' of git://amd64.org/linux/rric into perf/core
Ingo Molnar [Tue, 15 Nov 2011 10:05:18 +0000 (11:05 +0100)]
Merge branch 'core' of git://amd64.org/linux/rric into perf/core

13 years agoMerge branch 'urgent' of git://amd64.org/linux/rric into perf/urgent
Ingo Molnar [Tue, 15 Nov 2011 10:03:30 +0000 (11:03 +0100)]
Merge branch 'urgent' of git://amd64.org/linux/rric into perf/urgent

13 years agoevents: Don't divide events if it has field period
Andrew Vagin [Mon, 7 Nov 2011 12:54:12 +0000 (15:54 +0300)]
events: Don't divide events if it has field period

This patch solves the following problem:

Now some samples may be lost due to throttling. The number of samples is
restricted by sysctl_perf_event_sample_rate/HZ.  A trace event is
divided on some samples according to event's period.  I don't sure, that
we should generate more than one sample on each trace event. I think the
better way to use SAMPLE_PERIOD.

E.g.: I want to trace when a process sleeps. I created a process, which
sleeps for 1ms and for 4ms.  perf got 100 events in both cases.

swapper     0 [000]  1141.371830: sched_stat_sleep: comm=foo pid=1801 delay=1386750 [ns]
swapper     0 [000]  1141.369444: sched_stat_sleep: comm=foo pid=1801 delay=4499585 [ns]

In the first case a kernel want to send 4499585 events and
in the second case it wants to send 1386750 events.
perf-reports shows that process sleeps in both places equal time. It's
bug.

With this patch kernel generates one event on each "sleep" and the time
slice is saved in the field "period". Perf knows how handle it.

Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1320670457-2633428-3-git-send-email-avagin@openvz.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf: Carve out callchain functionality
Borislav Petkov [Sun, 16 Oct 2011 15:15:04 +0000 (17:15 +0200)]
perf: Carve out callchain functionality

Split the callchain code from the perf events core into
a new kernel/events/callchain.c file.

This simplifies a bit the big core.c

Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
[keep ctx recursion handling inline and use internal headers]
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1318778104-17152-1-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf/x86: Enable raw event access to Intel offcore events
Peter Zijlstra [Mon, 14 Nov 2011 09:03:25 +0000 (10:03 +0100)]
perf/x86: Enable raw event access to Intel offcore events

Now that the core offcore support is fixed up (thanks Stephane) and we
have sane generic events utilizing them, re-enable the raw access to
the feature as well.

Note that it doesn't matter if you use event 0x1b7 or 0x1bb to specify
an offcore event, either one works and neither guarantees you'll end
up on a particular offcore MSR.

Based on original patch from: Vince Weaver <vweaver1@eecs.utk.edu>.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Vince Weaver <vweaver1@eecs.utk.edu>.
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1108031200390.703@cl320.eecs.utk.edu
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf: Don't use -ENOSPC for out of PMU resources
Peter Zijlstra [Wed, 9 Nov 2011 16:56:37 +0000 (17:56 +0100)]
perf: Don't use -ENOSPC for out of PMU resources

People (Linus) objected to using -ENOSPC to signal not having enough
resources on the PMU to satisfy the request. Use -EINVAL.

Requested-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: David Daney <david.daney@cavium.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-xv8geaz2zpbjhlx0svmpp28n@git.kernel.org
[ merged to newer kernel, fixed up MIPS impact ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf: Do not set task_ctx pointer in cpuctx if there are no events in the context
Gleb Natapov [Sun, 23 Oct 2011 17:10:33 +0000 (19:10 +0200)]
perf: Do not set task_ctx pointer in cpuctx if there are no events in the context

Do not set task_ctx pointer during sched_in if there are no
events associated with the context.  Otherwise if during task
execution total number of events in the system will become zero
perf_event_context_sched_out() will not be called and cpuctx->task_ctx
will be left with a stale value.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20111023171033.GI17571@redhat.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf/x86: Fix PEBS instruction unwind
Peter Zijlstra [Fri, 7 Oct 2011 11:36:40 +0000 (13:36 +0200)]
perf/x86: Fix PEBS instruction unwind

Masami spotted that we always try to decode the instruction stream as
64bit instructions when running a 64bit kernel, this doesn't work for
ia32-compat proglets.

Use TIF_IA32 to detect if we need to use the 32bit instruction
decoder.

Reported-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: stable@kernel.org
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge branch 'rmobile-fixes-for-linus' of git://github.com/pmundt/linux-sh
Linus Torvalds [Mon, 14 Nov 2011 08:47:04 +0000 (06:47 -0200)]
Merge branch 'rmobile-fixes-for-linus' of git://github.com/pmundt/linux-sh

* 'rmobile-fixes-for-linus' of git://github.com/pmundt/linux-sh:
  ARM: mach-shmobile: cpuidle single/global and last_state fixes
  ARM: mach-shmobile: move helper macro PORTCR to sh_pfc.h
  ARM: mach-shmobile: move helper macro PORT_xx to sh_pfc.h
  ARM: mach-shmobile: move helper macro PORT_DATA_xx to sh_pfc.h
  ARM: mach-shmobile: ap4evb: remove white space from end of line
  ARM: mach-shmobile: clock-sh7372: remove un-necessary index
  ARM: mach-shmobile: kota2: add comment out separator
  ARM: mach-shmobile: sh73a0: add MMC data pin pull-up

13 years agoMerge branch 'sh-fixes-for-linus' of git://github.com/pmundt/linux-sh
Linus Torvalds [Mon, 14 Nov 2011 08:45:30 +0000 (06:45 -0200)]
Merge branch 'sh-fixes-for-linus' of git://github.com/pmundt/linux-sh

* 'sh-fixes-for-linus' of git://github.com/pmundt/linux-sh:
  mailmap: Fix up some renesas attributions
  sh: clkfwk: Kill off remaining debugfs cruft.
  drivers: sh: Kill off dead pathname for runtime PM stub.
  drivers: sh: Generalize runtime PM platform stub.
  sh: Wire up process_vm syscalls.
  sh: clkfwk: add clk_rate_mult_range_round()
  serial: sh-sci: Fix up SH-2A SCIF support.
  sh: Fix cached/uncaced address calculation in 29bit mode

13 years agoMerge git://github.com/rustyrussell/linux
Linus Torvalds [Mon, 14 Nov 2011 06:00:01 +0000 (04:00 -0200)]
Merge git://github.com/rustyrussell/linux

* git://github.com/rustyrussell/linux:
  virtio-pci: fix use after free

13 years agovirtio-pci: fix use after free
Michael S. Tsirkin [Mon, 7 Nov 2011 16:37:05 +0000 (18:37 +0200)]
virtio-pci: fix use after free

Commit 31a3ddda166cda86d2b5111e09ba4bda5239fae6 introduced
a use after free in virtio-pci. The main issue is
that the release method signals removal of the virtio device,
while remove signals removal of the pci device.

For example, on driver removal or hot-unplug,
virtio_pci_release_dev is called before virtio_pci_remove.
We then might get a crash as virtio_pci_remove tries to use the
device freed by virtio_pci_release_dev.

We allocate/free all resources together with the
pci device, so we can leave the release method empty.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: stable@kernel.org
13 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sun, 13 Nov 2011 19:09:55 +0000 (17:09 -0200)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/radeon/kms/combios: fix dynamic allocation of PM clock modes

13 years agoACPI / cpuidle: Remove acpi_idle_suspend (to fix suspend regression)
Rafael J. Wysocki [Sat, 12 Nov 2011 22:17:27 +0000 (23:17 +0100)]
ACPI / cpuidle: Remove acpi_idle_suspend (to fix suspend regression)

After commit e978aa7d7d57 ("cpuidle: Move dev->last_residency update to
driver enter routine; remove dev->last_state") setting acpi_idle_suspend
to 1 by acpi_processor_suspend() causes the ACPI cpuidle routines to
return error codes continuously, which in turn causes cpuidle to lock up
(hard).

However, acpi_idle_suspend doesn't appear to be useful for any
particular purpose (it's racy and doesn't really provide any real
protection), so it can be removed, which makes the problem go away.

Reported-and-tested-by: Tomas M. <tmezzadra@gmail.com>
Reported-and-tested-by: Ferenc Wagner <wferi@niif.hu>
Tested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agodrm/radeon/kms/combios: fix dynamic allocation of PM clock modes
Alex Deucher [Sat, 12 Nov 2011 16:57:29 +0000 (11:57 -0500)]
drm/radeon/kms/combios: fix dynamic allocation of PM clock modes

I missed the combios path when I updated the atombios pm code.

Reported by amarsh04 on IRC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Linus Torvalds [Sat, 12 Nov 2011 15:29:04 +0000 (13:29 -0200)]
Merge branch 'fixes' of git://git./linux/kernel/git/arm/arm-soc

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  arm/imx: fix imx6q mmc error when mounting rootfs
  arm/imx: fix AUTO_ZRELADDR selection
  arm/imx: fix the references to ARCH_MX3
  ARM: mx51/53: set pwm clock parent to ipg_perclk
  arm/tegra: enable headphone detection gpio on seaboard
  arm/dt: Fix ventana SDHCI power-gpios
  arm/tegra: Don't create duplicate gpio and pinmux devices
  ARM: at91: Fix USBA gadget registration
  atmel/spi: fix missing probe
  at91/yl-9200: Fix section mismatch
  at91: vmalloc fix missing AT91_VIRT_BASE define
  ARM: at91: usart: drop static map regs for dbgu
  ARM: picoxcell: add extra temp register to addruart
  ARM: msm: fix compilation flags for MSM_SCM
  arm/mxs: fix mmc device adding for mach-mx28evk
  ARM: mxc: Remove test_for_ltirq
  ARM:i.MX: fix build error in clock-mx51-mx53.c
  ARM:i.MX: fix build error in tzic/avic.c
  ARM: mxc: fix local timer interrupt handling
  msm: boards: Fix fallout from removal of machine_desc in fixup

13 years ago[CPUFREQ] db8500: fix build error due to undeclared i variable
Axel Lin [Fri, 4 Nov 2011 12:04:41 +0000 (20:04 +0800)]
[CPUFREQ] db8500: fix build error due to undeclared i variable

The variable i is removed by commit ded8433
"[CPUFREQ] db8500: remove unneeded for loop iteration over freq_table",
but current code to print available frequencies still uses the i variable.
Thus add the i variable back to fix below buld error:

  CC      drivers/cpufreq/db8500-cpufreq.o
drivers/cpufreq/db8500-cpufreq.c: In function 'db8500_cpufreq_init':
drivers/cpufreq/db8500-cpufreq.c:123: error: 'i' undeclared (first use in this function)
drivers/cpufreq/db8500-cpufreq.c:123: error: (Each undeclared identifier is reported only once
drivers/cpufreq/db8500-cpufreq.c:123: error: for each function it appears in.)
make[2]: *** [drivers/cpufreq/db8500-cpufreq.o] Error 1
make[1]: *** [drivers/cpufreq] Error 2
make: *** [drivers] Error 2

This patch also fixes using uninitialized i variable as array index.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Dave Jones <davej@redhat.com>
13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux...
Linus Torvalds [Sat, 12 Nov 2011 02:04:51 +0000 (00:04 -0200)]
Merge branch 'for-linus' of git://git./linux/kernel/git/geert/linux-m68k

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k: (29 commits)
  m68k/mac: Remove mac_irq_{en,dis}able() wrappers
  m68k/irq: Remove obsolete support for user vector interrupt fixups
  m68k/irq: Remove obsolete m68k irq framework
  m68k/q40: Convert Q40/Q60 to genirq
  m68k/sun3: Convert Sun3/3x to genirq
  m68k/sun3: Use the kstat_irqs_cpu() wrapper
  m68k/apollo: Convert Apollo to genirq
  m68k/vme: Convert VME to genirq
  m68k/hp300: Convert HP9000/300 and HP9000/400 to genirq
  m68k/mac: Optimize interrupts using chain handlers
  m68k/mac: Convert Mac to genirq
  m68k/amiga: Optimize interrupts using chain handlers
  m68k/amiga: Convert Amiga to genirq
  m68k/amiga: Refactor amiints.c
  m68k/atari: Remove code and comments about different irq types
  m68k/atari: Convert Atari to genirq
  m68k/irq: Add genirq support
  m68k/irq: Remove obsolete IRQ_FLG_* users
  m68k/irq: Rename {,__}m68k_handle_int()
  m68k/irq: Add m68k_setup_irq_controller()
  ...

13 years agoMerge branch 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Sat, 12 Nov 2011 02:03:50 +0000 (00:03 -0200)]
Merge branch 'v4l_for_linus' of git://git./linux/kernel/git/mchehab/linux-media

* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] v4l2-ctrl: Send change events to all fh for auto cluster slave controls
  [media] v4l2-event: Don't set sev->fh to NULL on unsubscribe
  [media] v4l2-event: Remove pending events from fh event queue when unsubscribing
  [media] v4l2-event: Deny subscribing with a type of V4L2_EVENT_ALL
  [media] MAINTAINERS: add a maintainer for s5p-mfc driver
  [media] v4l: s5p-mfc: fix reported capabilities
  [media] media: vb2: reset queued list on REQBUFS(0) call
  [media] media: vb2: set buffer length correctly for all buffer types
  [media] media: vb2: add a check for uninitialized buffer
  [media] mxl111sf: fix build warning
  [media] mxl111sf: remove pointless if condition in mxl111sf_config_spi
  [media] mxl111sf: check for errors after mxl111sf_write_reg in mxl111sf_idac_config
  [media] mxl111sf: fix return value of mxl111sf_idac_config
  [media] uvcvideo: GET_RES should only be checked for BITMAP type menu controls

13 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Sat, 12 Nov 2011 02:01:51 +0000 (00:01 -0200)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/kvm: Fix build failure with HV KVM and CBE
  powerpc/ps3: Fix lv1_gpu_attribute hcall
  powerpc/ps3: Fix PS3 repository build warnings
  powerpc/ps3: irq: Remove IRQF_DISABLED
  powerpc/irq: Remove IRQF_DISABLED
  powerpc/numa: NUMA topology support for PowerNV
  powerpc: Add System RAM to /proc/iomem
  powerpc: Add KVM as module to defconfigs
  powerpc/kvm: Fix build with older toolchains
  powerpc, tqm5200: update tqm5200_defconfig to fit for charon board.
  powerpc/5200: add support for charon board

13 years agoMerge branch 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild
Linus Torvalds [Sat, 12 Nov 2011 01:59:40 +0000 (23:59 -0200)]
Merge branch 'rc-fixes' of git://git./linux/kernel/git/mmarek/kbuild

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  kbuild: Fix missing system calls check on mips.

13 years agobma023: Add SFI translation for this device
William Douglas [Thu, 10 Nov 2011 13:50:38 +0000 (13:50 +0000)]
bma023: Add SFI translation for this device

This needed the sfi IRQ 0xFF fix to go in first. It simply plumbs in the
bma023 driver with the firmware naming of it.

Signed-off-by: William Douglas <william.douglas@intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agovrtc: change its year offset from 1960 to 1972
Feng Tang [Thu, 10 Nov 2011 13:23:39 +0000 (13:23 +0000)]
vrtc: change its year offset from 1960 to 1972

Real world year equals the value in vrtc YEAR register plus an offset.
We used 1960 as the offset to make leap year consistent, but for a
device's first use, its YEAR register is 0 and the system year will
be parsed as 1960 which is not a valid UNIX time and will cause many
applications to fail mysteriously. So we use 1972 instead to fix this
issue.

Updated patch which adds a sanity check suggested by Mathias

This isn't a change in behaviour for systems, because 1972 is the one we
actually use. It's the old version in upstream which is out of sync with
all devices.

Signed-off-by: Feng Tang <feng.tang@intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoce4100: fix a build error
Zhang Rui [Thu, 10 Nov 2011 13:21:10 +0000 (13:21 +0000)]
ce4100: fix a build error

Fix a build error. CE4100 with no serial errors because the alternate
function is only a prototype not a null function as intended.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'upstream-linus' of git://github.com/jgarzik/libata-dev
Linus Torvalds [Sat, 12 Nov 2011 01:55:01 +0000 (23:55 -0200)]
Merge branch 'upstream-linus' of git://github.com/jgarzik/libata-dev

* 'upstream-linus' of git://github.com/jgarzik/libata-dev:
  pata_of_platform: Don't use NO_IRQ
  [libata] ahci: Add ASMedia ASM1061 support
  [libata] Issue SRST to Sil3726 PMP
  sata_sis.c: trivial spelling fix
  ahci_platform: use dev_get_platdata()
  [libata] libata-scsi.c: Add function parameter documentation

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sat, 12 Nov 2011 01:47:06 +0000 (23:47 -0200)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: rename the option to nospace_cache
  Btrfs: handle bio_add_page failure gracefully in scrub
  Btrfs: fix deadlock caused by the race between relocation
  Btrfs: only map pages if we know we need them when reading the space cache
  Btrfs: fix orphan backref nodes
  Btrfs: Abstract similar code for btrfs_block_rsv_add{, _noflush}
  Btrfs: fix unreleased path in btrfs_orphan_cleanup()
  Btrfs: fix no reserved space for writing out inode cache
  Btrfs: fix nocow when deleting the item
  Btrfs: tweak the delayed inode reservations again
  Btrfs: rework error handling in btrfs_mount()
  Btrfs: close devices on all error paths in open_ctree()
  Btrfs: avoid null dereference and leaks when bailing from open_ctree()
  Btrfs: fix subvol_name leak on error in btrfs_mount()
  Btrfs: fix memory leak in btrfs_parse_early_options()
  Btrfs: fix our reservations for updating an inode when completing io
  Btrfs: fix oops on NULL trans handle in btrfs_truncate
  btrfs: fix double-free 'tree_root' in 'btrfs_mount()'

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Sat, 12 Nov 2011 01:45:42 +0000 (23:45 -0200)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: usb-audio: Use kmemdup rather than duplicating its implementation
  ALSA: hda - Re-enable the check NO_PRESENCE misc bit
  ALSA: vmaster - Free slave-links when freeing the master element
  ALSA: hda - Don't add elements of other codecs to vmaster slave
  ALSA: intel8x0: improve virtual environment detection
  ALSA: intel8x0: move virtual environment detection code into one place
  ALSA: snd_usb_audio: add Logitech HD Webcam c510 to quirk-384
  ALSA: hda - fix internal mic on Dell Vostro 3500 laptop
  ALSA: HDA: Remove quirk for Toshiba T110
  ALSA: usb-audio - Fix the missing volume quirks at delayed init
  ALSA: hda - Mute unused capture sources for Realtek codecs
  ALSA: intel8x0: Improve comments for VM optimization
  ASoC: Ensure we get an impedence reported for WM8958 jack detect
  ASoC: Don't use wm8994->control_data when requesting IRQs
  ASoC: Don't use wm8994->control_data in wm8994_readable_register()
  ASoC: Update git repository URL

13 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 12 Nov 2011 01:43:51 +0000 (23:43 -0200)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (42 commits)
  drm/radeon/kms/pm: switch to dynamically allocating clock mode array
  drm/radeon/kms: optimize r600_pm_profile_init
  drm/radeon/kms/pm: add a proper pm profile init function for fusion
  drm/radeon/kms: remove extraneous calls to radeon_pm_compute_clocks()
  drm/exynos: added padding to be 64-bit align.
  drm: fix kconfig unmet dependency warning
  drm: add some comments to drm_wait_vblank and drm_queue_vblank_event
  drm/radeon/benchmark: signedness bug in radeon_benchmark_move()
  drm: do not sleep on vblank while holding a mutex
  MAINTAINERS: exynos: Add EXYNOS DRM maintainer entry
  drm: try to restore previous CRTC config if mode set fails
  drm/radeon/kms: make an aux failure debug only
  drm: drop select of SLOW_WORK
  drm: serialize access to list of debugfs files
  drm/radeon/kms: fix use of vram scratch page on evergreen/ni
  drm/radeon: Make sure CS mutex is held across GPU reset.
  drm: Ensure string is null terminated.
  vmwgfx: Only allow 64x64 cursors
  vmwgfx: Initialize clip rect loop correctly in surface dirty
  vmwgfx: Close screen object system
  ...

13 years agoMerge git://github.com/herbertx/crypto
Linus Torvalds [Sat, 12 Nov 2011 01:40:02 +0000 (23:40 -0200)]
Merge git://github.com/herbertx/crypto

* git://github.com/herbertx/crypto:
  crypto: algapi - Fix build problem with NET disabled
  crypto: user - Fix rwsem leak in crypto_user

13 years agoMerge branch 'for-linus' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Sat, 12 Nov 2011 01:37:17 +0000 (23:37 -0200)]
Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: fix force shutdown handling in xfs_end_io
  xfs: constify xfs_item_ops
  xfs: Fix possible memory corruption in xfs_readlink

13 years agoMerge branch 'picoxcell-fixes' of git://github.com/jamieiles/linux-2.6-ji into fixes
Olof Johansson [Fri, 11 Nov 2011 18:34:20 +0000 (10:34 -0800)]
Merge branch 'picoxcell-fixes' of git://github.com/jamieiles/linux-2.6-ji into fixes

13 years agoMerge branch 'imx-fixes-for-arnd' of git://git.pengutronix.de/git/imx/linux-2.6 into...
Olof Johansson [Fri, 11 Nov 2011 16:53:46 +0000 (08:53 -0800)]
Merge branch 'imx-fixes-for-arnd' of git://git.pengutronix.de/git/imx/linux-2.6 into fixes

13 years agoarm/imx: fix imx6q mmc error when mounting rootfs
Shawn Guo [Thu, 10 Nov 2011 08:39:32 +0000 (16:39 +0800)]
arm/imx: fix imx6q mmc error when mounting rootfs

The following error is seen in some case when mounting rootfs from
SD/MMC cards.

  Waiting for root device /dev/mmcblk0p1...
  mmc1: host does not support reading read-only switch. assuming write-enable.
  mmc1: new high speed SDHC card at address b368
  mmcblk0: mmc1:b368 SDC   3.74 GiB
   mmcblk0: p1
  mmc1: Timeout waiting for hardware interrupt.
  mmcblk0: error -110 transferring data, sector 3678224, nr 40, cmd response 0x900, card status 0xc00
  end_request: I/O error, dev mmcblk0, sector 3678225
  Buffer I/O error on device mmcblk0p1, logical block 458754
  lost page write due to I/O error on mmcblk0p1

This patch fixes the problem by lowering the usdhc clock and correcting
watermark configuration.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Cc: Chris Ball <cjb@laptop.org>
Cc: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agoarm/imx: fix AUTO_ZRELADDR selection
Shawn Guo [Thu, 10 Nov 2011 08:39:31 +0000 (16:39 +0800)]
arm/imx: fix AUTO_ZRELADDR selection

The AUTO_ZRELADDR selection for ARCH_IMX_V4_V5 and ARCH_MX5 should
really be mutually exclusive to ZBOOT_ROM just like what ARCH_IMX_V6_V7
does.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Cc: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agoarm/imx: fix the references to ARCH_MX3
Shawn Guo [Fri, 11 Nov 2011 07:37:14 +0000 (15:37 +0800)]
arm/imx: fix the references to ARCH_MX3

The config symbol ARCH_MX3 has been removed by commit 'a89cf59
arm/imx: merge i.MX3 and i.MX6', and it should not be referenced
any more.

The patch also change ARCH_MX* to SOC_IMX* for other platforms.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Cc: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agoARM: mx51/53: set pwm clock parent to ipg_perclk
Richard Zhao [Tue, 8 Nov 2011 01:59:34 +0000 (09:59 +0800)]
ARM: mx51/53: set pwm clock parent to ipg_perclk

Signed-off-by: Richard Zhao <richard.zhao@linaro.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
13 years agobtrfs: rename the option to nospace_cache
David Sterba [Fri, 11 Nov 2011 15:14:57 +0000 (10:14 -0500)]
btrfs: rename the option to nospace_cache

Rename no_space_cache option to nospace_cache to be more consistent with
the rest, where the simple prefix 'no' is used to negate an option.

The option has been introduced during the -rc1 cycle and there are has not been
widely used, so it's safe.

Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
13 years agodrm/radeon/kms/pm: switch to dynamically allocating clock mode array
Alex Deucher [Fri, 4 Nov 2011 14:09:43 +0000 (10:09 -0400)]
drm/radeon/kms/pm: switch to dynamically allocating clock mode array

On newer chips the number of clock modes per power state varies.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agodrm/radeon/kms: optimize r600_pm_profile_init
Alex Deucher [Fri, 4 Nov 2011 14:09:42 +0000 (10:09 -0400)]
drm/radeon/kms: optimize r600_pm_profile_init

Avoid a lot of extra loops through the pm state array.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agodrm/radeon/kms/pm: add a proper pm profile init function for fusion
Alex Deucher [Fri, 4 Nov 2011 14:09:41 +0000 (10:09 -0400)]
drm/radeon/kms/pm: add a proper pm profile init function for fusion

The new power tables need to be handled differently when setting
up the profiles.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agodrm/radeon/kms: remove extraneous calls to radeon_pm_compute_clocks()
Alex Deucher [Fri, 4 Nov 2011 14:09:40 +0000 (10:09 -0400)]
drm/radeon/kms: remove extraneous calls to radeon_pm_compute_clocks()

It's already called via the DPMS functions.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agoBtrfs: handle bio_add_page failure gracefully in scrub
Arne Jansen [Fri, 11 Nov 2011 13:17:10 +0000 (08:17 -0500)]
Btrfs: handle bio_add_page failure gracefully in scrub

Currently scrub fails with ENOMEM when bio_add_page fails. Unfortunately
dm based targets accept only one page per bio, thus making scrub always
fails. This patch just submits the current bio when an error is encountered
and starts a new one.

Signed-off-by: Arne Jansen <sensille@gmx.net>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
13 years agodrm/exynos: added padding to be 64-bit align.
Inki Dae [Fri, 4 Nov 2011 12:09:50 +0000 (21:09 +0900)]
drm/exynos: added padding to be 64-bit align.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agodrm: fix kconfig unmet dependency warning
Randy Dunlap [Mon, 31 Oct 2011 19:51:30 +0000 (12:51 -0700)]
drm: fix kconfig unmet dependency warning

Fix kconfig unmet dependency warning.  BACKLIGHT_CLASS_DEVICE depends on
BACKLIGHT_LCD_SUPPORT, so select the latter along with the former.

warning: (DRM_RADEON_KMS && DRM_I915 && STUB_POULSBO && FB_BACKLIGHT && PANEL_SHARP_LS037V7DW01 && PANEL_ACX565AKM && USB_APPLEDISPLAY && FB_OLPC_DCON && ASUS_LAPTOP && SONY_LAPTOP && THINKPAD_ACPI && EEEPC_LAPTOP && ACPI_ASUS && ACPI_CMPC && SAMSUNG_Q10) selects BACKLIGHT_CLASS_DEVICE which has unmet direct dependencies (HAS_IOMEM && BACKLIGHT_LCD_SUPPORT)

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: David Airlie <airlied@linux.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agoMerge branch 'drm-nouveau-fixes' of git://git.freedesktop.org/git/nouveau/linux-2...
Dave Airlie [Fri, 11 Nov 2011 11:17:43 +0000 (11:17 +0000)]
Merge branch 'drm-nouveau-fixes' of git://git.freedesktop.org/git/nouveau/linux-2.6 into drm-fixes

* 'drm-nouveau-fixes' of git://git.freedesktop.org/git/nouveau/linux-2.6:
  drm/nouveau: Fix bandwidth calculation for DisplayPort
  drm/nouveau: by default use low bpp framebuffer on low memory cards
  drm/nv10: Change the BO size threshold determining the memory placement range.
  drm/nvc0: enable acceleration for nvc1 by default
  drm/nvc0/gr: fixup the mmio list register writes for 0xc1
  drm/nvc1: hacky workaround to fix accel issues
  drm/nvc0/gr: fix some bugs in grctx generation
  drm/nvc0: enable acceleration on 0xc8 by default
  drm/nvc0/vram: skip disabled PBFB subunits
  drm/nv40/pm: fix issues on igp chipsets, which don't have memory
  drm/nouveau: testing the wrong variable
  drm/nvc0/vram: storage type 0xc3 is not compressed
  drm/nv50: fix stability issue on NV86.
  drm/nouveau: initialize chan->fence.lock before use
  drm/nv50/vram: fix incorrect detection of bank count on newer chipsets
  drm/nv50/gr: typo fix, how about we not reset fifo during graph init?
  drm/nv50/bios: fixup mpll programming from the init table parser
  drm/nouveau: fix oops if i2c bus not found in nouveau_i2c_identify()
  drm: make sure drm_vblank_init() has been called before touching vbl_lock

13 years agodrm: add some comments to drm_wait_vblank and drm_queue_vblank_event
Ilija Hadzic [Mon, 31 Oct 2011 17:11:57 +0000 (13:11 -0400)]
drm: add some comments to drm_wait_vblank and drm_queue_vblank_event

during the review of the fix for locks problems in drm_wait_vblank,
a couple of false concerns were raised about how the drm_vblank_get
and drm_vblank_put are used in this function; it turned out that the
code is correct and that it cannot be simplified

add a few comments to explain non-obvious flows in the code,
to prevent "false alarms" in the future

v2: incorporate comments received from Daniel Vetter

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>