Commit Graph

588 Commits

Author SHA1 Message Date
Frederic Weisbecker
d6b17bebd7 perf: Provide a new deterministic events reordering algorithm
The current events reordering algorithm is based on a heuristic that
gets broken once we deal with a very fast flow of events.

Indeed the time period based flushing is not suitable anymore
in the following case, assuming we have a flush period of two
seconds.

    CPU 0           |        CPU 1
                    |
  cnt1 timestamps   |      cnt1 timestamps
                    |
    0               |         0
    1               |         1
    2               |         2
    3               |         3
    [...]           |        [...]
    4 seconds later

If we spend too much time to read the buffers (case of a lot of
events to record in each buffers or when we have a lot of CPU buffers
to read), in the next pass the CPU 0 buffer could contain a slice
of several seconds of events. We'll read them all and notice we've
reached the period to flush. In the above example we flush the first
half of the CPU 0 buffer, then we read the CPU 1 buffer where we
have events that were on the flush slice and then the reordering
fails.

It's simple to reproduce with:

	perf lock record perf bench sched messaging

To solve this, we use a new solution that doesn't rely on an
heuristical time slice period anymore but on a deterministic basis
based on how perf record does its job.

perf record saves the buffers through passes. A pass is a tour
on every buffers from every CPUs. This is made in order: for
each CPU we read the buffers of every counters. So the more
buffers we visit, the later will be the timstamps of their events.

When perf record finishes a pass it records a
PERF_RECORD_FINISHED_ROUND pseudo event.
We record the max timestamp t found in the pass n. Assuming these
timestamps are monotonic across cpus, we know that if a buffer
still has events with timestamps below t, they will be all available
and then read in the pass n + 1.
Hence when we start to read the pass n + 2, we can safely flush every
events with timestamps below t.

      ============ PASS n =================
         CPU 0         |   CPU 1
                       |
      cnt1 timestamps  |   cnt2 timestamps
            1          |         2
            2          |         3
            -          |         4  <--- max recorded

      ============ PASS n + 1 ==============
         CPU 0         |   CPU 1
                       |
      cnt1 timestamps  |   cnt2 timestamps
            3          |         5
            4          |         6
            5          |         7 <---- max recorded

        Flush every events below timestamp 4

      ============ PASS n + 2 ==============
         CPU 0         |   CPU 1
                       |
      cnt1 timestamps  |   cnt2 timestamps
            6          |         8
            7          |         9
            -          |         10

        Flush every events below timestamp 7
        etc...

It also works on perf.data versions that don't have
PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that
the events will be only flushed in the end of the perf.data
processing. It will then consume more memory and scale less with
large perf.data files.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-09 13:43:42 +02:00
Frederic Weisbecker
9840280757 perf: Introduce a new "round of buffers read" pseudo event
In order to provide a more rubust and deterministic reordering
algorithm, we need to know when we reach a point where we just
did a pass through over every counter buffers to read every thing
they had.

This patch introduces a new PERF_RECORD_FINISHED_ROUND pseudo event
that only consist in an event header and doesn't need to contain
anything.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-09 13:43:42 +02:00
Ingo Molnar
ed82702155 Merge branch 'perf' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 into perf/core 2010-05-08 10:02:57 +02:00
Arnaldo Carvalho de Melo
1cf4a0632c perf list: Improve the raw hw event descriptor documentation
It was x86 specific and imcomplete at that, improve the situation by
making it clear where the example provided applies and by adding the
URLs for the Intel and AMD manuals where this is discussed in depth.

Acked-by: Robert Richter <robert.richter@amd.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Frédéric 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: Tom Zanussi <tzanussi@gmail.com>
Cc: Robert Richter <robert.richter@amd.com>
Reported-by: Robert Richter <robert.richter@amd.com
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-07 14:07:05 -03:00
Peter Zijlstra
ab608344bc perf, x86: Improve the PEBS ABI
Rename perf_event_attr::precise to perf_event_attr::precise_ip and
widen it to 2 bits. This new field describes the required precision of
the PERF_SAMPLE_IP field:

  0 - SAMPLE_IP can have arbitrary skid
  1 - SAMPLE_IP must have constant skid
  2 - SAMPLE_IP requested to have 0 skid
  3 - SAMPLE_IP must have 0 skid

And modify the Intel PEBS code accordingly. The PEBS implementation
now supports up to precise_ip == 2, where we perform the IP fixup.

Also s/PERF_RECORD_MISC_EXACT/&_IP/ to clarify its meaning, this bit
should be set for each PERF_SAMPLE_IP field known to match the actual
instruction triggering the event.

This new scheme allows for a PEBS mode that uses the buffer for more
than a single event.

Signed-off-by: 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: Ingo Molnar <mingo@elte.hu>
2010-05-07 11:31:02 +02:00
Arnaldo Carvalho de Melo
9e32a3cb06 perf list: Add explanation about raw hardware event descriptors
Using explanation given by Ingo Molnar in the oprofile mailing list.

Suggested-by: Nick Black <dank@qemfd.net>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Nick Black <dank@qemfd.net>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-05 11:20:05 -03:00
Tom Zanussi
db620b1c2f perf/record: simplify TRACE_INFO tracepoint check
Fix a couple of inefficiencies and redundancies related to
have_tracepoints() and its use when checking whether to write
TRACE_INFO.

First, there's no need to use get_tracepoints_path() in
have_tracepoints() - we really just want the part that checks whether
any attributes correspondo to tracepoints.

Second, we really don't care about raw_samples per se - tracepoints
are always raw_samples.  In any case, the have_tracepoints() check
should be sufficient to decide whether or not to write TRACE_INFO.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>,
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1273030770.6383.6.camel@tropicana>
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-05 11:12:53 -03:00
Arnaldo Carvalho de Melo
9890948d85 perf report: Make dso__calc_col_width agree with hist_entry__dso_snprintf
The first was always using the ->long_name, while the later used
->short_name if verbose was not set, resulting in the dso column to be
much wider than needed most of the time.

Cc: Frédéric 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: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-05 09:49:48 -03:00
Ingo Molnar
c4f3b5a2d7 Merge branch 'perf' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 into perf/core 2010-05-04 18:31:47 +02:00
Anton Blanchard
02bf60aad7 perf: Fix performance issue with perf report
On a large machine we spend a lot of time in perf_header__find_attr when
running perf report.

If we are parsing a file without PERF_SAMPLE_ID then for each sample we call
perf_header__find_attr and loop through all counter IDs, never finding a match.
As the machine gets larger there are more per cpu counters and we spend an
awful lot of time in there.

The patch below initialises each sample id to -1ULL and checks for this in
perf_header__find_attr. We may need to do something more intelligent eventually
(eg a hash lookup from counter id to attr) but this at least fixes the most
common usage of perf report.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric B Munson <ebmunson@us.ibm.com>
Acked-by: Eric B Munson <ebmunson@us.ibm.com>
LKML-Reference: <20100504111915.GB14636@kryten>
Signed-off-by: Anton Blanchard <anton@samba.org>
--
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-04 10:54:09 -03:00
Tom Zanussi
63e0c7715a perf: record TRACE_INFO only if using tracepoints and SAMPLE_RAW
The current perf code implicitly assumes SAMPLE_RAW means tracepoints
are being used, but doesn't check for that.  It happily records the
TRACE_INFO even if SAMPLE_RAW is used without tracepoints, but when the
perf data is read it won't go any further when it finds TRACE_INFO but
no tracepoints, and displays misleading errors.

This adds a check for both in perf-record, and won't record TRACE_INFO
unless both are true.  This at least allows perf report -D to dump raw
events, and avoids triggering a misleading error condition in perf
trace.  It doesn't actually enable the non-tracepoint raw events to be
displayed in perf trace, since perf trace currently only deals with
tracepoint events.

Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1272865861.7932.16.camel@tropicana>
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-03 10:31:48 -03:00
Ingo Molnar
0806ebd974 Merge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing into perf/core 2010-05-03 08:29:35 +02:00
Tom Zanussi
454c407ec1 perf: add perf-inject builtin
Currently, perf 'live mode' writes build-ids at the end of the
session, which isn't actually useful for processing live mode events.

What would be better would be to have the build-ids sent before any of
the samples that reference them, which can be done by processing the
event stream and retrieving the build-ids on the first hit.  Doing
that in perf-record itself, however, is off-limits.

This patch introduces perf-inject, which does the same job while
leaving perf-record untouched.  Normal mode perf still records the
build-ids at the end of the session as it should, but for live mode,
perf-inject can be injected in between the record and report steps
e.g.:

perf record -o - ./hackbench 10 | perf inject -v -b | perf report -v -i -

perf-inject reads a perf-record event stream and repipes it to stdout.
At any point the processing code can inject other events into the
event stream - in this case build-ids (-b option) are read and
injected as needed into the event stream.

Build-ids are just the first user of perf-inject - potentially
anything that needs userspace processing to augment the trace stream
with additional information could make use of this facility.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1272696080-16435-3-git-send-email-tzanussi@gmail.com>
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-02 13:36:56 -03:00
Tom Zanussi
789688faef perf/live: don't synthesize build ids at the end of a live mode trace
It doesn't really make sense to record the build ids at the end of a
live mode session - live mode samples need that information during the
trace rather than at the end.

Leave event__synthesize_build_id() in place, however; we'll still be
using that to synthesize build ids in a more timely fashion in a
future patch.

Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1272696080-16435-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-02 12:04:05 -03:00
Arnaldo Carvalho de Melo
fb72014d98 perf tools: Don't use code surrounded by __KERNEL__
We need to refactor code to be explicitely shared by the kernel and at
least the tools/ userspace programs, so, till we do that, copy the bare
minimum bitmap/bitops code needed by tools/perf.

Reported-by: "H. Peter Anvin" <hpa@zytor.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-02 12:00:44 -03:00
Frederic Weisbecker
d00a47cce5 perf: Fix warning while reading ring buffer headers
commit e9e94e3bd8
"perf trace: Ignore "overwrite" field if present in
/events/header_page" makes perf trace launching spurious warnings
about unexpected tokens read:

	Warning: Error: expected type 6 but read 4

This change tries to handle the overcommit field in the header_page
file whenever this field is present or not.

The problem is that if this field is not present, we try to find it
and give up in the middle of the line when we realize we are actually
dealing with another field, which is the "data" one. And this failure
abandons the file pointer in the middle of the "data" description
line:

	field: u64 timestamp;	offset:0;	size:8;	signed:0;
	field: local_t commit;	offset:8;	size:8;	signed:1;
	field: char data;	offset:16;	size:4080;	signed:1;
                      ^^^
                      Here

What happens next is that we want to read this line to parse the data
field, but we fail because the pointer is not in the beginning of the
line.

We could probably fix that by rewinding the pointer. But in fact we
don't care much about these headers that only concern the ftrace
ring-buffer. We don't use them from perf.

Just skip this part of perf.data, but don't remove it from recording
to stay compatible with olders perf.data

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
2010-05-01 04:31:48 +02:00
Arnaldo Carvalho de Melo
5c0541d53e perf symbols: Add machine helper routines
Created when writing the first 'perf test' regression testing routine.

Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-29 15:25:23 -03:00
Arnaldo Carvalho de Melo
cbf6968098 perf machines: Make the machines class adopt the dsos__fprintf methods
Now those methods don't operate on a global list of dsos, but on lists
of machines, so make this clear by renaming the functions.

Cc: Avi Kivity <avi@redhat.com>
Cc: Frédéric 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: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-27 21:22:44 -03:00
Arnaldo Carvalho de Melo
d28c62232e perf machine: Adopt some map_groups functions
Those functions operated on members now grouped in 'struct machine', so
move those methods to this new class.

The changes made to 'perf probe' shows that using this abstraction
inserting probes on guests almost got supported for free.

Cc: Avi Kivity <avi@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-27 21:21:18 -03:00
Arnaldo Carvalho de Melo
48ea8f5470 perf machine: Pass buffer size to machine__mmap_name
Don't blindly assume that the size of the buffer is enough, use
snprintf.

Cc: Avi Kivity <avi@redhat.com>
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: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-27 21:19:05 -03:00
Arnaldo Carvalho de Melo
23346f21b2 perf tools: Rename "kernel_info" to "machine"
struct kernel_info and kerninfo__ are too vague, what they really
describe are machines, virtual ones or hosts.

There are more changes to introduce helpers to shorten function calls
and to make more clear what is really being done, but I left that for
subsequent patches.

Cc: Avi Kivity <avi@redhat.com>
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: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-27 21:17:50 -03:00
Ingo Molnar
462b04e28a Merge branch 'perf' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 into perf/core 2010-04-27 11:16:54 +02:00
Masami Hiramatsu
ef4a356574 perf probe: Add --max-probes option
Add --max-probes option to change the maximum limit of
findable probe points per event, since inlined function can be
expanded into thousands of probe points. Default value is 128.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Suggested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <20100421195640.24664.62984.stgit@localhost6.localdomain6>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-26 15:35:20 -03:00
Masami Hiramatsu
5d1ee0413c perf probe: Fix to exit callback soon after finding too many probe points
Fix to exit callback soon after finding too many probe points.
Don't try to continue searching because it already failed.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <20100421195632.24664.42598.stgit@localhost6.localdomain6>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-26 15:33:08 -03:00
Masami Hiramatsu
15eca306ec perf probe: Fix to use symtab only if no debuginfo
Fix perf probe to use symtab only if there is no debuginfo, because debuginfo
has more information than symtab.

If we can't find a function in debuginfo, we never find it in symtab.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <20100421195624.24664.46214.stgit@localhost6.localdomain6>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-26 15:32:37 -03:00
Masami Hiramatsu
0ab061cd52 perf tools: Initialize dso->node member in dso__new
If dso->node member is not initialized, it causes a segmentation fault when
adding to other lists.

It should be initilized in dso__new().

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: : <20100421195616.24664.89980.stgit@localhost6.localdomain6>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-26 15:31:32 -03:00
Frederic Weisbecker
c61e52ee70 perf: Generalize perf lock's sample event reordering to the session layer
The sample events recorded by perf record are not time ordered
because we have one buffer per cpu for each event (even demultiplexed
per task/per cpu for task bound events). But when we read trace events
we want them to be ordered by time because many state machines are
involved.

There are currently two ways perf tools deal with that:

- use -M to multiplex every buffers (perf sched, perf kmem)
  But this creates a lot of contention in SMP machines on
  record time.

- use a post-processing time reordering (perf timechart, perf lock)
  The reordering used by timechart is simple but doesn't scale well
  with huge flow of events, in terms of performance and memory use
  (unusable with perf lock for example).
  Perf lock has its own samples reordering that flushes its memory
  use in a regular basis and that uses a sorting based on the
  previous event queued (a new event to be queued is close to the
  previous one most of the time).

This patch proposes to export perf lock's samples reordering facility
to the session layer that reads the events. So if a tool wants to
get ordered sample events, it needs to set its
struct perf_event_ops::ordered_samples to true and that's it.

This prepares tracing based perf tools to get rid of the need to
use buffers multiplexing (-M) or to implement their own
reordering.

Also lower the flush period to 2 as it's sufficient already.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 03:49:58 +02:00
Stephane Eranian
5710fcad7c perf: Fix initialization bug in parse_single_tracepoint_event()
The parse_single_tracepoint_event() was setting some attributes
before it validated the event was indeed a tracepoint event. This
caused problems with other initialization routines like in the
builtin-top.c module whereby sample_period is not set if not 0.

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <4bcf232b.698fd80a.6fbe.ffffb737@mx.google.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
2010-04-24 03:24:09 +02:00
Ian Munsie
cd932c5939 perf: Move arch specific code into separate arch directory
The perf userspace tool included some architecture specific code to map
registers from the DWARF register number into the names used by the regs
and stack access API.

This moves the architecture specific code out into a separate
arch/x86 directory along with the infrastructure required to use it.

Signed-off-by: Ian Munsie <imunsie@au.ibm.com>
Acked-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
2010-04-22 13:48:31 +10:00
Frederic Weisbecker
6eca8cc35b perf: Fix perf probe build error
When we run into dry run mode, we want to make
write_kprobe_trace_event to succeed on writing the event. Let's
initialize it to 0.

Fixes the following build error:
	util/probe-event.c:1266: attention : «ret» may be used uninitialized in this function
	util/probe-event.c:1266: note: «ret» was declared here

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <1271808065-25290-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-04-21 09:39:52 +02:00
Zhang, Yanmin
a1645ce12a perf: 'perf kvm' tool for monitoring guest performance from host
Here is the patch of userspace perf tool.

Signed-off-by: Zhang Yanmin <yanmin_zhang@linux.intel.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
2010-04-19 12:37:24 +03:00
Ingo Molnar
b5a80b7e91 Merge branch 'perf' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 into perf/core 2010-04-15 09:16:51 +02:00
Ingo Molnar
84b13fd596 Merge branch 'perf/live' into perf/core
Conflicts:
	tools/perf/builtin-record.c

Merge reason: add the live tracing feature, resolve conflict.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-04-15 09:13:26 +02:00
Frederic Weisbecker
f921281930 perf: Make the trace events sample period default to 1
Trace events are mostly used for tracing and then require not to
be lost when possible. As opposite to hardware events that really
require to trigger after a given sample period, trace events mostly
need to trigger everytime.

It is a frustrating experience to trace with perf and realize we
lost a lot of events because we forgot the "-c 1" option.

Then default sample_period to 1 for trace events but let the user
override it.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2010-04-15 04:12:52 +02:00
Frederic Weisbecker
bdef3b02ce perf: Always record tracepoints raw samples from perf record
Trace events are mostly used for tracing rather than simple
counting. Don't bother anymore with adding -R when using them,
just record raw samples of trace events every time.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
2010-04-15 04:12:52 +02:00
Thomas Gleixner
a1e2f60e3e perf: Fix dynamic field detection
Checking if a tracing field is an array with a dynamic length
requires to check the field type and seek the "__data_loc"
string that prepends the actual type, as can be found in a trace
event format file:

	field:__data_loc char[] name;	offset:16;	size:4;	signed:1;

But we actually use strcmp() to check if the field type fully
matches "__data_loc", which may fail as we trip over the rest of
the type.

To fix this, use strncmp to only check if it starts with
"__data_loc".

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1271282283-23721-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-04-15 01:34:46 +02:00
Masami Hiramatsu
f6c903f585 perf probe: Show function entry line as probe-able
Function entry line should be shown as probe-able line,
because each function has declared line attribute.

LKML-Reference: <20100414224007.14630.96915.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:45:39 -03:00
Masami Hiramatsu
de1439d8a5 perf probe: Support DW_OP_plus_uconst in DW_AT_data_member_location
DW_OP_plus_uconst can be used for DW_AT_data_member_location.
This patch adds DW_OP_plus_uconst support when getting
structure member offset.

Commiter note:

Fixed up the size_t format specifier in one case:

cc1: warnings being treated as errors
util/probe-finder.c: In function ‘die_get_data_member_location’:
util/probe-finder.c:270: error: format ‘%d’ expects type ‘int’, but argument 4 has type ‘size_t’
make: *** [/home/acme/git/build/perf/util/probe-finder.o] Error 1

LKML-Reference: <20100414223958.14630.5230.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:44:00 -03:00
Masami Hiramatsu
dda4ab34fe perf probe: Fix line range to show end line
Line range should reject the range if the number of lines is 0
(e.g. "sched.c:1024+0"), and it should show the lines include
the end of line number (e.g. "sched.c:1024-2048" should show
2048th line).

LKML-Reference: <20100414223950.14630.42263.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:41:30 -03:00
Masami Hiramatsu
d3b63d7ae0 perf probe: Fix a bug that --line range can be overflow
Since line_finder.lno_s/e are signed int but line_range.start/end
are unsigned int, it is possible to be overflow when converting
line_range->start/end to line_finder->lno_s/e.
This changes line_range.start/end and line_list.line to signed int
and adds overflow checks when setting line_finder.lno_s/e.

LKML-Reference: <20100414223942.14630.72730.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:41:21 -03:00
Masami Hiramatsu
dd259c5db2 perf probe: Fix mis-estimation for shortening filename
Fix mis-estimation size for making a short filename.
Since the buffer size is 32 bytes and there are '@' prefix and
'\0' termination, maximum shorten filename length should be
30. This means, before searching '/', it should be 31 bytes.

LKML-Reference: <20100414223935.14630.11954.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:41:14 -03:00
Masami Hiramatsu
7ca5989dd0 perf probe: Fix to use correct debugfs path finder
Instead of using debugfs_path, use debugfs_find_mountpoint()
to find actual debugfs path.

LKML-Reference: <20100414223928.14630.38326.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Reported-by: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:41:06 -03:00
Masami Hiramatsu
02b95dadc8 perf probe: Remove xstrdup()/xstrndup() from util/probe-{event, finder}.c
Remove all xstr*dup() calls from util/probe-{event,finder}.c since
it may cause 'sudden death' in utility functions and it makes
reusing it from other code difficult.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171756.3790.89607.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:28:52 -03:00
Masami Hiramatsu
e334016f1d perf probe: Remove xzalloc() from util/probe-{event, finder}.c
Remove all xzalloc() calls from util/probe-{event,finder}.c since
it may cause 'sudden death' in utility functions and it makes
reusing it from other code difficult.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171749.3790.33303.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:28:45 -03:00
Masami Hiramatsu
146a143948 perf probe: Remove die() from probe-event code
Remove die() and DIE_IF() code from util/probe-event.c since
these 'sudden death' in utility functions make reusing it from
other code (especially tui/gui) difficult.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171742.3790.33650.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:28:36 -03:00
Masami Hiramatsu
b55a87ade3 perf probe: Remove die() from probe-finder code
Remove die() and DIE_IF() code from util/probe-finder.c since
these 'sudden death' in utility functions make reusing it from
other code (especially tui/gui) difficult.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171735.3790.88853.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:28:30 -03:00
Masami Hiramatsu
a34a985499 perf probe: Support DW_OP_call_frame_cfa in debuginfo
When building kernel without CONFIG_FRAME_POINTER, gcc uses
CFA (canonical frame address) for frame base. With this patch,
perf probe just gets CFI (call frame information) from debuginfo
and search corresponding CFA from the CFI. IOW, this allows
perf probe works correctly on the kernel without CONFIG_FRAME_POINTER.

<Before>
 ./perf probe -fn sched_slice:12 lw.weight
  Fatal: DW_OP 156 is not supported.
              (^^^ DW_OP_call_frame_cfa)

<After>
./perf probe -fn sched_slice:12 lw.weight
Add new event:
  probe:sched_slice    (on sched_slice:12 with weight=lw.weight)

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171728.3790.98217.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:28:20 -03:00
Masami Hiramatsu
11a1ca3554 perf probe: Support basic type casting
Add basic type casting for arguments to perf probe. This allows
users to specify the actual type of arguments. Of course, if
user sets invalid types, kprobe-tracer rejects that.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171722.3790.50372.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:28:09 -03:00
Masami Hiramatsu
4984912eb2 perf probe: Query basic types from debuginfo
Query the basic type information (byte-size and signed-flag) from
debuginfo and pass that to kprobe-tracer. This is especially useful
for tracing the members of data structure, because each member has
different byte-size on the memory.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171715.3790.23730.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:27:56 -03:00
Masami Hiramatsu
df0faf4be0 perf probe: Use the last field name as the argument name
Set the last field name to the argument name when the argument
is refering a data-structure member.

e.g.
 ./perf probe --add 'vfs_read file->f_mode'
 Add new event:
   probe:vfs_read       (on vfs_read with f_mode=file->f_mode)

 This probe records file->f_mode, but the argument name becomes "f_mode".

This enables perf-trace command to parse trace event format correctly.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100412171700.3790.72961.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-04-14 17:26:14 -03:00