Adding a generic way to use __output_copy function with specific copy
function via DEFINE_PERF_OUTPUT_COPY macro.
Using this to add new __output_copy_user function, that provides output
copy from user pointers. For x86 the copy_from_user_nmi function is used
and __copy_from_user_inatomic for the rest of the architectures.
This new function will be used in user stack dump on sample, coming in
next patches.
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Cc: "Frank Ch. Eigler" <fche@redhat.com>
Cc: Arun Sharma <asharma@fb.com>
Cc: Benjamin Redelings <benjamin.redelings@nescent.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Ulrich Drepper <drepper@gmail.com>
Link: http://lkml.kernel.org/r/1344345647-11536-4-git-send-email-jolsa@redhat.com
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Introducing PERF_SAMPLE_REGS_USER sample type bit to trigger the dump of
user level registers on sample. Registers we want to dump are specified
by sample_regs_user bitmask.
Only user level registers are dumped at the moment. Meaning the register
values of the user space context as it was before the user entered the
kernel for whatever reason (syscall, irq, exception, or a PMI happening
in userspace).
The layout of the sample_regs_user bitmap is described in
asm/perf_regs.h for archs that support register dump.
This is going to be useful to bring Dwarf CFI based stack unwinding on
top of samples.
Original-patch-by: Frederic Weisbecker <fweisbec@gmail.com>
[ Dump registers ABI specification. ]
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Suggested-by: Stephane Eranian <eranian@google.com>
Cc: "Frank Ch. Eigler" <fche@redhat.com>
Cc: Arun Sharma <asharma@fb.com>
Cc: Benjamin Redelings <benjamin.redelings@nescent.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Ulrich Drepper <drepper@gmail.com>
Link: http://lkml.kernel.org/r/1344345647-11536-3-git-send-email-jolsa@redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
A few events are interesting not only for a current task.
For example, sched_stat_* events are interesting for a task
which wakes up. For this reason, it will be good if such
events will be delivered to a target task too.
Now a target task can be set by using __perf_task().
The original idea and a draft patch belongs to Peter Zijlstra.
I need these events for profiling sleep times. sched_switch is used for
getting callchains and sched_stat_* is used for getting time periods.
These events are combined in user space, then it can be analyzed by
perf tools.
Inspired-by: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Arun Sharma <asharma@fb.com>
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1342016098-213063-1-git-send-email-avagin@openvz.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Originally from Peter Zijlstra. The helper migrates perf events
from one cpu to another cpu.
Signed-off-by: Zheng Yan <zheng.z.yan@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1339741902-8449-5-git-send-email-zheng.z.yan@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The rdpmc instruction is faster than the equivelant rdmsr call,
so use it when possible in the kernel.
The perfctr kernel patches did this, after extensive testing showed
rdpmc to always be faster (One can look in etc/costs in the perfctr-2.6
package to see a historical list of the overhead).
I have done some tests on a 3.2 kernel, the kernel module I used
was included in the first posting of this patch:
rdmsr rdpmc
Core2 T9900: 203.9 cycles 30.9 cycles
AMD fam0fh: 56.2 cycles 9.8 cycles
Atom 6/28/2: 129.7 cycles 50.6 cycles
The speedup of using rdpmc is large.
[ It's probably possible (and desirable) to do this without
requiring a new field in the hw_perf_event structure, but
the fixed events make this tricky. ]
Signed-off-by: Vince Weaver <vweaver1@eecs.utk.edu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1203011724030.26934@cl320.eecs.utk.edu
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We faced segmentation fault on perf top -G at very high sampling rate
due to a corrupted callchain. While the root cause was not revealed (I
failed to figure it out), this patch tries to protect us from the
segfault on such cases.
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Namhyung Kim <namhyung.kim@lge.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Sunjin Yang <fan4326@gmail.com>
Link: http://lkml.kernel.org/r/1338443007-24857-2-git-send-email-namhyung.kim@lge.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This reverts commit cb04ff9ac4 ("sched, perf: Use a single
callback into the scheduler").
Before this change was introduced, the process switch worked
like this (wrt. to perf event schedule):
schedule (prev, next)
- schedule out all perf events for prev
- switch to next
- schedule in all perf events for current (next)
After the commit, the process switch looks like:
schedule (prev, next)
- schedule out all perf events for prev
- schedule in all perf events for (next)
- switch to next
The problem is, that after we schedule perf events in, the pmu
is enabled and we can receive events even before we make the
switch to next - so "current" still being prev process (event
SAMPLE data are filled based on the value of the "current"
process).
Thats exactly what we see for test__PERF_RECORD test. We receive
SAMPLES with PID of the process that our tracee is scheduled
from.
Discussed with Peter Zijlstra:
> Bah!, yeah I guess reverting is the right thing for now. Sad
> though.
>
> So by having the two hooks we have a black-spot between them
> where we receive no events at all, this black-spot covers the
> hand-over of current and we thus don't receive the 'wrong'
> events.
>
> I rather liked we could do away with both that black-spot and
> clean up the code a little, but apparently people rely on it.
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: acme@redhat.com
Cc: paulus@samba.org
Cc: cjashfor@linux.vnet.ibm.com
Cc: fweisbec@gmail.com
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/20120523111302.GC1638@m.brq.redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We can easily use a single callback for both sched-in and sched-out. This
reduces the code footprint in the scheduler path as well as removes
the PMU black spot otherwise present between the out and in callback.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-o56ajxp1edwqg6x9d31wb805@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We always need to pass the last sample period to
perf_sample_data_init(), otherwise the event distribution will be
wrong. Thus, modifiyng the function interface with the required period
as argument. So basically a pattern like this:
perf_sample_data_init(&data, ~0ULL);
data.period = event->hw.last_period;
will now be like that:
perf_sample_data_init(&data, ~0ULL, event->hw.last_period);
Avoids unininitialized data.period and simplifies code.
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/1333390758-10893-3-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Complete the syscall-less self-profiling feature and address
all complaints, namely:
- capabilities, so we can detect what is actually available at runtime
Add a capabilities field to perf_event_mmap_page to indicate
what is actually available for use.
- on x86: RDPMC weirdness due to being 40/48 bits and not sign-extending
properly.
- ABI documentation as to how all this stuff works.
Also improve the documentation for the new features.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vweaver1@eecs.utk.edu>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1332433596.2487.33.camel@twins
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Adding sysfs group 'format' attribute for pmu device that
contains a syntax description on how to construct raw events.
The event configuration is described in following
struct pefr_event_attr attributes:
config
config1
config2
Each sysfs attribute within the format attribute group,
describes mapping of name and bitfield definition within
one of above attributes.
eg:
"/sys/...<dev>/format/event" contains "config:0-7"
"/sys/...<dev>/format/umask" contains "config:8-15"
"/sys/...<dev>/format/usr" contains "config:16"
the attribute value syntax is:
line: config ':' bits
config: 'config' | 'config1' | 'config2"
bits: bits ',' bit_term | bit_term
bit_term: VALUE '-' VALUE | VALUE
Adding format attribute definitions for x86 cpu pmus.
Acked-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/n/tip-vhdk5y2hyype9j63prymty36@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
With branch stack sampling, it is possible to filter by priv levels.
In system-wide mode, that means it is possible to capture only user
level branches. The builtin SW LBR filter needs to disassemble code
based on LBR captured addresses. For that, it needs to know the task
the addresses are associated with. Because of context switches, the
content of the branch stack buffer may contain addresses from
different tasks.
We need a callback on context switch to either flush the branch stack
or save it. This patch adds a new callback in struct pmu which is called
during context switches. The callback is called only when necessary.
That is when a system-wide context has, at least, one event which
uses PERF_SAMPLE_BRANCH_STACK. The callback is never called for
per-thread context.
In this version, the Intel x86 code simply flushes (resets) the LBR
on context switches (fills it with zeroes). Those zeroed branches are
then filtered out by the SW filter.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1328826068-11713-11-git-send-email-eranian@google.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch adds the ability to sample taken branches to the
perf_event interface.
The ability to capture taken branches is very useful for all
sorts of analysis. For instance, basic block profiling, call
counts, statistical call graph.
This new capability requires hardware assist and as such may
not be available on all HW platforms. On Intel x86 it is
implemented on top of the Last Branch Record (LBR) facility.
To enable taken branches sampling, the PERF_SAMPLE_BRANCH_STACK
bit must be set in attr->sample_type.
Sampled taken branches may be filtered by type and/or priv
levels.
The patch adds a new field, called branch_sample_type, to the
perf_event_attr structure. It contains a bitmask of filters
to apply to the sampled taken branches.
Filters may be implemented in HW. If the HW filter does not exist
or is not good enough, some arch may also implement a SW filter.
The following generic filters are currently defined:
- PERF_SAMPLE_USER
only branches whose targets are at the user level
- PERF_SAMPLE_KERNEL
only branches whose targets are at the kernel level
- PERF_SAMPLE_HV
only branches whose targets are at the hypervisor level
- PERF_SAMPLE_ANY
any type of branches (subject to priv levels filters)
- PERF_SAMPLE_ANY_CALL
any call branches (may incl. syscall on some arch)
- PERF_SAMPLE_ANY_RET
any return branches (may incl. syscall returns on some arch)
- PERF_SAMPLE_IND_CALL
indirect call branches
Obviously filter may be combined. The priv level bits are optional.
If not provided, the priv level of the associated event are used. It
is possible to collect branches at a priv level different from the
associated event. Use of kernel, hv priv levels is subject to permissions
and availability (hv).
The number of taken branch records present in each sample may vary based
on HW, the type of sampled branches, the executed code. Therefore
each sample contains the number of taken branches it contains.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1328826068-11713-2-git-send-email-eranian@google.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
So here's a boot tested patch on top of Jason's series that does
all the cleanups I talked about and turns jump labels into a
more intuitive to use facility. It should also address the
various misconceptions and confusions that surround jump labels.
Typical usage scenarios:
#include <linux/static_key.h>
struct static_key key = STATIC_KEY_INIT_TRUE;
if (static_key_false(&key))
do unlikely code
else
do likely code
Or:
if (static_key_true(&key))
do likely code
else
do unlikely code
The static key is modified via:
static_key_slow_inc(&key);
...
static_key_slow_dec(&key);
The 'slow' prefix makes it abundantly clear that this is an
expensive operation.
I've updated all in-kernel code to use this everywhere. Note
that I (intentionally) have not pushed through the rename
blindly through to the lowest levels: the actual jump-label
patching arch facility should be named like that, so we want to
decouple jump labels from the static-key facility a bit.
On non-jump-label enabled architectures static keys default to
likely()/unlikely() branches.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Jason Baron <jbaron@redhat.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: a.p.zijlstra@chello.nl
Cc: mathieu.desnoyers@efficios.com
Cc: davem@davemloft.net
Cc: ddaney.cavm@gmail.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20120222085809.GA26397@elte.hu
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Adding perf registration support for the ftrace function event,
so it is now possible to register it via perf interface.
The perf_event struct statically contains ftrace_ops as a handle
for function tracer. The function tracer is registered/unregistered
in open/close actions.
To be efficient, we enable/disable ftrace_ops each time the traced
process is scheduled in/out (via TRACE_REG_PERF_(ADD|DELL) handlers).
This way tracing is enabled only when the process is running.
Intentionally using this way instead of the event's hw state
PERF_HES_STOPPED, which would not disable the ftrace_ops.
It is now possible to use function trace within perf commands
like:
perf record -e ftrace:function ls
perf stat -e ftrace:function ls
Allowed only for root.
Link: http://lkml.kernel.org/r/1329317514-8131-6-git-send-email-jolsa@redhat.com
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
This patch fixes the sampling interrupt throttling mechanism.
It was broken in v3.2. Events were not being unthrottled. The
unthrottling mechanism required that events be checked at each
timer tick.
This patch solves this problem and also separates:
- unthrottling
- multiplexing
- frequency-mode period adjustments
Not all of them need to be executed at each timer tick.
This third version of the patch is based on my original patch +
PeterZ proposal (https://lkml.org/lkml/2012/1/7/87).
At each timer tick, for each context:
- if the current CPU has throttled events, we unthrottle events
- if context has frequency-based events, we adjust sampling periods
- if we have reached the jiffies interval, we multiplex (rotate)
We decoupled rotation (multiplexing) from frequency-mode sampling
period adjustments. They should not necessarily happen at the same
rate. Multiplexing is subject to jiffies_interval (currently at 1
but could be higher once the tunable is exposed via sysfs).
We have grouped frequency-mode adjustment and unthrottling into the
same routine to minimize code duplication. When throttled while in
frequency mode, we scan the events only once.
We have fixed the threshold enforcement code in __perf_event_overflow().
There was a bug whereby it would allow more than the authorized rate
because an increment of hwc->interrupts was not executed at the right
place.
The patch was tested with low sampling limit (2000) and fixed periods,
frequency mode, overcommitted PMU.
On a 2.1GHz AMD CPU:
$ cat /proc/sys/kernel/perf_event_max_sample_rate
2000
We set a rate of 3000 samples/sec (2.1GHz/3000 = 700000):
$ perf record -e cycles,cycles -c 700000 noploop 10
$ perf report -D | tail -21
Aggregated stats:
TOTAL events: 80086
MMAP events: 88
COMM events: 2
EXIT events: 4
THROTTLE events: 19996
UNTHROTTLE events: 19996
SAMPLE events: 40000
cycles stats:
TOTAL events: 40006
MMAP events: 5
COMM events: 1
EXIT events: 4
THROTTLE events: 9998
UNTHROTTLE events: 9998
SAMPLE events: 20000
cycles stats:
TOTAL events: 39996
THROTTLE events: 9998
UNTHROTTLE events: 9998
SAMPLE events: 20000
For 10s, the cap is 2x2000x10 = 40000 samples.
We get exactly that: 20000 samples/event.
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: <stable@kernel.org> # v3.2+
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120126160319.GA5655@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Extend the mmap control page with fields so that userspace can compute
time deltas relative to the provided time fields.
Currently only implemented for x86 with constant and nonstop TSC.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Arun Sharma <asharma@fb.com>
Link: http://lkml.kernel.org/n/tip-3u1jucza77j3wuvs0x2bic0f@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Put the logic to compute the event index into a per pmu method. This
is required because the x86 rules are weird and wonderful and don't
match the capabilities of the current scheme.
AFAIK only powerpc actually has a usable userspace read of the PMCs
but I'm not at all sure anybody actually used that.
ARM is restored to the default since it currently does not support
userspace access at all. And all software events are provided with a
method that reports their index as 0 (disabled).
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Michael Cree <mcree@orcon.net.nz>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: Anton Blanchard <anton@samba.org>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Richard Kuo <rkuo@codeaurora.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Arun Sharma <asharma@fb.com>
Link: http://lkml.kernel.org/n/tip-dfydxodki16lylkt3gl2j7cw@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This event counts the number of reference core cpu cycles.
Reference means that the event increments at a constant rate which
is not subject to core CPU frequency adjustments. The event may
not count when the processor is in halted (low power) state.
As such, it may not be equivalent to wall clock time. However,
when the processor is not halted state, the event keeps
a constant correlation with wall clock time.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1323559734-3488-3-git-send-email-eranian@google.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
jump_lable patching is very expensive operation that involves pausing all
cpus. The patching of perf_sched_events jump_label is easily controllable
from userspace by unprivileged user.
When te user runs a loop like this:
"while true; do perf stat -e cycles true; done"
... the performance of my test application that just increments a counter
for one second drops by 4%.
This is on a 16 cpu box with my test application using only one of
them. An impact on a real server doing real work will be worse.
Performance of KVM PMU drops nearly 50% due to jump_lable for "perf
record" since KVM PMU implementation creates and destroys perf event
frequently.
This patch introduces a way to rate limit jump_label patching and uses
it to fix the above problem.
I believe that as jump_label use will spread the problem will become more
common and thus solving it in a generic code is appropriate. Also fixing
it in the perf code would result in moving jump_label accounting logic to
perf code with all the ifdefs in case of JUMP_LABEL=n kernel. With this
patch all details are nicely hidden inside jump_label code.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Acked-by: Jason Baron <jbaron@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20111127155909.GO2557@redhat.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
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>
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>
The two new attributes exclude_guest and exclude_host can
bes used by user-space to tell the kernel to setup
performance counter to either only count while the CPU is in
guest or in host mode.
An additional check is also introduced to make sure
user-space does not try to exclude guest and host mode from
counting.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1317816084-18026-2-git-send-email-gleb@redhat.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The current cgroup context switch code was incorrect leading
to bogus counts. Furthermore, as soon as there was an active
cgroup event on a CPU, the context switch cost on that CPU
would increase by a significant amount as demonstrated by a
simple ping/pong example:
$ ./pong
Both processes pinned to CPU1, running for 10s
10684.51 ctxsw/s
Now start a cgroup perf stat:
$ perf stat -e cycles,cycles -A -a -G test -C 1 -- sleep 100
$ ./pong
Both processes pinned to CPU1, running for 10s
6674.61 ctxsw/s
That's a 37% penalty.
Note that pong is not even in the monitored cgroup.
The results shown by perf stat are bogus:
$ perf stat -e cycles,cycles -A -a -G test -C 1 -- sleep 100
Performance counter stats for 'sleep 100':
CPU1 <not counted> cycles test
CPU1 16,984,189,138 cycles # 0.000 GHz
The second 'cycles' event should report a count @ CPU clock
(here 2.4GHz) as it is counting across all cgroups.
The patch below fixes the bogus accounting and bypasses any
cgroup switches in case the outgoing and incoming tasks are
in the same cgroup.
With this patch the same test now yields:
$ ./pong
Both processes pinned to CPU1, running for 10s
10775.30 ctxsw/s
Start perf stat with cgroup:
$ perf stat -e cycles,cycles -A -a -G test -C 1 -- sleep 10
Run pong outside the cgroup:
$ /pong
Both processes pinned to CPU1, running for 10s
10687.80 ctxsw/s
The penalty is now less than 2%.
And the results for perf stat are correct:
$ perf stat -e cycles,cycles -A -a -G test -C 1 -- sleep 10
Performance counter stats for 'sleep 10':
CPU1 <not counted> cycles test # 0.000 GHz
CPU1 23,933,981,448 cycles # 0.000 GHz
Now perf stat reports the correct counts for
for the non cgroup event.
If we run pong inside the cgroup, then we also get the
correct counts:
$ perf stat -e cycles,cycles -A -a -G test -C 1 -- sleep 10
Performance counter stats for 'sleep 10':
CPU1 22,297,726,205 cycles test # 0.000 GHz
CPU1 23,933,981,448 cycles # 0.000 GHz
10.001457237 seconds time elapsed
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110825135803.GA4697@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This allows us to move duplicated code in <asm/atomic.h>
(atomic_inc_not_zero() for now) to <linux/atomic.h>
Signed-off-by: Arun Sharma <asharma@fb.com>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
KVM needs one-shot samples, since a PMC programmed to -X will fire after X
events and then again after 2^40 events (i.e. variable period).
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1309362157-6596-4-git-send-email-avi@redhat.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The perf_event overflow handler does not receive any caller-derived
argument, so many callers need to resort to looking up the perf_event
in their local data structure. This is ugly and doesn't scale if a
single callback services many perf_events.
Fix by adding a context parameter to perf_event_create_kernel_counter()
(and derived hardware breakpoints APIs) and storing it in the perf_event.
The field can be accessed from the callback as event->overflow_handler_context.
All callers are updated.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1309362157-6596-2-git-send-email-avi@redhat.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add a NODE level to the generic cache events which is used to measure
local vs remote memory accesses. Like all other cache events, an
ACCESS is HIT+MISS, if there is no way to distinguish between reads
and writes do reads only etc..
The below needs filling out for !x86 (which I filled out with
unsupported events).
I'm fairly sure ARM can leave it like that since it doesn't strike me as
an architecture that even has NUMA support. SH might have something since
it does appear to have some NUMA bits.
Sparc64, PowerPC and MIPS certainly want a good look there since they
clearly are NUMA capable.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: David Miller <davem@davemloft.net>
Cc: Anton Blanchard <anton@samba.org>
Cc: David Daney <ddaney@caviumnetworks.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1303508226.4865.8.camel@laptop
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch improves the code managing the extra shared registers
used for offcore_response events on Intel Nehalem/Westmere. The
idea is to use static allocation instead of dynamic allocation.
This simplifies greatly the get and put constraint routines for
those events.
The patch also renames per_core to shared_regs because the same
data structure gets used whether or not HT is on. When HT is
off, those events still need to coordination because they use
a extra MSR that has to be shared within an event group.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110606145703.GA7258@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Since only samples call perf_output_sample() its much saner (and more
correct) to put the sample logic in there than in the
perf_output_begin()/perf_output_end() pair.
Saves a useless argument, reduces conditionals and shrinks
struct perf_output_handle, win!
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-2crpvsx3cqu67q3zqjbnlpsc@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The nmi parameter indicated if we could do wakeups from the current
context, if not, we would set some state and self-IPI and let the
resulting interrupt do the wakeup.
For the various event classes:
- hardware: nmi=0; PMI is in fact an NMI or we run irq_work_run from
the PMI-tail (ARM etc.)
- tracepoint: nmi=0; since tracepoint could be from NMI context.
- software: nmi=[0,1]; some, like the schedule thing cannot
perform wakeups, and hence need 0.
As one can see, there is very little nmi=1 usage, and the down-side of
not using it is that on some platforms some software events can have a
jiffy delay in wakeup (when arch_irq_work_raise isn't implemented).
The up-side however is that we can remove the nmi parameter and save a
bunch of conditionals in fast paths.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Michael Cree <mcree@orcon.net.nz>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: Anton Blanchard <anton@samba.org>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Don Zickus <dzickus@redhat.com>
Link: http://lkml.kernel.org/n/tip-agjev8eu666tvknpb3iaj0fg@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reorder perf_event_context to remove 8 bytes of 64 bit alignment padding
shrinking its size to 192 bytes, allowing it to fit into a smaller slab
and use one fewer cache lines.
Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1307460819.1950.5.camel@castor.rsk
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Fix include/linux/perf_event.h comments to be consistent with
the actual #define names. This is trivial, but it can be a bit
confusing when first reading through the file.
Signed-off-by: Vince Weaver <vweaver1@eecs.utk.edu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus@samba.org
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1106031757090.29381@cl320.eecs.utk.edu
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add two generic hardware events: front-end and back-end stalled cycles.
These events measure conditions when the CPU is executing code but its
capabilities are not fully utilized. Understanding such situations and
analyzing them is an important sub-task of code optimization workflows.
Both events limit performance: most front end stalls tend to be caused
by branch misprediction or instruction fetch cachemisses, backend
stalls can be caused by various resource shortages or inefficient
instruction scheduling.
Front-end stalls are the more important ones: code cannot run fast
if the instruction stream is not being kept up.
An over-utilized back-end can cause front-end stalls and thus
has to be kept an eye on as well.
The exact composition is very program logic and instruction mix
dependent.
We use the terms 'stall', 'front-end' and 'back-end' loosely and
try to use the best available events from specific CPUs that
approximate these concepts.
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-7y40wib8n000io7hjpn1dsrm@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The new PERF_COUNT_HW_STALLED_CYCLES event tries to approximate
cycles the CPU does nothing useful, because it is stalled on a
cache-miss or some other condition.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-fue11vymwqsoo5to72jxxjyl@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Introduce:
static __always_inline bool static_branch(struct jump_label_key *key);
instead of the old JUMP_LABEL(key, label) macro.
In this way, jump labels become really easy to use:
Define:
struct jump_label_key jump_key;
Can be used as:
if (static_branch(&jump_key))
do unlikely code
enable/disale via:
jump_label_inc(&jump_key);
jump_label_dec(&jump_key);
that's it!
For the jump labels disabled case, the static_branch() becomes an
atomic_read(), and jump_label_inc()/dec() are simply atomic_inc(),
atomic_dec() operations. We show testing results for this change below.
Thanks to H. Peter Anvin for suggesting the 'static_branch()' construct.
Since we now require a 'struct jump_label_key *key', we can store a pointer into
the jump table addresses. In this way, we can enable/disable jump labels, in
basically constant time. This change allows us to completely remove the previous
hashtable scheme. Thanks to Peter Zijlstra for this re-write.
Testing:
I ran a series of 'tbench 20' runs 5 times (with reboots) for 3
configurations, where tracepoints were disabled.
jump label configured in
avg: 815.6
jump label *not* configured in (using atomic reads)
avg: 800.1
jump label *not* configured in (regular reads)
avg: 803.4
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110316212947.GA8792@redhat.com>
Signed-off-by: Jason Baron <jbaron@redhat.com>
Suggested-by: H. Peter Anvin <hpa@linux.intel.com>
Tested-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Jiri reported:
|
| - once an event is created by sys_perf_event_open, task context
| is created and it stays even if the event is closed, until the
| task is finished ... thats what I see in code and I assume it's
| correct
|
| - when the task opens event, perf_sched_events jump label is
| incremented and following callbacks are started from scheduler
|
| __perf_event_task_sched_in
| __perf_event_task_sched_out
|
| These callback *in/out set/unset cpuctx->task_ctx value to the
| task context.
|
| - close is called on event on CPU 0:
| - the task is scheduled on CPU 0
| - __perf_event_task_sched_in is called
| - cpuctx->task_ctx is set
| - perf_sched_events jump label is decremented and == 0
| - __perf_event_task_sched_out is not called
| - cpuctx->task_ctx on CPU 0 stays set
|
| - exit is called on CPU 1:
| - the task is scheduled on CPU 1
| - perf_event_exit_task is called
| - task_ctx_sched_out unsets cpuctx->task_ctx on CPU 1
| - put_ctx destroys the context
|
| - another call of perf_rotate_context on CPU 0 will use invalid
| task_ctx pointer, and eventualy panic.
|
Cure this the simplest possibly way by partially reverting the
jump_label optimization for the sched_out case.
Reported-and-tested-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: <stable@kernel.org> # .37+
LKML-Reference: <1301520405.4859.213.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch solves a stale pointer problem in
update_cgrp_time_from_cpuctx(). The cpuctx->cgrp
was not cleared on all possible event exit paths,
including:
close()
perf_release()
perf_release_kernel()
list_del_event()
This patch fixes list_del_event() to clear cpuctx->cgrp
when there are no cgroup events left in the context.
[ This second version makes the code compile when
CONFIG_CGROUP_PERF is not enabled. We unconditionally define
perf_cpu_context->cgrp. ]
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: peterz@infradead.org
Cc: perfmon2-devel@lists.sf.net
Cc: paulus@samba.org
Cc: davem@davemloft.net
LKML-Reference: <20110323150306.GA1580@quad>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Remove 8 bytes of alignment padding from perf_event_context on 64 bit
builds which shrinks its size to 192 bytes allowing it to fit into one
fewer cache lines and into a smaller slab.
Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1299512819.2039.5.camel@castor.rsk>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Change logs against Andi's original version:
- Extends perf_event_attr:config to config{,1,2} (Peter Zijlstra)
- Fixed a major event scheduling issue. There cannot be a ref++ on an
event that has already done ref++ once and without calling
put_constraint() in between. (Stephane Eranian)
- Use thread_cpumask for percore allocation. (Lin Ming)
- Use MSR names in the extra reg lists. (Lin Ming)
- Remove redundant "c = NULL" in intel_percore_constraints
- Fix comment of perf_event_attr::config1
Intel Nehalem/Westmere have a special OFFCORE_RESPONSE event
that can be used to monitor any offcore accesses from a core.
This is a very useful event for various tunings, and it's
also needed to implement the generic LLC-* events correctly.
Unfortunately this event requires programming a mask in a separate
register. And worse this separate register is per core, not per
CPU thread.
This patch:
- Teaches perf_events that OFFCORE_RESPONSE needs extra parameters.
The extra parameters are passed by user space in the
perf_event_attr::config1 field.
- Adds support to the Intel perf_event core to schedule per
core resources. This adds fairly generic infrastructure that
can be also used for other per core resources.
The basic code has is patterned after the similar AMD northbridge
constraints code.
Thanks to Stephane Eranian who pointed out some problems
in the original version and suggested improvements.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1299119690-13991-2-git-send-email-ming.m.lin@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
By pre-computing the maximum number of samples per tick we can avoid a
multiplication and a conditional since MAX_INTERRUPTS >
max_samples_per_tick.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This kernel patch adds the ability to filter monitoring based on
container groups (cgroups). This is for use in per-cpu mode only.
The cgroup to monitor is passed as a file descriptor in the pid
argument to the syscall. The file descriptor must be opened to
the cgroup name in the cgroup filesystem. For instance, if the
cgroup name is foo and cgroupfs is mounted in /cgroup, then the
file descriptor is opened to /cgroup/foo. Cgroup mode is
activated by passing PERF_FLAG_PID_CGROUP in the flags argument
to the syscall.
For instance to measure in cgroup foo on CPU1 assuming
cgroupfs is mounted under /cgroup:
struct perf_event_attr attr;
int cgroup_fd, fd;
cgroup_fd = open("/cgroup/foo", O_RDONLY);
fd = perf_event_open(&attr, cgroup_fd, 1, -1, PERF_FLAG_PID_CGROUP);
close(cgroup_fd);
Signed-off-by: Stephane Eranian <eranian@google.com>
[ added perf_cgroup_{exit,attach} ]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4d590250.114ddf0a.689e.4482@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Simple sysfs emumeration of the PMUs.
Use a "event_source" bus, and add PMU devices using their name.
Each PMU device has a type attribute which contrains the value needed
for perf_event_attr::type to identify this PMU.
This is the minimal stub needed to start using this interface,
we'll consider extending the sysfs usage later.
Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Greg KH <gregkh@suse.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101117222056.316982569@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Frederic Weisbecker