Per-thread events with a cpu filter, i.e., cpu != -1, were not
reporting correct timings when the thread never ran on the
monitored cpu. The time enabled was reported as a negative
value.
This patch fixes the problem by updating tstamp_stopped,
tstamp_running in event_sched_out() for events with filters and
which are marked as INACTIVE.
The function group_sched_out() is modified to systematically
call into event_sched_out() to avoid duplicating the timing
adjustment code twice.
With the patch, I now get:
$ task_cpu -i -e unhalted_core_cycles,unhalted_core_cycles
noploop 2 noploop for 2 seconds
CPU0 0 unhalted_core_cycles (ena=1,991,136,594, run=0)
CPU0 0 unhalted_core_cycles (ena=1,991,136,594, run=0)
CPU1 0 unhalted_core_cycles (ena=1,991,136,594, run=0)
CPU1 0 unhalted_core_cycles (ena=1,991,136,594, run=0)
CPU2 0 unhalted_core_cycles (ena=1,991,136,594, run=0)
CPU2 0 unhalted_core_cycles (ena=1,991,136,594, run=0)
CPU3 4,747,990,931 unhalted_core_cycles (ena=1,991,136,594, run=1,991,136,594)
CPU3 4,747,990,931 unhalted_core_cycles (ena=1,991,136,594, run=1,991,136,594)
Signed-off-by: Stephane Eranian <eranian@gmail.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus@samba.org
Cc: davem@davemloft.net
Cc: fweisbec@gmail.com
Cc: perfmon2-devel@lists.sf.net
Cc: eranian@google.com
LKML-Reference: <4c76802d.aae9d80a.115d.70fe@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Instead of hardcoding the number of contexts for the recursions
barriers, define a cpp constant to make the code more
self-explanatory.
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>
Now that software events don't have interrupt disabled anymore in
the event path, callchains can nest on any context. So seperating
nmi and others contexts in two buffers has become racy.
Fix this by providing one buffer per nesting level. Given the size
of the callchain entries (2040 bytes * 4), we now need to allocate
them dynamically.
v2: Fixed put_callchain_entry call after recursion.
Fix the type of the recursion, it must be an array.
v3: Use a manual pr cpu allocation (temporary solution until NMIs
can safely access vmalloc'ed memory).
Do a better separation between callchain reference tracking and
allocation. Make the "put" path lockless for non-release cases.
v4: Protect the callchain buffers with rcu.
v5: Do the cpu buffers allocations node affine.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Tested-by: Will Deacon <will.deacon@arm.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: Paul Mundt <lethal@linux-sh.org>
Cc: David Miller <davem@davemloft.net>
Cc: Borislav Petkov <bp@amd64.org>
Store the kernel and user contexts from the generic layer instead
of archs, this gathers some repetitive code.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Tested-by: Will Deacon <will.deacon@arm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: David Miller <davem@davemloft.net>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Borislav Petkov <bp@amd64.org>
- Most archs use one callchain buffer per cpu, except x86 that needs
to deal with NMIs. Provide a default perf_callchain_buffer()
implementation that x86 overrides.
- Centralize all the kernel/user regs handling and invoke new arch
handlers from there: perf_callchain_user() / perf_callchain_kernel()
That avoid all the user_mode(), current->mm checks and so...
- Invert some parameters in perf_callchain_*() helpers: entry to the
left, regs to the right, following the traditional (dst, src).
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Tested-by: Will Deacon <will.deacon@arm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: David Miller <davem@davemloft.net>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Borislav Petkov <bp@amd64.org>
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (27 commits)
sched: Use correct macro to display sched_child_runs_first in /proc/sched_debug
sched: No need for bootmem special cases
sched: Revert nohz_ratelimit() for now
sched: Reduce update_group_power() calls
sched: Update rq->clock for nohz balanced cpus
sched: Fix spelling of sibling
sched, cpuset: Drop __cpuexit from cpu hotplug callbacks
sched: Fix the racy usage of thread_group_cputimer() in fastpath_timer_check()
sched: run_posix_cpu_timers: Don't check ->exit_state, use lock_task_sighand()
sched: thread_group_cputime: Simplify, document the "alive" check
sched: Remove the obsolete exit_state/signal hacks
sched: task_tick_rt: Remove the obsolete ->signal != NULL check
sched: __sched_setscheduler: Read the RLIMIT_RTPRIO value lockless
sched: Fix comments to make them DocBook happy
sched: Fix fix_small_capacity
powerpc: Exclude arch_sd_sibiling_asym_packing() on UP
powerpc: Enable asymmetric SMT scheduling on POWER7
sched: Add asymmetric group packing option for sibling domain
sched: Fix capacity calculations for SMT4
sched: Change nohz idle load balancing logic to push model
...
Since now all modification to event->count (and ->prev_count
and ->period_left) are local to a cpu, change then to local64_t so we
avoid the LOCK'ed ops.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Only child counters adding back their values into the parent counter
are responsible for cross-cpu updates to event->count.
So if we pull that out into a new child_count variable, we get an
event->count that is only modified locally.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Create a helper function for those sites that want to read the event count.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Currently there are perf_buffer_alloc() + perf_buffer_init() + some
separate bits, fold it all into a single perf_buffer_alloc() and only
leave the attachment to the event separate.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Clarify some of the transactional group scheduling API details
and change it so that a successfull ->commit_txn also closes
the transaction.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1274803086.5882.1752.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add the capacility to track data mmap()s. This can be used together
with PERF_SAMPLE_ADDR for data profiling.
Signed-off-by: Anton Blanchard <anton@samba.org>
[Updated code for stable perf ABI]
Signed-off-by: Eric B Munson <ebmunson@us.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1274193049-25997-1-git-send-email-ebmunson@us.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
__DO_TRACE() already calls the callbacks under rcu_read_lock_sched(),
which is sufficient for our needs, avoid doing it again.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Inline perf_swevent_put_recursion_context into perf_tp_event(), this
shrinks the per trace template code footprint and saves a function
call.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
For people who otherwise get to write: cpu_clock(smp_processor_id()),
there is now: local_clock().
Also, as per suggestion from Andrew, provide some documentation on
the various clock interfaces, and minimize the unsigned long long vs
u64 mess.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jens Axboe <jaxboe@fusionio.com>
LKML-Reference: <1275052414.1645.52.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Drop this argument now that we always want to rewind only to the
state of the first caller.
It means frame pointers are not necessary anymore to reliably get
the source of an event. But this also means we need this helper
to be a macro now, as an inline function is not an option since
we need to know when to provide a default implentation.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Cc: David Miller <davem@davemloft.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Frederic reported that frequency driven swevents didn't work properly
and even caused a division-by-zero error.
It turns out there are two bugs, the division-by-zero comes from a
failure to deal with that in perf_calculate_period().
The other was more interesting and turned out to be a wrong comparison
in perf_adjust_period(). The comparison was between an s64 and u64 and
got implicitly converted to an unsigned comparison. The problem is
that period_left is typically < 0, so it ended up being always true.
Cure this by making the local period variables s64.
Reported-by: Frederic Weisbecker <fweisbec@gmail.com>
Tested-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: <stable@kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Frederic reported that because swevents handling doesn't disable IRQs
anymore, we can get a recursion of perf_adjust_period(), once from
overflow handling and once from the tick.
If both call ->disable, we get a double hlist_del_rcu() and trigger
a LIST_POISON2 dereference.
Since we don't actually need to stop/start a swevent to re-programm
the hardware (lack of hardware to program), simply nop out these
callbacks for the swevent pmu.
Reported-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1275557609.27810.35218.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
If a sample size crosses to the next page boundary, the copy
will be made in more than one step. However we forget to advance
the source offset for the next copy, leading to unexpected double
copies that completely mess up the traces.
This fixes various kinds of bad traces that have irrelevant
data inside, as an example:
geany-4979 [001] 5758.077775: sched_switch: prev_comm=! prev_pid=121
prev_prio=0 prev_state=S|D|Z|X|x ==> next_comm= next_pid=7497072
next_prio=0
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1274988898-5639-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The transactional API patch between the generic and model-specific
code introduced several important bugs with event scheduling, at
least on X86. If you had pinned events, e.g., watchdog, and were
over-committing the PMU, you would get bogus counts. The bug was
showing up on Intel CPU because events would move around more
often that on AMD. But the problem also existed on AMD, though
harder to expose.
The issues were:
- group_sched_in() was missing a cancel_txn() in the error path
- cpuc->n_added was not properly maintained, leading to missing
actions in hw_perf_enable(), i.e., n_running being 0. You cannot
update n_added until you know the transaction has succeeded. In
case of failed transaction n_added was not adjusted back.
- in case of failed transactions, event_sched_out() was called
and eventually invoked x86_disable_event() to touch the HW reg.
But with transactions, on X86, event_sched_in() does not touch
HW registers, it simply collects events into a list. Thus, you
could end up calling x86_disable_event() on a counter which
did not correspond to the current event when idx != -1.
The patch modifies the generic and X86 code to avoid all those problems.
First, we keep track of the number of events added last. In case the
transaction fails, we substract them from n_added. This approach is
necessary (as opposed to delaying updates to n_added) because not all
event updates use the transaction API, e.g., single events.
Second, we encapsulate the event_sched_in() and event_sched_out() in
group_sched_in() inside the transaction. That makes the operations
symmetrical and you can also detect that you are inside a transaction
and skip the HW reg access by checking cpuc->group_flag.
With this patch, you can now overcommit the PMU even with pinned
system-wide events present and still get valid counts.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1274796225.5882.1389.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Group siblings don't pin each-other or the parent, so when we destroy
events we must make sure to clean up all cross referencing pointers.
In particular, for destruction of a group leader we must be able to
find all its siblings and remove their reference to it.
This means that detaching an event from its context must not detach it
from the group, otherwise we can end up failing to clear all pointers.
Solve this by clearly separating the attachment to a context and
attachment to a group, and keep the group composed until we destroy
the events.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
In order to move toward separate buffer objects, rework the whole
perf_mmap_data construct to be a more self-sufficient entity, one
with its own lifetime rules.
This greatly sanitizes the whole output redirection code, which
was riddled with bugs and races.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: <stable@kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
once anon_inode_getfd() is called, you can't expect *anything* about
struct file that descriptor points to - another thread might be doing
whatever it likes with descriptor table at that point.
Cc: stable <stable@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Since we know tracepoints come from kernel context,
avoid conditionals that try and establish that very
fact.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.904944001@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reduce code and data by using the knowledge that for
!PERF_USE_VMALLOC data_order is always 0.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.795019386@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reduce the clutter in perf_output_copy() by keeping
an interator in perf_output_handle.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.742809176@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
RO mmap()s don't update the tail pointer, so
comparing against it for determining the written data
size doesn't really do any good.
Keep track of when we last did a wakeup, and compare
against that.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.684479310@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Since we want to ensure buffers only have a single
writer, we must avoid creating one with multiple.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.528215873@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Avoid the swevent hash-table by using per-tracepoint
hlists.
Also, avoid conditionals on the fast path by ordering
with probe unregister so that we should never get on
the callback path without the data being there.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <20100521090710.473188012@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
A writer that gets a reference to the buffer handle disables
preemption. When we put that reference, we check if we are
the outer most writer and if not, we simply return and defer
the head update to the outer most writer. The problem here
is that preemption is only reenabled by the outer most, that
produces preemption count imbalance for every nested writer
that exit.
So just don't forget to always re-enable preemption when we
put the buffer reference, whoever we are.
Fixes lots of sleeping in atomic warnings, visible with lock
events recording.
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: Robert Richter <robert.richter@amd.com>
The software events hlist doesn't fully comply with the new
rcu checks api.
We need to consider three different sides that access the hlist:
- the hlist allocation/release side. This side happens when an
events is created or released, accesses to the hlist are
serialized under the cpuctx mutex.
- the events insertion/removal in the hlist. This side is always
serialized against the above one. The hlist is always present
during such operations. This side happens when a software event
is scheduled in/out. The serialization that ensures the software
event is really attached to the context is made under the
ctx->lock.
- events triggering. This is the read side, it can happen
concurrently with any update side.
This patch deals with them one by one and anticipates with the
separate rcu mem space patches in preparation.
This patch fixes various annoying rcu warnings.
Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
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>
Since the x86 XCHG ins implies LOCK, avoid the use by
using a sequence count instead.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Since there is now only a single writer, we can use
local_t instead and avoid all these pesky LOCK insn.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Since we can now assume there is only a single writer
to each buffer, we can remove per-cpu lock thingy and
use a simply nest-count to the same effect.
This removes the need to disable IRQs.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Since we now have working per-task-per-cpu events for
a while, disallow mmap() on per-task inherited
events. Those things were a performance problem
anyway, and doing away with it allows us to optimize
the buffer somewhat by assuming there is only a
single writer.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Ensure cpu bound buffers live on the right NUMA node.
Suggested-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1274114880.5605.5236.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
In case the sampling buffer has no "payload" pages,
nr_pages is 0. The problem is that the error path in
perf_output_begin() skips to a label which assumes
perf_output_lock() has been issued which is not the
case. That triggers a WARN_ON() in
perf_output_unlock().
This patch fixes the problem by skipping
perf_output_unlock() in case data->nr_pages is 0.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <4bf13674.014fd80a.6c82.ffffb20c@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When we've got but a single event per tracepoint
there is no reason to try and multiplex it so don't.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Tested-by: Ingo Molnar <mingo@elte.hu>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Corey reported that the value scale times of group siblings are not
updated when the monitored task dies.
The problem appears to be that we only update the group leader's
time values, fix it by updating the whole group.
Reported-by: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: <stable@kernel.org> # .34.x
LKML-Reference: <1273588935.1810.6.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Both Stephane and Corey reported that PERF_FORMAT_GROUP didn't
work as expected if the task the counters were attached to quit
before the read() call.
The cause is that we unconditionally destroy the grouping when
we remove counters from their context. Fix this by splitting off
the group destroy from the list removal such that
perf_event_remove_from_context() does not do this and change
perf_event_release() to do so.
Reported-by: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: <stable@kernel.org> # .34.x
LKML-Reference: <1273571513.5605.3527.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This reverts commit 4fd38e4595.
It causes various crashes and hangs when events are activated.
The cause is not fully understood yet but we need to revert it
because the effects are severe.
Reported-by: Stephane Eranian <eranian@google.com>
Reported-by: Lin Ming <ming.m.lin@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Commit 6bde9b6ce0 ("perf: Add
group scheduling transactional APIs") added code to allow a
group to be scheduled in a single transaction. However, it
introduced a bug in handling events whose pmu does not implement
transactions -- at the end of scheduling in the events in the
group, in the non-transactional case the code now falls through
to the group_error label, and proceeds to unschedule all the
events in the group and return failure.
This fixes it by returning 0 (success) in the non-transactional
case.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: eranian@gmail.com
LKML-Reference: <20100508105800.GB10650@brick.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add group scheduling transactional APIs to struct pmu.
These APIs will be implemented in arch code, based on Peter's idea as
below.
> the idea behind hw_perf_group_sched_in() is to not perform
> schedulability tests on each event in the group, but to add the group
> as a whole and then perform one test.
>
> Of course, when that test fails, you'll have to roll-back the whole
> group again.
>
> So start_txn (or a better name) would simply toggle a flag in the pmu
> implementation that will make pmu::enable() not perform the
> schedulablilty test.
>
> Then commit_txn() will perform the schedulability test (so note the
> method has to have a !void return value.
>
> This will allow us to use the regular
> kernel/perf_event.c::group_sched_in() and all the rollback code.
> Currently each hw_perf_group_sched_in() implementation duplicates all
> the rolllback code (with various bugs).
->start_txn:
Start group events scheduling transaction, set a flag to make
pmu::enable() not perform the schedulability test, it will be performed
at commit time.
->commit_txn:
Commit group events scheduling transaction, perform the group
schedulability as a whole
->cancel_txn:
Stop group events scheduling transaction, clear the flag so
pmu::enable() will perform the schedulability test.
Reviewed-by: Stephane Eranian <eranian@google.com>
Reviewed-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Cc: David Miller <davem@davemloft.net>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1272002160.5707.60.camel@minggr.sh.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Stephane reported a lockdep warning while using PERF_FORMAT_GROUP.
The issue is that perf_event_read_group() takes faults while holding
the ctx->mutex, while perf_event_release_kernel() can be called from
munmap(). Which makes for an AB-BA deadlock.
Except we can never establish the deadlock because we'll only ever
call perf_event_release_kernel() after all file descriptors are dead
so there is no concurrency possible.
Reported-by: Stephane Eranian <eranian@google.com>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Both Stephane and Corey reported that PERF_FORMAT_GROUP didn't work
as expected if the task the counters were attached to quit before
the read() call.
The cause is that we unconditionally destroy the grouping when we
remove counters from their context. Fix this by only doing this when
we free the counter itself.
Reported-by: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1273160566.5605.404.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Below patch introduces perf_guest_info_callbacks and related
register/unregister functions. Add more PERF_RECORD_MISC_XXX bits
meaning guest kernel and guest user space.
Signed-off-by: Zhang Yanmin <yanmin_zhang@linux.intel.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
hlist helpers need to be available for all software events, not
only trace events.
Pull them out outside the ifdef CONFIG_EVENT_TRACING section.
Fixes:
kernel/perf_event.c:4573: error: implicit declaration of function 'swevent_hlist_put'
kernel/perf_event.c:4614: error: implicit declaration of function 'swevent_hlist_get'
kernel/perf_event.c:5534: error: implicit declaration of function 'swevent_hlist_release
Reported-by: Ingo Molnar <mingo@elte.hu>
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>
LKML-Reference: <1271281338-23491-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The cpu/task clock events implement their own version of exclusion
on top of exclude_user and exclude_kernel.
The result is that when the event triggered in the kernel but we
have exclude_kernel set, we try to rewind using task_pt_regs.
There are two side effects of this:
- we call task_pt_regs even on kernel threads, which doesn't give
us the desired result.
- if the event occured in the kernel, we shouldn't rewind to the
user context. We want to actually ignore the event.
get_irq_regs() will always give us the right interrupted context, so
use its result and submit it to perf_exclude_context() that knows
when an event must be ignored.
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: Ingo Molnar <mingo@elte.hu>
Each time a software event triggers, we need to walk through
the entire list of events from the current cpu and task contexts
to retrieve a running perf event that matches.
We also need to check a matching perf event is actually counting.
This walk is wasteful and makes the event fast path scaling
down with a growing number of events running on the same
contexts.
To solve this, we store the running perf events in a hashlist to
get an immediate access to them against their type:event_id when
they trigger.
v2: - Fix SWEVENT_HLIST_SIZE definition (and re-learn some basic
maths along the way)
- Only allocate hlist for online cpus, but keep track of the
refcount on offline possible cpus too, so that we allocate it
if needed when it becomes online.
- Drop the kref use as it's not adapted to our tricks anymore.
v3: - Fix bad refcount check (address instead of value). Thanks to
Eric Dumazet who spotted this.
- While exiting cpu, move the hlist release out of the IPI path
to lock the hlist mutex sanely.
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: Ingo Molnar <mingo@elte.hu>
Perf_event does not need seeking, so prevent it in order to
get rid of default_llseek, which uses the BKL.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
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>
[drop the nonseekable_open, not needed for anon inodes]
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Now that software events use perf_arch_fetch_caller_regs() too, we
need the stub version to be always built in for archs that don't
implement it.
Fixes the following build error in PARISC:
kernel/built-in.o: In function `perf_event_task_sched_out':
(.text.perf_event_task_sched_out+0x54): undefined reference to `perf_arch_fetch_caller_regs'
Reported-by: Ingo Molnar <mingo@elte.hu>
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>
perf sched record can deadlock a box should the holder of
handle->data->lock take an interrupt, and then attempt to
acquire an rq lock held by a CPU trying to acquire the
same lock. Disable interrupts.
CPU0 CPU1
sched event with rq->lock held
grab handle->data->lock
spin on handle->data->lock
interrupt
try to grab rq->lock
Reported-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Tested-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1269598293.6174.8.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Scheduler's task migration events don't work because they always
pass NULL regs perf_sw_event(). The event hence gets filtered
in perf_swevent_add().
Scheduler's context switches events use task_pt_regs() to get
the context when the event occured which is a wrong thing to
do as this won't give us the place in the kernel where we went
to sleep but the place where we left userspace. The result is
even more wrong if we switch from a kernel thread.
Use the hot regs snapshot for both events as they belong to the
non-interrupt/exception based events family. Unlike page faults
or so that provide the regs matching the exact origin of the event,
we need to save the current context.
This makes the task migration event working and fix the context
switch callchains and origin ip.
Example: perf record -a -e cs
Before:
10.91% ksoftirqd/0 0 [k] 0000000000000000
|
--- (nil)
perf_callchain
perf_prepare_sample
__perf_event_overflow
perf_swevent_overflow
perf_swevent_add
perf_swevent_ctx_event
do_perf_sw_event
__perf_sw_event
perf_event_task_sched_out
schedule
run_ksoftirqd
kthread
kernel_thread_helper
After:
23.77% hald-addon-stor [kernel.kallsyms] [k] schedule
|
--- schedule
|
|--60.00%-- schedule_timeout
| wait_for_common
| wait_for_completion
| blk_execute_rq
| scsi_execute
| scsi_execute_req
| sr_test_unit_ready
| |
| |--66.67%-- sr_media_change
| | media_changed
| | cdrom_media_changed
| | sr_block_media_changed
| | check_disk_change
| | cdrom_open
v2: Always build perf_arch_fetch_caller_regs() now that software
events need that too. They don't need it from modules, unlike trace
events, so we keep the EXPORT_SYMBOL in trace_event_perf.c
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: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
perf_arch_fetch_caller_regs() is exported for the overriden x86
version, but not for the generic weak version.
As a general rule, weak functions should not have their symbol
exported in the same file they are defined.
So let's export it on trace_event_perf.c as it is used by trace
events only.
This fixes:
ERROR: ".perf_arch_fetch_caller_regs" [fs/xfs/xfs.ko] undefined!
ERROR: ".perf_arch_fetch_caller_regs" [arch/powerpc/platforms/cell/spufs/spufs.ko] undefined!
-v2: And also only build it if trace events are enabled.
-v3: Fix changelog mistake
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1268697902-9518-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
perf_arch_fetch_caller_regs() is exported for the overriden x86
version, but not for the generic weak version.
As a general rule, weak functions should not have their symbol
exported in the same file they are defined.
So let's export it on trace_event_perf.c as it is used by trace
events only.
This fixes:
ERROR: ".perf_arch_fetch_caller_regs" [fs/xfs/xfs.ko] undefined!
ERROR: ".perf_arch_fetch_caller_regs" [arch/powerpc/platforms/cell/spufs/spufs.ko] undefined!
-v2: And also only build it if trace events are enabled.
-v3: Fix changelog mistake
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1268697902-9518-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch is an optimization in perf_event_task_sched_in() to avoid
scheduling the events twice in a row.
Without it, the perf_disable()/perf_enable() pair is invoked twice,
thereby pinned events counts while scheduling flexible events and we go
throuh hw_perf_enable() twice.
By encapsulating, the whole sequence into perf_disable()/perf_enable() we
ensure, hw_perf_enable() is going to be invoked only once because of the
refcount protection.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1268288765-5326-1-git-send-email-eranian@google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Anton Blanchard found that he could reliably make the kernel hit a
BUG_ON in the slab allocator by taking a cpu offline and then online
while a system-wide perf record session was running.
The reason is that when the cpu comes up, we completely reinitialize
the ctx field of the struct perf_cpu_context for the cpu. If there is
a system-wide perf record session running, then there will be a struct
perf_event that has a reference to the context, so its refcount will
be 2. (The perf_event has been removed from the context's group_entry
and event_entry lists by perf_event_exit_cpu(), but that doesn't
remove the perf_event's reference to the context and doesn't decrement
the context's refcount.)
When the cpu comes up, perf_event_init_cpu() gets called, and it calls
__perf_event_init_context() on the cpu's context. That resets the
refcount to 1. Then when the perf record session finishes and the
perf_event is closed, the refcount gets decremented to 0 and the
context gets kfreed after an RCU grace period. Since the context
wasn't kmalloced -- it's part of a per-cpu variable -- bad things
happen.
In fact we don't need to completely reinitialize the context when the
cpu comes up. It's sufficient to initialize the context once at boot,
but we need to do it for all possible cpus.
This moves the context initialization to happen at boot time. With
this, we don't trash the refcount and the context never gets kfreed,
and we don't hit the BUG_ON.
Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Tested-by: Anton Blanchard <anton@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Drop the obsolete "profile" naming used by perf for trace events.
Perf can now do more than simple events counting, so generalize
the API naming.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Jason Baron <jbaron@redhat.com>
We are taking a wrong regs snapshot when a trace event triggers.
Either we use get_irq_regs(), which gives us the interrupted
registers if we are in an interrupt, or we use task_pt_regs()
which gives us the state before we entered the kernel, assuming
we are lucky enough to be no kernel thread, in which case
task_pt_regs() returns the initial set of regs when the kernel
thread was started.
What we want is different. We need a hot snapshot of the regs,
so that we can get the instruction pointer to record in the
sample, the frame pointer for the callchain, and some other
things.
Let's use the new perf_fetch_caller_regs() for that.
Comparison with perf record -e lock: -R -a -f -g
Before:
perf [kernel] [k] __do_softirq
|
--- __do_softirq
|
|--55.16%-- __open
|
--44.84%-- __write_nocancel
After:
perf [kernel] [k] perf_tp_event
|
--- perf_tp_event
|
|--41.07%-- lock_acquire
| |
| |--39.36%-- _raw_spin_lock
| | |
| | |--7.81%-- hrtimer_interrupt
| | | smp_apic_timer_interrupt
| | | apic_timer_interrupt
The old case was producing unreliable callchains. Now having
right frame and instruction pointers, we have the trace we
want.
Also syscalls and kprobe events already have the right regs,
let's use them instead of wasting a retrieval.
v2: Follow the rename perf_save_regs() -> perf_fetch_caller_regs()
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Archs <linux-arch@vger.kernel.org>
Events that trigger overflows by interrupting a context can
use get_irq_regs() or task_pt_regs() to retrieve the state
when the event triggered. But this is not the case for some
other class of events like trace events as tracepoints are
executed in the same context than the code that triggered
the event.
It means we need a different api to capture the regs there,
namely we need a hot snapshot to get the most important
informations for perf: the instruction pointer to get the
event origin, the frame pointer for the callchain, the code
segment for user_mode() tests (we always use __KERNEL_CS as
trace events always occur from the kernel) and the eflags
for further purposes.
v2: rename perf_save_regs to perf_fetch_caller_regs as per
Masami's suggestion.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Archs <linux-arch@vger.kernel.org>
Try to avoid useless rotation and PMU disables.
[ Could be improved by keeping a nr_runnable count to better account
for the < PERF_STAT_INACTIVE counters ]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: paulus@samba.org
Cc: eranian@google.com
Cc: robert.richter@amd.com
Cc: fweisbec@gmail.com
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Currently we always call hw_perf_disable(), even if its already disabled,
this seems superflous, esp. since it cannot be made NMI safe (see further
patches).
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus@samba.org
Cc: eranian@google.com
Cc: robert.richter@amd.com
Cc: fweisbec@gmail.com
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Remove the hw_perf_event_*() hotplug hooks in favour of per PMU hotplug
notifiers. This has the advantage of reducing the static weak interface
as well as exposing all hotplug actions to the PMU.
Use this to fix x86 hotplug usage where we did things in ONLINE which
should have been done in UP_PREPARE or STARTING.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: paulus@samba.org
Cc: eranian@google.com
Cc: robert.richter@amd.com
Cc: fweisbec@gmail.com
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
LKML-Reference: <20100305154128.736225361@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This makes it easier to extend perf_sample_data and fixes a bug on arm
and sparc, which failed to set ->raw to NULL, which can cause crashes
when combined with PERF_SAMPLE_RAW.
It also optimizes PowerPC and tracepoint, because the struct
initialization is forced to zero out the whole structure.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Jean Pihet <jpihet@mvista.com>
Reviewed-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Jamie Iles <jamie.iles@picochip.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: stable@kernel.org
LKML-Reference: <20100304140100.315416040@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Passing the attribute to the low level IO functions allows all kinds
of cleanups, by sharing low level IO code without requiring
an own function for every piece of data.
Also drivers can extend the attributes with own data fields
and use that in the low level function.
Similar to sysdev_attributes and normal attributes.
This is a tree-wide sweep, converting everything in one go.
No functional changes in this patch other than passing the new
argument everywhere.
Tested on x86, the non x86 parts are uncompiled.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Make sure compiler won't do weird things with limits. E.g. fetching them
twice may return 2 different values after writable limits are implemented.
I.e. either use rlimit helpers added in commit 3e10e716ab ("resource:
add helpers for fetching rlimits") or ACCESS_ONCE if not applicable.
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The ANY flag can show SMT data of another task (like 'top'),
so we want to disable it when system-wide profiling is
disabled.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
A recent commit introduced a preemption warning for
perf_clock(), use raw_smp_processor_id() to avoid this, it
really doesn't matter which cpu we use here.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1267198583.22519.684.camel@laptop>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Since the cpu argument to hw_perf_group_sched_in() is always
smp_processor_id(), simplify the code a little by removing this argument
and using the current cpu where needed.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: David Miller <davem@davemloft.net>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1265890918.5396.3.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch adds correct AMD NorthBridge event scheduling.
NB events are events measuring L3 cache, Hypertransport traffic. They are
identified by an event code >= 0xe0. They measure events on the
Northbride which is shared by all cores on a package. NB events are
counted on a shared set of counters. When a NB event is programmed in a
counter, the data actually comes from a shared counter. Thus, access to
those counters needs to be synchronized.
We implement the synchronization such that no two cores can be measuring
NB events using the same counters. Thus, we maintain a per-NB allocation
table. The available slot is propagated using the event_constraint
structure.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4b703957.0702d00a.6bf2.7b7d@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
In certain situations, the kernel may need to stop and start the same
event rapidly. The current PMU callbacks do not distinguish between stop
and release (i.e., stop + free the resource). Thus, a counter may be
released, then it will be immediately re-acquired. Event scheduling will
again take place with no guarantee to assign the same counter. On some
processors, this may event yield to failure to assign the event back due
to competion between cores.
This patch is adding a new pair of callback to stop and restart a counter
without actually release the underlying counter resource. On stop, the
counter is stopped, its values saved and that's it. On start, the value
is reloaded and counter is restarted (on x86, actual restart is delayed
until perf_enable()).
Signed-off-by: Stephane Eranian <eranian@google.com>
[ added fallback to ->enable/->disable for all other PMUs
fixed x86_pmu_start() to call x86_pmu.enable()
merged __x86_pmu_disable into x86_pmu_stop() ]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4b703875.0a04d00a.7896.ffffb824@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
DaveM reported that currently perf interprets the pgoff value reported by
the MMAP events as a byte range, but the kernel reports it as a page
offset.
Since its broken (and unusable) anyway, change the kernel behaviour (ABI)
to report bytes indeed, avoiding the need for userspace to deal with
PAGE_SIZE things.
Reported-by: David Miller <davem@davemloft.net>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Commit 22e19085 ("Honour event state for aux stream data")
introduced a bug where we would drop FORK events.
The thing is that we deliver FORK events to the child process'
event, which at that time will be PERF_EVENT_STATE_INACTIVE
because the child won't be scheduled in (we're in the middle of
fork).
Solve this twice, change the event state filter to exclude only
disabled (STATE_OFF) or worse, and deliver FORK events to the
current (parent).
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Anton Blanchard <anton@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
LKML-Reference: <1266142324.5273.411.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Pretty much all of the calls do perf_disable/perf_enable cycles, pull
that out to cut back on hardware programming.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Change 'bp_len' type to __u64 to make it work across archs as
the s390 architecture watch point length can be upto 2^64.
reference:
http://lkml.org/lkml/2010/1/25/212
This is an ABI change that is not backward compatible with
the previous hardware breakpoint info layout integrated in this
development cycle, a rebuilt of perf tools is necessary for
versions based on 2.6.33-rc1 - 2.6.33-rc6 to work with a
kernel based on this patch.
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: "K. Prasad" <prasad@linux.vnet.ibm.com>
Cc: Maneesh Soni <maneesh@in.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin <schwidefsky@de.ibm.com>
LKML-Reference: <20100130045518.GA20776@in.ibm.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
One problem with frequency driven counters is that we cannot
predict the rate at which they trigger, therefore we have to
start them at period=1, this causes a ramp up effect. However,
if we fail to propagate the stable state on fork each new child
will have to ramp up again. This can lead to significant
artifacts in sample data.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: eranian@google.com
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1264752266.4283.2121.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
There was a bug in the old period code that caused intel_pmu_enable_all()
or native_write_msr_safe() to show up quite high in the profiles.
In staring at that code it made my head hurt, so I rewrote it in a
hopefully simpler fashion. Its now fully symetric between tick and
overflow driven adjustments and uses less data to boot.
The only complication is that it basically wants to do a u128 division.
The code approximates that in a rather simple truncate until it fits
fashion, taking care to balance the terms while truncating.
This version does not generate that sampling artefact.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Anton reported that perf record kept receiving events even after calling
ioctl(PERF_EVENT_IOC_DISABLE). It turns out that FORK,COMM and MMAP
events didn't respect the disabled state and kept flowing in.
Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Tested-by: Anton Blanchard <anton@samba.org>
LKML-Reference: <1263459187.4244.265.camel@laptop>
CC: stable@kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When a task gets scheduled in. We don't touch the cpu bound events
so the priority order becomes:
cpu pinned, cpu flexible, task pinned, task flexible.
So schedule out cpu flexibles when a new task context gets in
and correctly order the groups to schedule in:
task pinned, cpu flexible, task flexible.
Cpu pinned groups don't need to be touched at this time.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
We don't need to schedule in/out pinned events on task tick,
now that pinned and flexible groups can be scheduled separately.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Tune the scheduling helpers so that we can choose to schedule either
pinned and/or flexible groups from a context.
And while at it, refactor a bit the naming of these helpers to make
these more consistent and flexible.
There is no (intended) change in scheduling behaviour in this
patch.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
__perf_event_sched_out doesn't need to be globally available, make
it static.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Stephane Eranian