This patch improves event scheduling by maximizing the use of PMU
registers regardless of the order in which events are created in a group.
The algorithm takes into account the list of counter constraints for each
event. It assigns events to counters from the most constrained, i.e.,
works on only one counter, to the least constrained, i.e., works on any
counter.
Intel Fixed counter events and the BTS special event are also handled via
this algorithm which is designed to be fairly generic.
The patch also updates the validation of an event to use the scheduling
algorithm. This will cause early failure in perf_event_open().
The 2nd version of this patch follows the model used by PPC, by running
the scheduling algorithm and the actual assignment separately. Actual
assignment takes place in hw_perf_enable() whereas scheduling is
implemented in hw_perf_group_sched_in() and x86_pmu_enable().
Signed-off-by: Stephane Eranian <eranian@google.com>
[ fixup whitespace and style nits as well as adding is_x86_event() ]
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: <4b5430c6.0f975e0a.1bf9.ffff85fe@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When running perf across all cpus with backtracing (-a -g), sometimes we
get samples without associated backtraces:
23.44% init [kernel] [k] restore
11.46% init eeba0c [k] 0x00000000eeba0c
6.77% swapper [kernel] [k] .perf_ctx_adjust_freq
5.73% init [kernel] [k] .__trace_hcall_entry
4.69% perf libc-2.9.so [.] 0x0000000006bb8c
|
|--11.11%-- 0xfffa941bbbc
It turns out the backtrace code has a check for the idle task and the IP
sampling does not. This creates problems when profiling an interrupt
heavy workload (in my case 10Gbit ethernet) since we get no backtraces
for interrupts received while idle (ie most of the workload).
Right now x86 and sh check that current is not NULL, which should never
happen so remove that too.
Idle task's exclusion must be performed from the core code, on top
of perf_event_attr:exclude_idle.
Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mundt <lethal@linux-sh.org>
LKML-Reference: <20100118054707.GT12666@kryten>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
The check that ignores the debug and nmi stack frames is useless
now that we have a frame pointer that makes us start at the
right place. We don't anymore have to deal with these.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1262235183-5320-2-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Pass the frame pointer from the regs of the interrupted path
to dump_trace() while processing the stack trace.
Currently, dump_trace() takes the current bp and starts the
callchain from dump_trace() itself. This is wasteful because
we need to walk through the entire NMI/DEBUG stack before
retrieving the interrupted point.
We can fix that by just using the frame pointer from the
captured regs. It points exactly where we want to start.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1262235183-5320-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
It's just wasteful for stacktrace users like perf to walk
through every entries on the stack whereas these only accept
reliable ones, ie: that the frame pointer validates.
Since perf requires pure reliable stacktraces, it needs a stack
walker based on frame pointers-only to optimize the stacktrace
processing.
This might solve some near-lockup scenarios that can be triggered
by call-graph tracing timer events.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261024834-5336-2-git-send-regression-fweisbec@gmail.com>
[ v2: fix for modular builds and small detail tidyup ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The current print_context_stack helper that does the stack
walking job is good for usual stacktraces as it walks through
all the stack and reports even addresses that look unreliable,
which is nice when we don't have frame pointers for example.
But we have users like perf that only require reliable
stacktraces, and those may want a more adapted stack walker, so
lets make this function a callback in stacktrace_ops that users
can tune for their needs.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261024834-5336-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (57 commits)
x86, perf events: Check if we have APIC enabled
perf_event: Fix variable initialization in other codepaths
perf kmem: Fix unused argument build warning
perf symbols: perf_header__read_build_ids() offset'n'size should be u64
perf symbols: dsos__read_build_ids() should read both user and kernel buildids
perf tools: Align long options which have no short forms
perf kmem: Show usage if no option is specified
sched: Mark sched_clock() as notrace
perf sched: Add max delay time snapshot
perf tools: Correct size given to memset
perf_event: Fix perf_swevent_hrtimer() variable initialization
perf sched: Fix for getting task's execution time
tracing/kprobes: Fix field creation's bad error handling
perf_event: Cleanup for cpu_clock_perf_event_update()
perf_event: Allocate children's perf_event_ctxp at the right time
perf_event: Clean up __perf_event_init_context()
hw-breakpoints: Modify breakpoints without unregistering them
perf probe: Update perf-probe document
perf probe: Support --del option
trace-kprobe: Support delete probe syntax
...
Dumping the callchains from breakpoint events with perf gives strange
results:
3.75% perf [kernel] [k] _raw_read_unlock
|
--- _raw_read_unlock
perf_callchain
perf_prepare_sample
__perf_event_overflow
perf_swevent_overflow
perf_swevent_add
perf_bp_event
hw_breakpoint_exceptions_notify
notifier_call_chain
__atomic_notifier_call_chain
atomic_notifier_call_chain
notify_die
do_debug
debug
munmap
We are infected with all the debug stack. Like the nmi stack, the debug
stack is undesired as it is part of the profiling path, not helpful for
the user.
Ignore it.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: "K. Prasad" <prasad@linux.vnet.ibm.com>
The validate_event() was failing on valid event combinations. The
function was assuming that if x86_schedule_event() returned 0, it
meant error. But x86_schedule_event() returns the counter index and
0 is a perfectly valid value. An error is returned if the function
returns a negative value.
Furthermore, validate_event() was also failing for event groups
because the event->pmu was not set until after
hw_perf_event_init().
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: peterz@infradead.org
Cc: paulus@samba.org
Cc: perfmon2-devel@lists.sourceforge.net
Cc: eranian@gmail.com
LKML-Reference: <4b0bdf36.1818d00a.07cc.25ae@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
--
arch/x86/kernel/cpu/perf_event.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
There was namespace overlap due to a rename i did - this caused
the following build warning, reported by Stephen Rothwell against
linux-next x86_64 allmodconfig:
arch/x86/kernel/cpu/perf_event.c: In function 'intel_get_event_idx':
arch/x86/kernel/cpu/perf_event.c:1445: warning: 'event_constraint' is used uninitialized in this function
This is a real bug not just a warning: fix it by renaming the
global event-constraints table pointer to 'event_constraints'.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Stephane Eranian <eranian@gmail.com>
Cc: 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: <20091013144223.369d616d.sfr@canb.auug.org.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Refuse to add events when the group wouldn't fit onto the PMU
anymore.
Naive implementation.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@gmail.com>
LKML-Reference: <1254911461.26976.239.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
On some Intel processors, not all events can be measured in all
counters. Some events can only be measured in one particular
counter, for instance. Assigning an event to the wrong counter does
not crash the machine but this yields bogus counts, i.e., silent
error.
This patch changes the event to counter assignment logic to take
into account event constraints for Intel P6, Core and Nehalem
processors. There is no contraints on Intel Atom. There are
constraints on Intel Yonah (Core Duo) but they are not provided in
this patch given that this processor is not yet supported by
perf_events.
As a result of the constraints, it is possible for some event
groups to never actually be loaded onto the PMU if they contain two
events which can only be measured on a single counter. That
situation can be detected with the scaling information extracted
with read().
Signed-off-by: Stephane Eranian <eranian@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1254840129-6198-3-git-send-email-eranian@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Intel fixed counters do not support all the filters possible with a
generic counter. Thus, if a fixed counter event is passed but with
certain filters set, then the fixed_mode_idx() function must fail
and the event must be measured in a generic counter instead.
Reject filters are: inv, edge, cnt-mask.
Signed-off-by: Stephane Eranian <eranian@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1254840129-6198-2-git-send-email-eranian@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
- provide compatibility Kconfig entry for existing PERF_COUNTERS .config's
- provide courtesy copy of old perf_counter.h, for user-space projects
- small indentation fixups
- fix up MAINTAINERS
- fix small x86 printout fallout
- fix up small PowerPC comment fallout (use 'counter' as in register)
Reviewed-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Bye-bye Performance Counters, welcome Performance Events!
In the past few months the perfcounters subsystem has grown out its
initial role of counting hardware events, and has become (and is
becoming) a much broader generic event enumeration, reporting, logging,
monitoring, analysis facility.
Naming its core object 'perf_counter' and naming the subsystem
'perfcounters' has become more and more of a misnomer. With pending
code like hw-breakpoints support the 'counter' name is less and
less appropriate.
All in one, we've decided to rename the subsystem to 'performance
events' and to propagate this rename through all fields, variables
and API names. (in an ABI compatible fashion)
The word 'event' is also a bit shorter than 'counter' - which makes
it slightly more convenient to write/handle as well.
Thanks goes to Stephane Eranian who first observed this misnomer and
suggested a rename.
User-space tooling and ABI compatibility is not affected - this patch
should be function-invariant. (Also, defconfigs were not touched to
keep the size down.)
This patch has been generated via the following script:
FILES=$(find * -type f | grep -vE 'oprofile|[^K]config')
sed -i \
-e 's/PERF_EVENT_/PERF_RECORD_/g' \
-e 's/PERF_COUNTER/PERF_EVENT/g' \
-e 's/perf_counter/perf_event/g' \
-e 's/nb_counters/nb_events/g' \
-e 's/swcounter/swevent/g' \
-e 's/tpcounter_event/tp_event/g' \
$FILES
for N in $(find . -name perf_counter.[ch]); do
M=$(echo $N | sed 's/perf_counter/perf_event/g')
mv $N $M
done
FILES=$(find . -name perf_event.*)
sed -i \
-e 's/COUNTER_MASK/REG_MASK/g' \
-e 's/COUNTER/EVENT/g' \
-e 's/\<event\>/event_id/g' \
-e 's/counter/event/g' \
-e 's/Counter/Event/g' \
$FILES
... to keep it as correct as possible. This script can also be
used by anyone who has pending perfcounters patches - it converts
a Linux kernel tree over to the new naming. We tried to time this
change to the point in time where the amount of pending patches
is the smallest: the end of the merge window.
Namespace clashes were fixed up in a preparatory patch - and some
stylistic fallout will be fixed up in a subsequent patch.
( NOTE: 'counters' are still the proper terminology when we deal
with hardware registers - and these sed scripts are a bit
over-eager in renaming them. I've undone some of that, but
in case there's something left where 'counter' would be
better than 'event' we can undo that on an individual basis
instead of touching an otherwise nicely automated patch. )
Suggested-by: Stephane Eranian <eranian@google.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Paul Mackerras <paulus@samba.org>
Reviewed-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
