The loop in perf_ctx_adjust_freq checks the frequency of sampling
event counters, and adjusts the event interval and unthrottles the
event if required, and resets the interrupt count for the event.
However, at present it only looks at group leaders.

This means that a sampling event that is not a group leader will
eventually get throttled, once its interrupt count reaches
sysctl_perf_event_sample_rate/HZ --- and that is guaranteed to
happen, if the event is active for long enough, since the interrupt
count never gets reset.  Once it is throttled it never gets
unthrottled, so it basically just stops working at that point.

This fixes it by making perf_ctx_adjust_freq use ctx->event_list
rather than ctx->group_list.  The existing spin_lock/spin_unlock
around the loop makes it unnecessary to put rcu_read_lock/
rcu_read_unlock around the list_for_each_entry_rcu().
Reported-by: NMark W. Krentel <krentel@cs.rice.edu>
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <19157.26731.855609.165622@cargo.ozlabs.ibm.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Some architectures such as Sparc, ARM and MIPS (basically
everything with flush_dcache_page()) need to deal with dcache
aliases by carefully placing pages in both kernel and user maps.

These architectures typically have to use vmalloc_user() for this.

However, on other architectures, vmalloc() is not needed and has
the downsides of being more restricted and slower than regular
allocations.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NDavid Miller <davem@davemloft.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1254830228.21044.272.camel@laptop>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

While at it: we can traverse ctx->group_list to get all
group leader, it should be safe since we hold ctx->mutex.

Changlog v1->v2:

  - remove WARN_ON_ONCE() according to Peter Zijlstra's suggestion
Signed-off-by: NXiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Acked-by: NPeter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <4ABC5AF9.6060808@cn.fujitsu.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Paul Mackerras says:

 "Actually, looking at this more closely, it has to be a group
 leader anyway since it's at the top level of ctx->group_list.  In
 fact I see four places where we do:

  list_for_each_entry(event, &ctx->group_list, group_entry) {
	if (event == event->group_leader)
		...

 or the equivalent, three of which appear to have been introduced
 by afedadf2 ("perf_counter: Optimize sched in/out of counters")
 back in May by Peter Z.

 As far as I can see the if () is superfluous in each case (a
 singleton event will be a group of 1 and will have its
 group_leader pointing to itself)."

 [ See: http://marc.info/?l=linux-kernel&m=125361238901442&w=2 ]

And Peter Zijlstra points out this is a bugfix:

 "The intent was to call event_sched_{in,out}() for single event
  groups because that's cheaper than group_sched_{in,out}(),
  however..

  - as you noticed, I got the condition wrong, it should have read:

      list_empty(&event->sibling_list)

  - it failed to call group_can_go_on() which deals with ->exclusive.

  - it also doesn't call hw_perf_group_sched_in() which might break
    power."

 [ See: http://marc.info/?l=linux-kernel&m=125369523318583&w=2 ]

Changelog v1->v2:

 - Fix the title name according to Peter Zijlstra's suggestion

 - Remove the comments and WARN_ON_ONCE() as Peter Zijlstra's
   suggestion
Signed-off-by: NXiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Acked-by: NPeter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <4ABC5A55.7000208@cn.fujitsu.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

* mark struct vm_area_struct::vm_ops as const
* mark vm_ops in AGP code

But leave TTM code alone, something is fishy there with global vm_ops
being used.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

 - 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: NArjan van de Ven <arjan@linux.intel.com>
Acked-by: NPeter 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: NIngo 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: NStephane Eranian <eranian@google.com>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NPaul Mackerras <paulus@samba.org>
Reviewed-by: NArjan 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: NIngo Molnar <mingo@elte.hu>

In preparation to the renames, to avoid a namespace clash.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

This is in preparation of the big rename, but also makes sense
in a standalone way: 'list_entry' is a bad name as we already
have a list_entry() in list.h.

Also, the 'counter list' is too vague, it doesnt tell us the
purpose of that list.

Clarify these names to show that it's all about the group
hiearchy.
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

There is still some weird code in per_copy_attr(). Which supposedly
checks that all bytes trailing a struct are zero.

It doesn't seem to get pointer arithmetic right. Since it
increments an iterating pointer by sizeof(unsigned long) rather
than 1.
Signed-off-by: NIan Schram <ischram@telenet.be>
[ v2: clean up the messy PTR_ALIGN logic as well. ]
Signed-off-by: NPeter 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>
Cc: <stable@kernel.org> # for v2.6.31.x
LKML-Reference: <4AB3DEE2.3030600@telenet.be>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

perf timechart needs to know when a process forked, in order to be
able to visualize properly when tasks start.

This patch adds a time field to the event structure, and fills it
in appropriately.
Signed-off-by: NArjan van de Ven <arjan@linux.intel.com>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090912130341.51ad2de2@infradead.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Arjan complained about the suckyness of TSC on modern machines, and
asked if we could do something about that for PERF_SAMPLE_TIME.

Make cpu_clock() NMI safe by removing the spinlock and using
cmpxchg. This also makes it smaller and more robust.

Affects architectures that use HAVE_UNSTABLE_SCHED_CLOCK, i.e. IA64
and x86.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

/me dons the brown paper bag.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Draining the BTS buffer on a buffer overflow interrupt takes too
long resulting in a kernel lockup when tracing the kernel.

Restructure perf_counter sampling into sample creation and sample
output.

Prepare a single reference sample for BTS sampling and update the
from and to address fields when draining the BTS buffer. Drain the
entire BTS buffer between a single perf_output_begin() /
perf_output_end() pair.
Signed-off-by: NMarkus Metzger <markus.t.metzger@intel.com>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090915130023.A16204@sedona.ch.intel.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Currently we wake the mmap() consumer once every PAGE_SIZE of data
and/or once event wakeup_events when specified.

For high speed sampling this results in too many wakeups wrt. the
buffer size, hence change this.

We move the default wakeup limit to 1/4-th the buffer size, and
provide for means to manually specify this limit.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

We can have swcounter events that contribute more than a single
count per event, when used with a non-zero period, those can
generate multiple events, which is when we need throttling.

However, swcounter that contribute only a single count per event
can only come as fast as we can run code, hence don't throttle
them.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

If we pass a big size data over perf_counter_open() syscall,
the kernel will copy this data to a small buffer, it will
cause kernel crash.

This bug makes the kernel unsafe and non-root local user can
trigger it.
Signed-off-by: NXiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Acked-by: NPeter Zijlstra <peterz@infradead.org>
Acked-by: NPaul Mackerras <paulus@samba.org>
Cc: <stable@kernel.org>
LKML-Reference: <4AAF37D4.5010706@cn.fujitsu.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Before:

  $ perf sched record -f sleep 1
  Error: failed to mmap with 1 (Operation not permitted)

After:

  $ perf sched record -f sleep 1
  [ perf record: Captured and wrote 0.095 MB perf.data (~4161 samples) ]

Note, this is only allowed if perfcounter_paranoid is set to
the most permissive (non-default) value of -1.

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: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

We forget to release the fd in the PERF_FLAG_FD_OUTPUT
error path.

Reorganize the error flow here to be a clean fall-through
logic.

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: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

I want to sample inherited tracepoint workloads as a normal
user and the CAP_SYS_ADMIN check prevents me from doing that
right now.

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: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

We have a race in the swcounter stuff where we can start
counting a counter that has never been enabled, this leads to a
/0 situation.

The below avoids the /0 but doesn't close the race, this would
need a new counter state.

The race is due to perf_swcounter_is_counting() which cannot
discern between disabled due to scheduled out, and disabled for
any other reason.

Such a crash has been seen by Ingo:

[  967.092372] divide error: 0000 [#1] SMP
[  967.096499] last sysfs file: /sys/devices/system/cpu/cpu15/cache/index2/shared_cpu_map
[  967.104846] CPU 5
[  967.106965] Modules linked in:
[  967.110169] Pid: 3351, comm: hackbench Not tainted 2.6.31-rc8-tip-01158-gd940a54-dirty #1568 X8DTN
[  967.119456] RIP: 0010:[<ffffffff810c0aba>]  [<ffffffff810c0aba>] perf_swcounter_ctx_event+0x127/0x1af
[  967.129137] RSP: 0018:ffff8801a95abd70  EFLAGS: 00010046
[  967.134699] RAX: 0000000000000002 RBX: ffff8801bd645c00 RCX: 0000000000000002
[  967.142162] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff8801bd645d40
[  967.149584] RBP: ffff8801a95abdb0 R08: 0000000000000001 R09: ffff8801a95abe00
[  967.157042] R10: 0000000000000037 R11: ffff8801aa1245f8 R12: ffff8801a95abe00
[  967.164481] R13: ffff8801a95abe00 R14: ffff8801aa1c0e78 R15: 0000000000000001
[  967.171953] FS:  0000000000000000(0000) GS:ffffc90000a00000(0063) knlGS:00000000f7f486c0
[  967.180406] CS:  0010 DS: 002b ES: 002b CR0: 000000008005003b
[  967.186374] CR2: 000000004822c0ac CR3: 00000001b19a2000 CR4: 00000000000006e0
[  967.193770] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  967.201224] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  967.208692] Process hackbench (pid: 3351, threadinfo ffff8801a95aa000, task ffff8801a96b0000)
[  967.217607] Stack:
[  967.219711]  0000000000000000 0000000000000037 0000000200000001 ffffc90000a1107c
[  967.227296] <0> ffff8801a95abe00 0000000000000001 0000000000000001 0000000000000037
[  967.235333] <0> ffff8801a95abdf0 ffffffff810c0c20 0000000200a14f30 ffff8801a95abe40
[  967.243532] Call Trace:
[  967.246103]  [<ffffffff810c0c20>] do_perf_swcounter_event+0xde/0xec
[  967.252635]  [<ffffffff810c0ca7>] perf_tpcounter_event+0x79/0x7b
[  967.258957]  [<ffffffff81037f73>] ftrace_profile_sched_switch+0xc0/0xcb
[  967.265791]  [<ffffffff8155f22d>] schedule+0x429/0x4c4
[  967.271156]  [<ffffffff8100c01e>] int_careful+0xd/0x14
Reported-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1251472247.17617.74.camel@laptop>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Per-cpu counters are an ASLR information leak as they show
the execution other tasks do. Increase the paranoia level
to 1, which disallows per-cpu counters. (they still allow
counting/profiling of own tasks - and admin can profile
everything.)
Acked-by: NPeter 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: NIngo Molnar <mingo@elte.hu>

Provide the ability to configure a counter to send its output
to another (already existing) counter's output stream.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20090819092023.980284148@chello.nl>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Currently, if a group is created where the group leader is
initially disabled but a non-leader member is initially
enabled, and then the leader is subsequently enabled some time
later, the time_enabled for the non-leader member will reflect
the whole time since it was created, not just the time since
the leader was enabled.

This is incorrect, because all of the members are effectively
disabled while the leader is disabled, since none of the
members can go on the PMU if the leader can't.

Thus we have to update the ->tstamp_enabled for all the enabled
group members when a group leader is enabled, so that the
time_enabled computation only counts the time since the leader
was enabled.

Similarly, when disabling a group leader we have to update the
time_enabled and time_running for all of the group members.

Also, in update_counter_times, we have to treat a counter whose
group leader is disabled as being disabled.
Reported-by: NStephane Eranian <eranian@googlemail.com>
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: <stable@kernel.org>
LKML-Reference: <19091.29664.342227.445006@drongo.ozlabs.ibm.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

When you iterate a list, using the iterator is useful.

Before:

   ID: 5
   ID: 5
   ID: 5
   ID: 5
   EVNT: 0x40088b scale: nan ID: 5 CNT: 1006252 ID: 6 CNT: 1011090 ID: 7 CNT: 1011196 ID: 8 CNT: 1011095
   EVNT: 0x40088c scale: 1.000000 ID: 5 CNT: 2003065 ID: 6 CNT: 2011671 ID: 7 CNT: 2012620 ID: 8 CNT: 2013479
   EVNT: 0x40088c scale: 1.000000 ID: 5 CNT: 3002390 ID: 6 CNT: 3015996 ID: 7 CNT: 3018019 ID: 8 CNT: 3020006
   EVNT: 0x40088b scale: 1.000000 ID: 5 CNT: 4002406 ID: 6 CNT: 4021120 ID: 7 CNT: 4024241 ID: 8 CNT: 4027059

After:

   ID: 1
   ID: 2
   ID: 3
   ID: 4
   EVNT: 0x400889 scale: nan ID: 1 CNT: 1005270 ID: 2 CNT: 1009833 ID: 3 CNT: 1010065 ID: 4 CNT: 1010088
   EVNT: 0x400898 scale: nan ID: 1 CNT: 2001531 ID: 2 CNT: 2022309 ID: 3 CNT: 2022470 ID: 4 CNT: 2022627
   EVNT: 0x400888 scale: 0.489467 ID: 1 CNT: 3001261 ID: 2 CNT: 3027088 ID: 3 CNT: 3027941 ID: 4 CNT: 3028762
Reported-by: Nstephane eranian <eranian@googlemail.com>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey J Ashford <cjashfor@us.ibm.com>
Cc: perfmon2-devel <perfmon2-devel@lists.sourceforge.net>
LKML-Reference: <1250867976.7538.73.camel@twins>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

PARISC does not build:

/home/mingo/tip/kernel/perf_counter.c: In function 'perf_counter_index':
/home/mingo/tip/kernel/perf_counter.c:2016: error: 'PERF_COUNTER_INDEX_OFFSET' undeclared (first use in this function)
/home/mingo/tip/kernel/perf_counter.c:2016: error: (Each undeclared identifier is reported only once
/home/mingo/tip/kernel/perf_counter.c:2016: error: for each function it appears in.)

As PERF_COUNTER_INDEX_OFFSET is not defined.

Now, we could define it in the architecture - but lets also provide
a core default of 0 (which happens to be what all but one
architecture uses at the moment).

Architectures that need a different index offset should set this
value in their asm/perf_counter.h files.

Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-parisc@vger.kernel.org
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: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

In general, code in perf_counter.c that is called through an
IPI checks, for per-task counters, that the counter's task is
still the current task.  This is to handle the race condition
where the cpu switches from the task we want to another task in
the interval between sending the IPI and the IPI arriving and
being handled on the target CPU.

For some reason, __perf_counter_read is missing this check, yet
there is no reason why the race condition can't occur.  This
adds a check that the current task is the one we want.  If it
isn't, we just return.  In that case the counter->count value
should be up to date, since it will have been updated when the
counter was scheduled out, which must have happened since the
IPI was sent.

I don't have an example of an actual failure due to this race,
but it seems obvious that it could occur and we need to guard
against it.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <19076.63614.277861.368125@drongo.ozlabs.ibm.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

A bug in (9f498cc5: perf_counter: Full task tracing) makes
profiling multi-threaded apps it go belly up.

[ output as: (PID:TID):(PPID:PTID) ]

 # ./perf report -D | grep FORK
0x4b0 [0x18]: PERF_EVENT_FORK: (3237:3237):(3236:3236)
0xa10 [0x18]: PERF_EVENT_FORK: (3237:3238):(3236:3236)
0xa70 [0x18]: PERF_EVENT_FORK: (3237:3239):(3236:3236)
0xad0 [0x18]: PERF_EVENT_FORK: (3237:3240):(3236:3236)
0xb18 [0x18]: PERF_EVENT_FORK: (3237:3241):(3236:3236)

Shows us that the test (27d028de perf report: Update for the new
FORK/EXIT events) in builtin-report.c:

        /*
         * A thread clone will have the same PID for both
         * parent and child.
         */
        if (thread == parent)
                return 0;

Will clearly fail.

The problem is that perf_counter_fork() reports the actual
parent, instead of the cloning thread.

Fixing that (with the below patch), yields:

 # ./perf report -D | grep FORK
0x4c8 [0x18]: PERF_EVENT_FORK: (1590:1590):(1589:1589)
0xbd8 [0x18]: PERF_EVENT_FORK: (1590:1591):(1590:1590)
0xc80 [0x18]: PERF_EVENT_FORK: (1590:1592):(1590:1590)
0x3338 [0x18]: PERF_EVENT_FORK: (1590:1593):(1590:1590)
0x66b0 [0x18]: PERF_EVENT_FORK: (1590:1594):(1590:1590)

Which both makes more sense and doesn't confuse perf report
anymore.
Reported-by: NPekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: paulus@samba.org
Cc: Anton Blanchard <anton@samba.org>
Cc: Arjan van de Ven <arjan@infradead.org>
LKML-Reference: <1250172882.5241.62.camel@twins>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

perf_pending_counter() is called from IRQ context and will call
perf_counter_disable(), however perf_counter_disable() uses
smp_call_function_single() which doesn't fancy being used with
IRQs disabled due to IPI deadlocks.

Fix this by making it use the local __perf_counter_disable()
call and teaching the counter_sched_out() code about pending
disables as well.

This should cover the case where a counter migrates before the
pending queue gets processed.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey J Ashford <cjashfor@us.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: stephane eranian <eranian@googlemail.com>
LKML-Reference: <20090813103655.244097721@chello.nl>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Replace PERF_SAMPLE_GROUP with PERF_SAMPLE_READ and introduce
PERF_FORMAT_GROUP to deal with group reads in a more generic
way.

This allows you to get group reads out of read() as well.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey J Ashford <cjashfor@us.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: stephane eranian <eranian@googlemail.com>
LKML-Reference: <20090813103655.117411814@chello.nl>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

perf_swcounter_is_counting() uses a lock, which means we cannot
use swcounters from NMI or when holding that particular lock,
this is unintended.

The below removes the lock, this opens up race window, but not
worse than the swcounters already experience due to RCU
traversal of the context in perf_swcounter_ctx_event().

This also fixes the hard lockups while opening a lockdep
tracepoint counter.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Corey J Ashford <cjashfor@us.ibm.com>
LKML-Reference: <1250149915.10001.66.camel@twins>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Provide weak aliases for hw_perf_counter_setup_online(). This is
used by the BTS patches (for v2.6.32), but it interacts with
fixes so propagate this upstream. (it has no effect as of yet)

Also export perf_counter_output() to architecture code.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Raw tracepoint data contains various kernel internals and
data from other users, so restrict this to CAP_SYS_ADMIN.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1249896452.17467.75.camel@twins>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

PERF_SAMPLE_* output switches should unconditionally output the
correct format, as they are the only way to unambiguously parse
the PERF_EVENT_SAMPLE data.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1249896447.17467.74.camel@twins>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Implement a performance counter with:

    attr.type           = PERF_TYPE_HARDWARE
    attr.config         = PERF_COUNT_HW_BRANCH_INSTRUCTIONS
    attr.sample_period  = 1

Using branch trace store (BTS) on x86 hardware, if available.

The from and to address for each branch can be sampled using:

    PERF_SAMPLE_IP      for the from address
    PERF_SAMPLE_ADDR    for the to address

[ v2: address review feedback, fix bugs ]
Signed-off-by: NMarkus Metzger <markus.t.metzger@intel.com>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

While extending perfcounters with BTS hw-tracing, Markus
Metzger managed to trigger this warning:

   [  995.557128] WARNING: at kernel/perf_counter.c:1191 __perf_counter_task_sched_out+0x48/0x6b()

triggers because commit
9f498cc5 (perf_counter: Full
task tracing) removed clearing of tsk->perf_counter_ctxp out
from under ctx->lock which introduced a race (against
perf_lock_task_context).

Move it back and deal with the exit notification by explicitly
passing along the former task context.
Reported-by: NMarkus T Metzger <markus.t.metzger@intel.com>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1249667341.17467.5.camel@twins>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Based on Peter's comments, make tracepoint sampling generic
just like all the other sampling bits are. This is a rename
with no code changes:

- PERF_SAMPLE_TP_RECORD to PERF_SAMPLE_RAW
- struct perf_tracepoint_record to perf_raw_record

We want the system in place that transport tracepoints raw
samples events into the perf ring buffer to be generalized and
usable by any type of counter.

Reported-by; Peter Zijlstra <peterz@infradead.org>
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1249698400-5441-4-git-send-email-fweisbec@gmail.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Despite that the tracepoint record is always present when the
PERF_SAMPLE_TP_RECORD flag is set, gcc raises a warning,
thinking it might not be initialized:

  kernel/perf_counter.c: In function ‘perf_counter_output’:
  kernel/perf_counter.c:2650: warning: ‘tp’ may be used uninitialized in this function

Then, initialize it to NULL and always check if it's not NULL
before dereference it.
Reported-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1249698400-5441-2-git-send-email-fweisbec@gmail.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Reimplement the software counters to deal with fast moving
event sources (such as tracepoints). This means being able
to generate multiple overflows from a single 'event' as well
as support throttling.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

This patch implements the kernel side support for ftrace event
record sampling.

A new counter sampling attribute is added:

   PERF_SAMPLE_TP_RECORD

which requests ftrace events record sampling. In this case
if a PERF_TYPE_TRACEPOINT counter is active and a tracepoint
fires, we emit the tracepoint binary record to the
perfcounter event buffer, as a sample.

Result, after setting PERF_SAMPLE_TP_RECORD attribute from perf
record:

 perf record -f -F 1 -a -e workqueue:workqueue_execution
 perf report -D

 0x21e18 [0x48]: event: 9
 .
 . ... raw event: size 72 bytes
 .  0000:  09 00 00 00 01 00 48 00 d0 c7 00 81 ff ff ff ff  ......H........
 .  0010:  0a 00 00 00 0a 00 00 00 21 00 00 00 00 00 00 00  ........!......
 .  0020:  2b 00 01 02 0a 00 00 00 0a 00 00 00 65 76 65 6e  +...........eve
 .  0030:  74 73 2f 31 00 00 00 00 00 00 00 00 0a 00 00 00  ts/1...........
 .  0040:  e0 b1 31 81 ff ff ff ff                          .......
.
0x21e18 [0x48]: PERF_EVENT_SAMPLE (IP, 1): 10: 0xffffffff8100c7d0 period: 33

The raw ftrace binary record starts at offset 0020.

Translation:

 struct trace_entry {
	type		= 0x2b = 43;
	flags		= 1;
	preempt_count	= 2;
	pid		= 0xa = 10;
	tgid		= 0xa = 10;
 }

 thread_comm = "events/1"
 thread_pid  = 0xa = 10;
 func	    = 0xffffffff8131b1e0 = flush_to_ldisc()

What will come next?

 - Userspace support ('perf trace'), 'flight data recorder' mode
   for perf trace, etc.

 - The unconditional copy from the profiling callback brings
   some costs however if someone wants no such sampling to
   occur, and needs to be fixed in the future. For that we need
   to have an instant access to the perf counter attribute.
   This is a matter of a flag to add in the struct ftrace_event.

 - Take care of the events recursivity! Don't ever try to record
   a lock event for example, it seems some locking is used in
   the profiling fast path and lead to a tracing recursivity.
   That will be fixed using raw spinlock or recursivity
   protection.

 - [...]

 - Profit! :-)
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Gabriel Munteanu <eduard.munteanu@linux360.ro>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>