perf/ring_buffer: Document AUX API usage

In order to ensure safe AUX buffer management, we rely on the assumption
that pmu::stop() stops its ongoing AUX transaction and not just the hw.

This patch documents this requirement for the perf_aux_output_{begin,end}()
APIs.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1457098969-21595-4-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Free AUX pages in unmap path

Now that we can ensure that when ring buffer's AUX area is on the way
to getting unmapped new transactions won't start, we only need to stop
all events that can potentially be writing aux data to our ring buffer.

Having done that, we can safely free the AUX pages and corresponding
PMU data, as this time it is guaranteed to be the last aux reference
holder.

This partially reverts:

  57ffc5ca679 ("perf: Fix AUX buffer refcounting")

... which was made to defer deallocation that was otherwise possible
from an NMI context. Now it is no longer the case; the last call to
rb_free_aux() that drops the last AUX reference has to happen in
perf_mmap_close() on that AUX area.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/87d1qtz23d.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/ring_buffer: Refuse to begin AUX transaction after rb->aux_mmap_count drops

When ring buffer's AUX area is unmapped and rb->aux_mmap_count drops to
zero, new AUX transactions into this buffer can still be started,
even though the buffer in en route to deallocation.

This patch adds a check to perf_aux_output_begin() for rb->aux_mmap_count
being zero, in which case there is no point starting new transactions,
in other words, the ring buffers that pass a certain point in
perf_mmap_close will not have their events sending new data, which
clears path for freeing those buffers' pages right there and then,
provided that no active transactions are holding the AUX reference.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1457098969-21595-2-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Verify we have a single perf_hw_context PMU

There should (and can) only be a single PMU for perf_hw_context
events.

This is because of how we schedule events: once a hardware event fails to
schedule (the PMU is 'full') we stop trying to add more. The trivial
'fix' would break the Round-Robin scheduling we do.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Move Kconfig.perf and other perf configuration bits to events/Kconfig

Ingo says:

 "If we do a separate file we should have it in arch/x86/events/Kconfig
  (not in arch/x86/Kconfig.perf), and also move some of the other bits,
  such as PERF_EVENTS_AMD_POWER?"

Suggested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/msr: Add AMD IRPERF (Instructions Retired) performance counter

AMD Zeppelin (Family 17h, Model 00h) introduces an instructions
retired performance counter which is indicated by
CPUID.8000_0008H:EBX[1]. A dedicated Instructions Retired MSR register
(MSR 0xC000_000E9) increments once for every instruction retired.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jacob Shin <jacob.w.shin@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <rric@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1454056197-5893-3-git-send-email-ray.huang@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/msr: Add AMD PTSC (Performance Time-Stamp Counter) support

AMD Carrizo (Family 15h, Model 60h) introduces a time-stamp counter
which is indicated by CPUID.8000_0001H:ECX[27]. It increments at a 100
MHz rate in all P-states, and C states, S0, or S1. The frequency is
about 100MHz. This counter will be used to calculate processor power
and other parts. So add an interface into the MSR PMU to get the PTSC
counter value.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jacob Shin <jacob.w.shin@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <rric@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1454056197-5893-2-git-send-email-ray.huang@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/perf/intel/cstate: Modularize driver

Add the exit function and allow the driver to be built as a module.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20160320185623.658869675@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/perf/intel/cstate: Sanitize error handling

There is no point in WARN_ON() inside of a well known init function. We
already know the call stack and it's really not of critical importance whether
the registration of a PMU fails.

Aside of that for consistency reasons it's just pointless to try to register
another PMU if the first register attempt failed. There is also no value in
keeping one PMU if the second one can not be registered.

Make it consistent so we can finaly modularize the driver.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20160320185623.579794064@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/perf/intel/cstate: Sanitize probing

The whole probing functionality can simply be expressed with model matching
and a bunch of structures describing the variants. This is a first step to
make that driver modular.

While at it, get rid of completely pointless comments and name the enums so
they are self explaining.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[ Reworked probing to clear msr[].attr for all !present msrs. ]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20160320185623.500381872@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/perf/intel/cstate: Make cstate hotplug handling actually work

The current implementation aside of being an incomprehensible mess is broken.

  # cat /sys/bus/event_source/devices/cstate_core/cpumask
  0-17

That's on a quad socket machine with 72 physical cores! Qualitee stuff.

So it's not a surprise that event migration in case of CPU hotplug does not
work either.

  # perf stat -e cstate_core/c6-residency/ -C 1 sleep 60 &
  # echo 0 >/sys/devices/system/cpu/cpu1/online

Tracing cstate_pmu_event_update gives me:

 [001] cstate_pmu_event_update <-event_sched_out

After the fix it properly moves the event:

 [001] cstate_pmu_event_update <-event_sched_out
 [073] cstate_pmu_event_update <-__perf_event_read
 [073] cstate_pmu_event_update <-event_sched_out

The migration of pkg events does not work either. Not that I'm surprised.

I really could not be bothered to decode that loop mess and simply replaced it
by querying the proper cpumasks which give us the answer in a comprehensible
way.

This also requires to direct the event to the current active reader CPU in
cstate_pmu_event_init() otherwise the hotplug logic can't work.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[ Added event->cpu < 0 test to not explode]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20160320185623.422519970@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/perf/intel/rapl: Make the Intel RAPL PMU driver modular

By default, the RAPL driver will be built into the kernel. If it is
configured as a module, the supported CPU model can be auto loaded.

Also clean up the code of rapl_pmu_init().

Based-on-a-patch-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1458372050-2420-2-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/perf/intel/uncore: Make the Intel uncore PMU driver modular

By default, the uncore driver will be built into the kernel. If it is
configured as a module, the supported CPU model can be auto loaded.

This patch also cleans up the code of uncore_cpu_init() and
uncore_pci_init().

Based-on-a-patch-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1458462817-2475-1-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'perf/urgent' into perf/core, to fix up fixes before queueing up new changes

Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/amd/ibs: Fix pmu::stop() nesting

Patch 5a50f5291701 ("perf/x86/ibs: Fix race with IBS_STARTING state")
closed a big hole while opening another, smaller hole.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Fixes: 5a50f5291701 ("perf/x86/ibs: Fix race with IBS_STARTING state")
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Don't leak event in the syscall error path

In the error path, event_file not being NULL is used to determine
whether the event itself still needs to be free'd, so fix it up to
avoid leaking.

Reported-by: Leon Yu <chianglungyu@gmail.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Fixes: 130056275ade ("perf: Do not double free")
Link: http://lkml.kernel.org/r/87twk06yxp.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Fix time tracking bug with multiplexing

Stephane reported that commit:

  3cbaa5906967 ("perf: Fix ctx time tracking by introducing EVENT_TIME")

introduced a regression wrt. time tracking, as easily observed by:

> This patch introduce a bug in the time tracking of events when
> multiplexing is used.
>
> The issue is easily reproducible with the following perf run:
>
>  $ perf stat -a -C 0 -e branches,branches,branches,branches,branches,branches -I 1000
>      1.000730239            652,394      branches   (66.41%)
>      1.000730239            597,809      branches   (66.41%)
>      1.000730239            593,870      branches   (66.63%)
>      1.000730239            651,440      branches   (67.03%)
>      1.000730239            656,725      branches   (66.96%)
>      1.000730239      <not counted>      branches
>
> One branches event is shown as not having run. Yet, with
> multiplexing, all events should run especially with a 1s (-I 1000)
> interval. The delta for time_running comes out to 0. Yet, the event
> has run because the kernel is actually multiplexing the events. The
> problem is that the time tracking is the kernel and especially in
> ctx_sched_out() is wrong now.
>
> The problem is that in case that the kernel enters ctx_sched_out() with the
> following state:
>    ctx->is_active=0x7 event_type=0x1
>    Call Trace:
>     [<ffffffff813ddd41>] dump_stack+0x63/0x82
>     [<ffffffff81182bdc>] ctx_sched_out+0x2bc/0x2d0
>     [<ffffffff81183896>] perf_mux_hrtimer_handler+0xf6/0x2c0
>     [<ffffffff811837a0>] ? __perf_install_in_context+0x130/0x130
>     [<ffffffff810f5818>] __hrtimer_run_queues+0xf8/0x2f0
>     [<ffffffff810f6097>] hrtimer_interrupt+0xb7/0x1d0
>     [<ffffffff810509a8>] local_apic_timer_interrupt+0x38/0x60
>     [<ffffffff8175ca9d>] smp_apic_timer_interrupt+0x3d/0x50
>     [<ffffffff8175ac7c>] apic_timer_interrupt+0x8c/0xa0
>
> In that case, the test:
>       if (is_active & EVENT_TIME)
>
> will be false and the time will not be updated. Time must always be updated on
> sched out.

Fix this by always updating time if EVENT_TIME was set, as opposed to
only updating time when EVENT_TIME changed.

Reported-by: Stephane Eranian <eranian@google.com>
Tested-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: kan.liang@intel.com
Cc: namhyung@kernel.org
Fixes: 3cbaa5906967 ("perf: Fix ctx time tracking by introducing EVENT_TIME")
Link: http://lkml.kernel.org/r/20160329072644.GB3408@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-core-for-mingo-20160330' of git://git./linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements and fixes:

User visible changes:

  - Add support for skipping itrace instructions, useful to fast forward
    processor trace (Intel PT, BTS) to right after initialization code at the start
    of a workload (Andi Kleen)

  - Add support for backtraces in perl 'perf script's (Dima Kogan)

  - Add -U/-K (--all-user/--all-kernel) options to 'perf mem' (Jiri Olsa)

  - Make -f/--force option documentation consistent across tools (Jiri Olsa)

Infrastructure changes:

  - Add 'perf test' to check for event times (Jiri Olsa)

  - 'perf config' cleanups (Taeung Song)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-urgent-for-mingo-20160330' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fixes from Arnaldo Carvalho de Melo:

 - Fix determination of a callchain node's childlessness in
   the top/report TUI, which was preventing navigating some
   callchains, --stdio unnaffected (Andres Freund)

 - Fix jitdump's genelf assumption that PowerPC is big endian
   only (Anton Blanchard)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf jit: genelf makes assumptions about endian

Commit 9b07e27f88b9 ("perf inject: Add jitdump mmap injection support")
incorrectly assumed that PowerPC is big endian only.

Simplify things by consolidating the define of GEN_ELF_ENDIAN and checking
for __BYTE_ORDER == __BIG_ENDIAN.

The PowerPC checks were also incorrect, they do not match what gcc
emits. We should first look for __powerpc64__, then __powerpc__.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: Carl Love <cel@us.ibm.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
Fixes: 9b07e27f88b9 ("perf inject: Add jitdump mmap injection support")
Link: http://lkml.kernel.org/r/20160329175944.33a211cc@kryten
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf hists: Fix determination of a callchain node's childlessness

The 4b3a3212233a ("perf hists browser: Support flat callchains") commit
over-aggressively tried to optimize callchain_node__init_have_children().

That lead to --tui mode not allowing to expand call chain elements if a
call chain element had only one parent. That's why --inverted callgraphs
looked halfway sane, but plain ones didn't.

Revert that individual optimization, it wasn't really related to the
rest of the commit.

Signed-off-by: Andres Freund <andres@anarazel.de>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Fixes: 4b3a3212233a ("perf hists browser: Support flat callchains")
Link: http://lkml.kernel.org/r/20160330190245.GB13305@awork2.anarazel.de
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add support for skipping itrace instructions

When using 'perf script' to look at PT traces it is often useful to
ignore the initialization code at the beginning.

On larger traces which may have many millions of instructions in
initialization code doing that in a pipeline can be very slow, with perf
script spending a lot of CPU time calling printf and writing data.

This patch adds an extension to the --itrace argument that skips 'n'
events (instructions, branches or transactions) at the beginning. This
is much more efficient.

v2:
Add support for BTS (Adrian Hunter)
Document in itrace.txt
Fix branch check
Check transactions and instructions too

Committer note:

To test intel_pt one needs to make sure VT-x isn't active, i.e.
stopping KVM guests on the test machine, as described by Andi Kleen
at http://lkml.kernel.org/r/20160301234953.GD23621@tassilo.jf.intel.com

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1459187142-20035-1-git-send-email-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script perl: Perl scripts now get a backtrace, like the python ones

We have some infrastructure to use perl or python to analyze logs
generated by perf.  Prior to this patch, only the python tools had
access to backtrace information.  This patch makes this information
available to perl scripts as well.  Example:

  Let's look at malloc() calls made by the seq utility.  First we
  create a probe point:

      $ perf probe -x /lib/x86_64-linux-gnu/libc.so.6 malloc
      Added new events:
      ...

  Now we run seq, while monitoring malloc() calls with perf

      $ perf record --call-graph=dwarf -e probe_libc:malloc seq 5
      1
      2
      3
      4
      5
      [ perf record: Woken up 1 times to write data ]
      [ perf record: Captured and wrote 0.064 MB perf.data (6 samples) ]

  We can use perf to look at its log to see the malloc calls and the backtrace

      $ perf script
      seq 14195 [000] 1927993.748254: probe_libc:malloc: (7f9ff8edd320) bytes=0x22
                  7f9ff8edd320 malloc (/lib/x86_64-linux-gnu/libc-2.22.so)
                  7f9ff8e8eab0 set_binding_values.part.0 (/lib/x86_64-linux-gnu/libc-2.22.so)
                  7f9ff8e8eda1 __bindtextdomain (/lib/x86_64-linux-gnu/libc-2.22.so)
                        401b22 main (/usr/bin/seq)
                  7f9ff8e82610 __libc_start_main (/lib/x86_64-linux-gnu/libc-2.22.so)
                        402799 _start (/usr/bin/seq)
      ...

  We can also use the scripting facilities.  We create a skeleton perl
  script that simply prints out the events

      $ perf script -g perl
      generated Perl script: perf-script.pl

  We can then use this script to see the malloc() calls with a
  backtrace.  Prior to this patch, the backtrace was not available to
  the perl scripts.

      $ perf script -s perf-script.pl
      probe_libc::malloc  0 1927993.748254260  14195 seq   __probe_ip=140325052863264, bytes=34
              [7f9ff8edd320] malloc
              [7f9ff8e8eab0] set_binding_values.part.0
              [7f9ff8e8eda1] __bindtextdomain
              [401b22] main
              [7f9ff8e82610] __libc_start_main
              [402799] _start
      ...

Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: http://lkml.kernel.org/r/87mvphzld0.fsf@secretsauce.net
Signed-off-by: Dima Kogan <dima@secretsauce.net>

perf config: Rename 'v' to 'home' in set_buildid_dir()

Change the variable name 'v' to 'home' to make it more readable.

Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1459099340-16911-3-git-send-email-treeze.taeung@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf config: Rework buildid_dir_command_config to perf_buildid_config

To avoid repeated calling perf_config() remove
buildid_dir_command_config() and add new perf_buildid_config into
perf_default_config.

Because perf_config() is already called with perf_default_config at
main().

Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1459099340-16911-2-git-send-email-treeze.taeung@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf config: Remove duplicated set_buildid_dir calls

Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1459099340-16911-1-git-send-email-treeze.taeung@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Add test to check for event times

This test creates software event 'cpu-clock' attaches it in several ways
and checks that enabled and running times match.

Committer notes:

Testing it:

  [acme@jouet linux]$ perf test -v times
  44: Test events times                                        :
  --- start ---
  test child forked, pid 27170
  attaching to spawned child, enable on exec
    OK    : ena 307328, run 307328
  attaching to current thread as enabled
    OK    : ena 7826, run 7826
  attaching to current thread as disabled
    OK    : ena 738, run 738
  attaching to CPU 0 as enabled
    SKIP  : not enough rights
  attaching to CPU 0 as enabled
    SKIP  : not enough rights
  test child finished with -2
  ---- end ----
  Test events times: Skip
  [acme@jouet linux]$

  [root@jouet ~]# perf test times
  44: Test events times                                        : Ok
  [root@jouet ~]# perf test -v times
  44: Test events times                                        :
  --- start ---
  test child forked, pid 27306
  attaching to spawned child, enable on exec
    OK    : ena 479290, run 479290
  attaching to current thread as enabled
    OK    : ena 11356, run 11356
  attaching to current thread as disabled
    OK    : ena 987, run 987
  attaching to CPU 0 as enabled
    OK    : ena 3717, run 3717
  attaching to CPU 0 as enabled
    OK    : ena 2323, run 2323
  test child finished with 0
  ---- end ----
  Test events times: Ok
  [root@jouet ~]#

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1458823940-24583-7-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Make -f/--force option documentation consistent across tools

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1458823940-24583-6-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Make hists__collapse_insert_entry static

No need to export hists__collapse_insert_entry function.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1458823940-24583-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf mem: Add -U/-K (--all-user/--all-kernel) options

Add -U/-K (--all-user/--all-kernel) options to use the perf record
--all-user/--all-kernel options.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1458823940-24583-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'perf-urgent-for-mingo-20160329' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fix from Arnaldo Carvalho de Melo:

 - Add missing initialization of perf_sample.cpumode in synthesized samples,
   affects jitdump, records for pre-existing threads and records synthesized
   from processor trace data, noticed while testing intel_pt events with
   'perf script' (Arnaldo Carvalho de Melo)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tools: Add missing initialization of perf_sample.cpumode in synthesized samples

In 473398a21d28 ("perf tools: Add cpumode to struct perf_sample"), I
missed some places where perf_sample fields are directly initialized in
addition to what is done in perf_evsel__parse_sample(), namely when
synthesizing PERF_RECORD_{MMAP*,COMM,FORK,EXIT} for pre-existing threads
and also in intel_pt and intel_bts when synthesizing events from
processor trace, the jitdump code also was affected, fix it.

The problem was noticed with running:

  # perf record -e intel_pt//u true
  # perf script

Where the samples wouldn't get resolved because perf_sample.cpumode
would be left as zero, i.e. PERF_RECORD_MISC_CPUMODE_UNKNOWN, not
resolving as kernel, hypervisor or user cpu modes.

Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 473398a21d28 ("perf tools: Add cpumode to struct perf_sample")
Link: http://lkml.kernel.org/n/tip-n5sdauxgk24d5nun8kuuu2mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'perf-urgent-for-mingo-20160328' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fix from Arnaldo Carvalho de Melo:

  - Fix build break on PowerPC due to missing headers with prototypes for
    functions defined in tools/perf/arch/powerpc/util/header.c (Sukadev Bhattiprolu)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tools: Fix build break on powerpc

Commit 531d2410635c ("perf tools: Do not include stringify.h from the
kernel sources") seems to have accidentially removed the inclusion of
"util/header.h" from "arch/powerpc/util/header.c".

"util/header.h" provides the prototype for get_cpuid() and is needed to
build perf on Powerpc:

arch/powerpc/util/header.c:17:1: error: no previous prototype for 'get_cpuid' [-Werror=missing-prototypes]

Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 531d2410635c ("perf tools: Do not include stringify.h from the kernel sources")
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
[ Included "util.h" too, to get the scnprintf() prototype ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/x86: Move events_sysfs_show() outside CPU_SUP_INTEL

randconfig builds can sometimes disable CONFIG_CPU_SUP_INTEL while
enabling the AMD power reporting PMU driver, resulting in this
build failure:

  arch/x86/kernel/cpu/perf_event.h:663:31: error: 'events_sysfs_show' undeclared here (not in a function)

To fix it, move events_sysfs_show() outside of #ifdef CONFIG_CPU_SUP_INTEL.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: build test robot <lkp@intel.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sherry Hurwitz <sherry.hurwitz@amd.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@kernel.org
Cc: kbuild-all@01.org
Cc: linux-next@vger.kernel.org
Cc: spg_linux_kernel@amd.com
Link: http://lkml.kernel.org/r/1458875905-4278-1-git-send-email-ray.huang@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-urgent-for-mingo-20160324' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fixes from Arnaldo Carvalho de Melo:

 - Fix detached tarball building due to missing 'perf bench memcpy' headers
   (Arnaldo Carvalho de Melo)

 - Fix tarpkg build test error output redirection (Arnaldo Carvalho de Melo)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf bench: Fix detached tarball building due to missing 'perf bench memcpy' headers

A change on kernel files included by the 'perf bench memcpy' code grew some new
include deps, breaking the detached tarball build:

  $ make -C tools/perf build-test
  make: Entering directory '/home/acme/git/linux/tools/perf'
  - tarpkg: ./tests/perf-targz-src-pkg .
  tests/make:302: recipe for target 'tarpkg' failed
  make[1]: *** [tarpkg] Error 2
  Makefile:102: recipe for target 'build-test' failed
  make: *** [build-test] Error 2
  make: Leaving directory '/home/acme/git/linux/tools/perf'
  $ cat tools/perf/tarpkg
  ./tests/perf-targz-src-pkg .
    PERF_VERSION = 4.5.g05f5ec
    PERF_VERSION = 4.5.g05f5ec
  In file included from bench/mem-memcpy-x86-64-asm.S:9:0:
  bench/../../../arch/x86/lib/memcpy_64.S:5:29: fatal error: asm/cpufeatures.h: No such file or directory
  compilation terminated.
  mv: cannot stat ‘bench/.mem-memcpy-x86-64-asm.o.tmp’: No such file or directory
  make[5]: *** [bench/mem-memcpy-x86-64-asm.o] Error 1
  make[5]: *** Waiting for unfinished jobs....
  make[4]: *** [bench] Error 2
  make[4]: *** Waiting for unfinished jobs....
  make[3]: *** [perf-in.o] Error 2
  make[3]: *** Waiting for unfinished jobs....
  make[2]: *** [all] Error 2
  $

Add arch/*/include/asm/*features.h to tools/perf/MANIFEST so that we can
continue to use detached tarballs to build perf.

Now it builds ok, doing it manually:

  $ make help | grep perf
    perf-tar-src-pkg    - Build perf-4.5.0.tar source tarball
    perf-targz-src-pkg  - Build perf-4.5.0.tar.gz source tarball
    perf-tarbz2-src-pkg - Build perf-4.5.0.tar.bz2 source tarball
    perf-tarxz-src-pkg  - Build perf-4.5.0.tar.xz source tarball
  $ ls -la perf-4.5.0.tar
  ls: cannot access perf-4.5.0.tar: No such file or directory
  $ make perf-tar-src-pkg
    TAR
    PERF_VERSION = 4.5.g32c25b
  $ ls -la perf-4.5.0.tar
  -rw-rw-r--. 1 acme acme 6318080 Mar 24 11:52 perf-4.5.0.tar
  $ mv perf-4.5.0.tar /tmp
  $ cd /tmp
  $ tar xf perf-4.5.0.tar
  $ cd perf-4.5.0/tools/perf
  $ make > /dev/null
  PERF_VERSION = 4.5.g32c25b
  $ ls -la perf
  -rwxrwxr-x. 1 acme acme 14046416 Mar 24 11:53 perf
  $ ./perf --version
  perf version 4.5.g32c25b
  $ perf bench
  Usage:
perf bench [<common options>] <collection> <benchmark> [<options>]

        # List of all available benchmark collections:

         sched: Scheduler and IPC benchmarks
           mem: Memory access benchmarks
          numa: NUMA scheduling and MM benchmarks
         futex: Futex stressing benchmarks
           all: All benchmarks

  $ perf bench mem

        # List of available benchmarks for collection 'mem':

        memcpy: Benchmark for memcpy() functions
        memset: Benchmark for memset() functions
           all: Run all memory access benchmarks

  $ perf bench mem memcpy
  # Running 'mem/memcpy' benchmark:
  # function 'default' (Default memcpy() provided by glibc)
  # Copying 1MB bytes ...

        15.024038 GB/sec
  # function 'x86-64-unrolled' (unrolled memcpy() in arch/x86/lib/memcpy_64.S)
  # Copying 1MB bytes ...

        17.438616 GB/sec
  # function 'x86-64-movsq' (movsq-based memcpy() in arch/x86/lib/memcpy_64.S)
  # Copying 1MB bytes ...

        25.040064 GB/sec
  # function 'x86-64-movsb' (movsb-based memcpy() in arch/x86/lib/memcpy_64.S)
  # Copying 1MB bytes ...

        25.040064 GB/sec
  $

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-2c2sncwffuabw58fj1pw86gu@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Fix tarpkg build test error output redirection

So we need to trow away just stdout, leaving stderr to be caught by
the build tests infrastructure, so that we can see what went wrong
when the tarpkg build test fails:

  $ make -C tools/perf build-test
  make: Entering directory '/home/acme/git/linux/tools/perf'
  - tarpkg: ./tests/perf-targz-src-pkg .
  tests/make:302: recipe for target 'tarpkg' failed
  make[1]: *** [tarpkg] Error 2
  Makefile:102: recipe for target 'build-test' failed
  make: *** [build-test] Error 2
  make: Leaving directory '/home/acme/git/linux/tools/perf'
  $ cat tools/perf/tarpkg
  ./tests/perf-targz-src-pkg .
    PERF_VERSION = 4.5.g05f5ec
    PERF_VERSION = 4.5.g05f5ec
  In file included from bench/mem-memcpy-x86-64-asm.S:9:0:
  bench/../../../arch/x86/lib/memcpy_64.S:5:29: fatal error: asm/cpufeatures.h: No such file or directory
  compilation terminated.
  mv: cannot stat ‘bench/.mem-memcpy-x86-64-asm.o.tmp’: No such file or directory
  make[5]: *** [bench/mem-memcpy-x86-64-asm.o] Error 1
  make[5]: *** Waiting for unfinished jobs....
  make[4]: *** [bench] Error 2
  make[4]: *** Waiting for unfinished jobs....
  make[3]: *** [perf-in.o] Error 2
  make[3]: *** Waiting for unfinished jobs....
  make[2]: *** [all] Error 2
  $

So the test flow is:

1. Run: 'make -C tools/perf build-test'

2. One of its tests failed, in this case, the 'tarpkg' one

3. Look at what went wrong, by looking at the output of that test, in
   tools/perf/tarpkg

Admittedly, this should be shortcircuited to showing what went wrong directly
from the 'make build-test' step, but lets first fix this tarpkg one and the
problem it spotted, which should be fixed by adding some extra file to the
tools/perf/MANIFEST so that detached tarballs continue being self contained and
build successfully.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-ynld6egoxolmftcddpnd7oh6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'perf-core-for-mingo-20160323' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/core improvements and fixes:

User visible fixes:

 - Fix documentation of :ppp modifier in 'perf list' (Andi Kleen)

 - Fix silly nodes bitfield bits/bytes length assertion in 'perf bench numa' (Jakub Jelen)

 - Remove redundant CPU output in libtraceevent (Steven Rostedt)

 - Remove 'core_id' check in topology 'perf test' (Sukadev Bhattiprolu)

Infrastructure changes/fixes:

 - Record text offset in dso to calculate objdump address, to use with
   modules in addition to vDSO symbol address calculations (Wang Nan)

 - Move utilities.mak from perf to tools/scripts/ (Arnaldo Carvalho de Melo)

 - Add cpumode to the perf_sample struct, this way we don't need to pass
   the union event to the machine and thread resolving routines, shortening
   function signatures and allowing the future introduction of a way
   to use tracepoint events instead of the unavailable HW cycles counter on
   powerpc guests in perf kvm by just hooking on perf_evsel__parse_sample,
   at the end (Arnaldo Carvalho de Melo)

 - Remove/unexport die() related infrastructure, that at some point will
   finally be removed (Arnaldo Carvalho de Melo)

 - Adopt linux/stringify.h from the kernel sources, not to touch this
   kernel header from tools/ (Arnaldo Carvalho de Melo)

 - Stop using strbuf for things we can instead trivially use libc's asprintf()
   (Arnaldo Carvalho de Melo)

 - Ditch tools/lib/util/abspath.c, its only exported function was used at just
   one place and can be replaced by libc's realpath() (Arnaldo Carvalho de Melo)

 - Use strerror_r() in the llvm infrastructure, tread safe, its what is used
   elsewhere in tools/perf/ (Arnaldo Carvalho de Melo)

Cleanups:

 - Removed misplaced or needless __maybe_unused/export (Arnaldo Carvalho de Melo)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf llvm: Use strerror_r instead of the thread unsafe strerror one

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-5njrq9dltckgm624omw9ljgu@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf llvm: Use realpath to canonicalize paths

To kill the last user of make_nonrelative_path(), that gets ditched,
one more panicking function killed.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-3hu56rvyh4q5gxogovb6ko8a@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Unexport some methods unused outside strbuf.c

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-nq1wvtky4mpu0nupjyar7sbw@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: No need to use formatting strbuf method

We have addch() for chars, add() for fixed size data, and addstr() for
variable length strings, use them.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-0ap02fn2xtvpduj2j6b2o1j4@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf help: Use asprintf instead of adhoc equivalents

That doesn't chekcs malloc return and that, when using strbuf, if it
can't grow, just explodes away via die().

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-vr8qsjbwub7e892hpa9msz95@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove unused perf_pathdup, xstrdup functions

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-s87zi5d03m6rz622y1z6rlsa@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Do not include stringify.h from the kernel sources

Use instead the copy just made to tools/include/linux/.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-q736w12nwy98x5ox2hamp5ow@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools include: Copy linux/stringify.h from the kernel

There is code in tools/ that is directly including this file from the
kernel, and this is verboten for a while, copy it so that the next csets
can fix this situation.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-e0r3nks2uai020ndghvxv5qw@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Remove redundant CPU output

Commit a6745330789f ("tools lib traceevent: Split pevent_print_event()
into specific functionality functions") broke apart the function
pevent_print_event() into three functions.

The first function prints the comm, pid and CPU, the second prints the
timestamp.

But that commit added the printing of the CPU in the timestamp function,
which now causes pevent_print_event() to duplicate the CPU output.

Remove the redundant printing of the record's CPU from the timestamp
function.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Fixes: a6745330789f ("tools lib traceevent: Split pevent_print_event() into specific functionality functions")
Link: http://lkml.kernel.org/r/20160323101628.459375d2@gandalf.local.home
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove needless 'extern' from function prototypes

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-w246stf7ponfamclsai6b9zo@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Simplify die() mechanism

This should die altogether, but for now lets remove a bit of this stuff,
as it is not used at all.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-ade3n99xscldhg5mx2vzd8p3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove unused DIE_IF macro

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-elxg25jd4dhwod4wqbko87qh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script: Remove lots of unused arguments

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf thread: Rename perf_event__preprocess_sample_addr to thread__resolve

Since none of the perf_event fields are used anymore, just the
perf_sample ones, and since this resolves to (map, symbol) from data
structures within struct thread, rename it to thread__resolve and make
the argument ordering similar to the one in machine__resolve().

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-2b33hs9bp550tezzlhl4kejh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf machine: Rename perf_event__preprocess_sample to machine__resolve

Since we only deal with fields in the passed struct perf_sample move
this method to struct machine, that is where the perf_sample fields
will be resolved to a struct addr_location, i.e. thread, map, symbol,
etc.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-a1ww2lbm2vbuqsv4p7ilubu9@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add cpumode to struct perf_sample

To avoid parsing event->header.misc in many locations.

This will also allow setting perf.sample.{ip,cpumode} in a single place,
from tracepoint fields, as needed by 'perf kvm' with PPC guests, where
the guest hardware counters is not available at the host.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-qp3yradhyt6q3wl895b1aat0@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Forward the perf_sample in the dwarf unwind test

It _will_ be used, no sense in receiving it and nor fowarding it along.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-ht8v5et209wuoh5o6nh9pzyq@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove misplaced __maybe_unused

All over the tree.

Cc: David Ahern <dsahern@gmail.com>
cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/n/tip-8nzhnokxyp8y4v7gf0j00oyb@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf list: Fix documentation of :ppp

Correctly document what is implemented for :ppp on Intel CPUs in recent
kernels.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1458575793-12091-2-git-send-email-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf bench numa: Fix assertion for nodes bitfield

Comparing bits and bytes in numa benchmark assertion

I hit the issue on two socket Power8 machine presenting its numa nodes
as 0,1,16,17 (according to numactl). Therefore I got error (and hang of
parent process):

    perf: bench/numa.c:296: bind_to_memnode: Assertion `!(g->p.nr_nodes > (int)sizeof(nodemask))' failed.

This is obviously false positive. We can fit all the 18 nodes into
bitfield of 8 bytes (long on 64b architecture).

Signed-off-by: Jakub Jelen <jakuje@gmail.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jakub Jelen <jjelen@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: trivial@kernel.org
Link: http://lkml.kernel.org/r/1458388687-24421-1-git-send-email-jakuje@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/x86/intel/rapl: Add missing Broadwell models

Added Broadwell-H and Broadwell-Server.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: bp@alien8.de
Link: http://lkml.kernel.org/r/1458517938-25308-1-git-send-email-srinivas.pandruvada@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/uncore: Remove ev_sel_ext bit support for PCU

The ev_sel_ext in PCU_MSR_PMON_CTL is locked on some CPU models, so despite
it being documented in the SDM, if we write 1 to that bit then we can get a #GP
fault.

Which #GP the perf fuzzer happily triggered in Peter Zijlstra's testing.

Also, there are no public events which use that bit, so remove ev_sel_ext
bit support for PCU.

Reported-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1458500301-3594-1-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/amd/power: Add AMD accumulated power reporting mechanism

Introduce an AMD accumlated power reporting mechanism for the Family
15h, Model 60h processor that can be used to calculate the average
power consumed by a processor during a measurement interval. The
feature support is indicated by CPUID Fn8000_0007_EDX[12].

This feature will be implemented both in hwmon and perf. The current
design provides one event to report per package/processor power
consumption by counting each compute unit power value.

Here the gory details of how the computation is done:

* Tsample: compute unit power accumulator sample period
* Tref: the PTSC counter period (PTSC: performance timestamp counter)
* N: the ratio of compute unit power accumulator sample period to the
  PTSC period

* Jmax: max compute unit accumulated power which is indicated by
  MSR_C001007b[MaxCpuSwPwrAcc]

* Jx/Jy: compute unit accumulated power which is indicated by
  MSR_C001007a[CpuSwPwrAcc]

* Tx/Ty: the value of performance timestamp counter which is indicated
  by CU_PTSC MSR_C0010280[PTSC]
* PwrCPUave: CPU average power

i. Determine the ratio of Tsample to Tref by executing CPUID Fn8000_0007.
N = value of CPUID Fn8000_0007_ECX[CpuPwrSampleTimeRatio[15:0]].

ii. Read the full range of the cumulative energy value from the new
    MSR MaxCpuSwPwrAcc.
Jmax = value returned.

iii. At time x, software reads CpuSwPwrAcc and samples the PTSC.
Jx = value read from CpuSwPwrAcc and Tx = value read from PTSC.

iv. At time y, software reads CpuSwPwrAcc and samples the PTSC.
Jy = value read from CpuSwPwrAcc and Ty = value read from PTSC.

v. Calculate the average power consumption for a compute unit over
time period (y-x). Unit of result is uWatt:

if (Jy < Jx) // Rollover has occurred
Jdelta = (Jy + Jmax) - Jx
else
Jdelta = Jy - Jx
PwrCPUave = N * Jdelta * 1000 / (Ty - Tx)

Simple example:

  root@hr-zp:/home/ray/tip# ./tools/perf/perf stat -a -e 'power/power-pkg/' make -j4
    CHK     include/config/kernel.release
    CHK     include/generated/uapi/linux/version.h
    CHK     include/generated/utsrelease.h
    CHK     include/generated/timeconst.h
    CHK     include/generated/bounds.h
    CHK     include/generated/asm-offsets.h
    CALL    scripts/checksyscalls.sh
    CHK     include/generated/compile.h
    SKIPPED include/generated/compile.h
    Building modules, stage 2.
  Kernel: arch/x86/boot/bzImage is ready  (#40)
    MODPOST 4225 modules

   Performance counter stats for 'system wide':

              183.44 mWatts power/power-pkg/

       341.837270111 seconds time elapsed

  root@hr-zp:/home/ray/tip# ./tools/perf/perf stat -a -e 'power/power-pkg/' sleep 10

   Performance counter stats for 'system wide':

                0.18 mWatts power/power-pkg/

        10.012551815 seconds time elapsed

Suggested-by: Peter Zijlstra <peterz@infradead.org>
Suggested-by: Ingo Molnar <mingo@kernel.org>
Suggested-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Richter <rric@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: jacob.w.shin@gmail.com
Link: http://lkml.kernel.org/r/1457502306-2559-1-git-send-email-ray.huang@amd.com
[ Fixed the modular build. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/cpufeature, perf/x86: Add AMD Accumulated Power Mechanism feature flag

AMD CPU family 15h model 0x60 introduces a mechanism for measuring
accumulated power. It is used to report the processor power consumption
and support for it is indicated by CPUID Fn8000_0007_EDX[12].

Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andreas Herrmann <herrmann.der.user@googlemail.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hector Marco-Gisbert <hecmargi@upv.es>
Cc: Jacob Shin <jacob.w.shin@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Kristen Carlson Accardi <kristen@linux.intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <rric@kernel.org>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Wan Zongshun <Vincent.Wan@amd.com>
Cc: spg_linux_kernel@amd.com
Link: http://lkml.kernel.org/r/1452739808-11871-4-git-send-email-ray.huang@amd.com
[ Resolved conflict and moved the synthetic CPUID slot to 19. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Document some hotplug bits

Document some of the hotplug notifier usage.

Requested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/amd: Add support for new IOMMU performance events

This patch adds new IOMMU performance event based on
the information in table 74 of the AMD I/O Virtualization Technology
(IOMMU) Specification (Document Id: 4882, Rev 2.62, Feb 2015)

Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Joerg Roedel <jroedel@suse.de>
Acked-by: Joerg Roedel <jroedel@suse.de>
Cc: <acme@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://support.amd.com/TechDocs/48882_IOMMU.pdf
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/amd: Move nodes_per_socket into bsp_init_amd()

nodes_per_socket is static and it needn't be initialized many
times during every CPU core init. So move its initialization into
bsp_init_amd().

Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andreas Herrmann <herrmann.der.user@googlemail.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hector Marco-Gisbert <hecmargi@upv.es>
Cc: Jacob Shin <jacob.w.shin@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <rric@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: spg_linux_kernel@amd.com
Link: http://lkml.kernel.org/r/1452739808-11871-2-git-send-email-ray.huang@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/cqm: Factor out some common code

Having the same code twice (and once quite ugly) is fragile.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/mbm: Add support for MBM counter overflow handling

This patch adds a per package timer which periodically updates the
memory bandwidth counters for the events that are currently active.

Current patch has a periodic timer every 1s since the SDM guarantees
that the counter will not overflow in 1s but this time can be definitely
improved by calibrating on the system. The overflow is really a function
of the max memory b/w that the socket can support, max counter value and
scaling factor.

Signed-off-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: fenghua.yu@intel.com
Cc: h.peter.anvin@intel.com
Cc: ravi.v.shankar@intel.com
Cc: vikas.shivappa@intel.com
Link: http://lkml.kernel.org/r/013b756c5006b1c4ca411f3ecf43ed52f19fbf87.1457723885.git.tony.luck@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/mbm: Implement RMID recycling

RMID could be allocated or deallocated as part of RMID recycling.

When an RMID is allocated for MBM event, the MBM counter needs to be
initialized because next time we read the counter we need the previous
value to account for total bytes that went to the memory controller.

Similarly, when RMID is deallocated we need to update the ->count
variable.

Signed-off-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: fenghua.yu@intel.com
Cc: h.peter.anvin@intel.com
Cc: ravi.v.shankar@intel.com
Cc: vikas.shivappa@intel.com
Link: http://lkml.kernel.org/r/1457652732-4499-6-git-send-email-vikas.shivappa@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/mbm: Add memory bandwidth monitoring event management

Includes all the core infrastructure to measure the total_bytes and
bandwidth.

We have per socket counters for both total system wide L3 external
bytes and local socket memory-controller bytes. The OS does MSR writes
to MSR_IA32_QM_EVTSEL and MSR_IA32_QM_CTR to read the counters and
uses the IA32_PQR_ASSOC_MSR to associate the RMID with the task. The
tasks have a common RMID for CQM (cache quality of service monitoring)
and MBM. Hence most of the scheduling code is reused from CQM.

Signed-off-by: Tony Luck <tony.luck@intel.com>
[ Restructured rmid_read to not have an obvious hole, removed MBM_CNTR_MAX as its unused. ]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: fenghua.yu@intel.com
Cc: h.peter.anvin@intel.com
Cc: ravi.v.shankar@intel.com
Cc: vikas.shivappa@intel.com
Link: http://lkml.kernel.org/r/abd7aac9a18d93b95b985b931cf258df0164746d.1457723885.git.tony.luck@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/mbm: Add Intel Memory B/W Monitoring enumeration and init

The MBM init patch enumerates the Intel MBM (Memory b/w monitoring)
and initializes the perf events and datastructures for monitoring the
memory b/w.

Its based on original patch series by Tony Luck and Kanaka Juvva.

Memory bandwidth monitoring (MBM) provides OS/VMM a way to monitor
bandwidth from one level of cache to another. The current patches
support L3 external bandwidth monitoring. It supports both 'local
bandwidth' and 'total bandwidth' monitoring for the socket. Local
bandwidth measures the amount of data sent through the memory controller
on the socket and total b/w measures the total system bandwidth.

Extending the cache quality of service monitoring (CQM) we add two
more events to the perf infrastructure:

  intel_cqm_llc/local_bytes - bytes sent through local socket memory controller
  intel_cqm_llc/total_bytes - total L3 external bytes sent

The tasks are associated with a Resouce Monitoring ID (RMID) just like
in CQM and OS uses a MSR write to indicate the RMID of the task during
scheduling.

Signed-off-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: fenghua.yu@intel.com
Cc: h.peter.anvin@intel.com
Cc: ravi.v.shankar@intel.com
Cc: vikas.shivappa@intel.com
Link: http://lkml.kernel.org/r/1457652732-4499-4-git-send-email-vikas.shivappa@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/cqm: Fix CQM memory leak and notifier leak

Fixes the hotcpu notifier leak and other global variable memory leaks
during CQM (cache quality of service monitoring) initialization.

Signed-off-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: fenghua.yu@intel.com
Cc: h.peter.anvin@intel.com
Cc: ravi.v.shankar@intel.com
Cc: vikas.shivappa@intel.com
Link: http://lkml.kernel.org/r/1457652732-4499-3-git-send-email-vikas.shivappa@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/cqm: Fix CQM handling of grouping events into a cache_group

Currently CQM (cache quality of service monitoring) is grouping all
events belonging to same PID to use one RMID. However its not counting
all of these different events. Hence we end up with a count of zero
for all events other than the group leader.

The patch tries to address the issue by keeping a flag in the
perf_event.hw which has other CQM related fields. The field is updated
at event creation and during grouping.

Signed-off-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
[peterz: Changed hw_perf_event::is_group_event to an int]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: fenghua.yu@intel.com
Cc: h.peter.anvin@intel.com
Cc: ravi.v.shankar@intel.com
Cc: vikas.shivappa@intel.com
Link: http://lkml.kernel.org/r/1457652732-4499-2-git-send-email-vikas.shivappa@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Fix Undefined behaviour in rb_alloc()

Sasha reported:

 [ 3494.030114] UBSAN: Undefined behaviour in kernel/events/ring_buffer.c:685:22
 [ 3494.030647] shift exponent -1 is negative

Andrey spotted that this is because:

  It happens if nr_pages = 0:
     rb->page_order = ilog2(nr_pages);

Fix it by making both assignments conditional on nr_pages; since
otherwise they should both be 0 anyway, and will be because of the
kzalloc() used to allocate the structure.

Reported-by: Sasha Levin <sasha.levin@oracle.com>
Reported-by: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20160129141751.GA407@worktop
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/BTS: Fix RCU usage

This splat reminds us:

[ 8166.045595] [ INFO: suspicious RCU usage. ]

[ 8166.168972]  [<ffffffff81127837>] lockdep_rcu_suspicious+0xe7/0x120
[ 8166.175966]  [<ffffffff811e0bae>] perf_callchain+0x23e/0x250
[ 8166.182280]  [<ffffffff811dda3d>] perf_prepare_sample+0x27d/0x350
[ 8166.189082]  [<ffffffff8100f503>] intel_pmu_drain_bts_buffer+0x133/0x200

... that as the core code does, one should hold rcu_read_lock() over that
entire BTS event-output generation sequence as well.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Fix dynamic interrupt throttle

There were two problems with the dynamic interrupt throttle mechanism,
both triggered by the same action.

When you (or perf_fuzzer) write a huge value into
/proc/sys/kernel/perf_event_max_sample_rate the computed
perf_sample_allowed_ns becomes 0. This effectively disables the whole
dynamic throttle.

This is fixed by ensuring update_perf_cpu_limits() never sets the
value to 0. However, we allow disabling of the dynamic throttle by
writing 100 to /proc/sys/kernel/perf_cpu_time_max_percent. This will
generate a warning in dmesg.

The second problem is that by setting the max_sample_rate to a huge
number, the adaptive process can take a few tries, since it halfs the
limit each time. Change that to directly compute a new value based on
the observed duration.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/ibs: Add IBS interrupt to the dynamic throttle

Interrupt throttling is normally only done against
sysctl_perf_event_sample_rate. This means that if that number is too
high (for whatever reason) you can lock up your machine.

We have, however, a dynamic throttling scheme too, but for that to
work, we need to add a callback to the interrupt handler, IBS did not
have this, so add it.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/ibs: Fix race with IBS_STARTING state

While tracing the IBS bits I saw the NMI hitting between clearing
IBS_STARTING and the actual MSR writes to disable the counter.

Since IBS_STARTING was cleared, the handler assumed these were spurious
NMIs and because STOPPING wasn't set yet either, insta-triggered an
"Unknown NMI".

Cure this by clearing IBS_STARTING after disabling the hardware.

Tested-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/ibs: Fix IBS throttle

When the IBS IRQ handler get a !0 return from perf_event_overflow;
meaning it should throttle the event, it only disables it, it doesn't
call perf_ibs_stop().

This confuses the state machine, as we'll use pmu::start() ->
perf_ibs_start() to unthrottle.

Tested-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vince@deater.net>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: dvyukov@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Link: http://lkml.kernel.org/r/20160311142346.GE6344@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Fix the unthrottle logic

Its possible to IOC_PERIOD while the event is throttled, this would
re-start the event and the next tick would then try to unthrottle it,
and find the event wasn't actually stopped anymore.

This would tickle a WARN in the x86-pmu code which isn't expecting to
start a !stopped event.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: dvyukov@google.com
Cc: oleg@redhat.com
Cc: panand@redhat.com
Cc: sasha.levin@oracle.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160310143924.GR6356@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/oprofile/nmi: Add missing hotplug FROZEN handling

We really do not want to keep that nmi enabled across suspend/resume.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86/hpet: Use proper mask to modify hotplug action

Magic hex constants are a guarantee for wreckage when the defines change.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86/apic/uv: Fix the hotplug notifier

The notifier is missing the CPU_DOWN_FAILED transition. That leaves the
heartbeat disabled when CPU_DOWN_PREPARE fails.

It also does not handle the FROZEN transition variants. That might not be an
issue for UV, but it's inconsistent.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Dimitri Sivanich <sivanich@sgi.com>

x86/apb/timer: Use proper mask to modify hotplug action

Magic hex constants are a guarantee for wreckage when the defines change.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86/topology: Use total_cpus not nr_cpu_ids for logical packages

nr_cpu_ids can be limited on the command line via nr_cpus=. That can break the
logical package management because it results in a smaller number of packages,
but the cpus to online are occupying the full package space as the hyper
threads are enumerated after the physical cores typically.

total_cpus is the real possible cpu space not limited by nr_cpus command line
and gives us the proper number of packages.

Reported-by: Mike Galbraith <umgwanakikbuti@gmail.com>
Fixes: 1f12e32f4cd5 ("x86/topology: Create logical package id")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Xiong Zhou <jencce.kernel@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Andreas Herrmann <aherrmann@suse.com>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1603181254330.3978@nanos

x86/topology: Fix Intel HT disable

As per the comment in the code; due to BIOS it is sometimes impossible to know
if there actually are smp siblings until the machine is fully enumerated. So
we rather overestimate the number of possible packages.

Fixes: 1f12e32f4cd5 ("x86/topology: Create logical package id")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: aherrmann@suse.com
Cc: jencce.kernel@gmail.com
Cc: bp@alien8.de
Cc: Mike Galbraith <umgwanakikbuti@gmail.com>
Link: http://lkml.kernel.org/r/20160318150538.611014173@infradead.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86/topology: Fix logical package mapping

That first branch testing pkg against __max_logical_packages is wrong,
because if the first pkg id is larger, then the find_first_zero will
find us logical package id 0. However, if the second pkg id is indeed
0, we'll again claim it without testing if it was already taken.

Also, it fails to print the mapping.

Fixes: 1f12e32f4cd5 ("x86/topology: Create logical package id")
Reported-by: Xiong Zhou <jencce.kernel@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: aherrmann@suse.com
Cc: bp@alien8.de
Cc: Mike Galbraith <umgwanakikbuti@gmail.com>
Link: http://lkml.kernel.org/r/20160317095220.GO6344@twins.programming.kicks-ass.net
Link: http://lkml.kernel.org/r/20160318150538.482393396@infradead.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

perf symbols: Record text offset in dso to calculate objdump address

Store DSO's .text offset into DSO, used for VDSOs and will also be used for
other needs, like handling kernel modules.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Cody P Schafer <dev@codyps.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kirill Smelkov <kirr@nexedi.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456479154-136027-2-git-send-email-wangnan0@huawei.com
[ Extracted from larger patch ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Move utilities.mak from perf to tools/scripts/

As it is used by several other tools, better move it outside tools/perf.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-34s9kue3xq9w5mijdmfrfx8s@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

x86/irq: Cure live lock in fixup_irqs()

Harry reported, that he's able to trigger a system freeze with cpu hot
unplug. The freeze turned out to be a live lock caused by recent changes in
irq_force_complete_move().

When fixup_irqs() and from there irq_force_complete_move() is called on the
dying cpu, then all other cpus are in stop machine an wait for the dying cpu
to complete the teardown. If there is a move of an interrupt pending then
irq_force_complete_move() sends the cleanup IPI to the cpus in the old_domain
mask and waits for them to clear the mask. That's obviously impossible as
those cpus are firmly stuck in stop machine with interrupts disabled.

I should have known that, but I completely overlooked it being concentrated on
the locking issues around the vectors. And the existance of the call to
__irq_complete_move() in the code, which actually sends the cleanup IPI made
it reasonable to wait for that cleanup to complete. That call was bogus even
before the recent changes as it was just a pointless distraction.

We have to look at two cases:

1) The move_in_progress flag of the interrupt is set

   This means the ioapic has been updated with the new vector, but it has not
   fired yet. In theory there is a race:

   set_ioapic(new_vector) <-- Interrupt is raised before update is effective,
          i.e. it's raised on the old vector.

   So if the target cpu cannot handle that interrupt before the old vector is
   cleaned up, we get a spurious interrupt and in the worst case the ioapic
   irq line becomes stale, but my experiments so far have only resulted in
   spurious interrupts.

   But in case of cpu hotplug this should be a non issue because if the
   affinity update happens right before all cpus rendevouz in stop machine,
   there is no way that the interrupt can be blocked on the target cpu because
   all cpus loops first with interrupts enabled in stop machine, so the old
   vector is not yet cleaned up when the interrupt fires.

   So the only way to run into this issue is if the delivery of the interrupt
   on the apic/system bus would be delayed beyond the point where the target
   cpu disables interrupts in stop machine. I doubt that it can happen, but at
   least there is a theroretical chance. Virtualization might be able to
   expose this, but AFAICT the IOAPIC emulation is not as stupid as the real
   hardware.

   I've spent quite some time over the weekend to enforce that situation,
   though I was not able to trigger the delayed case.

2) The move_in_progress flag is not set and the old_domain cpu mask is not
   empty.

   That means, that an interrupt was delivered after the change and the
   cleanup IPI has been sent to the cpus in old_domain, but not all CPUs have
   responded to it yet.

In both cases we can assume that the next interrupt will arrive on the new
vector, so we can cleanup the old vectors on the cpus in the old_domain cpu
mask.

Fixes: 98229aa36caa "x86/irq: Plug vector cleanup race"
Reported-by: Harry Junior <harryjr@outlook.fr>
Tested-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Joe Lawrence <joe.lawrence@stratus.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1603140931430.3657@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86/tsc: Prevent NULL pointer deref in calibrate_delay_is_known()

The topology_core_cpumask is used to find a neighbour cpu in
calibrate_delay_is_known(). It might not be allocated at the first invocation
of that function on the boot cpu, when CONFIG_CPUMASK_OFFSTACK is set.

The mask is allocated later in native_smp_prepare_cpus. As a consequence the
underlying find_next_bit() call dereferences a NULL pointer.

Add a proper check to prevent this.

Fixes: c25323c07345 "x86/tsc: Use topology functions"
Reported-and-tested-by: Richard W.M. Jones <rjones@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Josh Boyer <jwboyer@fedoraproject.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1603180843270.3978@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

x86/apic: Fix suspicious RCU usage in smp_trace_call_function_interrupt()

Since 4.4, I've been able to trigger this occasionally:

===============================
[ INFO: suspicious RCU usage. ]
4.5.0-rc7-think+ #3 Not tainted
Cc: Andi Kleen <ak@linux.intel.com>
Link: http://lkml.kernel.org/r/20160315012054.GA17765@codemonkey.org.uk
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
-------------------------------
./arch/x86/include/asm/msr-trace.h:47 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

RCU used illegally from idle CPU!
rcu_scheduler_active = 1, debug_locks = 1
RCU used illegally from extended quiescent state!
no locks held by swapper/3/0.

stack backtrace:
CPU: 3 PID: 0 Comm: swapper/3 Not tainted 4.5.0-rc7-think+ #3
 ffffffff92f821e0 1f3e5c340597d7fc ffff880468e07f10 ffffffff92560c2a
 ffff880462145280 0000000000000001 ffff880468e07f40 ffffffff921376a6
 ffffffff93665ea0 0000cc7c876d28da 0000000000000005 ffffffff9383dd60
Call Trace:
 <IRQ>  [<ffffffff92560c2a>] dump_stack+0x67/0x9d
 [<ffffffff921376a6>] lockdep_rcu_suspicious+0xe6/0x100
 [<ffffffff925ae7a7>] do_trace_write_msr+0x127/0x1a0
 [<ffffffff92061c83>] native_apic_msr_eoi_write+0x23/0x30
 [<ffffffff92054408>] smp_trace_call_function_interrupt+0x38/0x360
 [<ffffffff92d1ca60>] trace_call_function_interrupt+0x90/0xa0
 <EOI>  [<ffffffff92ac5124>] ? cpuidle_enter_state+0x1b4/0x520

Move the entering_irq() call before ack_APIC_irq(), because entering_irq()
tells the RCU susbstems to end the extended quiescent state, so that the
following trace call in ack_APIC_irq() works correctly.

Suggested-by: Andi Kleen <ak@linux.intel.com>
Fixes: 4787c368a9bc "x86/tracing: Add irq_enter/exit() in smp_trace_reschedule_interrupt()"
Signed-off-by: Dave Jones <davej@codemonkey.org.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org

x86/iopl: Fix iopl capability check on Xen PV

iopl(3) is supposed to work if iopl is already 3, even if
unprivileged.  This didn't work right on Xen PV.  Fix it.

Reviewewd-by: Jan Beulich <JBeulich@suse.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jan Beulich <JBeulich@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/8ce12013e6e4c0a44a97e316be4a6faff31bd5ea.1458162709.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/iopl/64: Properly context-switch IOPL on Xen PV

On Xen PV, regs->flags doesn't reliably reflect IOPL and the
exit-to-userspace code doesn't change IOPL.  We need to context
switch it manually.

I'm doing this without going through paravirt because this is
specific to Xen PV.  After the dust settles, we can merge this with
the 32-bit code, tidy up the iopl syscall implementation, and remove
the set_iopl pvop entirely.

Fixes XSA-171.

Reviewewd-by: Jan Beulich <JBeulich@suse.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jan Beulich <JBeulich@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/693c3bd7aeb4d3c27c92c622b7d0f554a458173c.1458162709.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

selftests/x86: Add an iopl test

This exercises two cases that are known to be buggy on Xen PV right
now.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jan Beulich <JBeulich@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/61afe904c95c92abb29cd075b51e10e7feb0f774.1458162709.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'x86/cleanups' into x86/urgent

Pull in some merge window leftovers.

Signed-off-by: Ingo Molnar <mingo@kernel.org>

x86/mm, x86/mce: Fix return type/value for memcpy_mcsafe()

Returning a 'bool' was very unpopular. Doubly so because the
code was just wrong (returning zero for true, one for false;
great for shell programming, not so good for C).

Change return type to "int". Keep zero as the success indicator
because it matches other similar code and people may be more
comfortable writing:

if (memcpy_mcsafe(to, from, count)) {
printk("Sad panda, copy failed\n");
...
}

Make the failure return value -EFAULT for now.

Reported by: Mika Penttilä <mika.penttila@nextfour.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: mika.penttila@nextfour.com
Fixes: 92b0729c34ca ("x86/mm, x86/mce: Add memcpy_mcsafe()")
Link: http://lkml.kernel.org/r/695f14233fa7a54fcac4406c706d7fec228e3f4c.1457993040.git.tony.luck@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'linus' into x86/urgent, to pick up dependencies for a fix

Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'smp-hotplug-for-linus' of git://git./linux/kernel/git/tip/tip

Pull cpu hotplug updates from Thomas Gleixner:
 "This is the first part of the ongoing cpu hotplug rework:

   - Initial implementation of the state machine

   - Runs all online and prepare down callbacks on the plugged cpu and
     not on some random processor

   - Replaces busy loop waiting with completions

   - Adds tracepoints so the states can be followed"

More detailed commentary on this work from an earlier email:
 "What's wrong with the current cpu hotplug infrastructure?

   - Asymmetry

     The hotplug notifier mechanism is asymmetric versus the bringup and
     teardown.  This is mostly caused by the notifier mechanism.

   - Largely undocumented dependencies

     While some notifiers use explicitely defined notifier priorities,
     we have quite some notifiers which use numerical priorities to
     express dependencies without any documentation why.

   - Control processor driven

     Most of the bringup/teardown of a cpu is driven by a control
     processor.  While it is understandable, that preperatory steps,
     like idle thread creation, memory allocation for and initialization
     of essential facilities needs to be done before a cpu can boot,
     there is no reason why everything else must run on a control
     processor.  Before this patch series, bringup looks like this:

       Control CPU                     Booting CPU

       do preparatory steps
       kick cpu into life

                                       do low level init

       sync with booting cpu           sync with control cpu

       bring the rest up

   - All or nothing approach

     There is no way to do partial bringups.  That's something which is
     really desired because we waste e.g.  at boot substantial amount of
     time just busy waiting that the cpu comes to life.  That's stupid
     as we could very well do preparatory steps and the initial IPI for
     other cpus and then go back and do the necessary low level
     synchronization with the freshly booted cpu.

   - Minimal debuggability

     Due to the notifier based design, it's impossible to switch between
     two stages of the bringup/teardown back and forth in order to test
     the correctness.  So in many hotplug notifiers the cancel
     mechanisms are either not existant or completely untested.

   - Notifier [un]registering is tedious

     To [un]register notifiers we need to protect against hotplug at
     every callsite.  There is no mechanism that bringup/teardown
     callbacks are issued on the online cpus, so every caller needs to
     do it itself.  That also includes error rollback.

  What's the new design?

     The base of the new design is a symmetric state machine, where both
     the control processor and the booting/dying cpu execute a well
     defined set of states.  Each state is symmetric in the end, except
     for some well defined exceptions, and the bringup/teardown can be
     stopped and reversed at almost all states.

     So the bringup of a cpu will look like this in the future:

       Control CPU                     Booting CPU

       do preparatory steps
       kick cpu into life

                                       do low level init

       sync with booting cpu           sync with control cpu

                                       bring itself up

     The synchronization step does not require the control cpu to wait.
     That mechanism can be done asynchronously via a worker or some
     other mechanism.

     The teardown can be made very similar, so that the dying cpu cleans
     up and brings itself down.  Cleanups which need to be done after
     the cpu is gone, can be scheduled asynchronously as well.

  There is a long way to this, as we need to refactor the notion when a
  cpu is available.  Today we set the cpu online right after it comes
  out of the low level bringup, which is not really correct.

  The proper mechanism is to set it to available, i.e. cpu local
  threads, like softirqd, hotplug thread etc. can be scheduled on that
  cpu, and once it finished all booting steps, it's set to online, so
  general workloads can be scheduled on it.  The reverse happens on
  teardown.  First thing to do is to forbid scheduling of general
  workloads, then teardown all the per cpu resources and finally shut it
  off completely.

  This patch series implements the basic infrastructure for this at the
  core level.  This includes the following:

   - Basic state machine implementation with well defined states, so
     ordering and prioritization can be expressed.

   - Interfaces to [un]register state callbacks

     This invokes the bringup/teardown callback on all online cpus with
     the proper protection in place and [un]installs the callbacks in
     the state machine array.

     For callbacks which have no particular ordering requirement we have
     a dynamic state space, so that drivers don't have to register an
     explicit hotplug state.

     If a callback fails, the code automatically does a rollback to the
     previous state.

   - Sysfs interface to drive the state machine to a particular step.

     This is only partially functional today.  Full functionality and
     therefor testability will be achieved once we converted all
     existing hotplug notifiers over to the new scheme.

   - Run all CPU_ONLINE/DOWN_PREPARE notifiers on the booting/dying
     processor:

       Control CPU                     Booting CPU

       do preparatory steps
       kick cpu into life

                                       do low level init

       sync with booting cpu           sync with control cpu
       wait for boot
                                       bring itself up

                                       Signal completion to control cpu

     In a previous step of this work we've done a full tree mechanical
     conversion of all hotplug notifiers to the new scheme.  The balance
     is a net removal of about 4000 lines of code.

     This is not included in this series, as we decided to take a
     different approach.  Instead of mechanically converting everything
     over, we will do a proper overhaul of the usage sites one by one so
     they nicely fit into the symmetric callback scheme.

     I decided to do that after I looked at the ugliness of some of the
     converted sites and figured out that their hotplug mechanism is
     completely buggered anyway.  So there is no point to do a
     mechanical conversion first as we need to go through the usage
     sites one by one again in order to achieve a full symmetric and
     testable behaviour"

* 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (23 commits)
  cpu/hotplug: Document states better
  cpu/hotplug: Fix smpboot thread ordering
  cpu/hotplug: Remove redundant state check
  cpu/hotplug: Plug death reporting race
  rcu: Make CPU_DYING_IDLE an explicit call
  cpu/hotplug: Make wait for dead cpu completion based
  cpu/hotplug: Let upcoming cpu bring itself fully up
  arch/hotplug: Call into idle with a proper state
  cpu/hotplug: Move online calls to hotplugged cpu
  cpu/hotplug: Create hotplug threads
  cpu/hotplug: Split out the state walk into functions
  cpu/hotplug: Unpark smpboot threads from the state machine
  cpu/hotplug: Move scheduler cpu_online notifier to hotplug core
  cpu/hotplug: Implement setup/removal interface
  cpu/hotplug: Make target state writeable
  cpu/hotplug: Add sysfs state interface
  cpu/hotplug: Hand in target state to _cpu_up/down
  cpu/hotplug: Convert the hotplugged cpu work to a state machine
  cpu/hotplug: Convert to a state machine for the control processor
  cpu/hotplug: Add tracepoints
  ...

Merge branch 'irq-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq updates from Thomas Gleixner:
 "The 4.6 pile of irq updates contains:

   - Support for IPI irqdomains to support proper integration of IPIs to
     and from coprocessors.  The first user of this new facility is
     MIPS.  The relevant MIPS patches come with the core to avoid merge
     ordering issues and have been acked by Ralf.

   - A new command line option to set the default interrupt affinity
     mask at boot time.

   - Support for some more new ARM and MIPS interrupt controllers:
     tango, alpine-msix and bcm6345-l1

   - Two small cleanups for x86/apic which we merged into irq/core to
     avoid yet another branch in x86 with two tiny commits.

   - The usual set of updates, cleanups in drivers/irqchip.  Mostly in
     the area of ARM-GIC, arada-37-xp and atmel chips.  Nothing
     outstanding here"

* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (56 commits)
  irqchip/irq-alpine-msi: Release the correct domain on error
  irqchip/mxs: Fix error check of of_io_request_and_map()
  irqchip/sunxi-nmi: Fix error check of of_io_request_and_map()
  genirq: Export IRQ functions for module use
  irqchip/gic/realview: Support more RealView DCC variants
  Documentation/bindings: Document the Alpine MSIX driver
  irqchip: Add the Alpine MSIX interrupt controller
  irqchip/gic-v3: Always return IRQ_SET_MASK_OK_DONE in gic_set_affinity
  irqchip/gic-v3-its: Mark its_init() and its children as __init
  irqchip/gic-v3: Remove gic_root_node variable from the ITS code
  irqchip/gic-v3: ACPI: Add redistributor support via GICC structures
  irqchip/gic-v3: Add ACPI support for GICv3/4 initialization
  irqchip/gic-v3: Refactor gic_of_init() for GICv3 driver
  x86/apic: Deinline _flat_send_IPI_mask, save ~150 bytes
  x86/apic: Deinline __default_send_IPI_*, save ~200 bytes
  dt-bindings: interrupt-controller: Add SoC-specific compatible string to Marvell ODMI
  irqchip/mips-gic: Add new DT property to reserve IPIs
  MIPS: Delete smp-gic.c
  MIPS: Make smp CMP, CPS and MT use the new generic IPI functions
  MIPS: Add generic SMP IPI support
  ...