OSDN Git Service

sagit-ice-cold/kernel_xiaomi_msm8998.git
9 years agoperf thread_map: Introduce thread_map__reset function
Jiri Olsa [Fri, 26 Jun 2015 09:29:06 +0000 (11:29 +0200)]
perf thread_map: Introduce thread_map__reset function

We need to reset newly allocated 'struct thread_map_data' entries,
because we will introduce new comm memeber, which will get set later or
not at all.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435310967-14570-2-git-send-email-jolsa@kernel.org
[ Use sizeof(map->map[0]) to be independent of the array entry type ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf trace: Validate syscall list passed via -e argument
Arnaldo Carvalho de Melo [Thu, 25 Jun 2015 22:32:33 +0000 (19:32 -0300)]
perf trace: Validate syscall list passed via -e argument

The 'trace' tool was accepting any names passed and just looking if
syscalls returned via the raw_syscalls:* tracepoints were in that list,
leading to it accepting perf events and then never finding any, as those
are not valid syscall names, confusing users.

Fix it by checking each entry in the list using audit_name_to_syscall,
telling the user which entries are invalid and suggesting where to look
for valid syscall names.

E.g:

  [root@zoo ~]# trace -e open,foo,bar,close,baz
  Error: Invalid syscall bar, baz, foo
  Hint:  try 'perf list syscalls:sys_enter_*'
  Hint:  and: 'man syscalls'
  [root@zoo ~]#

Reported-by: Flavio Leitner <fbl@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/n/tip-4g1i3m1z6fzsrznn2umi02wa@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoMerge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Fri, 26 Jun 2015 08:38:11 +0000 (10:38 +0200)]
Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/core improvements and refactorings from Arnaldo Carvalho de Melo:

Infrastructure changes:

  - Reference count the cpu_map and thread_map classes. (Jiri Olsa)

  - Set evsel->{cpus,threads} from the evlist, if not set,
    allowing the generalization of some 'perf stat' functions that
    previously were accessing private static evlist variable. (Jiri Olsa)

  - Delete an unnecessary check before the calling
    free_event_desc() (Markus Elfring)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Fri, 26 Jun 2015 08:37:15 +0000 (10:37 +0200)]
Merge tag 'perf-urgent-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fix from Arnaldo Carvalho de Melo:

  - Fix to show proper convergence stats in 'perf bench numa' (Srikar Dronamraju)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf tools: Make perf_evsel__(nr_)cpus generic
Jiri Olsa [Mon, 22 Jun 2015 22:36:08 +0000 (00:36 +0200)]
perf tools: Make perf_evsel__(nr_)cpus generic

Because we now propagate all evlist's cpu_maps and thread_map objects
through all evsels, the perf_evsel__(nr_)cpus no longer need to be
specific to stat object and check evlist and target objects.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-8-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf evlist: Propagate thread maps through the evlist
Jiri Olsa [Mon, 22 Jun 2015 22:36:07 +0000 (00:36 +0200)]
perf evlist: Propagate thread maps through the evlist

Propagate evlist's thread_map object through all the evsel objects.

It'll be handy to access evsel's threads directly in following patches.
The reason is there's no link from evsel to evlist which hold threads
map now and evlist is not always available.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-7-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf evlist: Propagate cpu maps to evsels in an evlist
Jiri Olsa [Mon, 22 Jun 2015 22:36:06 +0000 (00:36 +0200)]
perf evlist: Propagate cpu maps to evsels in an evlist

Propagate evlist's cpu_map object through all the evsel objects, while
keeping already configured evsel->cpus.

It'll be handy to access evsel's cpus directly in following patches.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-6-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Add reference counting for thread_map object
Jiri Olsa [Mon, 22 Jun 2015 22:36:05 +0000 (00:36 +0200)]
perf tools: Add reference counting for thread_map object

Adding reference counting for thread_map object, so it could be easily
shared among other objects.

Using thread_map__put instead thread_map__delete and making
thread_map__delete static.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-5-git-send-email-jolsa@kernel.org
[ Adjustments to move it ahead of the "comm" patches ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Add reference counting for cpu_map object
Jiri Olsa [Mon, 22 Jun 2015 22:36:04 +0000 (00:36 +0200)]
perf tools: Add reference counting for cpu_map object

Adding refference counting for cpu_map object, so it could be easily
shared among other objects.

Using cpu_map__put instead cpu_map__delete and making cpu_map__delete
static.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf header: Delete an unnecessary check before the calling free_event_desc()
Markus Elfring [Thu, 25 Jun 2015 15:12:32 +0000 (17:12 +0200)]
perf header: Delete an unnecessary check before the calling free_event_desc()

The free_event_desc() function tests whether its argument is NULL and
then returns immediately. Thus the test around the call is not needed.

This issue was detected by using the Coccinelle software.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Cc: Julia Lawall <julia.lawall@lip6.fr>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: kernel-janitors@vger.kernel.org
Link: http://lkml.kernel.org/r/558C2ABA.3000603@users.sourceforge.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Future-proof thread_map allocation size calculation
Arnaldo Carvalho de Melo [Thu, 25 Jun 2015 17:48:49 +0000 (14:48 -0300)]
perf tools: Future-proof thread_map allocation size calculation

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/20150625174840.GH3253@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf bench numa: Fix to show proper convergence stats
Srikar Dronamraju [Wed, 24 Jun 2015 11:10:04 +0000 (16:40 +0530)]
perf bench numa: Fix to show proper convergence stats

With commit: e1e455f4f4d3 (perf tools: Work around lack of sched_getcpu
in glibc < 2.6), perf_bench numa mem with -c or -m option is not able to
correctly calculate convergence.

With the above commit, sched_getcpu always seems to return -1. The
intention of commit e1e455f was to add a sched_getcpu in glibc < 2.6.
Hence keep the sched_getcpu definition under an ifdef.

This regression happened occurred between v4.0 and v4.1

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Vinson Lee <vlee@twitter.com>
Fixes:  e1e455f4f4d3 ("perf tools: Work around lack of sched_getcpu in glibc < 2.6")
Link: http://lkml.kernel.org/r/20150624111004.GA5220@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoMerge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Thu, 25 Jun 2015 07:28:42 +0000 (09:28 +0200)]
Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

User visible changes:

  - Move toggling event logic from 'perf top' and into hists browser, allowing
    freeze/unfreeze with event lists with more than one entry (Namhyung Kim)

  - Add missing newlines when dumping PERF_RECORD_FINISHED_ROUND and
    showing the Aggregated stats in 'perf report -D' (Adrian Hunter)

Infrastructure changes:

  - Allow auxtrace data alignment (Adrian Hunter)

  - Allow events with dot (Andi Kleen)

  - Fix failure to 'perf probe' events on arm (He Kuang)

  - Add testing for Makefile.perf (Jiri Olsa)

  - Add test for make install with prefix (Jiri Olsa)

  - Fix single target build dependency check (Jiri Olsa)

  - Access thread_map entries via accessors, prep patch to hold more info per
    entry, for ongoing 'perf stat --per-thread' work (Jiri Olsa)

  - Use __weak definition from compiler.h (Sukadev Bhattiprolu)

  - Split perf_pmu__new_alias() (Sukadev Bhattiprolu)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf tools: Allow auxtrace data alignment
Adrian Hunter [Fri, 29 May 2015 13:33:38 +0000 (16:33 +0300)]
perf tools: Allow auxtrace data alignment

Allow auxtrace data to be a multiple of something other than page size.
That is needed for BTS where the buffer contains 24-byte records.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1432906425-9911-11-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf thread_map: Change map entries into a struct
Jiri Olsa [Mon, 22 Jun 2015 22:36:02 +0000 (00:36 +0200)]
perf thread_map: Change map entries into a struct

We need to store command names with the pid. Changing map entries to be
a struct holding pid. Process name is coming in shortly.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-2-git-send-email-jolsa@kernel.org
[ Split providing the set/get accessors from transforming the entries structs ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf thread_map: Don't access the array entries directly
Jiri Olsa [Mon, 22 Jun 2015 22:36:02 +0000 (00:36 +0200)]
perf thread_map: Don't access the array entries directly

Instead provide a method to set the array entries, and another to access
the contents.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.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/1435012588-9007-2-git-send-email-jolsa@kernel.org
[ Split providing the set/get accessors from transforming the entries structs ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf probe: Fix failure to probe events on arm
He Kuang [Thu, 18 Jun 2015 02:49:10 +0000 (02:49 +0000)]
perf probe: Fix failure to probe events on arm

Fix failure to probe events on arm, the problem was introduced by commit
5a51fcd1f30c ("perf probe: Skip kernel symbols which is out of .text").

For some architectures, the '_etext' label is not in the .text section
(in the .notes section for arm/arm64).  Labels out of the .text section
are not loaded as symbols and we get a zero value when looking up its
addresses, which causes all events to be wrongly skipped.

This patch skips checking the text address range when failing to get the
address of '_etext' and thus fixes the problem.

The problem can be reproduced on arm as follows:

  # perf probe --add='generic_perform_write'
  generic_perform_write+0 is out of .text, skip it.
  Probe point 'generic_perform_write' not found.
    Error: Failed to add events.

After this patch:

  # perf probe --add='generic_perform_write'
  Added new event:
    probe:generic_perform_write (on generic_perform_write)

  You can now use it in all perf tools, such as:

    perf record -e probe:generic_perform_write -aR sleep 1

Signed-off-by: He Kuang <hekuang@huawei.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1434595750-129791-1-git-send-email-hekuang@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Print a newline before dumping Aggregated stats
Adrian Hunter [Tue, 23 Jun 2015 07:52:49 +0000 (10:52 +0300)]
perf tools: Print a newline before dumping Aggregated stats

When dumping events with 'perf report -D' the event print always starts
with a newline (see dump_event()).

Do the same with the "Aggregated stats" print so that it is not jammed
up against the last event print.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1435045969-15999-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf session: Print a newline when dumping PERF_RECORD_FINISHED_ROUND
Adrian Hunter [Tue, 23 Jun 2015 07:52:48 +0000 (10:52 +0300)]
perf session: Print a newline when dumping PERF_RECORD_FINISHED_ROUND

With 'perf report -D' the PERF_RECORD_FINISHED_ROUND event was printed
without a newline, resulting in:

0x91a18 [0x8]: PERF_RECORD_FINISHED_ROUNDAggregated stats

Other events print their details, but PERF_RECORD_FINISHED_ROUND doesn't
have any so just add a print for a newline.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1435045969-15999-1-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Allow events with dot
Andi Kleen [Wed, 10 Jun 2015 07:25:11 +0000 (00:25 -0700)]
perf tools: Allow events with dot

The Intel events use a dot to separate event name and unit mask.  Allow
dot in names in the scanner, and remove special handling of dot as EOF.
Also remove the hack in jevents to replace dot with underscore. This way
dotted events can be specified directly by the user.

I'm not fully sure this change to the scanner is correct (what was the
dot special case good for?), but I haven't found anything that breaks
with it so far at least.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/1433921123-25327-8-git-send-email-sukadev@linux.vnet.ibm.com
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf pmu: Split perf_pmu__new_alias()
Sukadev Bhattiprolu [Wed, 10 Jun 2015 07:25:08 +0000 (00:25 -0700)]
perf pmu: Split perf_pmu__new_alias()

Separate the event parsing code in perf_pmu__new_alias() out into a
separate function __perf_pmu__new_alias() so that code can be called
indepdently.

This is based on an earlier patch from Andi Kleen.

Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/1433921123-25327-5-git-send-email-sukadev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf pmu: Use __weak definition from <linux/compiler.h>
Sukadev Bhattiprolu [Wed, 10 Jun 2015 07:25:07 +0000 (00:25 -0700)]
perf pmu: Use __weak definition from <linux/compiler.h>

Jiri Olsa pointed out, that the <linux/compiler.h> defines the attribute
'__weak'. We might as well use that.

Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/1433921123-25327-4-git-send-email-sukadev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoMerge branch 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 23 Jun 2015 02:42:56 +0000 (19:42 -0700)]
Merge branch 'irq-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq updates from Thomas Gleixner:
 "The irq departement delivers:

   - plug a potential race related to chained interrupt handlers

   - core updates which address the needs of the x86 irqdomain conversion

   - new irqchip callback to support affinity settings for VCPUs

   - the usual pile of updates to interrupt chip drivers

   - a few helper functions to allow further cleanups and
     simplifications

  I have a largish pile of coccinelle scripted/verified cleanups and
  simplifications pending on top of that, but I prefer to send that
  towards the end of the merge window when the arch/driver changes have
  hit your tree to avoid API change wreckage as far as possible"

* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (34 commits)
  genirq: Remove bogus restriction in irq_move_mask_irq()
  irqchip: atmel-aic5: Add sama5d2 support
  irq: spear-shirq: Fix race in installing chained IRQ handler
  irq: irq-keystone: Fix race in installing chained IRQ handler
  gpio: gpio-tegra: Fix race in installing chained IRQ handler
  gpio: gpio-mxs: Fix race in installing chained IRQ handler
  gpio: gpio-mxc: Fix race in installing chained IRQ handler
  ARM: gemini: Fix race in installing GPIO chained IRQ handler
  GPU: ipu: Fix race in installing IPU chained IRQ handler
  ARM: sa1100: convert SA11x0 related code to use new chained handler helper
  irq: Add irq_set_chained_handler_and_data()
  irqchip: exynos-combiner: Save IRQ enable set on suspend
  genirq: Introduce helper function irq_data_get_affinity_mask()
  genirq: Introduce helper function irq_data_get_node()
  genirq: Introduce struct irq_common_data to host shared irq data
  genirq: Prevent crash in irq_move_irq()
  genirq: Enhance irq_data_to_desc() to support hierarchy irqdomain
  irqchip: gic: Simplify gic_configure_irq by using IRQCHIP_SET_TYPE_MASKED
  irqchip: renesas: intc-irqpin: Improve binding documentation
  genirq: Set IRQCHIP_SKIP_SET_WAKE for no_irq_chip
  ...

9 years agoMerge branch 'timers-nohz-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 23 Jun 2015 02:20:04 +0000 (19:20 -0700)]
Merge branch 'timers-nohz-for-linus' of git://git./linux/kernel/git/tip/tip

Pull NOHZ updates from Thomas Gleixner:
 "A few updates to the nohz infrastructure:

   - recursion protection for context tracking

   - make the TIF_NOHZ inheritance smarter

   - isolate cpus which belong to the NOHZ full set"

* 'timers-nohz-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  nohz: Set isolcpus when nohz_full is set
  nohz: Add tick_nohz_full_add_cpus_to() API
  context_tracking: Inherit TIF_NOHZ through forks instead of context switches
  context_tracking: Protect against recursion

9 years agoMerge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 23 Jun 2015 01:57:44 +0000 (18:57 -0700)]
Merge branch 'timers-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer updates from Thomas Gleixner:
 "A rather largish update for everything time and timer related:

   - Cache footprint optimizations for both hrtimers and timer wheel

   - Lower the NOHZ impact on systems which have NOHZ or timer migration
     disabled at runtime.

   - Optimize run time overhead of hrtimer interrupt by making the clock
     offset updates smarter

   - hrtimer cleanups and removal of restrictions to tackle some
     problems in sched/perf

   - Some more leap second tweaks

   - Another round of changes addressing the 2038 problem

   - First step to change the internals of clock event devices by
     introducing the necessary infrastructure

   - Allow constant folding for usecs/msecs_to_jiffies()

   - The usual pile of clockevent/clocksource driver updates

  The hrtimer changes contain updates to sched, perf and x86 as they
  depend on them plus changes all over the tree to cleanup API changes
  and redundant code, which got copied all over the place.  The y2038
  changes touch s390 to remove the last non 2038 safe code related to
  boot/persistant clock"

* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (114 commits)
  clocksource: Increase dependencies of timer-stm32 to limit build wreckage
  timer: Minimize nohz off overhead
  timer: Reduce timer migration overhead if disabled
  timer: Stats: Simplify the flags handling
  timer: Replace timer base by a cpu index
  timer: Use hlist for the timer wheel hash buckets
  timer: Remove FIFO "guarantee"
  timers: Sanitize catchup_timer_jiffies() usage
  hrtimer: Allow hrtimer::function() to free the timer
  seqcount: Introduce raw_write_seqcount_barrier()
  seqcount: Rename write_seqcount_barrier()
  hrtimer: Fix hrtimer_is_queued() hole
  hrtimer: Remove HRTIMER_STATE_MIGRATE
  selftest: Timers: Avoid signal deadlock in leap-a-day
  timekeeping: Copy the shadow-timekeeper over the real timekeeper last
  clockevents: Check state instead of mode in suspend/resume path
  selftests: timers: Add leap-second timer edge testing to leap-a-day.c
  ntp: Do leapsecond adjustment in adjtimex read path
  time: Prevent early expiry of hrtimers[CLOCK_REALTIME] at the leap second edge
  ntp: Introduce and use SECS_PER_DAY macro instead of 86400
  ...

9 years agoMerge branch 'x86-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 23 Jun 2015 00:59:09 +0000 (17:59 -0700)]
Merge branch 'x86-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 core updates from Ingo Molnar:
 "There were so many changes in the x86/asm, x86/apic and x86/mm topics
  in this cycle that the topical separation of -tip broke down somewhat -
  so the result is a more traditional architecture pull request,
  collected into the 'x86/core' topic.

  The topics were still maintained separately as far as possible, so
  bisectability and conceptual separation should still be pretty good -
  but there were a handful of merge points to avoid excessive
  dependencies (and conflicts) that would have been poorly tested in the
  end.

  The next cycle will hopefully be much more quiet (or at least will
  have fewer dependencies).

  The main changes in this cycle were:

   * x86/apic changes, with related IRQ core changes: (Jiang Liu, Thomas
     Gleixner)

     - This is the second and most intrusive part of changes to the x86
       interrupt handling - full conversion to hierarchical interrupt
       domains:

          [IOAPIC domain]   -----
                                 |
          [MSI domain]      --------[Remapping domain] ----- [ Vector domain ]
                                 |   (optional)          |
          [HPET MSI domain] -----                        |
                                                         |
          [DMAR domain]     -----------------------------
                                                         |
          [Legacy domain]   -----------------------------

       This now reflects the actual hardware and allowed us to distangle
       the domain specific code from the underlying parent domain, which
       can be optional in the case of interrupt remapping.  It's a clear
       separation of functionality and removes quite some duct tape
       constructs which plugged the remap code between ioapic/msi/hpet
       and the vector management.

     - Intel IOMMU IRQ remapping enhancements, to allow direct interrupt
       injection into guests (Feng Wu)

   * x86/asm changes:

     - Tons of cleanups and small speedups, micro-optimizations.  This
       is in preparation to move a good chunk of the low level entry
       code from assembly to C code (Denys Vlasenko, Andy Lutomirski,
       Brian Gerst)

     - Moved all system entry related code to a new home under
       arch/x86/entry/ (Ingo Molnar)

     - Removal of the fragile and ugly CFI dwarf debuginfo annotations.
       Conversion to C will reintroduce many of them - but meanwhile
       they are only getting in the way, and the upstream kernel does
       not rely on them (Ingo Molnar)

     - NOP handling refinements. (Borislav Petkov)

   * x86/mm changes:

     - Big PAT and MTRR rework: making the code more robust and
       preparing to phase out exposing direct MTRR interfaces to drivers -
       in favor of using PAT driven interfaces (Toshi Kani, Luis R
       Rodriguez, Borislav Petkov)

     - New ioremap_wt()/set_memory_wt() interfaces to support
       Write-Through cached memory mappings.  This is especially
       important for good performance on NVDIMM hardware (Toshi Kani)

   * x86/ras changes:

     - Add support for deferred errors on AMD (Aravind Gopalakrishnan)

       This is an important RAS feature which adds hardware support for
       poisoned data.  That means roughly that the hardware marks data
       which it has detected as corrupted but wasn't able to correct, as
       poisoned data and raises an APIC interrupt to signal that in the
       form of a deferred error.  It is the OS's responsibility then to
       take proper recovery action and thus prolonge system lifetime as
       far as possible.

     - Add support for Intel "Local MCE"s: upcoming CPUs will support
       CPU-local MCE interrupts, as opposed to the traditional system-
       wide broadcasted MCE interrupts (Ashok Raj)

     - Misc cleanups (Borislav Petkov)

   * x86/platform changes:

     - Intel Atom SoC updates

  ... and lots of other cleanups, fixlets and other changes - see the
  shortlog and the Git log for details"

* 'x86-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (222 commits)
  x86/hpet: Use proper hpet device number for MSI allocation
  x86/hpet: Check for irq==0 when allocating hpet MSI interrupts
  x86/mm/pat, drivers/infiniband/ipath: Use arch_phys_wc_add() and require PAT disabled
  x86/mm/pat, drivers/media/ivtv: Use arch_phys_wc_add() and require PAT disabled
  x86/platform/intel/baytrail: Add comments about why we disabled HPET on Baytrail
  genirq: Prevent crash in irq_move_irq()
  genirq: Enhance irq_data_to_desc() to support hierarchy irqdomain
  iommu, x86: Properly handle posted interrupts for IOMMU hotplug
  iommu, x86: Provide irq_remapping_cap() interface
  iommu, x86: Setup Posted-Interrupts capability for Intel iommu
  iommu, x86: Add cap_pi_support() to detect VT-d PI capability
  iommu, x86: Avoid migrating VT-d posted interrupts
  iommu, x86: Save the mode (posted or remapped) of an IRTE
  iommu, x86: Implement irq_set_vcpu_affinity for intel_ir_chip
  iommu: dmar: Provide helper to copy shared irte fields
  iommu: dmar: Extend struct irte for VT-d Posted-Interrupts
  iommu: Add new member capability to struct irq_remap_ops
  x86/asm/entry/64: Disentangle error_entry/exit gsbase/ebx/usermode code
  x86/asm/entry/32: Shorten __audit_syscall_entry() args preparation
  x86/asm/entry/32: Explain reloading of registers after __audit_syscall_entry()
  ...

9 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 23 Jun 2015 00:51:59 +0000 (17:51 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 warning fixlet from Ingo Molnar:
 "A build fix for certain (rare) variants of binutils that did not make
  it into v4.1"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/boot: Fix overflow warning with 32-bit binutils

9 years agoMerge branch 'x86-microcode-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 23 Jun 2015 00:46:14 +0000 (17:46 -0700)]
Merge branch 'x86-microcode-for-linus' of git://git./linux/kernel/git/tip/tip

Pul x86 microcode updates from Ingo Molnar:
 "x86 microcode loader updates from Borislav Petkov:

   - early parsing of the built-in microcode

   - cleanups

   - misc smaller fixes"

* 'x86-microcode-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/microcode: Correct CPU family related variable types
  x86/microcode: Disable builtin microcode loading on 32-bit for now
  x86/microcode/intel: Rename get_matching_sig()
  x86/microcode/intel: Simplify get_matching_sig()
  x86/microcode/intel: Simplify update_match_cpu()
  x86/microcode/intel: Rename get_matching_microcode
  x86/cpu/microcode: Zap changelog
  x86/microcode: Parse built-in microcode early
  x86/microcode/intel: Remove unused @rev arg of get_matching_sig()
  x86/microcode/intel: Get rid of revision_is_newer()

9 years agoMerge branch 'x86-kdump-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 23 Jun 2015 00:40:55 +0000 (17:40 -0700)]
Merge branch 'x86-kdump-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 kdump updates from Ingo Molnar:
 "Three kdump robustness related improvements (Joerg Roedel)"

* 'x86-kdump-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/crash: Allocate enough low memory when crashkernel=high
  x86/swiotlb: Try coherent allocations with __GFP_NOWARN
  swiotlb: Warn on allocation failure in swiotlb_alloc_coherent()

9 years agoMerge branch 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 23 Jun 2015 00:16:11 +0000 (17:16 -0700)]
Merge branch 'x86-fpu-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 FPU updates from Ingo Molnar:
 "This tree contains two main changes:

   - The big FPU code rewrite: wide reaching cleanups and reorganization
     that pulls all the FPU code together into a clean base in
     arch/x86/fpu/.

     The resulting code is leaner and faster, and much easier to
     understand.  This enables future work to further simplify the FPU
     code (such as removing lazy FPU restores).

     By its nature these changes have a substantial regression risk: FPU
     code related bugs are long lived, because races are often subtle
     and bugs mask as user-space failures that are difficult to track
     back to kernel side backs.  I'm aware of no unfixed (or even
     suspected) FPU related regression so far.

   - MPX support rework/fixes.  As this is still not a released CPU
     feature, there were some buglets in the code - should be much more
     robust now (Dave Hansen)"

* 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (250 commits)
  x86/fpu: Fix double-increment in setup_xstate_features()
  x86/mpx: Allow 32-bit binaries on 64-bit kernels again
  x86/mpx: Do not count MPX VMAs as neighbors when unmapping
  x86/mpx: Rewrite the unmap code
  x86/mpx: Support 32-bit binaries on 64-bit kernels
  x86/mpx: Use 32-bit-only cmpxchg() for 32-bit apps
  x86/mpx: Introduce new 'directory entry' to 'addr' helper function
  x86/mpx: Add temporary variable to reduce masking
  x86: Make is_64bit_mm() widely available
  x86/mpx: Trace allocation of new bounds tables
  x86/mpx: Trace the attempts to find bounds tables
  x86/mpx: Trace entry to bounds exception paths
  x86/mpx: Trace #BR exceptions
  x86/mpx: Introduce a boot-time disable flag
  x86/mpx: Restrict the mmap() size check to bounds tables
  x86/mpx: Remove redundant MPX_BNDCFG_ADDR_MASK
  x86/mpx: Clean up the code by not passing a task pointer around when unnecessary
  x86/mpx: Use the new get_xsave_field_ptr()API
  x86/fpu/xstate: Wrap get_xsave_addr() to make it safer
  x86/fpu/xstate: Fix up bad get_xsave_addr() assumptions
  ...

9 years agoMerge branch 'x86-efi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 23 Jun 2015 00:10:44 +0000 (17:10 -0700)]
Merge branch 'x86-efi-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 EFI updates from Ingo Molnar:
 "EFI changes:

   - Use idiomatic negative error values in efivar_create_sysfs_entry()
     instead of returning '1' to indicate error (Dan Carpenter)

   - Implement new support to expose the EFI System Resource Tables in
     sysfs, which provides information for performing firmware updates
     (Peter Jones)

   - Documentation cleanup in the EFI handover protocol section which
     falsely claimed that 'cmdline_size' needed to be filled out by the
     boot loader (Alex Smith)

   - Align the order of SMBIOS tables in /sys/firmware/efi/systab to
     match the way that we do things for ACPI and add documentation to
     Documentation/ABI (Jean Delvare)"

* 'x86-efi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  efi: Work around ia64 build problem with ESRT driver
  efi: Add 'systab' information to Documentation/ABI
  efi: dmi: List SMBIOS3 table before SMBIOS table
  efi/esrt: Fix some compiler warnings
  x86, doc: Remove cmdline_size from list of fields to be filled in for EFI handover
  efi: Add esrt support
  efi: efivar_create_sysfs_entry() should return negative error codes

9 years agoMerge branch 'x86-debug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 23 Jun 2015 00:09:32 +0000 (17:09 -0700)]
Merge branch 'x86-debug-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 debugging documentation updates from Ingo Molnar:
 "Documentation updates about x86 kernel stacks"

* 'x86-debug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/Documentation: Adapt Ingo's explanation on printing backtraces
  x86/Documentation: Remove STACKFAULT_STACK bulletpoint
  x86/Documentation: Move kernel-stacks doc one level up

9 years agoMerge branch 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 22 Jun 2015 23:43:01 +0000 (16:43 -0700)]
Merge branch 'x86-cpu-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 CPU features from Ingo Molnar:
 "Various CPU feature support related changes: in particular the
  /proc/cpuinfo model name sanitization change should be monitored, it
  has a chance to break stuff.  (but really shouldn't and there are no
  regression reports)"

* 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/cpu/amd: Give access to the number of nodes in a physical package
  x86/cpu: Trim model ID whitespace
  x86/cpu: Strip any /proc/cpuinfo model name field whitespace
  x86/cpu/amd: Set X86_FEATURE_EXTD_APICID for future processors
  x86/gart: Check for GART support before accessing GART registers

9 years agoMerge branch 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 22 Jun 2015 23:23:00 +0000 (16:23 -0700)]
Merge branch 'x86-cleanups-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 cleanups from Ingo Molnar:
 "Misc cleanups"

* 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm: Clean up types in xlate_dev_mem_ptr() some more
  x86: Deinline dma_free_attrs()
  x86: Deinline dma_alloc_attrs()
  x86: Remove unused TI_cpu
  x86: Merge common 32-bit values in asm-offsets.c

9 years agoMerge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 22 Jun 2015 22:52:04 +0000 (15:52 -0700)]
Merge branch 'sched-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler updates from Ingo Molnar:
 "The main changes are:

   - lockless wakeup support for futexes and IPC message queues
     (Davidlohr Bueso, Peter Zijlstra)

   - Replace spinlocks with atomics in thread_group_cputimer(), to
     improve scalability (Jason Low)

   - NUMA balancing improvements (Rik van Riel)

   - SCHED_DEADLINE improvements (Wanpeng Li)

   - clean up and reorganize preemption helpers (Frederic Weisbecker)

   - decouple page fault disabling machinery from the preemption
     counter, to improve debuggability and robustness (David
     Hildenbrand)

   - SCHED_DEADLINE documentation updates (Luca Abeni)

   - topology CPU masks cleanups (Bartosz Golaszewski)

   - /proc/sched_debug improvements (Srikar Dronamraju)"

* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (79 commits)
  sched/deadline: Remove needless parameter in dl_runtime_exceeded()
  sched: Remove superfluous resetting of the p->dl_throttled flag
  sched/deadline: Drop duplicate init_sched_dl_class() declaration
  sched/deadline: Reduce rq lock contention by eliminating locking of non-feasible target
  sched/deadline: Make init_sched_dl_class() __init
  sched/deadline: Optimize pull_dl_task()
  sched/preempt: Add static_key() to preempt_notifiers
  sched/preempt: Fix preempt notifiers documentation about hlist_del() within unsafe iteration
  sched/stop_machine: Fix deadlock between multiple stop_two_cpus()
  sched/debug: Add sum_sleep_runtime to /proc/<pid>/sched
  sched/debug: Replace vruntime with wait_sum in /proc/sched_debug
  sched/debug: Properly format runnable tasks in /proc/sched_debug
  sched/numa: Only consider less busy nodes as numa balancing destinations
  Revert 095bebf61a46 ("sched/numa: Do not move past the balance point if unbalanced")
  sched/fair: Prevent throttling in early pick_next_task_fair()
  preempt: Reorganize the notrace definitions a bit
  preempt: Use preempt_schedule_context() as the official tracing preemption point
  sched: Make preempt_schedule_context() function-tracing safe
  x86: Remove cpu_sibling_mask() and cpu_core_mask()
  x86: Replace cpu_**_mask() with topology_**_cpumask()
  ...

9 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 22 Jun 2015 22:45:41 +0000 (15:45 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "These are the left over fixes from the v4.1 cycle"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf tools: Fix build breakage if prefix= is specified
  perf/x86: Honor the architectural performance monitoring version
  perf/x86/intel: Fix PMI handling for Intel PT
  perf/x86/intel/bts: Fix DS area sharing with x86_pmu events
  perf/x86: Add more Broadwell model numbers
  perf: Fix ring_buffer_attach() RCU sync, again

9 years agoMerge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 22 Jun 2015 22:19:21 +0000 (15:19 -0700)]
Merge branch 'perf-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf updates from Ingo Molnar:
 "Kernel side changes mostly consist of work on x86 PMU drivers:

   - x86 Intel PT (hardware CPU tracer) improvements (Alexander
     Shishkin)

   - x86 Intel CQM (cache quality monitoring) improvements (Thomas
     Gleixner)

   - x86 Intel PEBSv3 support (Peter Zijlstra)

   - x86 Intel PEBS interrupt batching support for lower overhead
     sampling (Zheng Yan, Kan Liang)

   - x86 PMU scheduler fixes and improvements (Peter Zijlstra)

  There's too many tooling improvements to list them all - here are a
  few select highlights:

  'perf bench':

      - Introduce new 'perf bench futex' benchmark: 'wake-parallel', to
        measure parallel waker threads generating contention for kernel
        locks (hb->lock). (Davidlohr Bueso)

  'perf top', 'perf report':

      - Allow disabling/enabling events dynamicaly in 'perf top':
        a 'perf top' session can instantly become a 'perf report'
        one, i.e. going from dynamic analysis to a static one,
        returning to a dynamic one is possible, to toogle the
        modes, just press 'f' to 'freeze/unfreeze' the sampling. (Arnaldo Carvalho de Melo)

      - Make Ctrl-C stop processing on TUI, allowing interrupting the load of big
        perf.data files (Namhyung Kim)

  'perf probe': (Masami Hiramatsu)

      - Support glob wildcards for function name
      - Support $params special probe argument: Collect all function arguments
      - Make --line checks validate C-style function name.
      - Add --no-inlines option to avoid searching inline functions
      - Greatly speed up 'perf probe --list' by caching debuginfo.
      - Improve --filter support for 'perf probe', allowing using its arguments
        on other commands, as --add, --del, etc.

  'perf sched':

      - Add option in 'perf sched' to merge like comms to lat output (Josef Bacik)

  Plus tons of infrastructure work - in particular preparation for
  upcoming threaded perf report support, but also lots of other work -
  and fixes and other improvements.  See (much) more details in the
  shortlog and in the git log"

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (305 commits)
  perf tools: Configurable per thread proc map processing time out
  perf tools: Add time out to force stop proc map processing
  perf report: Fix sort__sym_cmp to also compare end of symbol
  perf hists browser: React to unassigned hotkey pressing
  perf top: Tell the user how to unfreeze events after pressing 'f'
  perf hists browser: Honour the help line provided by builtin-{top,report}.c
  perf hists browser: Do not exit when 'f' is pressed in 'report' mode
  perf top: Replace CTRL+z with 'f' as hotkey for enable/disable events
  perf annotate: Rename source_line_percent to source_line_samples
  perf annotate: Display total number of samples with --show-total-period
  perf tools: Ensure thread-stack is flushed
  perf top: Allow disabling/enabling events dynamicly
  perf evlist: Add toggle_enable() method
  perf trace: Fix race condition at the end of started workloads
  perf probe: Speed up perf probe --list by caching debuginfo
  perf probe: Show usage even if the last event is skipped
  perf tools: Move libtraceevent dynamic list to separated LDFLAGS variable
  perf tools: Fix a problem when opening old perf.data with different byte order
  perf tools: Ignore .config-detected in .gitignore
  perf probe: Fix to return error if no probe is added
  ...

9 years agoMerge branch 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 22 Jun 2015 21:54:22 +0000 (14:54 -0700)]
Merge branch 'locking-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull locking updates from Ingo Molnar:
 "The main changes are:

   - 'qspinlock' support, enabled on x86: queued spinlocks - these are
     now the spinlock variant used by x86 as they outperform ticket
     spinlocks in every category.  (Waiman Long)

   - 'pvqspinlock' support on x86: paravirtualized variant of queued
     spinlocks.  (Waiman Long, Peter Zijlstra)

   - 'qrwlock' support, enabled on x86: queued rwlocks.  Similar to
     queued spinlocks, they are now the variant used by x86:

       CONFIG_ARCH_USE_QUEUED_SPINLOCKS=y
       CONFIG_QUEUED_SPINLOCKS=y
       CONFIG_ARCH_USE_QUEUED_RWLOCKS=y
       CONFIG_QUEUED_RWLOCKS=y

   - various lockdep fixlets

   - various locking primitives cleanups, further WRITE_ONCE()
     propagation"

* 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (24 commits)
  locking/lockdep: Remove hard coded array size dependency
  locking/qrwlock: Don't contend with readers when setting _QW_WAITING
  lockdep: Do not break user-visible string
  locking/arch: Rename set_mb() to smp_store_mb()
  locking/arch: Add WRITE_ONCE() to set_mb()
  rtmutex: Warn if trylock is called from hard/softirq context
  arch: Remove __ARCH_HAVE_CMPXCHG
  locking/rtmutex: Drop usage of __HAVE_ARCH_CMPXCHG
  locking/qrwlock: Rename QUEUE_RWLOCK to QUEUED_RWLOCKS
  locking/pvqspinlock: Rename QUEUED_SPINLOCK to QUEUED_SPINLOCKS
  locking/pvqspinlock: Replace xchg() by the more descriptive set_mb()
  locking/pvqspinlock, x86: Enable PV qspinlock for Xen
  locking/pvqspinlock, x86: Enable PV qspinlock for KVM
  locking/pvqspinlock, x86: Implement the paravirt qspinlock call patching
  locking/pvqspinlock: Implement simple paravirt support for the qspinlock
  locking/qspinlock: Revert to test-and-set on hypervisors
  locking/qspinlock: Use a simple write to grab the lock
  locking/qspinlock: Optimize for smaller NR_CPUS
  locking/qspinlock: Extract out code snippets for the next patch
  locking/qspinlock: Add pending bit
  ...

9 years agoMerge branch 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 22 Jun 2015 21:01:01 +0000 (14:01 -0700)]
Merge branch 'core-rcu-for-linus' of git://git./linux/kernel/git/tip/tip

Pull RCU updates from Ingo Molnar:

 - Continued initialization/Kconfig updates: hide most Kconfig options
   from unsuspecting users.

   There's now a single high level configuration option:

        *
        * RCU Subsystem
        *
        Make expert-level adjustments to RCU configuration (RCU_EXPERT) [N/y/?] (NEW)

   Which if answered in the negative, leaves us with a single
   interactive configuration option:

        Offload RCU callback processing from boot-selected CPUs (RCU_NOCB_CPU) [N/y/?] (NEW)

   All the rest of the RCU options are configured automatically.  Later
   on we'll remove this single leftover configuration option as well.

 - Remove all uses of RCU-protected array indexes: replace the
   rcu_[access|dereference]_index_check() APIs with READ_ONCE() and
   rcu_lockdep_assert()

 - RCU CPU-hotplug cleanups

 - Updates to Tiny RCU: a race fix and further code shrinkage.

 - RCU torture-testing updates: fixes, speedups, cleanups and
   documentation updates.

 - Miscellaneous fixes

 - Documentation updates

* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (60 commits)
  rcutorture: Allow repetition factors in Kconfig-fragment lists
  rcutorture: Display "make oldconfig" errors
  rcutorture: Update TREE_RCU-kconfig.txt
  rcutorture: Make rcutorture scripts force RCU_EXPERT
  rcutorture: Update configuration fragments for rcutree.rcu_fanout_exact
  rcutorture: TASKS_RCU set directly, so don't explicitly set it
  rcutorture: Test SRCU cleanup code path
  rcutorture: Replace barriers with smp_store_release() and smp_load_acquire()
  locktorture: Change longdelay_us to longdelay_ms
  rcutorture: Allow negative values of nreaders to oversubscribe
  rcutorture: Exchange TREE03 and TREE08 NR_CPUS, speed up CPU hotplug
  rcutorture: Exchange TREE03 and TREE04 geometries
  locktorture: fix deadlock in 'rw_lock_irq' type
  rcu: Correctly handle non-empty Tiny RCU callback list with none ready
  rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
  rcu: Further shrink Tiny RCU by making empty functions static inlines
  rcu: Conditionally compile RCU's eqs warnings
  rcu: Remove prompt for RCU implementation
  rcu: Make RCU able to tolerate undefined CONFIG_RCU_KTHREAD_PRIO
  rcu: Make RCU able to tolerate undefined CONFIG_RCU_FANOUT_LEAF
  ...

9 years agoMerge branch 'for-linus-1' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Mon, 22 Jun 2015 19:51:21 +0000 (12:51 -0700)]
Merge branch 'for-linus-1' of git://git./linux/kernel/git/viro/vfs

Pull vfs updates from Al Viro:
 "In this pile: pathname resolution rewrite.

   - recursion in link_path_walk() is gone.

   - nesting limits on symlinks are gone (the only limit remaining is
     that the total amount of symlinks is no more than 40, no matter how
     nested).

   - "fast" (inline) symlinks are handled without leaving rcuwalk mode.

   - stack footprint (independent of the nesting) is below kilobyte now,
     about on par with what it used to be with one level of nested
     symlinks and ~2.8 times lower than it used to be in the worst case.

   - struct nameidata is entirely private to fs/namei.c now (not even
     opaque pointers are being passed around).

   - ->follow_link() and ->put_link() calling conventions had been
     changed; all in-tree filesystems converted, out-of-tree should be
     able to follow reasonably easily.

     For out-of-tree conversions, see Documentation/filesystems/porting
     for details (and in-tree filesystems for examples of conversion).

  That has sat in -next since mid-May, seems to survive all testing
  without regressions and merges clean with v4.1"

* 'for-linus-1' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (131 commits)
  turn user_{path_at,path,lpath,path_dir}() into static inlines
  namei: move saved_nd pointer into struct nameidata
  inline user_path_create()
  inline user_path_parent()
  namei: trim do_last() arguments
  namei: stash dfd and name into nameidata
  namei: fold path_cleanup() into terminate_walk()
  namei: saner calling conventions for filename_parentat()
  namei: saner calling conventions for filename_create()
  namei: shift nameidata down into filename_parentat()
  namei: make filename_lookup() reject ERR_PTR() passed as name
  namei: shift nameidata inside filename_lookup()
  namei: move putname() call into filename_lookup()
  namei: pass the struct path to store the result down into path_lookupat()
  namei: uninline set_root{,_rcu}()
  namei: be careful with mountpoint crossings in follow_dotdot_rcu()
  Documentation: remove outdated information from automount-support.txt
  get rid of assorted nameidata-related debris
  lustre: kill unused helper
  lustre: kill unused macro (LOOKUP_CONTINUE)
  ...

9 years agoperf build: Fix single target build dependency check
Jiri Olsa [Mon, 22 Jun 2015 12:50:52 +0000 (14:50 +0200)]
perf build: Fix single target build dependency check

Currently if we build a single target like:

  $ touch util/map.c && make util/map.o

It will not rebuild util/map.o if it already exists and util/map.c is
modified.

The reason is that the top-level 'Makefile' processes util/map.o as an
implicit rule and if util/map.o exists make considers the 'util/map.o'
target as done and will not nest into Makefile.perf.

Adding FORCE for '%', because that's what we want to nest into
Makefile.perf for any target.

Adding Makefile into phony targets, because make tries to rebuild it and
it's also resolved as '%' target.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Lukas Wunner <lukas@wunner.de>
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/1434977452-32520-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tests: Add test for make install with prefix
Jiri Olsa [Mon, 22 Jun 2015 12:50:51 +0000 (14:50 +0200)]
perf tests: Add test for make install with prefix

Lukas Wunner reported issue (and fix[1]) with 'make install prefix=...'.
Adding automated test for this, so it wouldn't happen again.

[1]: 75e84ab906ef ("perf tools: Fix build breakage if prefix= is specified")

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Lukas Wunner <lukas@wunner.de>
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/1434977452-32520-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tests: Add testing for Makefile.perf
Jiri Olsa [Mon, 22 Jun 2015 12:50:50 +0000 (14:50 +0200)]
perf tests: Add testing for Makefile.perf

Currently we test only builds through top level Makefile, but seems like
there's a bunch of users using Makefile.perf directly.

Changing the make suite to be run for Makefile.perf as well.  It takes
now considerable amount of time, but hopefully we catch more issues.

Also fixing the output indentation for make_kernelsrc and
make_kernelsrc_tools tests.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Lukas Wunner <lukas@wunner.de>
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/1434977452-32520-2-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf top: Move toggling event logic into hists browser
Namhyung Kim [Sun, 21 Jun 2015 03:41:16 +0000 (12:41 +0900)]
perf top: Move toggling event logic into hists browser

Current 'f' key action to enable/disable events won't work if there're
more than one event since perf_evsel_menu__run() doesn't return the key.
So move it to the hists browser loop so that it can be processed as like
other key action, and it's more natural to handle it there IMHO.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1434858076-6533-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoMerge branches 'x86/apic', 'x86/asm', 'x86/mm' and 'x86/platform' into x86/core,...
Ingo Molnar [Mon, 22 Jun 2015 07:15:03 +0000 (09:15 +0200)]
Merge branches 'x86/apic', 'x86/asm', 'x86/mm' and 'x86/platform' into x86/core, to merge last updates

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoLinux 4.1
Linus Torvalds [Mon, 22 Jun 2015 05:05:43 +0000 (22:05 -0700)]
Linux 4.1

9 years agoclocksource: Increase dependencies of timer-stm32 to limit build wreckage
Paul Gortmaker [Sat, 20 Jun 2015 23:02:32 +0000 (19:02 -0400)]
clocksource: Increase dependencies of timer-stm32 to limit build wreckage

This driver leaks out into arch/parisc builds that don't have
CONFIG_GENERIC_CLOCKEVENTS, leading to the following (truncated)
wreckage:

  CC      drivers/clocksource/timer-stm32.o
drivers/clocksource/timer-stm32.c:38:28: error: field 'evtdev' has incomplete type
drivers/clocksource/timer-stm32.c:44:19: warning: 'enum clock_event_mode' declared inside parameter list
drivers/clocksource/timer-stm32.c:44:19: warning: its scope is only this definition or declaration, which is probably not what you want
drivers/clocksource/timer-stm32.c:43:62: error: parameter 1 ('mode') has incomplete type
drivers/clocksource/timer-stm32.c:43:13: error: function declaration isn't a prototype
drivers/clocksource/timer-stm32.c: In function 'stm32_clock_event_set_mode':
drivers/clocksource/timer-stm32.c:47:3: error: type defaults to 'int' in declaration of '__mptr'
drivers/clocksource/timer-stm32.c:47:3: warning: initialization from incompatible pointer type
drivers/clocksource/timer-stm32.c:51:7: error: 'CLOCK_EVT_MODE_PERIODIC' undeclared (first use in this function)
drivers/clocksource/timer-stm32.c:51:7: note: each undeclared identifier is reported only once for each function it appears in
drivers/clocksource/timer-stm32.c:56:7: error: 'CLOCK_EVT_MODE_ONESHOT' undeclared (first use in this function)

Tighten up the dependencies to limit where it gets built by copying
the style of the Kconfig line for CLKSRC_EFM32 a few lines above.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Chanwoo Choi <cw00.choi@samsung.com>
Cc: Maxime Coquelin <mcoquelin.stm32@gmail.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: http://lkml.kernel.org/r/1434841352-24300-1-git-send-email-paul.gortmaker@windriver.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agox86/hpet: Use proper hpet device number for MSI allocation
Thomas Gleixner [Sun, 21 Jun 2015 14:21:50 +0000 (16:21 +0200)]
x86/hpet: Use proper hpet device number for MSI allocation

hpet_assign_irq() is called with hpet_device->num as "hardware
interrupt number", but hpet_device->num is initialized after the
interrupt has been assigned, so it's always 0. As a consequence only
the first MSI allocation succeeds, the following ones fail because the
"hardware interrupt number" already exists.

Move the initialization of dev->num and other fields before the call
to hpet_assign_irq(), which is the ordering before the offending
commit which introduced that regression.

Fixes: "3cb96f0c9733 x86/hpet: Enhance HPET IRQ to support hierarchical irqdomains"
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1506211635010.4107@nanos
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
Linus Torvalds [Sun, 21 Jun 2015 00:26:01 +0000 (17:26 -0700)]
Merge git://git./linux/kernel/git/nab/target-pending

Pull scsi target fixes from Nicholas Bellinger:
 "Apologies for the late pull request.

  Here are the outstanding target-pending fixes for v4.1 code.

  The series contains three patches from Sagi + Co that address a few
  iser-target issues that have been uncovered during recent testing at
  Mellanox.

  Patch #1 has a v3.16+ stable tag, and #2-3 have v3.10+ stable tags"

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  iser-target: Fix possible use-after-free
  iser-target: release stale iser connections
  iser-target: Fix variable-length response error completion

9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 20 Jun 2015 20:54:22 +0000 (13:54 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "A smattering of fixes,

  mgag200:
      don't accept modes that aren't aligned properly as hw can't do it

  i915:
      two regression fixes

  radeon:
      one query to allow userspace fixes
      one oops fixer for older hw with new options enabled"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/radeon: don't probe MST on hw we don't support it on
  drm/radeon: Add RADEON_INFO_VA_UNMAP_WORKING query
  drm/mgag200: Reject non-character-cell-aligned mode widths
  Revert "drm/i915: Don't skip request retirement if the active list is empty"
  drm/i915: Always reset vma->ggtt_view.pages cache on unbinding

9 years agoMerge branch 'irq/for-x86' into irq/core
Thomas Gleixner [Sat, 20 Jun 2015 17:14:31 +0000 (19:14 +0200)]
Merge branch 'irq/for-x86' into irq/core

Get the infrastructure patches which are required for x86/apic into core

9 years agogenirq: Remove bogus restriction in irq_move_mask_irq()
Thomas Gleixner [Sat, 20 Jun 2015 10:05:40 +0000 (12:05 +0200)]
genirq: Remove bogus restriction in irq_move_mask_irq()

If an interrupt is marked with the no balancing flag, we still allow
setting the affinity for such an interrupt from the kernel itself, but
for interrupts which move the affinity from interrupt context via
irq_move_mask_irq() this runs into a check for the no balancing flag,
which in turn ends up with an endless storm of stack dumps because the
move pending flag is not reset.

Allow the move for interrupts which have the no balancing flag set and
clear the move pending bit before checking for interrupts with the per
cpu flag set.

Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1506201002570.4107@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agox86/hpet: Check for irq==0 when allocating hpet MSI interrupts
Jiang Liu [Sat, 20 Jun 2015 09:50:50 +0000 (11:50 +0200)]
x86/hpet: Check for irq==0 when allocating hpet MSI interrupts

irq == 0 is not a valid irq for a irqdomain MSI allocation, but hpet
code checks only for negative return values.

Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Link: http://lkml.kernel.org/r/558447AF.30703@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agoMerge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Fri, 19 Jun 2015 23:11:11 +0000 (01:11 +0200)]
Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

User visible changes:

  - Replace CTRL+z with 'f' as hotkey for enable/disable events (Arnaldo Carvalho de Melo)

  - Do not exit when 'f' is pressed in 'report' mode (Arnaldo Carvalho de Melo)

  - Tell the user how to unfreeze events after pressing 'f' in 'perf top' (Arnaldo Carvalho de Melo)

  - React to unassigned hotkey pressing in 'top/report' (Arnaldo Carvalho de Melo)

  - Display total number of samples with --show-total-period in 'annotate' (Martin LiÅ¡ka)

  - Add timeout to make procfs mmap processing more robust (Kan Liang)

  - Fix sort__sym_cmp to also compare end of symbol (Yannick Brosseau)

Infrastructure changes:

  - Ensure thread-stack is flushed (Adrian Hunter)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Fri, 19 Jun 2015 23:07:02 +0000 (01:07 +0200)]
Merge tag 'perf-urgent-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fix from Arnaldo Carvalho de Melo:

  - Fix build breakage if prefix= is specified, this introduced
    a regression for a build idiom used by the Debian "linux-tools"
    package, that does:

       MAKE_PERF := $(MAKE) prefix=/usr V=1 ARCH=$(KERNEL_ARCH_PERF) ...

    Fix it. (Lukas Wunner)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf tools: Configurable per thread proc map processing time out
Kan Liang [Wed, 17 Jun 2015 13:51:11 +0000 (09:51 -0400)]
perf tools: Configurable per thread proc map processing time out

The time out to limit the individual proc map processing was hard code
to 500ms. This patch introduce a new option --proc-map-timeout to make
the time limit configurable.

Signed-off-by: Kan Liang <kan.liang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ying Huang <ying.huang@intel.com>
Link: http://lkml.kernel.org/r/1434549071-25611-2-git-send-email-kan.liang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Add time out to force stop proc map processing
Kan Liang [Wed, 17 Jun 2015 13:51:10 +0000 (09:51 -0400)]
perf tools: Add time out to force stop proc map processing

System wide sampling like 'perf top' or 'perf record -a' read all
threads /proc/xxx/maps before sampling. If there are any threads which
generating a keeping growing huge maps, perf will do infinite loop
during synthesizing. Nothing will be sampled.

This patch fixes this issue by adding per-thread timeout to force stop
this kind of endless proc map processing.

PERF_RECORD_MISC_PROC_MAP_PARSE_TIME_OUT is introduced to indicate that
the mmap record are truncated by time out. User will get warning
notification when truncated mmap records are detected.

Reported-by: Ying Huang <ying.huang@intel.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ying Huang <ying.huang@intel.com>
Link: http://lkml.kernel.org/r/1434549071-25611-1-git-send-email-kan.liang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf report: Fix sort__sym_cmp to also compare end of symbol
Yannick Brosseau [Wed, 17 Jun 2015 23:41:10 +0000 (16:41 -0700)]
perf report: Fix sort__sym_cmp to also compare end of symbol

When using a map file from a JIT, due to memory reuse, we can obtain
multiple symbols with the same start address but a different length.

The symbols__find does check for the end so not doing it in
sort__sym_cmp was causing the hist_entry in the annotate part of a
report to match to the wrong entry, causing a fatal error.

Signed-off-by: Yannick Brosseau <scientist@fb.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: kernel-team@fb.com
Link: http://lkml.kernel.org/r/1434584470-17771-1-git-send-email-scientist@fb.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf hists browser: React to unassigned hotkey pressing
Arnaldo Carvalho de Melo [Fri, 19 Jun 2015 20:49:29 +0000 (17:49 -0300)]
perf hists browser: React to unassigned hotkey pressing

When that happens we were just ignoring the key press, now this
message is presented in the bottom line (the help line):

  "Press '?' for help on key bindings"

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-iyma2j5kj3q9i1stl4mfh90n@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf top: Tell the user how to unfreeze events after pressing 'f'
Arnaldo Carvalho de Melo [Fri, 19 Jun 2015 20:35:25 +0000 (17:35 -0300)]
perf top: Tell the user how to unfreeze events after pressing 'f'

When the user presses 'f' to disable events the visual cues are, well,
the percentages not changing and the number of events freezing.

Be more explicit by changing the help line at the bottom of the screen
to show the following messages when 'f' is pressed:

  "Press 'f' again to re-enable the events"

And then, when 'f' is pressed again:

  "Press 'f' to disable the events or 'h'

Suggested-by: Ingo Molnar <mingo@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-uhiswg9a9rxm5gxg7ptjskjn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf hists browser: Honour the help line provided by builtin-{top,report}.c
Arnaldo Carvalho de Melo [Fri, 19 Jun 2015 20:30:20 +0000 (17:30 -0300)]
perf hists browser: Honour the help line provided by builtin-{top,report}.c

The hists_browser was replacing whatever helpline provided by 'top' or
'report' with a static "Press '?' for help on key bindings", fix it.

Now the message passed by top appears at the bottom of the screen:

"For a higher level overview, try: perf top --sort comm,dso"

As well the message that will be added when the user presses 'f' to
disable the events, something along the lines of "press f again to
re-enable...".

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-dacaja70mbfz3a0yj1n180gx@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf hists browser: Do not exit when 'f' is pressed in 'report' mode
Arnaldo Carvalho de Melo [Fri, 19 Jun 2015 19:59:43 +0000 (16:59 -0300)]
perf hists browser: Do not exit when 'f' is pressed in 'report' mode

The 'f' hotkey is only used when in 'top', dynamic mode, to
enable/disable events, currently not making sense in the 'report',
static mode, where we can't go from showing the histogram entries
created from a perf.data file to adding more events after recreating the
evlist created from the perf.data file, albeit possible, this is not
implemented right now.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-lholzf472pu98dkkijggwx2m@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf top: Replace CTRL+z with 'f' as hotkey for enable/disable events
Arnaldo Carvalho de Melo [Fri, 19 Jun 2015 19:56:04 +0000 (16:56 -0300)]
perf top: Replace CTRL+z with 'f' as hotkey for enable/disable events

I.e. 'freeze'/'unfreeze', this is because CTRL+z has a well known
action, i.e. suspend the app, perf needs to follow that convention, that
will be done on a separate patch, tho.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-oedcl6ovohara4koig14ayip@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Fix build breakage if prefix= is specified
Lukas Wunner [Thu, 18 Jun 2015 11:00:32 +0000 (13:00 +0200)]
perf tools: Fix build breakage if prefix= is specified

Invoking Makefile.perf with prefix= breaks the build since Makefile.perf
hands that variable down to Makefile.build where it overrides

    prefix       := $(subst ./,,$(OUTPUT)$(dir)/)

leading to errors like this:

    No rule to make target '/usrabspath.o', needed by '/usrlibperf-in.o'

Signed-off-by: Lukas Wunner <lukas@wunner.de>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Fixes: c819e2cf2eb6f65d3208d195d7a0edef6108d5
Link: http://lkml.kernel.org/r/5582c48a.84a22b0a.a918.5285SMTPIN_ADDED_MISSING@mx.google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf annotate: Rename source_line_percent to source_line_samples
Arnaldo Carvalho de Melo [Fri, 19 Jun 2015 19:36:12 +0000 (16:36 -0300)]
perf annotate: Rename source_line_percent to source_line_samples

To better reflect the purpose of this struct, that is to hold
info about samples, its total number and is percentage.

Cc: Martin Liska <mliska@suse.cz>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/n/tip-6bf8gwcl975uurl0ttpvtk69@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf annotate: Display total number of samples with --show-total-period
Martin Liška [Fri, 19 Jun 2015 19:10:43 +0000 (16:10 -0300)]
perf annotate: Display total number of samples with --show-total-period

To compare two records on an instruction base, with --show-total-period
option provided, display total number of samples that belong to a line
in assembly language.

New hot key 't' is introduced for 'perf annotate' TUI.

Signed-off-by: Martin Liska <mliska@suse.cz>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/5583E26D.1040407@suse.cz
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Ensure thread-stack is flushed
Adrian Hunter [Fri, 29 May 2015 13:33:30 +0000 (16:33 +0300)]
perf tools: Ensure thread-stack is flushed

The thread-stack represents a thread's current stack.  When a thread
exits there can still be many functions on the stack e.g. exit() can be
called many levels deep, so all the callers will never return.  To get
that information output, the thread-stack must be flushed.

Previously it was assumed the thread-stack would be flushed when the
struct thread was deleted.  With thread ref-counting it is no longer
clear when that will be, if ever. So instead explicitly flush all the
thread-stacks at the end of a session.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1432906425-9911-3-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 19 Jun 2015 17:36:50 +0000 (07:36 -1000)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Michael Turquette:
 "Very late clk regression fixes for the ARM-based AT91 platform.

  These went unnoticed by me until recently, hence the late pull
  request"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: at91: fix h32mx prototype inclusion in pmc header
  clk: at91: trivial: typo in peripheral clock description
  clk: at91: fix PERIPHERAL_MAX_SHIFT definition
  clk: at91: pll: fix input range validity check

9 years agoMerge tag 'sound-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Fri, 19 Jun 2015 17:34:14 +0000 (07:34 -1000)]
Merge tag 'sound-4.1' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Nothing looks scary, just a few usual HD-audio regression fixes and
  fixup, in addition to a minor Kconfig dependency fix for the old MIPS
  drivers"

* tag 'sound-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Fix unused label skip_i915
  ALSA: hda - Fix noisy outputs on Dell XPS13 (2015 model)
  ALSA: mips: let SND_SGI_O2 select SND_PCM
  ALSA: hda - Fix audio crackles on Dell Latitude E7x40
  ALSA: hda - adding a DAC/pin preference map for a HP Envy TS machine

9 years agoMerge branch 'ccf/atmel-fixes-for-4.1' of https://github.com/bbrezillon/linux-at91...
Michael Turquette [Fri, 19 Jun 2015 14:37:14 +0000 (07:37 -0700)]
Merge branch 'ccf/atmel-fixes-for-4.1' of https://github.com/bbrezillon/linux-at91 into clk-fixes

9 years agox86/boot: Fix overflow warning with 32-bit binutils
Borislav Petkov [Fri, 19 Jun 2015 11:49:06 +0000 (13:49 +0200)]
x86/boot: Fix overflow warning with 32-bit binutils

When building the kernel with 32-bit binutils built with support
only for the i386 target, we get the following warning:

  arch/x86/kernel/head_32.S:66: Warning: shift count out of range (32 is not between 0 and 31)

The problem is that in that case, binutils' internal type
representation is 32-bit wide and the shift range overflows.

In order to fix this, manipulate the shift expression which
creates the 4GiB constant to not overflow the shift count.

Suggested-by: Michael Matz <matz@suse.de>
Reported-and-tested-by: Enrico Mioso <mrkiko.rs@gmail.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoclk: at91: fix h32mx prototype inclusion in pmc header
Nicolas Ferre [Thu, 28 May 2015 13:07:21 +0000 (15:07 +0200)]
clk: at91: fix h32mx prototype inclusion in pmc header

Trivial fix that prevents to compile this pmc clock driver if h32mx clock is
present but smd clock isn't.

Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Acked-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Fixes: bcc5fd49a0fd ("clk: at91: add a driver for the h32mx clock")
Cc: <stable@vger.kernel.org> # 3.18+
9 years agoclk: at91: trivial: typo in peripheral clock description
Nicolas Ferre [Wed, 17 Jun 2015 13:22:51 +0000 (15:22 +0200)]
clk: at91: trivial: typo in peripheral clock description

Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
9 years agotimer: Minimize nohz off overhead
Thomas Gleixner [Tue, 26 May 2015 22:50:35 +0000 (22:50 +0000)]
timer: Minimize nohz off overhead

If nohz is disabled on the kernel command line the [hr]timer code
still calls wake_up_nohz_cpu() and tick_nohz_full_cpu(), a pretty
pointless exercise. Cache nohz_active in [hr]timer per cpu bases and
avoid the overhead.

Before:
  48.10%  hog       [.] main
  15.25%  [kernel]  [k] _raw_spin_lock_irqsave
   9.76%  [kernel]  [k] _raw_spin_unlock_irqrestore
   6.50%  [kernel]  [k] mod_timer
   6.44%  [kernel]  [k] lock_timer_base.isra.38
   3.87%  [kernel]  [k] detach_if_pending
   3.80%  [kernel]  [k] del_timer
   2.67%  [kernel]  [k] internal_add_timer
   1.33%  [kernel]  [k] __internal_add_timer
   0.73%  [kernel]  [k] timerfn
   0.54%  [kernel]  [k] wake_up_nohz_cpu

After:
  48.73%  hog       [.] main
  15.36%  [kernel]  [k] _raw_spin_lock_irqsave
   9.77%  [kernel]  [k] _raw_spin_unlock_irqrestore
   6.61%  [kernel]  [k] lock_timer_base.isra.38
   6.42%  [kernel]  [k] mod_timer
   3.90%  [kernel]  [k] detach_if_pending
   3.76%  [kernel]  [k] del_timer
   2.41%  [kernel]  [k] internal_add_timer
   1.39%  [kernel]  [k] __internal_add_timer
   0.76%  [kernel]  [k] timerfn

We probably should have a cached value for nohz full in the per cpu
bases as well to avoid the cpumask check. The base cache line is hot
already, the cpumask not necessarily.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Link: http://lkml.kernel.org/r/20150526224512.207378134@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agotimer: Reduce timer migration overhead if disabled
Thomas Gleixner [Tue, 26 May 2015 22:50:33 +0000 (22:50 +0000)]
timer: Reduce timer migration overhead if disabled

Eric reported that the timer_migration sysctl is not really nice
performance wise as it needs to check at every timer insertion whether
the feature is enabled or not. Further the check does not live in the
timer code, so we have an extra function call which checks an extra
cache line to figure out that it is disabled.

We can do better and store that information in the per cpu (hr)timer
bases. I pondered to use a static key, but that's a nightmare to
update from the nohz code and the timer base cache line is hot anyway
when we select a timer base.

The old logic enabled the timer migration unconditionally if
CONFIG_NO_HZ was set even if nohz was disabled on the kernel command
line.

With this modification, we start off with migration disabled. The user
visible sysctl is still set to enabled. If the kernel switches to NOHZ
migration is enabled, if the user did not disable it via the sysctl
prior to the switch. If nohz=off is on the kernel command line,
migration stays disabled no matter what.

Before:
  47.76%  hog       [.] main
  14.84%  [kernel]  [k] _raw_spin_lock_irqsave
   9.55%  [kernel]  [k] _raw_spin_unlock_irqrestore
   6.71%  [kernel]  [k] mod_timer
   6.24%  [kernel]  [k] lock_timer_base.isra.38
   3.76%  [kernel]  [k] detach_if_pending
   3.71%  [kernel]  [k] del_timer
   2.50%  [kernel]  [k] internal_add_timer
   1.51%  [kernel]  [k] get_nohz_timer_target
   1.28%  [kernel]  [k] __internal_add_timer
   0.78%  [kernel]  [k] timerfn
   0.48%  [kernel]  [k] wake_up_nohz_cpu

After:
  48.10%  hog       [.] main
  15.25%  [kernel]  [k] _raw_spin_lock_irqsave
   9.76%  [kernel]  [k] _raw_spin_unlock_irqrestore
   6.50%  [kernel]  [k] mod_timer
   6.44%  [kernel]  [k] lock_timer_base.isra.38
   3.87%  [kernel]  [k] detach_if_pending
   3.80%  [kernel]  [k] del_timer
   2.67%  [kernel]  [k] internal_add_timer
   1.33%  [kernel]  [k] __internal_add_timer
   0.73%  [kernel]  [k] timerfn
   0.54%  [kernel]  [k] wake_up_nohz_cpu

Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agotimer: Stats: Simplify the flags handling
Thomas Gleixner [Tue, 26 May 2015 22:50:31 +0000 (22:50 +0000)]
timer: Stats: Simplify the flags handling

Simplify the handling of the flag storage for the timer statistics. No
intermediate storage anymore. Just hand over the flags field.

I left the printout of 'deferrable' for now because changing this
would be an ABI update and I have no idea how strong people feel about
that. OTOH, I wonder whether we should kill the whole timer stats
stuff because all of that information can be retrieved via ftrace/perf
as well.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Link: http://lkml.kernel.org/r/20150526224512.046626248@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agotimer: Replace timer base by a cpu index
Thomas Gleixner [Tue, 26 May 2015 22:50:29 +0000 (22:50 +0000)]
timer: Replace timer base by a cpu index

Instead of storing a pointer to the per cpu tvec_base we can simply
cache a CPU index in the timer_list and use that to get hold of the
correct per cpu tvec_base. This is only used in lock_timer_base() and
the slightly larger code is peanuts versus the spinlock operation and
the d-cache foot print of the timer wheel.

Aside of that this allows to get rid of following nuisances:

 - boot_tvec_base

   That statically allocated 4k bss data is just kept around so the
   timer has a home when it gets statically initialized. It serves no
   other purpose.

   With the CPU index we assign the timer to CPU0 at static
   initialization time and therefor can avoid the whole boot_tvec_base
   dance.  That also simplifies the init code, which just can use the
   per cpu base.

   Before:
     text    data     bss     dec     hex filename
    17491    9201    4160   30852    7884 ../build/kernel/time/timer.o
   After:
     text    data     bss     dec     hex filename
    17440    9193       0   26633    6809 ../build/kernel/time/timer.o

 - Overloading the base pointer with various flags

   The CPU index has enough space to hold the flags (deferrable,
   irqsafe) so we can get rid of the extra masking and bit fiddling
   with the base pointer.

As a benefit we reduce the size of struct timer_list on 64 bit
machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list,
which is a real win as we have tons of them embedded in other structs.

This changes also the newly added deferrable printout of the timer
start trace point to capture and print all timer->flags, which allows
us to decode the target cpu of the timer as well.

We might have used bitfields for this, but that would change the
static initializers and the init function for no value to accomodate
big endian bitfields.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Badhri Jagan Sridharan <Badhri@google.com>
Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agotimer: Use hlist for the timer wheel hash buckets
Thomas Gleixner [Tue, 26 May 2015 22:50:28 +0000 (22:50 +0000)]
timer: Use hlist for the timer wheel hash buckets

This reduces the size of struct tvec_base by 50% and results in
slightly smaller code as well.

Before:
   struct tvec_base: size: 8256, cachelines: 129

   text    data     bss     dec     hex filename
  17698   13297    8256   39251    9953 ../build/kernel/time/timer.o

After:
  struct tvec_base: 4160, cachelines: 65

   text    data     bss     dec     hex filename
  17491    9201    4160   30852    7884 ../build/kernel/time/timer.o

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Link: http://lkml.kernel.org/r/20150526224511.854731214@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agotimer: Remove FIFO "guarantee"
Thomas Gleixner [Tue, 26 May 2015 22:50:26 +0000 (22:50 +0000)]
timer: Remove FIFO "guarantee"

The FIFO guarantee is only there if two timers are queued into the
same bucket at the same jiffie on the same cpu:

 - The slack value depends on the delta between expiry and enqueue
   time, so the resulting expiry time can be different for timers
   which are queued in different jiffies.

 - Timers which are queued into the secondary array end up after a
   later queued timer which was queued into the primary array due to
   cascading.

 - Timers can end up on different cpus due to the NOHZ target moving
   around. Obviously there is no guarantee of expiry ordering between
   cpus.

So anything which relies on FIFO behaviour of the timer wheel is
broken already.

This is a preparatory patch for converting the timer wheel to hlist
which reduces the memory foot print of the wheel by 50%.

It's a seperate patch so any (unlikely to happen) regression caused by
this can be identified clearly.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Cc: George Spelvin <linux@horizon.com>
Link: http://lkml.kernel.org/r/20150526224511.757520403@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agotimers: Sanitize catchup_timer_jiffies() usage
Thomas Gleixner [Tue, 26 May 2015 22:50:24 +0000 (22:50 +0000)]
timers: Sanitize catchup_timer_jiffies() usage

catchup_timer_jiffies() has been applied blindly to several functions
without looking for possible better ways to do it.

1) internal_add_timer()

   Move the update to base->all_timers before we actually insert the
   timer into the wheel.

2) detach_if_pending()

   Again the update to base->all_timers allows us to explicitely do
   the timer_jiffies update in place, if this was the last timer which
   got removed.

3) __run_timers()

   We only check on entry, which is silly, because base->timer_jiffies
   can be behind - especially on NOHZ kernels - and if there is a
   single deferrable timer somewhere between base->timer_jiffies and
   jiffies we expire it and then loop until base->timer_jiffies ==
   jiffies.

   Move it into the loop.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joonwoo Park <joonwoop@codeaurora.org>
Cc: Wenbo Wang <wenbo.wang@memblaze.com>
Link: http://lkml.kernel.org/r/20150526224511.662994644@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agoclk: at91: fix PERIPHERAL_MAX_SHIFT definition
Boris Brezillon [Thu, 28 May 2015 12:01:08 +0000 (14:01 +0200)]
clk: at91: fix PERIPHERAL_MAX_SHIFT definition

Fix the PERIPHERAL_MAX_SHIFT definition (3 instead of 4) and adapt the
round_rate and set_rate logic accordingly.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Reported-by: "Wu, Songjun" <Songjun.Wu@atmel.com>
9 years agoclk: at91: pll: fix input range validity check
Boris Brezillon [Fri, 27 Mar 2015 22:53:15 +0000 (23:53 +0100)]
clk: at91: pll: fix input range validity check

The PLL impose a certain input range to work correctly, but it appears that
this input range does not apply on the input clock (or parent clock) but
on the input clock after it has passed the PLL divisor.
Fix the implementation accordingly.

Cc: <stable@vger.kernel.org> # v3.14+
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Reported-by: Jonas Andersson <jonas@microbit.se>
9 years agosched/deadline: Remove needless parameter in dl_runtime_exceeded()
Zhiqiang Zhang [Mon, 15 Jun 2015 03:15:20 +0000 (11:15 +0800)]
sched/deadline: Remove needless parameter in dl_runtime_exceeded()

Sine commit 269ad8015a6b ("sched/deadline: Avoid double-accounting in
case of missed deadlines), parameter 'rq' is no longer used, so
remove it.

Signed-off-by: Zhiqiang Zhang <zhangzhiqiang.zhang@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <juri.lelli@gmail.com>
Cc: <luca.abeni@unitn.it>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.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/1434338120-43773-1-git-send-email-zhangzhiqiang.zhang@huawei.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched: Remove superfluous resetting of the p->dl_throttled flag
Wanpeng Li [Wed, 13 May 2015 06:01:06 +0000 (14:01 +0800)]
sched: Remove superfluous resetting of the p->dl_throttled flag

Resetting the p->dl_throttled flag in rt_mutex_setprio() (for a task that is going
to be boosted) is superfluous, as the natural place to do so is in
replenish_dl_entity().

If the task was on the runqueue and it is boosted by a DL task, it will be enqueued
back with ENQUEUE_REPLENISH flag set, which can guarantee that dl_throttled is
reset in replenish_dl_entity().

This patch drops the resetting of throttled status in function rt_mutex_setprio().

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juri Lelli <juri.lelli@arm.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/1431496867-4194-6-git-send-email-wanpeng.li@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/deadline: Drop duplicate init_sched_dl_class() declaration
Wanpeng Li [Wed, 13 May 2015 06:01:05 +0000 (14:01 +0800)]
sched/deadline: Drop duplicate init_sched_dl_class() declaration

There are two init_sched_dl_class() declarations, this patch drops
the duplicate.

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juri Lelli <juri.lelli@arm.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/1431496867-4194-5-git-send-email-wanpeng.li@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/deadline: Reduce rq lock contention by eliminating locking of non-feasible...
Wanpeng Li [Wed, 13 May 2015 06:01:03 +0000 (14:01 +0800)]
sched/deadline: Reduce rq lock contention by eliminating locking of non-feasible target

This patch adds a check that prevents futile attempts to move DL tasks
to a CPU with active tasks of equal or earlier deadline. The same
behavior as commit 80e3d87b2c55 ("sched/rt: Reduce rq lock contention
by eliminating locking of non-feasible target") for rt class.

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juri Lelli <juri.lelli@arm.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/1431496867-4194-3-git-send-email-wanpeng.li@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/deadline: Make init_sched_dl_class() __init
Wanpeng Li [Wed, 13 May 2015 06:01:02 +0000 (14:01 +0800)]
sched/deadline: Make init_sched_dl_class() __init

It's a bootstrap function, make init_sched_dl_class() __init.

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juri Lelli <juri.lelli@arm.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/1431496867-4194-2-git-send-email-wanpeng.li@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/deadline: Optimize pull_dl_task()
Wanpeng Li [Wed, 13 May 2015 06:01:01 +0000 (14:01 +0800)]
sched/deadline: Optimize pull_dl_task()

pull_dl_task() uses pick_next_earliest_dl_task() to select a migration
candidate; this is sub-optimal since the next earliest task -- as per
the regular runqueue -- might not be migratable at all. This could
result in iterating the entire runqueue looking for a task.

Instead iterate the pushable queue -- this queue only contains tasks
that have at least 2 cpus set in their cpus_allowed mask.

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
[ Improved the changelog. ]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juri Lelli <juri.lelli@arm.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/1431496867-4194-1-git-send-email-wanpeng.li@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/preempt: Add static_key() to preempt_notifiers
Peter Zijlstra [Mon, 8 Jun 2015 14:00:30 +0000 (16:00 +0200)]
sched/preempt: Add static_key() to preempt_notifiers

Avoid touching the curr->preempt_notifier cacheline when not needed.

Provides a small improvement on pipe-bench:

  taskset 01 perf stat --repeat 10 -- perf bench sched pipe

before:

 Performance counter stats for 'perf bench sched pipe' (10 runs):

      12385.016204      task-clock (msec)         #    1.001 CPUs utilized            ( +-  0.34% )
         2,000,023      context-switches          #    0.161 M/sec                    ( +-  0.00% )
                 0      cpu-migrations            #    0.000 K/sec
               175      page-faults               #    0.014 K/sec                    ( +-  0.26% )
    41,376,162,250      cycles                    #    3.341 GHz                      ( +-  0.11% )
    17,389,139,321      stalled-cycles-frontend   #   42.03% frontend cycles idle     ( +-  0.25% )
   <not supported>      stalled-cycles-backend
    68,788,588,003      instructions              #    1.66  insns per cycle
                                                  #    0.25  stalled cycles per insn  ( +-  0.02% )
    13,449,387,620      branches                  # 1085.940 M/sec                    ( +-  0.02% )
        20,880,690      branch-misses             #    0.16% of all branches          ( +-  0.98% )

      12.372646094 seconds time elapsed                                          ( +-  0.34% )

after:

 Performance counter stats for 'perf bench sched pipe' (10 runs):

      12180.936528      task-clock (msec)         #    1.001 CPUs utilized            ( +-  0.33% )
         2,000,077      context-switches          #    0.164 M/sec                    ( +-  0.00% )
                 0      cpu-migrations            #    0.000 K/sec
               174      page-faults               #    0.014 K/sec                    ( +-  0.27% )
    40,691,545,577      cycles                    #    3.341 GHz                      ( +-  0.06% )
    16,446,333,371      stalled-cycles-frontend   #   40.42% frontend cycles idle     ( +-  0.18% )
   <not supported>      stalled-cycles-backend
    68,570,100,387      instructions              #    1.69  insns per cycle
                                                  #    0.24  stalled cycles per insn  ( +-  0.01% )
    13,389,740,014      branches                  # 1099.237 M/sec                    ( +-  0.01% )
        20,175,440      branch-misses             #    0.15% of all branches          ( +-  0.52% )

      12.169253010 seconds time elapsed                                          ( +-  0.33% )

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/preempt: Fix preempt notifiers documentation about hlist_del() within unsafe...
Mathieu Desnoyers [Sun, 17 May 2015 16:53:10 +0000 (12:53 -0400)]
sched/preempt: Fix preempt notifiers documentation about hlist_del() within unsafe iteration

preempt_notifier_unregister() documents:

  "This is safe to call from within a preemption notifier."

However, both fire_sched_in_preempt_notifiers() and
fire_sched_out_preempt_notifiers() are using hlist_for_each_entry(),
which is not safe against entry removal during iteration.

Inspection of the KVM code does not reveal any use of
preempt_notifier_unregister() within the preempt notifiers.

Therefore, fix the comment.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.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/1431881590-1456-1-git-send-email-mathieu.desnoyers@efficios.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/stop_machine: Fix deadlock between multiple stop_two_cpus()
Peter Zijlstra [Fri, 5 Jun 2015 15:30:23 +0000 (17:30 +0200)]
sched/stop_machine: Fix deadlock between multiple stop_two_cpus()

Jiri reported a machine stuck in multi_cpu_stop() with
migrate_swap_stop() as function and with the following src,dst cpu
pairs: {11,  4} {13, 11} { 4, 13}

                        4       11      13

cpuM: queue(4 ,13)
                        *Ma
cpuN: queue(13,11)
                                *N      Na
                        *M              Mb
cpuO: queue(11, 4)
                        *O      Oa
                                *Nb
                        *Ob

Where *X denotes the cpu running the queueing of cpu-X and X[ab] denotes
the first/second queued work.

You'll observe the top of the workqueue for each cpu: 4,11,13 to be work
from cpus: M, O, N resp. IOW. deadlock.

Do away with the queueing trickery and introduce lg_double_lock() to
lock both CPUs and fully serialize the stop_two_cpus() callers instead
of the partial (and buggy) serialization we have now.

Reported-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20150605153023.GH19282@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/debug: Add sum_sleep_runtime to /proc/<pid>/sched
Srikar Dronamraju [Mon, 8 Jun 2015 08:10:41 +0000 (13:40 +0530)]
sched/debug: Add sum_sleep_runtime to /proc/<pid>/sched

When CONFIG_SCHEDSTATS is enabled, /proc/<pid>/sched prints almost all
sched statistics except sum_sleep_runtime. Since sum_sleep_runtime is
a good info to collect, add this it to /proc/<pid>/sched.

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.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/1433751041-11724-4-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/debug: Replace vruntime with wait_sum in /proc/sched_debug
Srikar Dronamraju [Mon, 8 Jun 2015 08:10:40 +0000 (13:40 +0530)]
sched/debug: Replace vruntime with wait_sum in /proc/sched_debug

Within runnable tasks in /proc/sched_debug, vruntime is printed twice,
once as tree-key and again as exec-runtime.

Since exec-runtime isnt populated in !CONFIG_SCHEDSTATS, use this field
to print wait_sum.

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.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/1433751041-11724-3-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agosched/debug: Properly format runnable tasks in /proc/sched_debug
Srikar Dronamraju [Mon, 8 Jun 2015 08:10:39 +0000 (13:40 +0530)]
sched/debug: Properly format runnable tasks in /proc/sched_debug

With !CONFIG_SCHEDSTATS, runnable tasks in /proc/sched_debug has too
many columns than required. Fix this by printing appropriate columns.

While at this, print sum_exec_runtime, since this information is
available even in !CONFIG_SCHEDSTATS case.

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.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/1433751041-11724-2-git-send-email-srikar@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agolocking/lockdep: Remove hard coded array size dependency
George Beshers [Thu, 18 Jun 2015 15:25:13 +0000 (10:25 -0500)]
locking/lockdep: Remove hard coded array size dependency

An apparent oversight left a hardcoded '4' in place when
LOCKSTAT_POINTS was introduced.

The contention_point[] and contending_point[] arrays in the
structs lock_class and lock_class_stats need to be the same
size for the loops in lock_stats() to be correct.

This patch allows LOCKSTAT_POINTS to be changed without
affecting the correctness of the code.

Signed-off-by: George Beshers <gbeshers@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agolocking/qrwlock: Don't contend with readers when setting _QW_WAITING
Waiman Long [Tue, 9 Jun 2015 15:19:13 +0000 (11:19 -0400)]
locking/qrwlock: Don't contend with readers when setting _QW_WAITING

The current cmpxchg() loop in setting the _QW_WAITING flag for writers
in queue_write_lock_slowpath() will contend with incoming readers
causing possibly extra cmpxchg() operations that are wasteful. This
patch changes the code to do a byte cmpxchg() to eliminate contention
with new readers.

A multithreaded microbenchmark running 5M read_lock/write_lock loop
on a 8-socket 80-core Westmere-EX machine running 4.0 based kernel
with the qspinlock patch have the following execution times (in ms)
with and without the patch:

With R:W ratio = 5:1

Threads    w/o patch with patch % change
-------    --------- ---------- --------
   2      990     895   -9.6%
   3     2136    1912  -10.5%
   4     3166    2830  -10.6%
   5     3953    3629   -8.2%
   6     4628    4405   -4.8%
   7     5344    5197   -2.8%
   8     6065    6004   -1.0%
   9     6826    6811   -0.2%
  10     7599    7599    0.0%
  15     9757    9766   +0.1%
  20    13767   13817   +0.4%

With small number of contending threads, this patch can improve
locking performance by up to 10%. With more contending threads,
however, the gain diminishes.

Signed-off-by: Waiman Long <Waiman.Long@hp.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Douglas Hatch <doug.hatch@hp.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott J Norton <scott.norton@hp.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1433863153-30722-3-git-send-email-Waiman.Long@hp.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86: Honor the architectural performance monitoring version
Palik, Imre [Mon, 8 Jun 2015 12:46:49 +0000 (14:46 +0200)]
perf/x86: Honor the architectural performance monitoring version

Architectural performance monitoring, version 1, doesn't support fixed counters.

Currently, even if a hypervisor advertises support for architectural
performance monitoring version 1, perf may still try to use the fixed
counters, as the constraints are set up based on the CPU model.

This patch ensures that perf honors the architectural performance monitoring
version returned by CPUID, and it only uses the fixed counters for version 2
and above.

(Some of the ideas in this patch came from Peter Zijlstra.)

Signed-off-by: Imre Palik <imrep@amazon.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Anthony Liguori <aliguori@amazon.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1433767609-1039-1-git-send-email-imrep.amz@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel: Fix PMI handling for Intel PT
Alexander Shishkin [Tue, 9 Jun 2015 10:03:26 +0000 (13:03 +0300)]
perf/x86/intel: Fix PMI handling for Intel PT

Intel PT is a separate PMU and it is not using any of the x86_pmu
code paths, which means in particular that the active_events counter
remains intact when new PT events are created.

However, PT uses the generic x86_pmu PMI handler for its PMI handling needs.

The problem here is that the latter checks active_events and in case of it
being zero, exits without calling the actual x86_pmu.handle_nmi(), which
results in unknown NMI errors and massive data loss for PT.

The effect is not visible if there are other perf events in the system
at the same time that keep active_events counter non-zero, for instance
if the NMI watchdog is running, so one needs to disable it to reproduce
the problem.

At the same time, the active_events counter besides doing what the name
suggests also implicitly serves as a PMC hardware and DS area reference
counter.

This patch adds a separate reference counter for the PMC hardware, leaving
active_events for actually counting the events and makes sure it also
counts PT and BTS events.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: adrian.hunter@intel.com
Link: http://lkml.kernel.org/r/87k2v92t0s.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel/bts: Fix DS area sharing with x86_pmu events
Alexander Shishkin [Thu, 11 Jun 2015 12:13:56 +0000 (15:13 +0300)]
perf/x86/intel/bts: Fix DS area sharing with x86_pmu events

Currently, the intel_bts driver relies on the DS area allocated by the x86_pmu
code in its event_init() path, which is a bug: creating a BTS event while
no x86_pmu events are present results in a NULL pointer dereference.

The same DS area is also used by PEBS sampling, which makes it quite a bit
trickier to have a separate one for intel_bts' purposes.

This patch makes intel_bts driver use the same DS allocation and reference
counting code as x86_pmu to make sure it is always present when either
intel_bts or x86_pmu need it.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: adrian.hunter@intel.com
Link: http://lkml.kernel.org/r/1434024837-9916-2-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86: Add more Broadwell model numbers
Andi Kleen [Thu, 11 Jun 2015 20:52:22 +0000 (13:52 -0700)]
perf/x86: Add more Broadwell model numbers

This patch adds additional model numbers for Broadwell to perf.
Support for Broadwell with Iris Pro (Intel Core i7-57xxC)
and support for Broadwell Server Xeon.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1434055942-28253-1-git-send-email-andi@firstfloor.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>