platform/kernel/linux-rpi.git
8 years agoperf hists: Add __hist_entry__snprintf function
Jiri Olsa [Thu, 22 Sep 2016 15:36:29 +0000 (17:36 +0200)]
perf hists: Add __hist_entry__snprintf function

Add __hist_entry__snprintf() to take a perf_hpp_list as an argument
instead of using he->hists->hpp_list.

This way we can display arbitrary list of entries regardless of the
hists setup, which will be useful in the upcoming c2c patch series.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1474558645-19956-2-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Add sink configuration for cs_etm PMU
Mathieu Poirier [Fri, 16 Sep 2016 15:50:04 +0000 (09:50 -0600)]
perf tools: Add sink configuration for cs_etm PMU

Using the PMU::set_drv_config() callback to enable the CoreSight sink
that will be used for the trace session.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-8-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Add PMU configuration to tools
Mathieu Poirier [Fri, 16 Sep 2016 15:50:03 +0000 (09:50 -0600)]
perf tools: Add PMU configuration to tools

Now that the required mechanic is there to deal with PMU specific
configuration, add the functionality to the tools where events can be
selected.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-7-git-send-email-mathieu.poirier@linaro.org
[ Fix the build on XSI-compliant systems, using str_error_r() to make sure we return a string, not an integer ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf pmu: Push configuration down to PMU driver
Mathieu Poirier [Fri, 16 Sep 2016 15:50:02 +0000 (09:50 -0600)]
perf pmu: Push configuration down to PMU driver

This patch adds a PMU callback and the required mechanic so that drivers
can process the command line configuration elements found in
evsel::config_terms.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-6-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Add coresight etm PMU record capabilities
Mathieu Poirier [Fri, 16 Sep 2016 15:50:00 +0000 (09:50 -0600)]
perf tools: Add coresight etm PMU record capabilities

Coresight ETMs are IP blocks used to perform HW assisted tracing on a
CPU core.  This patch introduce the required auxiliary API functions
allowing the perf core to interact with a tracer.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-4-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Make coresight PMU listable
Mathieu Poirier [Fri, 16 Sep 2016 15:49:59 +0000 (09:49 -0600)]
perf tools: Make coresight PMU listable

Adding the required mechanic allowing 'perf list pmu' to discover
coresight ETM/PTM tracers.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-3-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Confine __get_cpuid() to x86 architecture
Mathieu Poirier [Fri, 16 Sep 2016 15:49:58 +0000 (09:49 -0600)]
perf tools: Confine __get_cpuid() to x86 architecture

The __get_cpuid() test is only valid when compiling for x86.  When
compiling for other architectures like ARM/ARM64 the test fails event if
the functionality is not required.

This patch isolate the build-in feature check to x86 platform, allowing
the compilation and usage of PMUs that use the AUXTRACE infrastructure
on other architectures (i.e ARM CoreSight).

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1474041004-13956-2-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hists: Use bigger buffer for stdio headers
Jiri Olsa [Mon, 19 Sep 2016 13:09:13 +0000 (15:09 +0200)]
perf hists: Use bigger buffer for stdio headers

With node column on big CPUs servers we can run out of stdio header
space quite soon. Enlarging header buffer.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1474290610-23241-5-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf evsel: Remove superfluous initialization of weight
Jiri Olsa [Mon, 19 Sep 2016 13:09:11 +0000 (15:09 +0200)]
perf evsel: Remove superfluous initialization of weight

Removing superfluous initialization of weight, it's already set to 0 via
memset.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1474290610-23241-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoMerge tag 'perf-core-for-mingo-20160920' of git://git.kernel.org/pub/scm/linux/kernel...
Ingo Molnar [Tue, 20 Sep 2016 21:32:02 +0000 (23:32 +0200)]
Merge tag 'perf-core-for-mingo-20160920' 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:

- Support event group view with hierarchy mode in 'perf top' and 'perf report'
  (Namhyung Kim)

  e.g.:

  $ perf record -e '{cycles,instructions}' make
  $ perf report --hierarchy --stdio
  ...
  #               Overhead  Command / Shared Object / Symbol
  # ......................  ..................................
  ...
      25.74%  27.18%        sh
         19.96%  24.14%        libc-2.24.so
            9.55%  14.64%        [.] __strcmp_sse2
            1.54%   0.00%        [.] __tfind
            1.07%   1.13%        [.] _int_malloc
            0.95%   0.00%        [.] __strchr_sse2
            0.89%   1.39%        [.] __tsearch
            0.76%   0.00%        [.] strlen

- Fix the dwarf regs table for x86_64, adding a missing % to the "%di"
  register, noticed with a failing 'perf test bpf' (Arnaldo Carvalho de Melo)

- Fix handling of mmap parameters in the 'perf trace' beautifier in
  architectures that don't have the same mappings as x86_64 (Wang Nan)

- Handle hugetbl mappings in older systems running new kernels (Wang Nan)

- Resolve 'call' operands in 'annotate', that when using /proc/kcore
  were appearing just as hexadecimal addresses, to function names
  (Arnaldo Carvalho de Melo)

- Fix width computation for srcline sort entry (Jiri Olsa)

- Do not ignore call instruction with indirect target in 'annotate'
  (Ravi Bangoria)

- Handle MADV_FREE in the madvise 'trace' beautifier (Wang Nan)

- Fix build of 'perf trace' mman beautifier in !x86_64 (Wang Nan)

Infrastructure changes:

- Add infrastructure for PMU specific configuration, allowing to pass
  config variables directly to the kernel PMU driver, prefixing those
  variables with a '@', part of a larger series to support Coresight (Mathieu Poirier)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf symbols: Do not open device files
Jiri Olsa [Tue, 20 Sep 2016 16:12:45 +0000 (18:12 +0200)]
perf symbols: Do not open device files

The dso__read_binary_type_filename gets the dso's file name to open. We
need to check it for regular file before trying to open it, otherwise we
might get stuck with device file.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20160920161245.GA8995@krava
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hists: Factor out hists__reset_column_width()
Namhyung Kim [Tue, 20 Sep 2016 05:30:24 +0000 (14:30 +0900)]
perf hists: Factor out hists__reset_column_width()

The stdio and tui has same code to reset hpp format column width.
Factor it out as a new function.

Suggested-and-Acked-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20160920053025.13989-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf ui/tui: Reset output width for hierarchy
Namhyung Kim [Tue, 20 Sep 2016 05:30:23 +0000 (14:30 +0900)]
perf ui/tui: Reset output width for hierarchy

When --hierarchy option is used, each entry has its own hpp_list to show
the result.  But it missed to update width of each column.

Before:

  - 46.29% 48.12%        netctl-auto
     + 31.44% 29.25%        [kernel.vmlinux]
     + 8.52% 11.55%        libc-2.22.so
     + 5.19% 6.91%        bash
  + 10.75% 11.83%        wpa_cli
  + 8.25% 2.23%        swapper
  + 6.45% 5.40%        tr
  + 4.81% 8.09%        awk
  + 4.15% 2.85%        firefox
  + 3.86% 2.53%        sh

After:

  -  46.29%  48.12%        netctl-auto
      +  31.44%  29.25%        [kernel.vmlinux]
      +   8.52%  11.55%        libc-2.22.so
      +   5.19%   6.91%        bash
  +  10.75%  11.83%        wpa_cli
  +   8.25%   2.23%        swapper
  +   6.45%   5.40%        tr
  +   4.81%   8.09%        awk
  +   4.15%   2.85%        firefox
  +   3.86%   2.53%        sh

Committer note:

Full testing instructions:

1) Record with an event group:

  $ perf record -e '{cycles,instructions}' make -j4

2) Use report in hierarchy mode, to get a few expanded trees on
   the same screen, use --percent-limit:

  $ perf report --hierarchy --percent-limit 0.5

Samples: 103K of event 'anon group { cycles:u, instructions:u }',
Event count (approx.): 57317631725
         Overhead        Command / Shared Object / Symbol        ◆
-  58.89%  55.12%        cc1                                     ▒
   -  50.26%  48.10%        cc1                                  ▒
          3.61%   5.13%        [.] _cpp_lex_token                ▒
          2.58%   0.78%        [.] ht_lookup_with_hash           ▒
          1.31%   1.30%        [.] ggc_internal_alloc            ▒
          1.08%   2.25%        [.] get_combined_adhoc_loc        ▒
          1.01%   1.95%        [.] ira_init                      ▒
          0.96%   1.78%        [.] linemap_position_for_column   ▒
          0.65%   1.01%        [.] cpp_get_token_with_location   ▒
   -   7.52%   6.58%        libc-2.23.so                         ▒
          1.70%   1.78%        [.] _int_malloc                   ▒
          0.69%   0.75%        [.] _int_free                     ▒
          0.67%   0.42%        [.] malloc_consolidate            ▒
   -   0.58%   0.42%        ld-2.23.so                           ▒
                               no entry >= 0.50%                 ▒
   -   0.52%   0.03%        [kernel.vmlinux]                     ▒
                               no entry >= 0.50%                 ▒

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Fixes: 1b2dbbf41a0f ("perf hists: Use own hpp_list for hierarchy mode")
Link: http://lkml.kernel.org/r/20160920053025.13989-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf annotate: Resolve 'call' operands to function names
Arnaldo Carvalho de Melo [Mon, 19 Sep 2016 20:26:11 +0000 (17:26 -0300)]
perf annotate: Resolve 'call' operands to function names

Before this patch the '_raw_spin_lock_irqsave' and 'update_rq_clock' operands
were appearing just as hexadecimal numbers:

  update_blocked_averages  /proc/kcore
       │       push   %r12
       │       push   %rbx
       │       and    $0xfffffffffffffff0,%rsp
       │       sub    $0x40,%rsp
       │       add    -0x662cac00(,%rdi,8),%rax
       │       mov    %rax,%rbx
       │       mov    %rax,%rdi
       │       mov    %rax,0x38(%rsp)
       │     → callq  _raw_spin_lock_irqsave
       │       mov    %rbx,%rdi
       │       mov    %rax,0x30(%rsp)
       │     → callq  update_rq_clock
       │       mov    0x8d0(%rbx),%rax
       │       lea    0x8d0(%rbx),%r11

To check that all is right one can always use the 'o' hotkey and see
the original objdump -dS output, that for this case is:

  update_blocked_averages  /proc/kcore
       │ffffffff990d5489:   push   %r12
       │ffffffff990d548b:   push   %rbx
       │ffffffff990d548c:   and    $0xfffffffffffffff0,%rsp
       │ffffffff990d5490:   sub    $0x40,%rsp
       │ffffffff990d5494:   add    -0x662cac00(,%rdi,8),%rax
       │ffffffff990d549c:   mov    %rax,%rbx
       │ffffffff990d549f:   mov    %rax,%rdi
       │ffffffff990d54a2:   mov    %rax,0x38(%rsp)
       │ffffffff990d54a7: → callq  0xffffffff997eb7a0
       │ffffffff990d54ac:   mov    %rbx,%rdi
       │ffffffff990d54af:   mov    %rax,0x30(%rsp)
       │ffffffff990d54b4: → callq  0xffffffff990c7720
       │ffffffff990d54b9:   mov    0x8d0(%rbx),%rax
       │ffffffff990d54c0:   lea    0x8d0(%rbx),%r11

Use the 'h' hotkey to see a list of available hotkeys.

More work needed to cover operands for other instructions, such as 'mov',
that can resolve variable names, etc.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Chris Riyder <chris.ryder@arm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Markus Trippelsdorf <markus@trippelsdorf.de>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Taeung Song <treeze.taeung@gmail.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-xqgtw9mzmzcjgwkis9kiiv1p@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf annotate: Pass the symbol's map/dso to the instruction parsers
Arnaldo Carvalho de Melo [Mon, 19 Sep 2016 20:18:16 +0000 (17:18 -0300)]
perf annotate: Pass the symbol's map/dso to the instruction parsers

So that things like:

       → callq  0xffffffff993e3230

found while disassembling /proc/kcore can be beautified by later
patches, that will resolve that address to a function, looking it up in
/proc/kallsyms.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Chris Riyder <chris.ryder@arm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Markus Trippelsdorf <markus@trippelsdorf.de>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Taeung Song <treeze.taeung@gmail.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-p76myuke4j7gplg54amaklxk@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf annotate: Do not ignore call instruction with indirect target
Ravi Bangoria [Fri, 19 Aug 2016 12:59:35 +0000 (18:29 +0530)]
perf annotate: Do not ignore call instruction with indirect target

Do not ignore call instruction with indirect target when its already
identified as a call. This is an extension of commit e8ea1561952b ("perf
annotate: Use raw form for register indirect call instructions") to
generalize annotation for all instructions with indirect calls.

This is needed for certain powerpc call instructions that use address in
a register (such as bctrl, btarl, ...).

Apart from that, when kcore is used to disassemble function, all call
instructions were ignored. This patch will fix it as a side effect by
not ignoring them. For example,

Before (with kcore):
       mov    %r13,%rdi
       callq  0xffffffff811a7e70
     ^ jmpq   64
       mov    %gs:0x7ef41a6e(%rip),%al

After (with kcore):
       mov    %r13,%rdi
     > callq  0xffffffff811a7e70
     ^ jmpq   64
       mov    %gs:0x7ef41a6e(%rip),%al

Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
[Suggested about 'bctrl' instruction]
Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Chris Riyder <chris.ryder@arm.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Markus Trippelsdorf <markus@trippelsdorf.de>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Taeung Song <treeze.taeung@gmail.com>
Link: http://lkml.kernel.org/r/1471611578-11255-5-git-send-email-ravi.bangoria@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hists: Fix width computation for srcline sort entry
Jiri Olsa [Mon, 19 Sep 2016 13:10:10 +0000 (15:10 +0200)]
perf hists: Fix width computation for srcline sort entry

Adding header size to width computation for srcline sort entry,
because it's possible to get empty data with ':0' which set width
of 2 which is lower than width needed to display column header.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1474290610-23241-62-git-send-email-jolsa@kernel.org
[ Added declaration to sort.h ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf/x86/intel/pt: Add support for PTWRITE and power event tracing
Alexander Shishkin [Fri, 16 Sep 2016 13:48:19 +0000 (16:48 +0300)]
perf/x86/intel/pt: Add support for PTWRITE and power event tracing

The Intel PT facility grew some new functionality:

  * PTWRITE packet carries the payload of the new PTWRITE instruction
    that can be used to instrument Intel PT traces with user-supplied
    data. Packets of this type are only generated if 'ptwrite' capability
    is set and PTWEn bit is set in the event attribute's config. Flow
    update packets (FUP) can be generated on PTWRITE packets if FUPonPTW
    config bit is set. Setting these bits is not allowed if 'ptwrite'
    capability is not set.

  * PWRE, PWRX, MWAIT, EXSTOP packets communicate core power management
    events. These depend on 'power_event_tracing' capability and are
    enabled by setting PwrEvtEn bit in the event attribute.

Extend the driver capabilities and provide the proper sanity checks in the
event validation function.

[ tglx: Massaged changelog ]

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: vince@deater.net
Cc: eranian@google.com
Cc: Adrian Hunter <adrian.hunter@intel.com>
Link: http://lkml.kernel.org/r/20160916134819.1978-1-alexander.shishkin@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
8 years agotools include: Add mman macros needed by perf for all arch
Wang Nan [Wed, 14 Sep 2016 10:57:29 +0000 (10:57 +0000)]
tools include: Add mman macros needed by perf for all arch

Some macros required by tools/perf/trace/beauty/mmap.c is not support
for all architectures. For example, MAP_32BIT is defined on x86 only,
alpha doesn't define MADV_HWPOISON and MADV_SOFT_OFFLINE.

This patch regenerates mman.h for each arch, defines these missing
macros for perf. For missing MADV_*, fall back to asm-generic/mman-common
because they are in a 'case ...' statement. For flags, define it to 0.

Following is the script to generate this patch:

 macros=`cat $0 | awk 'V==1 {print}; /^# start macro list/ {V=1}'`
 rm `find ./tools/arch/ -name mman.h`
 for arch in `ls tools/arch`
 do
   [ -d tools/arch/$arch/include/uapi/asm ] || mkdir -p tools/arch/$arch/include/uapi/asm
   src=arch/$arch/include/uapi/asm/mman.h
   target=tools/arch/$arch/include/uapi/asm/mman.h.tmp
   real_target=tools/arch/$arch/include/uapi/asm/mman.h
   guard="TOOLS_ARCH_"`echo $arch | awk '{print toupper($0)}'`_UAPI_ASM_MMAN_FIX_H
   rm -f $target

   [ -f $src ] &&
   for m in $macros
   do
     if grep '#define[ \t]*'$m $src > /dev/null 2>&1
     then
       grep -h '#define[ \t]*'$m $src | sed 's/[ \t]*\/\*.*$//g' >> $target
     fi
   done

   if [ -f $src ]
   then
      grep '#include <asm-generic' $src >> $target
   else
      echo "#include <asm-generic/mman.h>" >> $target
   fi

   touch $real_target
   for m in $macros
   do
     if cat << EOF | gcc -Itools/arch/$arch/include -Itools/arch/$arch/include/uapi -Iinclude/ -Iinclude/uapi -E - | grep $m > /dev/null 2>&1
 #include <uapi/asm/mman.h.tmp>
 #include <uapi/linux/mman.h>
 $m
 EOF
   then
     echo "Fixing $m for $arch"
     echo "/* $m is undefined on $arch, fix it for perf */" >> $target
     if echo $m | grep '^MADV_' > /dev/null 2>&1
     then
       grep -h '#define[ \t]*'$m include/uapi/asm-generic/mman-common.h | sed 's/[ \t]*\/\*.*$//g' >> $target
     else
       echo "#define $m 0" >> $target
     fi
   fi
   done

   real_target=tools/arch/$arch/include/uapi/asm/mman.h
   echo '#ifndef '$guard > $real_target
   echo '#define '$guard >> $real_target
   cat $target | sed 's|asm-generic|uapi/asm-generic|g' >> $real_target
   echo '#endif' >> $real_target
   rm $target
   echo "$real_target"
 done

 exit 0
 # Following macros are extracted from:
 # tools/perf/trace/beauty/mmap.c
 #
 # start macro list
 MADV_DODUMP
 MADV_DOFORK
 MADV_DONTDUMP
 MADV_DONTFORK
 MADV_DONTNEED
 MADV_FREE
 MADV_HUGEPAGE
 MADV_HWPOISON
 MADV_MERGEABLE
 MADV_NOHUGEPAGE
 MADV_NORMAL
 MADV_RANDOM
 MADV_REMOVE
 MADV_SEQUENTIAL
 MADV_SOFT_OFFLINE
 MADV_UNMERGEABLE
 MADV_WILLNEED
 MAP_32BIT
 MAP_ANONYMOUS
 MAP_DENYWRITE
 MAP_EXECUTABLE
 MAP_FILE
 MAP_FIXED
 MAP_GROWSDOWN
 MAP_HUGETLB
 MAP_LOCKED
 MAP_NONBLOCK
 MAP_NORESERVE
 MAP_POPULATE
 MAP_PRIVATE
 MAP_SHARED
 MAP_STACK
 MAP_UNINITIALIZED
 MREMAP_FIXED
 MREMAP_MAYMOVE
 PROT_EXEC
 PROT_GROWSDOWN
 PROT_GROWSUP
 PROT_NONE
 PROT_READ
 PROT_SEM
 PROT_WRITE

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Kim Phillips <kim.phillips@arm.com>
Tested-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Fixes: 277cf08f3feb ("perf trace beauty mmap: Fix defines for non !x86_64")
Link: http://lkml.kernel.org/r/1473850649-83389-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf trace beauty mmap: Add missing MADV_FREE
Wang Nan [Wed, 14 Sep 2016 10:57:28 +0000 (10:57 +0000)]
perf trace beauty mmap: Add missing MADV_FREE

tools/perf/trace/beauty/mmap.c forgets to check MADV_FREE.
This patch fixes it.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1473850649-83389-2-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Add infrastructure for PMU specific configuration
Mathieu Poirier [Tue, 6 Sep 2016 16:37:15 +0000 (10:37 -0600)]
perf tools: Add infrastructure for PMU specific configuration

This patch adds PMU driver specific configuration to the parser
infrastructure by preceding any term with the '@' letter.  As such doing
something like:

perf record -e some_event/@cfg1,@cfg2=config/ ...

will see 'cfg1' and 'cfg2=config' being added to the list of evsel
config terms.  Token 'cfg1' and 'cfg2=config' are not processed in user
space and are meant to be interpreted by the PMU driver.

First the lexer/parser are supplemented with the required definitions to
recognise the driver specific configuration.  From there they are simply
added to the list of event terms.  The bulk of the work is done in
function "parse_events_add_pmu()" where driver config event terms are
added to a new list of driver config terms, which in turn spliced with
the event's new driver configuration list.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1473179837-3293-4-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf report: Enable group view with hierarchy
Namhyung Kim [Tue, 13 Sep 2016 07:45:52 +0000 (16:45 +0900)]
perf report: Enable group view with hierarchy

Now that all the missing pieces are implemented, let's enable it.  An
example output below:

  $ perf record -e '{cycles,instructions}' make
  $ perf report --hierarchy --stdio
  ...
  #               Overhead  Command / Shared Object / Symbol
  # ......................  ..................................
  #
  ...
      25.74%  27.18%        sh
         19.96%  24.14%        libc-2.24.so
            9.55%  14.64%        [.] __strcmp_sse2
            1.54%   0.00%        [.] __tfind
            1.07%   1.13%        [.] _int_malloc
            0.95%   0.00%        [.] __strchr_sse2
            0.89%   1.39%        [.] __tsearch
            0.76%   0.00%        [.] strlen
  ...

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Requested-by: Andi Kleen <andi@firstfloor.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20160913074552.13284-8-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf ui/stdio: Rename print_hierarchy_header()
Namhyung Kim [Tue, 13 Sep 2016 07:45:50 +0000 (16:45 +0900)]
perf ui/stdio: Rename print_hierarchy_header()

Now the hists__fprintf_hierarchy_headers() is a simple wrapper passing
field separator.  Let's do it directly.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20160913074552.13284-6-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf ui/stdio: Always reset output width for hierarchy
Namhyung Kim [Tue, 13 Sep 2016 07:45:49 +0000 (16:45 +0900)]
perf ui/stdio: Always reset output width for hierarchy

When the --hierarchy option is used, each entry has its own hpp_list to
show the result.  But it is not updating the width of each column for
perf-top.  The perf-report command has no problem since it resets it
during header display.

  $ sudo perf top --hierarchy --stdio

   PerfTop:     160 irqs/sec  kernel:38.8%  exact: 100.0%
                                     [4000Hz cycles:pp],  (all, 12 CPUs)
  ----------------------------------------------------------------------

   52.32%     perf
      24.74%     [.] __symbols__insert
      5.62%     [.] rb_next
      5.14%     [.] dso__load_sym

Move the code into hists__fprintf() so that it can be called always.
Also it'd be better to put similar code together.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Fixes: 1b2dbbf41a0f ("perf hists: Use own hpp_list for hierarchy mode")
Link: http://lkml.kernel.org/r/20160913074552.13284-5-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hist: Initialize hierarchy tree explicitly
Namhyung Kim [Tue, 13 Sep 2016 07:45:48 +0000 (16:45 +0900)]
perf hist: Initialize hierarchy tree explicitly

The hroot_in and hroot_out are roots of hierarchy trees of hist entries.

But when a hist entry is initialized by copying existing template entry,
it sometimes has non-empty tree and copies it incorrectly.  This is a
problem especially when an event group is used since it creates dummy
entries from already-processed entries in other event members.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20160913074552.13284-4-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hists: Introduce hists__link_hierarchy()
Namhyung Kim [Tue, 13 Sep 2016 07:45:47 +0000 (16:45 +0900)]
perf hists: Introduce hists__link_hierarchy()

The hists__link_hierarchy() is to support hierarchy reports with an
event group.  When it matches the leader event and the other members
(using hists__match_hierarchy()), it also needs to link unmatched member
entries with a dummy leader event so that it can show up in the output.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20160913074552.13284-3-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hists: Introduce hists__match_hierarchy()
Namhyung Kim [Tue, 13 Sep 2016 07:45:46 +0000 (16:45 +0900)]
perf hists: Introduce hists__match_hierarchy()

The hists__match_hierarchy() is to find matching hist entries in a
group.  A matching entry has the same values for all sort keys given.

With an event group (e.g.: -e "{cycles,instructions}"), a leader event
should show other members in a group.  So each entry in the leader
should be able to find its pair entries which have same values.

With hierarchy mode, it needs to search all matching children in a
hierarchy.

An example output looks like:

  #               Overhead  Command / Shared Object / Symbol
  # ......................  ..................................
  #
      25.74%  27.18%        sh
         19.96%  24.14%        libc-2.24.so
            9.55%  14.64%        [.] __strcmp_sse2
            1.54%   0.00%        [.] __tfind
            1.07%   1.13%        [.] _int_malloc
  ...

In the above example, two overheads are shown - one for the leader and
another for the other group member.  They were matched since their
command, dso and symbol have the same values.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20160913074552.13284-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf build: Compare mman.h related headers against kernel originals
Wang Nan [Mon, 12 Sep 2016 12:54:31 +0000 (12:54 +0000)]
perf build: Compare mman.h related headers against kernel originals

As with other cloned headers, compare the newly introduced mman related
headers against their source copy in kernel tree.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1473684871-209320-4-git-send-email-wangnan0@huawei.com
[ Added -I to ignore the uapi/ difference ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Do hugetlb handling in more systems
Arnaldo Carvalho de Melo [Mon, 12 Sep 2016 19:47:57 +0000 (16:47 -0300)]
perf tools: Do hugetlb handling in more systems

The csets:

  0ac3348e5024 ("perf tools: Recognize hugetlb mapping as anon mapping")
  d7e404af115b ("perf record: Mark MAP_HUGETLB when synthesizing mmap events")

Added code conditional on MAP_HUGETLB, to make it build in older systems
where that define wasn't available. Now that we grabbed copies of
uapi/linux/mmap.h to have all those definitions in tools/, use it so
that we can support building the tools for older systems (without the
MAP_HUGETLB define in its libc headers) using new kernels that support
such maps.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/n/tip-wv6oqbfkpxbix4umj2kcfmaz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf trace beauty mmap: Fix defines for non !x86_64
Arnaldo Carvalho de Melo [Mon, 12 Sep 2016 19:27:32 +0000 (16:27 -0300)]
perf trace beauty mmap: Fix defines for non !x86_64

Several defines have different values in different arches, so we can't
just define it to the x86_64 value, use uapi/linux/mmap.h that was
recently introduced to reliably find those, not using possibly outdated
libc headers.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/n/tip-4eajp5yp8i2fuw44n7jmcg5t@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agotools include: Add uapi mman.h for each architecture
Wang Nan [Mon, 12 Sep 2016 12:54:29 +0000 (12:54 +0000)]
tools include: Add uapi mman.h for each architecture

Some mmap related macros have different values for different
architectures. This patch introduces uapi mman.h for each
architectures.

Three headers are cloned from kernel include to tools/include:

 tools/include/uapi/asm-generic/mman-common.h
 tools/include/uapi/asm-generic/mman.h
 tools/include/uapi/linux/mman.h

The main part of this patch is generated by following script:

 macros=`cat $0 | awk 'V==1 {print}; /^# start macro list/ {V=1}'`
 for arch in `ls tools/arch`
 do
   [ -d tools/arch/$arch/include/uapi/asm ] || mkdir -p tools/arch/$arch/include/uapi/asm
   src=arch/$arch/include/uapi/asm/mman.h
   target=tools/arch/$arch/include/uapi/asm/mman.h
   guard="TOOLS_ARCH_"`echo $arch | awk '{print toupper($0)}'`_UAPI_ASM_MMAN_FIX_H
   echo '#ifndef '$guard > $target
   echo '#define '$guard >> $target

   [ -f $src ] &&
   for m in $macros
   do
     if grep '#define[ \t]*'$m $src > /dev/null 2>&1
     then
       grep -h '#define[ \t]*'$m $src | sed 's/[ \t]*\/\*.*$//g' >> $target
     fi
   done

   if [ -f $src ]
   then
      grep '#include <asm-generic' $src >> $target
   else
      echo "#include <asm-generic/mman.h>" >> $target
   fi
   echo '#endif' >> $target
   echo "$target"
 done

 exit 0
 # Following macros are extracted from:
 # tools/perf/trace/beauty/mmap.c
 #
 # start macro list
 MADV_DODUMP
 MADV_DOFORK
 MADV_DONTDUMP
 MADV_DONTFORK
 MADV_DONTNEED
 MADV_HUGEPAGE
 MADV_HWPOISON
 MADV_MERGEABLE
 MADV_NOHUGEPAGE
 MADV_NORMAL
 MADV_RANDOM
 MADV_REMOVE
 MADV_SEQUENTIAL
 MADV_SOFT_OFFLINE
 MADV_UNMERGEABLE
 MADV_WILLNEED
 MAP_32BIT
 MAP_ANONYMOUS
 MAP_DENYWRITE
 MAP_EXECUTABLE
 MAP_FILE
 MAP_FIXED
 MAP_GROWSDOWN
 MAP_HUGETLB
 MAP_LOCKED
 MAP_NONBLOCK
 MAP_NORESERVE
 MAP_POPULATE
 MAP_PRIVATE
 MAP_SHARED
 MAP_STACK
 MAP_UNINITIALIZED
 MREMAP_FIXED
 MREMAP_MAYMOVE
 PROT_EXEC
 PROT_GROWSDOWN
 PROT_GROWSUP
 PROT_NONE
 PROT_READ
 PROT_SEM
 PROT_WRITE

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1473684871-209320-2-git-send-email-wangnan0@huawei.com
[ Added new files to tools/perf/MANIFEST to fix the detached tarball build, add mman.h for ARC ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf hists browser: Fix event group display
Namhyung Kim [Mon, 12 Sep 2016 06:19:52 +0000 (15:19 +0900)]
perf hists browser: Fix event group display

Milian reported that the event group on TUI shows duplicated overhead.
This was due to a bug on calculating hpp->buf position.  The
hpp_advance() was called from __hpp__slsmg_color_printf() on TUI but
it's already called from the hpp__call_print_fn macro in __hpp__fmt().
The end result is that the print function returns number of bytes it
printed but the buffer advanced twice of the length.

This is generally not a problem since it doesn't need to access the
buffer again.  But with event group, overhead needs to be printed
multiple times and hist_entry__snprintf_alignment() tries to fill the
space with buffer after it printed.  So it (brokenly) showed the last
overhead again.

The bug was there from the beginning, but I think it's only revealed
when the alignment function was added.

Reported-by: Milian Wolff <milian.wolff@kdab.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Fixes: 89fee7094323 ("perf hists: Do column alignment on the format iterator")
Link: http://lkml.kernel.org/r/20160912061958.16656-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf probe: Fix dwarf regs table for x86_64
Arnaldo Carvalho de Melo [Fri, 9 Sep 2016 15:00:08 +0000 (12:00 -0300)]
perf probe: Fix dwarf regs table for x86_64

In 293d5b439483 ("perf probe: Support probing on offline cross-arch binary")
DWARF register tables were introduced for many architectures, with the one for
the "dx" register being broken for x86_64, which got noticed by the 'perf test
bpf' testcase, that has this difference from a successful run to one that
fails, with the aforementioned patch:

  -Writing event: p:perf_bpf_probe/func _text+5197232 f_mode=+68(%di):x32 offset=%si:s64 orig=dx:s32
  -Failed to write event: Invalid argument
  -bpf_probe: failed to apply perf probe eventsFailed to add events selected by BPF
  +Writing event: p:perf_bpf_probe/func _text+5197232 f_mode=+68(%di):x32 offset=%si:s64 orig=%dx:s32

Add the missing '%' to '%dx' to fix this.

Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: 293d5b439483 ("perf probe: Support probing on offline cross-arch binary")
Link: https://lkml.kernel.org/r/20160909145955.GC32585@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf/x86/intel/uncore: Add Skylake server uncore support
Kan Liang [Tue, 16 Aug 2016 20:09:50 +0000 (16:09 -0400)]
perf/x86/intel/uncore: Add Skylake server uncore support

This patch implements the uncore monitoring driver for Skylake server.
The uncore subsystem in Skylake server is similar to previous
server. There are some differences in config register encoding and pci
device IDs. Besides, Skylake introduces many new boxes to reflect the
MESH architecture changes.

The control registers for IIO and UPI have been extended to 64 bit. This
patch also introduces event_mask_ext to handle the high 32 bit mask.

The CHA box number could vary for different machines. This patch gets
the CHA box number by counting the CHA register space during
initialization at runtime.

Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1471378190-17276-3-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/rapl: Enable Apollo Lake RAPL support
Harry Pan [Thu, 8 Sep 2016 09:08:57 +0000 (17:08 +0800)]
perf/x86/rapl: Enable Apollo Lake RAPL support

This patch enables RAPL counters (energy consumption counters)
support for Intel Apollo Lake (Goldmont) processors (Model 92):

RAPL of Goldmont, unlikes ESU increment of Silvermont/Airmont,
it likes the Haswell microarchitecture in 1/2^ESU joules and
supports power domains in PP0/PP1/PKG/RAM.

ESU and power domains refer to Intel Software Developers' Manual,
Vol. 3C, Order No. 325384, Table 35-12.

Usage example:

  $ perf list
  $ perf stat -a -e power/energy-cores/,power/energy-pkg/ sleep 10

Signed-off-by: Harry Pan <harry.pan@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: bp@alien8.de
Cc: gs0622@gmail.com
Cc: hpa@zytor.com
Cc: srinivas.pandruvada@linux.intel.com
Link: http://lkml.kernel.org/r/1473325738-730-1-git-send-email-harry.pan@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoMerge branch 'perf/urgent' into perf/core, to pick up fixes
Ingo Molnar [Sat, 10 Sep 2016 09:17:54 +0000 (11:17 +0200)]
Merge branch 'perf/urgent' into perf/core, to pick up fixes

Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/intel: Fix PEBSv3 record drain
Peter Zijlstra [Wed, 7 Sep 2016 12:42:55 +0000 (14:42 +0200)]
perf/x86/intel: Fix PEBSv3 record drain

Alexander hit the WARN_ON_ONCE(!event) on his Skylake while running
the perf fuzzer.

This means the PEBSv3 record included a status bit for an inactive
event, something that _should_ not happen.

Move the code that filters the status bits against our known PEBS
events up a spot to guarantee we only deal with events we know about.

Further add "continue" statements to the WARN_ON_ONCE()s such that
we'll not die nor generate silly events in case we ever do hit them
again.

Reported-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Tested-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vince@deater.net>
Cc: stable@vger.kernel.org
Fixes: a3d86542de88 ("perf/x86/intel/pebs: Add PEBSv3 decoding")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/intel/bts: Kill a silly warning
Alexander Shishkin [Tue, 6 Sep 2016 13:23:53 +0000 (16:23 +0300)]
perf/x86/intel/bts: Kill a silly warning

At the moment, intel_bts will WARN() out if there is more than one
event writing to the same ring buffer, via SET_OUTPUT, and will only
send data from one event to a buffer.

There is no reason to have this warning in, so kill it.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160906132353.19887-6-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/intel/bts: Fix BTS PMI detection
Alexander Shishkin [Tue, 6 Sep 2016 13:23:52 +0000 (16:23 +0300)]
perf/x86/intel/bts: Fix BTS PMI detection

Since BTS doesn't have a dedicated PMI status bit, the driver needs to
take extra care to check for the condition that triggers it to avoid
spurious NMI warnings.

Regardless of the local BTS context state, the only way of knowing that
the NMI is ours is to compare the write pointer against the interrupt
threshold.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160906132353.19887-5-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/intel/bts: Fix confused ordering of PMU callbacks
Alexander Shishkin [Tue, 6 Sep 2016 13:23:51 +0000 (16:23 +0300)]
perf/x86/intel/bts: Fix confused ordering of PMU callbacks

The intel_bts driver is using a CPU-local 'started' variable to order
callbacks and PMIs and make sure that AUX transactions don't get messed
up. However, the ordering rules in regard to this variable is a complete
mess, which recently resulted in perf_fuzzer-triggered warnings and
panics.

The general ordering rule that is patch is enforcing is that this
cpu-local variable be set only when the cpu-local AUX transaction is
active; consequently, this variable is to be checked before the AUX
related bits can be touched.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160906132353.19887-4-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/core: Fix aux_mmap_count vs aux_refcount order
Alexander Shishkin [Tue, 6 Sep 2016 13:23:50 +0000 (16:23 +0300)]
perf/core: Fix aux_mmap_count vs aux_refcount order

The order of accesses to ring buffer's aux_mmap_count and aux_refcount
has to be preserved across the users, namely perf_mmap_close() and
perf_aux_output_begin(), otherwise the inversion can result in the latter
holding the last reference to the aux buffer and subsequently free'ing
it in atomic context, triggering a warning.

> ------------[ cut here ]------------
> WARNING: CPU: 0 PID: 257 at kernel/events/ring_buffer.c:541 __rb_free_aux+0x11a/0x130
> CPU: 0 PID: 257 Comm: stopbug Not tainted 4.8.0-rc1+ #2596
> Call Trace:
>  [<ffffffff810f3e0b>] __warn+0xcb/0xf0
>  [<ffffffff810f3f3d>] warn_slowpath_null+0x1d/0x20
>  [<ffffffff8121182a>] __rb_free_aux+0x11a/0x130
>  [<ffffffff812127a8>] rb_free_aux+0x18/0x20
>  [<ffffffff81212913>] perf_aux_output_begin+0x163/0x1e0
>  [<ffffffff8100c33a>] bts_event_start+0x3a/0xd0
>  [<ffffffff8100c42d>] bts_event_add+0x5d/0x80
>  [<ffffffff81203646>] event_sched_in.isra.104+0xf6/0x2f0
>  [<ffffffff8120652e>] group_sched_in+0x6e/0x190
>  [<ffffffff8120694e>] ctx_sched_in+0x2fe/0x5f0
>  [<ffffffff81206ca0>] perf_event_sched_in+0x60/0x80
>  [<ffffffff81206d1b>] ctx_resched+0x5b/0x90
>  [<ffffffff81207281>] __perf_event_enable+0x1e1/0x240
>  [<ffffffff81200639>] event_function+0xa9/0x180
>  [<ffffffff81202000>] ? perf_cgroup_attach+0x70/0x70
>  [<ffffffff8120203f>] remote_function+0x3f/0x50
>  [<ffffffff811971f3>] flush_smp_call_function_queue+0x83/0x150
>  [<ffffffff81197bd3>] generic_smp_call_function_single_interrupt+0x13/0x60
>  [<ffffffff810a6477>] smp_call_function_single_interrupt+0x27/0x40
>  [<ffffffff81a26ea9>] call_function_single_interrupt+0x89/0x90
>  [<ffffffff81120056>] finish_task_switch+0xa6/0x210
>  [<ffffffff81120017>] ? finish_task_switch+0x67/0x210
>  [<ffffffff81a1e83d>] __schedule+0x3dd/0xb50
>  [<ffffffff81a1efe5>] schedule+0x35/0x80
>  [<ffffffff81128031>] sys_sched_yield+0x61/0x70
>  [<ffffffff81a25be5>] entry_SYSCALL_64_fastpath+0x18/0xa8
> ---[ end trace 6235f556f5ea83a9 ]---

This patch puts the checks in perf_aux_output_begin() in the same order
as that of perf_mmap_close().

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160906132353.19887-3-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/core: Fix a race between mmap_close() and set_output() of AUX events
Alexander Shishkin [Tue, 6 Sep 2016 13:23:49 +0000 (16:23 +0300)]
perf/core: Fix a race between mmap_close() and set_output() of AUX events

In the mmap_close() path we need to stop all the AUX events that are
writing data to the AUX area that we are unmapping, before we can
safely free the pages. To determine if an event needs to be stopped,
we're comparing its ->rb against the one that's getting unmapped.
However, a SET_OUTPUT ioctl may turn up inside an AUX transaction
and swizzle event::rb to some other ring buffer, but the transaction
will keep writing data to the old ring buffer until the event gets
scheduled out. At this point, mmap_close() will skip over such an
event and will proceed to free the AUX area, while it's still being
used by this event, which will set off a warning in the mmap_close()
path and cause a memory corruption.

To avoid this, always stop an AUX event before its ->rb is updated;
this will release the (potentially) last reference on the AUX area
of the buffer. If the event gets restarted, its new ring buffer will
be used. If another SET_OUTPUT comes and switches it back to the
old ring buffer that's getting unmapped, it's also fine: this
ring buffer's aux_mmap_count will be zero and AUX transactions won't
start any more.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20160906132353.19887-2-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/amd/uncore: Prevent use after free
Sebastian Andrzej Siewior [Fri, 9 Sep 2016 16:08:23 +0000 (18:08 +0200)]
perf/x86/amd/uncore: Prevent use after free

The resent conversion of the cpu hotplug support in the uncore driver
introduced a regression due to the way the callbacks are invoked at
initialization time.

The old code called the prepare/starting/online function on each online cpu
as a block. The new code registers the hotplug callbacks in the core for
each state. The core invokes the callbacks at each registration on all
online cpus.

The code implicitely relied on the prepare/starting/online callbacks being
called as combo on a particular cpu, which was not obvious and completely
undocumented.

The resulting subtle wreckage happens due to the way how the uncore code
manages shared data structures for cpus which share an uncore resource in
hardware. The sharing is determined in the cpu starting callback, but the
prepare callback allocates per cpu data for the upcoming cpu because
potential sharing is unknown at this point. If the starting callback finds
a online cpu which shares the hardware resource it takes a refcount on the
percpu data of that cpu and puts the own data structure into a
'free_at_online' pointer of that shared data structure. The online callback
frees that.

With the old model this worked because in a starting callback only one non
unused structure (the one of the starting cpu) was available. The new code
allocates the data structures for all cpus when the prepare callback is
registered.

Now the starting function iterates through all online cpus and looks for a
data structure (skipping its own) which has a matching hardware id. The id
member of the data structure is initialized to 0, but the hardware id can
be 0 as well. The resulting wreckage is:

  CPU0 finds a matching id on CPU1, takes a refcount on CPU1 data and puts
  its own data structure into CPU1s data structure to be freed.

  CPU1 skips CPU0 because the data structure is its allegedly unsued own.
  It finds a matching id on CPU2, takes a refcount on CPU1 data and puts
  its own data structure into CPU2s data structure to be freed.

  ....

Now the online callbacks are invoked.

  CPU0 has a pointer to CPU1s data and frees the original CPU0 data. So
  far so good.

  CPU1 has a pointer to CPU2s data and frees the original CPU1 data, which
  is still referenced by CPU0 ---> Booom

So there are two issues to be solved here:

1) The id field must be initialized at allocation time to a value which
   cannot be a valid hardware id, i.e. -1

   This prevents the above scenario, but now CPU1 and CPU2 both stick their
   own data structure into the free_at_online pointer of CPU0. So we leak
   CPU1s data structure.

2) Fix the memory leak described in #1

   Instead of having a single pointer, use a hlist to enqueue the
   superflous data structures which are then freed by the first cpu
   invoking the online callback.

Ideally we should know the sharing _before_ invoking the prepare callback,
but that's way beyond the scope of this bug fix.

[ tglx: Rewrote changelog ]

Fixes: 96b2bd3866a0 ("perf/x86/amd/uncore: Convert to hotplug state machine")
Reported-and-tested-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Borislav Petkov <bp@suse.de>
Link: http://lkml.kernel.org/r/20160909160822.lowgmkdwms2dheyv@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
8 years agoMerge tag 'perf-core-for-mingo-20160908' of git://git.kernel.org/pub/scm/linux/kernel...
Ingo Molnar [Fri, 9 Sep 2016 05:46:13 +0000 (07:46 +0200)]
Merge tag 'perf-core-for-mingo-20160908' 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:

- Add branch stack / basic block info to 'perf annotate --stdio', where for
  each branch, we add an asm comment after the instruction with information on
  how often it was taken and predicted. See example with color output at:

    http://vger.kernel.org/~acme/perf/annotate_basic_blocks.png

  (Peter Zijlstra)

- Only open an evsel in CPUs in its cpu map, fixing some use cases in
  systems with multiple PMUs with different CPU maps (Mark Rutland)

- Fix handling of huge TLB maps, recognizing it as anonymous (Wang Nan)

Infrastructure changes:

- Remove the symbol filtering code, i.e. the callbacks passed to all functions
  that could end up loading a DSO symtab, simplifying the code, eventually
  allowing what we should have had since day one: removing the 'map' parameter
  from dso__load() functions (Arnaldo Carvalho de Melo)

Arch specific build fixes:

- Fix detached tarball build on powerpc, where we were still accessing a
  file outside tools/ (Ravi Bangoria)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf powerpc: Fix build-test failure
Ravi Bangoria [Wed, 31 Aug 2016 08:03:11 +0000 (13:33 +0530)]
perf powerpc: Fix build-test failure

'make -C tools/perf build-test' is failing with below log for poewrpc.

  In file included from /tmp/tmp.3eEwmGlYaF/perf-4.8.0-rc4/tools/perf/perf.h:15:0,
                   from util/cpumap.h:8,
                   from util/env.c:1:
  /tmp/tmp.3eEwmGlYaF/perf-4.8.0-rc4/tools/perf/perf-sys.h:23:56:
  fatal error: ../../arch/powerpc/include/uapi/asm/unistd.h: No such file or directory
  compilation terminated.

I bisected it and found it's failing from commit ad430729ae00 ("Remove:
kernel unistd*h files from perf's MANIFEST, not used").

Header file '../../arch/powerpc/include/uapi/asm/unistd.h' is included
only for powerpc in tools/perf/perf-sys.h.

By looking closly at commit history, I found little weird thing:

Commit f2d9cae9ea9e ("perf powerpc: Use uapi/unistd.h to fix build
error") replaced 'asm/unistd.h' with 'uapi/asm/unistd.h'

Commit d2709c7ce4c5 ("perf: Make perf build for x86 with UAPI
disintegration applied") removes all arch specific 'uapi/asm/unistd.h'
for all archs and adds generic <asm/unistd.h>.

Commit f0b9abfb0446 ("Merge branch 'linus' into perf/core") again
includes 'uapi/asm/unistd.h' for powerpc. Don't know how exactly this
happened as this change is not part of commit also.

Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1472630591-5089-1-git-send-email-ravi.bangoria@linux.vnet.ibm.com
Fixes: ad430729ae00 ("Remove: kernel unistd*h files from perf's MANIFEST, not used")
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf pmu: Support alternative sysfs cpumask
Mark Rutland [Thu, 8 Sep 2016 10:21:52 +0000 (11:21 +0100)]
perf pmu: Support alternative sysfs cpumask

The perf tools can read a cpumask file for a PMU, describing a subset of
CPUs which that PMU covers. So far this has only been used to cater for
uncore PMUs, which in practice happen to only have a single CPU
described in the mask.

Until recently, the perf tools only correctly handled cpumask containing
a single CPU, and only when monitoring in system-wide mode. For example,
prior to commit 00e727bb389359c8 ("perf stat: Balance opening and
reading events"), a mask with more than a single CPU could cause perf
stat to hang. When a CPU PMU covers a subset of CPUs, but lacks a
cpumask, perf record will fail to open events (on the cores the PMU does
not support), and gives up.

For systems with heterogeneous CPUs such as ARM big.LITTLE systems, this
presents a problem. We have a PMU for each microarchitecture (e.g. a big
PMU and a little PMU), and would like to expose a cpumask for each (so
as to allow perf record and other tools to do the right thing). However,
doing so kernel-side will cause old perf binaries to not function (e.g.
hitting the issue solved by 00e727bb389359c8), and thus commits the
cardinal sin of breaking (existing) userspace.

To address this chicken-and-egg problem, this patch adds support got a
new file, cpus, which is largely identical to the existing cpumask file.
A kernel can expose this file, knowing that new perf binaries will
correctly support it, while old perf binaries will not look for it (and
thus will not be broken).

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Link: http://lkml.kernel.org/r/1473330112-28528-8-git-send-email-mark.rutland@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf evlist: Only open events on CPUs an evsel permits
Mark Rutland [Thu, 8 Sep 2016 10:21:51 +0000 (11:21 +0100)]
perf evlist: Only open events on CPUs an evsel permits

In systems with heterogeneous CPU PMUs, it's possible for each evsel to
cover a distinct set of CPUs, and hence the cpu_map associated with each
evsel may have a distinct idx<->id mapping. Any of these may be distinct
from the evlist's cpu map.

Events can be tied to the same fd so long as they use the same per-cpu
ringbuffer (i.e. so long as they are on the same CPU). To acquire the
correct FDs, we must compare the Linux logical IDs rather than the evsel
or evlist indices.

This path adds logic to perf_evlist__mmap_per_evsel to handle this,
translating IDs as required. As PMUs may cover a subset of CPUs from the
evlist, we skip the CPUs a PMU cannot handle.

Without this patch, perf record may try to mmap erroneous FDs on
heterogeneous systems, and will bail out early rather than running the
workload.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Link: http://lkml.kernel.org/r/1473330112-28528-7-git-send-email-mark.rutland@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf annotate: Add branch stack / basic block
Peter Zijlstra [Mon, 5 Sep 2016 19:08:12 +0000 (16:08 -0300)]
perf annotate: Add branch stack / basic block

I wanted to know the hottest path through a function and figured the
branch-stack (LBR) information should be able to help out with that.

The below uses the branch-stack to create basic blocks and generate
statistics from them.

        from    to              branch_i
        * ----> *
                |
                | block
                v
                * ----> *
                from    to      branch_i+1

The blocks are broken down into non-overlapping ranges, while tracking
if the start of each range is an entry point and/or the end of a range
is a branch.

Each block iterates all ranges it covers (while splitting where required
to exactly match the block) and increments the 'coverage' count.

For the range including the branch we increment the taken counter, as
well as the pred counter if flags.predicted.

Using these number we can find if an instruction:

 - had coverage; given by:

        br->coverage / br->sym->max_coverage

   This metric ensures each symbol has a 100% spot, which reflects the
   observation that each symbol must have a most covered/hottest
   block.

 - is a branch target: br->is_target && br->start == add

 - for targets, how much of a branch's coverages comes from it:

target->entry / branch->coverage

 - is a branch: br->is_branch && br->end == addr

 - for branches, how often it was taken:

        br->taken / br->coverage

   after all, all execution that didn't take the branch would have
   incremented the coverage and continued onward to a later branch.

 - for branches, how often it was predicted:

        br->pred / br->taken

The coverage percentage is used to color the address and asm sections;
for low (<1%) coverage we use NORMAL (uncolored), indicating that these
instructions are not 'important'. For high coverage (>75%) we color the
address RED.

For each branch, we add an asm comment after the instruction with
information on how often it was taken and predicted.

Output looks like (sans color, which does loose a lot of the
information :/)

$ perf record --branch-filter u,any -e cycles:p ./branches 27
$ perf annotate branches

 Percent | Source code & Disassembly of branches for cycles:pu (217 samples)
---------------------------------------------------------------------------------
         : branches():
    0.00 :   40057a:       push   %rbp
    0.00 :   40057b:       mov    %rsp,%rbp
    0.00 :   40057e:       sub    $0x20,%rsp
    0.00 :   400582:       mov    %rdi,-0x18(%rbp)
    0.00 :   400586:       mov    %rsi,-0x20(%rbp)
    0.00 :   40058a:       mov    -0x18(%rbp),%rax
    0.00 :   40058e:       mov    %rax,-0x10(%rbp)
    0.00 :   400592:       movq   $0x0,-0x8(%rbp)
    0.00 :   40059a:       jmpq   400656 <branches+0xdc>
    1.84 :   40059f:       mov    -0x10(%rbp),%rax # +100.00%
    3.23 :   4005a3:       and    $0x1,%eax
    1.84 :   4005a6:       test   %rax,%rax
    0.00 :   4005a9:       je     4005bf <branches+0x45> # -54.50% (p:42.00%)
    0.46 :   4005ab:       mov    0x200bbe(%rip),%rax        # 601170 <acc>
   12.90 :   4005b2:       add    $0x1,%rax
    2.30 :   4005b6:       mov    %rax,0x200bb3(%rip)        # 601170 <acc>
    0.46 :   4005bd:       jmp    4005d1 <branches+0x57> # -100.00% (p:100.00%)
    0.92 :   4005bf:       mov    0x200baa(%rip),%rax        # 601170 <acc> # +49.54%
   13.82 :   4005c6:       sub    $0x1,%rax
    0.46 :   4005ca:       mov    %rax,0x200b9f(%rip)        # 601170 <acc>
    2.30 :   4005d1:       mov    -0x10(%rbp),%rax # +50.46%
    0.46 :   4005d5:       mov    %rax,%rdi
    0.46 :   4005d8:       callq  400526 <lfsr> # -100.00% (p:100.00%)
    0.00 :   4005dd:       mov    %rax,-0x10(%rbp) # +100.00%
    0.92 :   4005e1:       mov    -0x18(%rbp),%rax
    0.00 :   4005e5:       and    $0x1,%eax
    0.00 :   4005e8:       test   %rax,%rax
    0.00 :   4005eb:       je     4005ff <branches+0x85> # -100.00% (p:100.00%)
    0.00 :   4005ed:       mov    0x200b7c(%rip),%rax        # 601170 <acc>
    0.00 :   4005f4:       shr    $0x2,%rax
    0.00 :   4005f8:       mov    %rax,0x200b71(%rip)        # 601170 <acc>
    0.00 :   4005ff:       mov    -0x10(%rbp),%rax # +100.00%
    7.37 :   400603:       and    $0x1,%eax
    3.69 :   400606:       test   %rax,%rax
    0.00 :   400609:       jne    400612 <branches+0x98> # -59.25% (p:42.99%)
    1.84 :   40060b:       mov    $0x1,%eax
   14.29 :   400610:       jmp    400617 <branches+0x9d> # -100.00% (p:100.00%)
    1.38 :   400612:       mov    $0x0,%eax # +57.65%
   10.14 :   400617:       test   %al,%al # +42.35%
    0.00 :   400619:       je     40062f <branches+0xb5> # -57.65% (p:100.00%)
    0.46 :   40061b:       mov    0x200b4e(%rip),%rax        # 601170 <acc>
    2.76 :   400622:       sub    $0x1,%rax
    0.00 :   400626:       mov    %rax,0x200b43(%rip)        # 601170 <acc>
    0.46 :   40062d:       jmp    400641 <branches+0xc7> # -100.00% (p:100.00%)
    0.92 :   40062f:       mov    0x200b3a(%rip),%rax        # 601170 <acc> # +56.13%
    2.30 :   400636:       add    $0x1,%rax
    0.92 :   40063a:       mov    %rax,0x200b2f(%rip)        # 601170 <acc>
    0.92 :   400641:       mov    -0x10(%rbp),%rax # +43.87%
    2.30 :   400645:       mov    %rax,%rdi
    0.00 :   400648:       callq  400526 <lfsr> # -100.00% (p:100.00%)
    0.00 :   40064d:       mov    %rax,-0x10(%rbp) # +100.00%
    1.84 :   400651:       addq   $0x1,-0x8(%rbp)
    0.92 :   400656:       mov    -0x8(%rbp),%rax
    5.07 :   40065a:       cmp    -0x20(%rbp),%rax
    0.00 :   40065e:       jb     40059f <branches+0x25> # -100.00% (p:100.00%)
    0.00 :   400664:       nop
    0.00 :   400665:       leaveq
    0.00 :   400666:       retq

(Note: the --branch-filter u,any was used to avoid spurious target and
branch points due to interrupts/faults, they show up as very small -/+
annotations on 'weird' locations)

Committer note:

Please take a look at:

  http://vger.kernel.org/~acme/perf/annotate_basic_blocks.png

To see the colors.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: David Carrillo-Cisneros <davidcc@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
[ Moved sym->max_coverage to 'struct annotate', aka symbol__annotate(sym) ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf record: Mark MAP_HUGETLB when synthesizing mmap events
Wang Nan [Tue, 6 Sep 2016 04:58:29 +0000 (04:58 +0000)]
perf record: Mark MAP_HUGETLB when synthesizing mmap events

When synthesizing mmap events, add MAP_HUGETLB map flag if the source of
mapping is file in hugetlbfs.

After this patch, perf can identify hugetlb mapping even if perf is
started after the mapping of huge pages (like with 'perf top').

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Reviewed-by: Nilay Vaish <nilayvaish@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Hou Pengyang <houpengyang@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/1473137909-142064-4-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agotools lib api fs: Add hugetlbfs filesystem detector
Wang Nan [Tue, 6 Sep 2016 04:58:28 +0000 (04:58 +0000)]
tools lib api fs: Add hugetlbfs filesystem detector

Detect hugetlbfs. hugetlbfs__mountpoint() will be used during recording
to help identifying hugetlb mmaps: which should be recognized as anon
mapping.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Reviewed-by: Nilay Vaish <nilayvaish@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Hou Pengyang <houpengyang@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/1473137909-142064-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf tools: Recognize hugetlb mapping as anon mapping
Wang Nan [Tue, 6 Sep 2016 04:58:27 +0000 (04:58 +0000)]
perf tools: Recognize hugetlb mapping as anon mapping

Hugetlbfs mapping should be recognized as anon mapping so user has a
chance to create /tmp/perf-<pid>.map file for symbol resolving. This
patch utilizes MAP_HUGETLB to identify hugetlb mapping.

After this patch, if perf is started before a program starts using huge
pages (so perf gets MMAP2 events from kernel), perf is able to recognize
hugetlb mapping as anon mapping.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/1473137909-142064-2-git-send-email-wangnan0@huawei.com
Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf/x86/intel/cqm: Check cqm/mbm enabled state in event init
Jiri Olsa [Mon, 5 Sep 2016 15:30:07 +0000 (17:30 +0200)]
perf/x86/intel/cqm: Check cqm/mbm enabled state in event init

Yanqiu Zhang reported kernel panic when using mbm event
on system where CQM is detected but without mbm event
support, like with perf:

  # perf stat -e 'intel_cqm/event=3/' -a

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000020
  IP: [<ffffffff8100d64c>] update_sample+0xbc/0xe0
  ...
   <IRQ>
   [<ffffffff8100d688>] __intel_mbm_event_init+0x18/0x20
   [<ffffffff81113d6b>] flush_smp_call_function_queue+0x7b/0x160
   [<ffffffff81114853>] generic_smp_call_function_single_interrupt+0x13/0x60
   [<ffffffff81052017>] smp_call_function_interrupt+0x27/0x40
   [<ffffffff816fb06c>] call_function_interrupt+0x8c/0xa0
  ...

The reason is that we currently allow to init mbm event
even if mbm support is not detected.  Adding checks for
both cqm and mbm events and support into cqm's event_init.

Fixes: 33c3cc7acfd9 ("perf/x86/mbm: Add Intel Memory B/W Monitoring enumeration and init")
Reported-by: Yanqiu Zhang <yanqzhan@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/1473089407-21857-1-git-send-email-jolsa@kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
8 years agoperf symbols: Remove symbol_filter_t machinery
Arnaldo Carvalho de Melo [Thu, 1 Sep 2016 22:25:52 +0000 (19:25 -0300)]
perf symbols: Remove symbol_filter_t machinery

We're not using it anymore, few users were, but we really could do
without it, simplify lots of functions by removing it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-1zng8wdznn00iiz08bb7q3vn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf test vmlinux: Remove dead symbol_filter_t code
Arnaldo Carvalho de Melo [Thu, 1 Sep 2016 22:22:02 +0000 (19:22 -0300)]
perf test vmlinux: Remove dead symbol_filter_t code

We don't need to initialize that area as we're not using it afterwards,
leftover, ditch it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-jb2un8buy4rqawz73mcdm1sn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf machine: Remove machine->symbol_filter and friends
Arnaldo Carvalho de Melo [Thu, 1 Sep 2016 21:53:58 +0000 (18:53 -0300)]
perf machine: Remove machine->symbol_filter and friends

Including machines__set_symbol_filter(), not used anymore.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-7o1qgmrpvzuis4a9f0t8mnri@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf top: Remove old kernel-only symbol filter
Arnaldo Carvalho de Melo [Thu, 1 Sep 2016 21:47:15 +0000 (18:47 -0300)]
perf top: Remove old kernel-only symbol filter

Not needed, we already have code to prune aliases.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-1ysyce7qjgui93gi1efbjwhf@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoperf symbols: Mark if a symbol is idle in the library
Arnaldo Carvalho de Melo [Thu, 1 Sep 2016 20:54:31 +0000 (17:54 -0300)]
perf symbols: Mark if a symbol is idle in the library

This was being done just in 'perf top', but grouping idle symbols should
be useful in other places as well, so remove one more symbol_filter_t
user by moving this to the symbol library.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-5r7xitjkzjr9jak1zy3d8u5l@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
8 years agoMerge tag 'perf-core-for-mingo-20160901' of git://git.kernel.org/pub/scm/linux/kernel...
Ingo Molnar [Mon, 5 Sep 2016 13:15:49 +0000 (15:15 +0200)]
Merge tag 'perf-core-for-mingo-20160901' 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:

- Support generating cross arch probes, i.e. if you specify a vmlinux
  file for different arch than the one in the host machine,

 $ perf probe --definition function_name args

  will generate the probe definition string needed to append to the
  target machine /sys/kernel/debug/tracing/kprobes_events file, using
  scripting (Masami Hiramatsu).

- Make 'perf probe' skip the function prologue in uprobes if program
  compiled without optimization, using the same strategy as gdb and
  systemtap uses, fixing a bug where:

$ perf probe -x ./test 'foo i'

  When 'foo(42)' was used on the "./test" executable would produce i=0
  instead of the expected i=42 (Ravi Bangoria)

- Demangle symbols for synthesized @plt entries too (Millian Wolff)

Documentation changes:

- Show default report configuration in 'perf config' example
  and docs (Millian Wolff)

Infrastructure changes:

- Make 'perf test vmlinux' tolerate the symbol aliasing pruning done when
  loading kallsyms and vmlinux (Arnaldo Carvalho de Melo)

- Improve output of 'perf test vmlinux' test, to help identify on the verbose
  output which lines are warning and which are errors (Arnaldo Carvalho de Melo)

- Prep work to stop having to pass symbol_filter_t to lots of functions,
  simplifying symtab loading routines (Arnaldo Carvalho de Melo)

- Honor symbol_conf.allow_aliases when loading kallsyms as well, it was using
  it only when loading vmlinux files (Arnaldo Carvalho de Melo)

- Fixup symbol->end before doing alias pruning when loading symbol tables
  (Arnaldo Carvalho de Melo)

- Fix error handling of lzma kernel module decompression (Shawn Lin)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/core: Don't pass PERF_EF_START to the PMU ->start callback
Will Deacon [Mon, 15 Aug 2016 10:42:45 +0000 (11:42 +0100)]
perf/core: Don't pass PERF_EF_START to the PMU ->start callback

PERF_EF_START is a flag to indicate to the PMU ->add() callback that, as
well as claiming the PMU resources required by the event being added,
it should also start the PMU.

Passing this flag to the ->start() callback doesn't make sense, because
->start() always tries to start the PMU. Remove it.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: mark.rutland@arm.com
Link: http://lkml.kernel.org/r/1471257765-29662-1-git-send-email-will.deacon@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/intel/uncore: Handle non-standard counter offset
Stephane Eranian [Tue, 16 Aug 2016 20:09:49 +0000 (16:09 -0400)]
perf/x86/intel/uncore: Handle non-standard counter offset

The offset of the counters for UPI and M2M boxes on Skylake server is
non-standard (8 bytes apart).

This patch introduces a custom flag UNCORE_BOX_FLAG_CTL_OFFS8 to
specially handle it.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1471378190-17276-2-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/x86/intel/uncore: Remove hard-coded implementation for Node ID mapping location
Kan Liang [Tue, 16 Aug 2016 20:09:48 +0000 (16:09 -0400)]
perf/x86/intel/uncore: Remove hard-coded implementation for Node ID mapping location

The method to build PCI bus to socket mapping is similar among
platforms. However, the PCI location which stores Node ID mapping could
vary between different platforms. For example, the Node ID mapping address
on Skylake server is different from the previous platform. Also, to
build the mapping for the PCI bus without UBOX, it has to start from
bus 0 on Skylake server.

This patch removes the current hardcoded implementation and adds
three parameters for snbep_pci2phy_map_init(). This way the Node ID mapping
address and bus searching direction can be configured according to
different platforms.

Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Nilay Vaish <nilayvaish@gmail.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1471378190-17276-1-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoMerge branch 'perf/urgent' into perf/core, to pick up fixed and resolve conflicts
Ingo Molnar [Mon, 5 Sep 2016 10:09:59 +0000 (12:09 +0200)]
Merge branch 'perf/urgent' into perf/core, to pick up fixed and resolve conflicts

 Conflicts:
kernel/events/core.c

Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoperf/core: Remove WARN from perf_event_read()
Peter Zijlstra [Tue, 30 Aug 2016 08:15:03 +0000 (10:15 +0200)]
perf/core: Remove WARN from perf_event_read()

This effectively reverts commit:

  71e7bc2bab77 ("perf/core: Check return value of the perf_event_read() IPI")

... and puts in a comment explaining why we ignore the return value.

Reported-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David Carrillo-Cisneros <davidcc@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 71e7bc2bab77 ("perf/core: Check return value of the perf_event_read() IPI")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
8 years agoLinux 4.8-rc5
Linus Torvalds [Sun, 4 Sep 2016 21:31:46 +0000 (14:31 -0700)]
Linux 4.8-rc5

8 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Sep 2016 15:45:41 +0000 (08:45 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fix from Thomas Gleixner:
 "A single fix for an AMD erratum so machines without a BIOS fix work"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/AMD: Apply erratum 665 on machines without a BIOS fix

8 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Sep 2016 15:43:45 +0000 (08:43 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Thomas Gleixner:
 "Two fixlet from the timers departement:

   - A fix for scheduler stalls in the tick idle code affecting
     NOHZ_FULL kernels

   - A trivial compile fix"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick/nohz: Fix softlockup on scheduler stalls in kvm guest
  clocksource/drivers/atmel-pit: Fix compilation error

8 years agoMerge tag 'dm-4.8-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Sun, 4 Sep 2016 00:29:58 +0000 (17:29 -0700)]
Merge tag 'dm-4.8-fixes-4' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - a stable fix in both DM crypt and DM log-writes for too large bios
   (as generated by bcache)

 - two other stable fixes for DM log-writes

 - a stable fix for a DM crypt bug that could result in freeing pointers
   from uninitialized memory in the tfm allocation error path

 - a DM bufio cleanup to discontinue using create_singlethread_workqueue()

* tag 'dm-4.8-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm bufio: remove use of deprecated create_singlethread_workqueue()
  dm crypt: fix free of bad values after tfm allocation failure
  dm crypt: fix error with too large bios
  dm log writes: fix check of kthread_run() return value
  dm log writes: fix bug with too large bios
  dm log writes: move IO accounting earlier to fix error path

8 years agoMerge branch 'for-linus-4.8' of git://git.kernel.org/pub/scm/linux/kernel/git/mason...
Linus Torvalds [Sat, 3 Sep 2016 19:40:45 +0000 (12:40 -0700)]
Merge branch 'for-linus-4.8' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "I'm still prepping a set of fixes for btrfs fsync, just nailing down a
  hard to trigger memory corruption.  For now, these are tested and ready."

* 'for-linus-4.8' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: fix one bug that process may endlessly wait for ticket in wait_reserve_ticket()
  Btrfs: fix endless loop in balancing block groups
  Btrfs: kill invalid ASSERT() in process_all_refs()

8 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Sat, 3 Sep 2016 19:31:37 +0000 (12:31 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:
 "arm64 and arm/perf fixes:

   - arm64 fix: debug exception unmasking on the CPU resume path

   - ARM PMU fixes: memory leak on error path and NULL pointer
     dereference"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: kernel: Fix unmasked debug exceptions when restoring mdscr_el1
  drivers/perf: arm_pmu: Fix NULL pointer dereference during probe
  drivers/perf: arm_pmu: Fix leak in error path

8 years agoMerge tag 'char-misc-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 3 Sep 2016 18:38:43 +0000 (11:38 -0700)]
Merge tag 'char-misc-4.8-rc5' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are a number of small driver fixes for 4.8-rc5.

  The largest thing here is deleting an obsolete driver,
  drivers/misc/bh1780gli.c, as the functionality of it was replaced by
  an iio driver a while ago.

  The other fixes are things that have been reported, or reverts of
  broken stuff (the binder change).  All of these changes have been in
  linux-next for a while with no reported issues"

* tag 'char-misc-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  thunderbolt: Don't declare Falcon Ridge unsupported
  thunderbolt: Add support for INTEL_FALCON_RIDGE_2C controller.
  thunderbolt: Fix resume quirk for Falcon Ridge 4C.
  lkdtm: Mark lkdtm_rodata_do_nothing() notrace
  mei: me: disable driver on SPT SPS firmware
  Revert "android: binder: fix dangling pointer comparison"
  drivers/iio/light/Kconfig: SENSORS_BH1780 cleanup
  android: binder: fix dangling pointer comparison
  misc: delete bh1780 driver

8 years agoMerge tag 'driver-core-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 3 Sep 2016 18:36:55 +0000 (11:36 -0700)]
Merge tag 'driver-core-4.8-rc5' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fixes from Greg KH:
 "Here are three small fixes for 4.8-rc5.

  One for sysfs, one for kernfs, and one documentation fix, all for
  reported issues.  All of these have been in linux-next for a while"

* tag 'driver-core-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  sysfs: correctly handle read offset on PREALLOC attrs
  documentation: drivers/core/of: fix name of of_node symlink
  kernfs: don't depend on d_find_any_alias() when generating notifications

8 years agoMerge tag 'staging-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 3 Sep 2016 18:33:33 +0000 (11:33 -0700)]
Merge tag 'staging-4.8-rc5' of git://git./linux/kernel/git/gregkh/staging

Pull staging/IIO driver fixes from Greg KH:
 "Here are a number of small fixes for staging and IIO drivers that
  resolve reported problems.

  Full details are in the shortlog.  All of these have been in
  linux-next with no reported issues"

* tag 'staging-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging: (35 commits)
  arm: dts: rockchip: add reset node for the exist saradc SoCs
  arm64: dts: rockchip: add reset saradc node for rk3368 SoCs
  iio: adc: rockchip_saradc: reset saradc controller before programming it
  iio: accel: kxsd9: Fix raw read return
  iio: adc: ti_am335x_adc: Increase timeout value waiting for ADC sample
  iio: adc: ti_am335x_adc: Protect FIFO1 from concurrent access
  include/linux: fix excess fence.h kernel-doc notation
  staging: wilc1000: correctly check if associatedsta has not been found
  staging: wilc1000: NULL dereference on error
  staging: wilc1000: txq_event: Fix coding error
  MAINTAINERS: Add file patterns for ion device tree bindings
  MAINTAINERS: Update maintainer entry for wilc1000
  iio: chemical: atlas-ph-sensor: fix typo in val assignment
  iio: fix sched WARNING "do not call blocking ops when !TASK_RUNNING"
  staging: comedi: ni_mio_common: fix AO inttrig backwards compatibility
  staging: comedi: dt2811: fix a precedence bug
  staging: comedi: adv_pci1760: Do not return EINVAL for CMDF_ROUND_DOWN.
  staging: comedi: ni_mio_common: fix wrong insn_write handler
  staging: comedi: comedi_test: fix timer race conditions
  staging: comedi: daqboard2000: bug fix board type matching code
  ...

8 years agoMerge tag 'tty-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sat, 3 Sep 2016 18:29:31 +0000 (11:29 -0700)]
Merge tag 'tty-4.8-rc5' of git://git./linux/kernel/git/gregkh/tty

Pull serial driver fixes from Greg KH:
 "Here are some small serial driver fixes for 4.8-rc5.  One fixes an
  oft-reported build issue with the fintek driver, another reverts a
  patch that was causing problems, one fixes a crash, and some new
  device ids were added.

  All of these have been in linux-next for a while"

* tag 'tty-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: 8250: added acces i/o products quad and octal serial cards
  serial: 8250_mid: fix divide error bug if baud rate is 0
  Revert "tty/serial/8250: use mctrl_gpio helpers"
  8250/fintek: rename IRQ_MODE macro

8 years agoMerge tag 'usb-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 3 Sep 2016 18:24:23 +0000 (11:24 -0700)]
Merge tag 'usb-4.8-rc5' of git://git./linux/kernel/git/gregkh/usb

Pull USB/PHY fixes from Greg KH:
 "Here are some USB and PHY driver fixes for 4.8-rc5

  Nothing major, lots of little fixes for reported bugs, and a build fix
  for a missing .h file that the phy drivers needed.  All of these have
  been in linux-next for a while with no reported issues"

* tag 'usb-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (24 commits)
  usb: musb: Fix locking errors for host only mode
  usb: dwc3: gadget: always decrement by 1
  usb: dwc3: debug: fix ep name on trace output
  usb: gadget: udc: core: don't starve DMA resources
  USB: serial: option: add WeTelecom 0x6802 and 0x6803 products
  USB: avoid left shift by -1
  USB: fix typo in wMaxPacketSize validation
  usb: gadget: Add the gserial port checking in gs_start_tx()
  usb: dwc3: gadget: don't rely on jiffies while holding spinlock
  usb: gadget: fsl_qe_udc: signedness bug in qe_get_frame()
  usb: gadget: function: f_rndis: socket buffer may be NULL
  usb: gadget: function: f_eem: socket buffer may be NULL
  usb: renesas_usbhs: gadget: fix return value check in usbhs_mod_gadget_probe()
  usb: dwc2: Add reset control to dwc2
  usb: dwc3: core: allow device to runtime_suspend several times
  usb: dwc3: pci: runtime_resume child device
  USB: serial: option: add WeTelecom WM-D200
  usb: chipidea: udc: don't touch DP when controller is in host mode
  USB: serial: mos7840: fix non-atomic allocation in write path
  USB: serial: mos7720: fix non-atomic allocation in write path
  ...

8 years agodevpts: return NULL pts 'priv' entry for non-devpts nodes
Linus Torvalds [Sat, 3 Sep 2016 18:02:50 +0000 (11:02 -0700)]
devpts: return NULL pts 'priv' entry for non-devpts nodes

In commit 8ead9dd54716 ("devpts: more pty driver interface cleanups") I
made devpts_get_priv() just return the dentry->fs_data directly.  And
because I thought it wouldn't happen, I added a warning if you ever saw
a pts node that wasn't on devpts.

And no, that warning never triggered under any actual real use, but you
can trigger it by creating nonsensical pts nodes by hand.

So just revert the warning, and make devpts_get_priv() return NULL for
that case like it used to.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: stable@vger.kernel.org # 4.6+
Cc: Eric W Biederman" <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 3 Sep 2016 04:05:38 +0000 (21:05 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A collection of fixes for the nvme over fabrics code"

* 'for-linus' of git://git.kernel.dk/linux-block:
  nvme-rdma: Get rid of redundant defines
  nvme-rdma: Get rid of duplicate variable
  nvme: fabrics drivers don't need the nvme-pci driver
  nvme-fabrics: get a reference when reusing a nvme_host structure
  nvme-fabrics: change NQN UUID to big-endian format
  nvme-loop: set sqsize to 0-based value, per spec
  nvme-rdma: fix sqsize/hsqsize per spec
  fabrics: define admin sqsize min default, per spec
  nvmet-rdma: +1 to *queue_size from hsqsize/hrqsize
  nvmet-rdma: Fix use after free
  nvme-rdma: initialize ret to zero to avoid returning garbage

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Fri, 2 Sep 2016 22:33:54 +0000 (15:33 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/linux-security

Pull TPM bugfix from James Morris.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  tpm: invalid self test error message

8 years agotpm: invalid self test error message
Jarkko Sakkinen [Thu, 1 Sep 2016 23:36:58 +0000 (02:36 +0300)]
tpm: invalid self test error message

The driver emits invalid self test error message even though the init
succeeds.

Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Fixes: cae8b441fc20 ("tpm: Factor out common startup code")
Reviewed-by: James Morris <james.l.morris@oracle.com>
Signed-off-by: James Morris <james.l.morris@oracle.com>
8 years agoMerge tag 'acpi-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 2 Sep 2016 22:16:04 +0000 (15:16 -0700)]
Merge tag 'acpi-4.8-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fixes ffrom Rafael Wysocki:
 "Two stable-candidate fixes for the ACPI early device probing code
  added during the 4.4 cycle, one fixing a typo in a stub macro used
  when CONFIG_ACPI is unset and one that prevents sleeping functions
  from being called under a spinlock (Lorenzo Pieralisi)"

* tag 'acpi-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / drivers: replace acpi_probe_lock spinlock with mutex
  ACPI / drivers: fix typo in ACPI_DECLARE_PROBE_ENTRY macro

8 years agoMerge tag 'pm-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Fri, 2 Sep 2016 22:07:41 +0000 (15:07 -0700)]
Merge tag 'pm-4.8-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "This includes a stable-candidate cpufreq-dt driver problem fix and
  annotations of tracepoints in the runtime PM framework.

  Specifics:

   - Fix the definition of the cpufreq-dt driver's machines table
     introduced during the 4.7 cycle that should be NULL-terminated, but
     the termination entry is missing from it (Wei Yongjun).

   - Annotate tracepoints in the runtime PM framework's core so as to
     allow the functions containing them to be called from the idle code
     path without causing RCU to complain about illegal usage (Paul
     McKenney)"

* tag 'pm-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM / runtime: Add _rcuidle suffix to allow rpm_idle() use from idle
  PM / runtime: Add _rcuidle suffix to allow rpm_resume() to be called from idle
  cpufreq: dt: Add terminate entry for of_device_id tables

8 years agoMerge branches 'pm-cpufreq-fixes' and 'pm-core-fixes'
Rafael J. Wysocki [Fri, 2 Sep 2016 20:38:30 +0000 (22:38 +0200)]
Merge branches 'pm-cpufreq-fixes' and 'pm-core-fixes'

* pm-cpufreq-fixes:
  cpufreq: dt: Add terminate entry for of_device_id tables

* pm-core-fixes:
  PM / runtime: Add _rcuidle suffix to allow rpm_idle() use from idle
  PM / runtime: Add _rcuidle suffix to allow rpm_resume() to be called from idle

8 years agoACPI / drivers: replace acpi_probe_lock spinlock with mutex
Lorenzo Pieralisi [Tue, 16 Aug 2016 15:59:53 +0000 (16:59 +0100)]
ACPI / drivers: replace acpi_probe_lock spinlock with mutex

Commit e647b532275b ("ACPI: Add early device probing infrastructure")
introduced code that allows inserting driver specific
struct acpi_probe_entry probe entries into ACPI linker sections
(one per-subsystem, eg irqchip, clocksource) that are then walked
to retrieve the data and function hooks required to probe the
respective kernel components.

Probing for all entries in a section is triggered through
the __acpi_probe_device_table() function, that in turn, according
to the table ID a given probe entry reports parses the table
with the function retrieved from the respective section structures
(ie struct acpi_probe_entry). Owing to the current ACPI table
parsing implementation, the __acpi_probe_device_table() function
has to share global variables with the acpi_match_madt() function, so
in order to guarantee mutual exclusion locking is required
between the two functions.

Current kernel code implements the locking through the acpi_probe_lock
spinlock; this has the side effect of requiring all code called
within the lock (ie struct acpi_probe_entry.probe_{table/subtbl} hooks)
not to sleep.

However, kernel subsystems that make use of the early probing
infrastructure are relying on kernel APIs that may sleep (eg
irq_domain_alloc_fwnode(), among others) in the function calls
pointed at by struct acpi_probe_entry.{probe_table/subtbl} entries
(eg gic_v2_acpi_init()), which is a bug.

Since __acpi_probe_device_table() is called from context
that is allowed to sleep the acpi_probe_lock spinlock can be replaced
with a mutex; this fixes the issue whilst still guaranteeing
mutual exclusion.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Fixes: e647b532275b (ACPI: Add early device probing infrastructure)
Cc: 4.4+ <stable@vger.kernel.org> # 4.4+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
8 years agoACPI / drivers: fix typo in ACPI_DECLARE_PROBE_ENTRY macro
Lorenzo Pieralisi [Tue, 16 Aug 2016 15:59:52 +0000 (16:59 +0100)]
ACPI / drivers: fix typo in ACPI_DECLARE_PROBE_ENTRY macro

When the ACPI_DECLARE_PROBE_ENTRY macro was added in
commit e647b532275b ("ACPI: Add early device probing infrastructure"),
a stub macro adding an unused entry was added for the !CONFIG_ACPI
Kconfig option case to make sure kernel code making use of the
macro did not require to be guarded within CONFIG_ACPI in order to
be compiled.

The stub macro was never used since all kernel code that defines
ACPI_DECLARE_PROBE_ENTRY entries is currently guarded within
CONFIG_ACPI; it contains a typo that should be nonetheless fixed.

Fix the typo in the stub (ie !CONFIG_ACPI) ACPI_DECLARE_PROBE_ENTRY()
macro so that it can actually be used if needed.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Fixes: e647b532275b (ACPI: Add early device probing infrastructure)
Cc: 4.4+ <stable@vger.kernel.org> # 4.4+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
8 years agox86/AMD: Apply erratum 665 on machines without a BIOS fix
Emanuel Czirai [Fri, 2 Sep 2016 05:35:50 +0000 (07:35 +0200)]
x86/AMD: Apply erratum 665 on machines without a BIOS fix

AMD F12h machines have an erratum which can cause DIV/IDIV to behave
unpredictably. The workaround is to set MSRC001_1029[31] but sometimes
there is no BIOS update containing that workaround so let's do it
ourselves unconditionally. It is simple enough.

[ Borislav: Wrote commit message. ]

Signed-off-by: Emanuel Czirai <icanrealizeum@gmail.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Yaowu Xu <yaowu@google.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/20160902053550.18097-1-bp@alien8.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
8 years agox86/paravirt: Do not trace _paravirt_ident_*() functions
Steven Rostedt [Wed, 25 May 2016 17:47:26 +0000 (13:47 -0400)]
x86/paravirt: Do not trace _paravirt_ident_*() functions

Łukasz Daniluk reported that on a RHEL kernel that his machine would lock up
after enabling function tracer. I asked him to bisect the functions within
available_filter_functions, which he did and it came down to three:

  _paravirt_nop(), _paravirt_ident_32() and _paravirt_ident_64()

It was found that this is only an issue when noreplace-paravirt is added
to the kernel command line.

This means that those functions are most likely called within critical
sections of the funtion tracer, and must not be traced.

In newer kenels _paravirt_nop() is defined within gcc asm(), and is no
longer an issue.  But both _paravirt_ident_{32,64}() causes the
following splat when they are traced:

 mm/pgtable-generic.c:33: bad pmd ffff8800d2435150(0000000001d00054)
 mm/pgtable-generic.c:33: bad pmd ffff8800d3624190(0000000001d00070)
 mm/pgtable-generic.c:33: bad pmd ffff8800d36a5110(0000000001d00054)
 mm/pgtable-generic.c:33: bad pmd ffff880118eb1450(0000000001d00054)
 NMI watchdog: BUG: soft lockup - CPU#2 stuck for 22s! [systemd-journal:469]
 Modules linked in: e1000e
 CPU: 2 PID: 469 Comm: systemd-journal Not tainted 4.6.0-rc4-test+ #513
 Hardware name: Hewlett-Packard HP Compaq Pro 6300 SFF/339A, BIOS K01 v02.05 05/07/2012
 task: ffff880118f740c0 ti: ffff8800d4aec000 task.ti: ffff8800d4aec000
 RIP: 0010:[<ffffffff81134148>]  [<ffffffff81134148>] queued_spin_lock_slowpath+0x118/0x1a0
 RSP: 0018:ffff8800d4aefb90  EFLAGS: 00000246
 RAX: 0000000000000000 RBX: 0000000000000000 RCX: ffff88011eb16d40
 RDX: ffffffff82485760 RSI: 000000001f288820 RDI: ffffea0000008030
 RBP: ffff8800d4aefb90 R08: 00000000000c0000 R09: 0000000000000000
 R10: ffffffff821c8e0e R11: 0000000000000000 R12: ffff880000200fb8
 R13: 00007f7a4e3f7000 R14: ffffea000303f600 R15: ffff8800d4b562e0
 FS:  00007f7a4e3d7840(0000) GS:ffff88011eb00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f7a4e3f7000 CR3: 00000000d3e71000 CR4: 00000000001406e0
 Call Trace:
   _raw_spin_lock+0x27/0x30
   handle_pte_fault+0x13db/0x16b0
   handle_mm_fault+0x312/0x670
   __do_page_fault+0x1b1/0x4e0
   do_page_fault+0x22/0x30
   page_fault+0x28/0x30
   __vfs_read+0x28/0xe0
   vfs_read+0x86/0x130
   SyS_read+0x46/0xa0
   entry_SYSCALL_64_fastpath+0x1e/0xa8
 Code: 12 48 c1 ea 0c 83 e8 01 83 e2 30 48 98 48 81 c2 40 6d 01 00 48 03 14 c5 80 6a 5d 82 48 89 0a 8b 41 08 85 c0 75 09 f3 90 8b 41 08 <85> c0 74 f7 4c 8b 09 4d 85 c9 74 08 41 0f 18 09 eb 02 f3 90 8b

Reported-by: Łukasz Daniluk <lukasz.daniluk@intel.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMerge branch 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszer...
Linus Torvalds [Fri, 2 Sep 2016 16:32:15 +0000 (09:32 -0700)]
Merge branch 'overlayfs-linus' of git://git./linux/kernel/git/mszeredi/vfs

Pull overlayfs fixes from Miklos Szeredi:
 "Most of this is regression fixes for posix acl behavior introduced in
  4.8-rc1 (these were caught by the pjd-fstest suite).  The are also
  miscellaneous fixes marked as stable material and cleanups.

  Other than overlayfs code, it touches <linux/fs.h> to add a constant
  with which to disable posix acl caching.  No changes needed to the
  actual caching code, it automatically does the right thing, although
  later we may want to optimize this case.

  I'm now testing overlayfs with the following test suites to catch
  regressions:

   - unionmount-testsuite
   - xfstests
   - pjd-fstest"

* 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs:
  ovl: update doc
  ovl: listxattr: use strnlen()
  ovl: Switch to generic_getxattr
  ovl: copyattr after setting POSIX ACL
  ovl: Switch to generic_removexattr
  ovl: Get rid of ovl_xattr_noacl_handlers array
  ovl: Fix OVL_XATTR_PREFIX
  ovl: fix spelling mistake: "directries" -> "directories"
  ovl: don't cache acl on overlay layer
  ovl: use cached acl on underlying layer
  ovl: proper cleanup of workdir
  ovl: remove posix_acl_default from workdir
  ovl: handle umask and posix_acl_default correctly on creation
  ovl: don't copy up opaqueness

8 years agoarm64: kernel: Fix unmasked debug exceptions when restoring mdscr_el1
James Morse [Fri, 26 Aug 2016 15:03:42 +0000 (16:03 +0100)]
arm64: kernel: Fix unmasked debug exceptions when restoring mdscr_el1

Changes to make the resume from cpu_suspend() code behave more like
secondary boot caused debug exceptions to be unmasked early by
__cpu_setup(). We then go on to restore mdscr_el1 in cpu_do_resume(),
potentially taking break or watch points based on uninitialised registers.

Mask debug exceptions in cpu_do_resume(), which is specific to resume
from cpu_suspend(). Debug exceptions will be restored to their original
state by local_dbg_restore() in cpu_suspend(), which runs after
hw_breakpoint_restore() has re-initialised the other registers.

Reported-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Fixes: cabe1c81ea5b ("arm64: Change cpu_resume() to enable mmu early then access sleep_sp by va")
Cc: <stable@vger.kernel.org> # 4.7+
Signed-off-by: James Morse <james.morse@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
8 years agodrivers/perf: arm_pmu: Fix NULL pointer dereference during probe
Stefan Wahren [Sat, 27 Aug 2016 16:19:50 +0000 (16:19 +0000)]
drivers/perf: arm_pmu: Fix NULL pointer dereference during probe

Patch 7f1d642fbb5c ("drivers/perf: arm-pmu: Fix handling of SPI lacking
interrupt-affinity property") unintended also fixes perf_event support
for bcm2835 which doesn't have PMU interrupts. Unfortunately this change
introduce a NULL pointer dereference on bcm2835, because irq_is_percpu
always expected to be called with a valid IRQ. So fix this regression
by validating the IRQ before.

Tested-by: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Fixes: 7f1d642fbb5c ("drivers/perf: arm-pmu: Fix handling of SPI lacking "interrupt-affinity" property")
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
8 years agodrivers/perf: arm_pmu: Fix leak in error path
Stefan Wahren [Sat, 27 Aug 2016 16:19:49 +0000 (16:19 +0000)]
drivers/perf: arm_pmu: Fix leak in error path

In case of a IRQ type mismatch in of_pmu_irq_cfg() the
device node for interrupt affinity isn't freed. So fix this
issue by calling of_node_put().

Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Fixes: fa8ad7889d83 ("arm: perf: factor arm_pmu core out to drivers")
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
8 years agoMerge tag 'dmaengine-fix-4.8-rc5' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Fri, 2 Sep 2016 14:58:31 +0000 (07:58 -0700)]
Merge tag 'dmaengine-fix-4.8-rc5' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fixes from Vinod Koul:
 "The fixes this time are all in drivers:

   - possible NULL dereference in img-mdc
   - correct device identity for free_irq in at_xdmac
   - missing of_node_put() in fsl probe
   - fix debug log and hotchain corner case for pxa-dma
   - fix checking hardware bits in isr in usb dmac"

* tag 'dmaengine-fix-4.8-rc5' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: img-mdc: fix a possible NULL dereference
  dmaengine: at_xdmac: fix to pass correct device identity to free_irq()
  dmaengine: fsl_raid: add missing of_node_put() in fsl_re_probe()
  dmaengine: pxa_dma: fix debug message
  dmaengine: pxa_dma: fix hotchain corner case
  dmaengine: usb-dmac: check CHCR.DE bit in usb_dmac_isr_channel()

8 years agoMerge tag 'drm-fixes-for-4.8-rc5' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 2 Sep 2016 14:53:00 +0000 (07:53 -0700)]
Merge tag 'drm-fixes-for-4.8-rc5' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Contains fixes for imx, amdgpu, vc4, msm and one nouveau ACPI fix"

* tag 'drm-fixes-for-4.8-rc5' of git://people.freedesktop.org/~airlied/linux:
  drm/amdgpu: record error code when ring test failed
  drm/amd/amdgpu: compute ring test fail during S4 on CI
  drm/amd/amdgpu: sdma resume fail during S4 on CI
  drm/nouveau/acpi: use DSM if bridge does not support D3cold
  drm/imx: fix crtc vblank state regression
  drm/imx: Add active plane reconfiguration support
  drm/msm: protect against faults from copy_from_user() in submit ioctl
  drm/msm: fix use of copy_from_user() while holding spinlock
  drm/vc4: Fix oops when userspace hands in a bad BO.
  drm/vc4: Fix overflow mem unreferencing when the binner runs dry.
  drm/vc4: Free hang state before destroying BO cache.
  drm/vc4: Fix handling of a pm_runtime_get_sync() success case.
  drm/vc4: Use drm_malloc_ab to fix large rendering jobs.
  drm/vc4: Use drm_free_large() on handles to match its allocation.

8 years agotick/nohz: Fix softlockup on scheduler stalls in kvm guest
Wanpeng Li [Fri, 2 Sep 2016 06:38:23 +0000 (14:38 +0800)]
tick/nohz: Fix softlockup on scheduler stalls in kvm guest

tick_nohz_start_idle() is prevented to be called if the idle tick can't
be stopped since commit 1f3b0f8243cb934 ("tick/nohz: Optimize nohz idle
enter"). As a result, after suspend/resume the host machine, full dynticks
kvm guest will softlockup:

 NMI watchdog: BUG: soft lockup - CPU#0 stuck for 26s! [swapper/0:0]
 Call Trace:
  default_idle+0x31/0x1a0
  arch_cpu_idle+0xf/0x20
  default_idle_call+0x2a/0x50
  cpu_startup_entry+0x39b/0x4d0
  rest_init+0x138/0x140
  ? rest_init+0x5/0x140
  start_kernel+0x4c1/0x4ce
  ? set_init_arg+0x55/0x55
  ? early_idt_handler_array+0x120/0x120
  x86_64_start_reservations+0x24/0x26
  x86_64_start_kernel+0x142/0x14f

In addition, cat /proc/stat | grep cpu in guest or host:

cpu  398 16 5049 15754 5490 0 1 46 0 0
cpu0 206 5 450 0 0 0 1 14 0 0
cpu1 81 0 3937 3149 1514 0 0 9 0 0
cpu2 45 6 332 6052 2243 0 0 11 0 0
cpu3 65 2 328 6552 1732 0 0 11 0 0

The idle and iowait states are weird 0 for cpu0(housekeeping).

The bug is present in both guest and host kernels, and they both have
cpu0's idle and iowait states issue, however, host kernel's suspend/resume
path etc will touch watchdog to avoid the softlockup.

- The watchdog will not be touched in tick_nohz_stop_idle path (need be
  touched since the scheduler stall is expected) if idle_active flags are
  not detected.
- The idle and iowait states will not be accounted when exit idle loop
  (resched or interrupt) if idle start time and idle_active flags are
  not set.

This patch fixes it by reverting commit 1f3b0f8243cb934 since can't stop
idle tick doesn't mean can't be idle.

Fixes: 1f3b0f8243cb934 ("tick/nohz: Optimize nohz idle enter")
Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com>
Cc: Sanjeev Yadav<sanjeev.yadav@spreadtrum.com>
Cc: Gaurav Jindal<gaurav.jindal@spreadtrum.com>
Cc: stable@vger.kernel.org
Cc: kvm@vger.kernel.org
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Link: http://lkml.kernel.org/r/1472798303-4154-1-git-send-email-wanpeng.li@hotmail.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
8 years agoMerge tag 'drm-vc4-fixes-2016-08-29' of https://github.com/anholt/linux into drm...
Dave Airlie [Fri, 2 Sep 2016 05:55:15 +0000 (15:55 +1000)]
Merge tag 'drm-vc4-fixes-2016-08-29' of https://github.com/anholt/linux into drm-fixes

This pull request brings in fixes for VC4 3D in 4.8, most of which are
covered by testcases.

* tag 'drm-vc4-fixes-2016-08-29' of https://github.com/anholt/linux:
  drm/vc4: Fix oops when userspace hands in a bad BO.
  drm/vc4: Fix overflow mem unreferencing when the binner runs dry.
  drm/vc4: Free hang state before destroying BO cache.
  drm/vc4: Fix handling of a pm_runtime_get_sync() success case.
  drm/vc4: Use drm_malloc_ab to fix large rendering jobs.
  drm/vc4: Use drm_free_large() on handles to match its allocation.

8 years agoMerge tag 'imx-drm-fixes-2016-08-30' of git://git.pengutronix.de/git/pza/linux into...
Dave Airlie [Fri, 2 Sep 2016 05:48:38 +0000 (15:48 +1000)]
Merge tag 'imx-drm-fixes-2016-08-30' of git://git.pengutronix.de/git/pza/linux into drm-fixes

imx-drm atomic modeset regression fixes

- add active plane reconfiguration support
- add back crtc vblank state reporting

* tag 'imx-drm-fixes-2016-08-30' of git://git.pengutronix.de/git/pza/linux:
  drm/imx: fix crtc vblank state regression
  drm/imx: Add active plane reconfiguration support

8 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 2 Sep 2016 03:32:18 +0000 (20:32 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "A collection of small fixes for various SoC vendor clk drivers"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: rockchip: mark aclk_emmc_noc as a critical clock on rk3399
  clk: tegra: remove TEGRA_PLL_USE_LOCK for PLLD/PLLD2
  clk: rockchip: fix incorrect GATE bits for {c, g}pll_aclk_perihp_src on rk3399
  clk: rockchip: fix incorrect aclk_emmc source gate bits on rk3399
  clk: renesas: r8a7795: Fix SD clocks
  clk: rockchip: fix rk3399 aclk_vio gate bit
  clk: sunxi-ng: Fix inverted test condition in ccu_helper_wait_for_lock

8 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 2 Sep 2016 01:23:22 +0000 (18:23 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge fixes from Andrew Morton:
 "14 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  rapidio/tsi721: fix incorrect detection of address translation condition
  rapidio/documentation/mport_cdev: add missing parameter description
  kernel/fork: fix CLONE_CHILD_CLEARTID regression in nscd
  MAINTAINERS: Vladimir has moved
  mm, mempolicy: task->mempolicy must be NULL before dropping final reference
  printk/nmi: avoid direct printk()-s from __printk_nmi_flush()
  treewide: remove references to the now unnecessary DEFINE_PCI_DEVICE_TABLE
  drivers/scsi/wd719x.c: remove last declaration using DEFINE_PCI_DEVICE_TABLE
  mm, vmscan: only allocate and reclaim from zones with pages managed by the buddy allocator
  lib/test_hash.c: fix warning in preprocessor symbol evaluation
  lib/test_hash.c: fix warning in two-dimensional array init
  kconfig: tinyconfig: provide whole choice blocks to avoid warnings
  kexec: fix double-free when failing to relocate the purgatory
  mm, oom: prevent premature OOM killer invocation for high order request

8 years agorapidio/tsi721: fix incorrect detection of address translation condition
Alexandre Bounine [Thu, 1 Sep 2016 23:15:18 +0000 (16:15 -0700)]
rapidio/tsi721: fix incorrect detection of address translation condition

Fix incorrect condition to identify involvment of a address translation
mechanism.

This bug results in NULL pointer kernel crash dump in cases when mapping
of inbound RapidIO address range is requested within existing aprture.

Link: http://lkml.kernel.org/r/20160901173144.2983-1-alexandre.bounine@idt.com
Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Andre van Herk <andre.van.herk@prodrive-technologies.com>
Cc: Barry Wood <barry.wood@idt.com>
Cc: <stable@vger.kernel.org> [4.6+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agorapidio/documentation/mport_cdev: add missing parameter description
Alexandre Bounine [Thu, 1 Sep 2016 23:15:15 +0000 (16:15 -0700)]
rapidio/documentation/mport_cdev: add missing parameter description

Add missing description for rio_mport_cdev driver parameter
'dma_timeout'.

This patch is applicable to kernel versions starting from v4.6.

Link: http://lkml.kernel.org/r/20160901173104.2928-1-alexandre.bounine@idt.com
Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Andre van Herk <andre.van.herk@prodrive-technologies.com>
Cc: Barry Wood <barry.wood@idt.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agokernel/fork: fix CLONE_CHILD_CLEARTID regression in nscd
Michal Hocko [Thu, 1 Sep 2016 23:15:13 +0000 (16:15 -0700)]
kernel/fork: fix CLONE_CHILD_CLEARTID regression in nscd

Commit fec1d0115240 ("[PATCH] Disable CLONE_CHILD_CLEARTID for abnormal
exit") has caused a subtle regression in nscd which uses
CLONE_CHILD_CLEARTID to clear the nscd_certainly_running flag in the
shared databases, so that the clients are notified when nscd is
restarted.  Now, when nscd uses a non-persistent database, clients that
have it mapped keep thinking the database is being updated by nscd, when
in fact nscd has created a new (anonymous) one (for non-persistent
databases it uses an unlinked file as backend).

The original proposal for the CLONE_CHILD_CLEARTID change claimed
(https://lkml.org/lkml/2006/10/25/233):

: The NPTL library uses the CLONE_CHILD_CLEARTID flag on clone() syscalls
: on behalf of pthread_create() library calls.  This feature is used to
: request that the kernel clear the thread-id in user space (at an address
: provided in the syscall) when the thread disassociates itself from the
: address space, which is done in mm_release().
:
: Unfortunately, when a multi-threaded process incurs a core dump (such as
: from a SIGSEGV), the core-dumping thread sends SIGKILL signals to all of
: the other threads, which then proceed to clear their user-space tids
: before synchronizing in exit_mm() with the start of core dumping.  This
: misrepresents the state of process's address space at the time of the
: SIGSEGV and makes it more difficult for someone to debug NPTL and glibc
: problems (misleading him/her to conclude that the threads had gone away
: before the fault).
:
: The fix below is to simply avoid the CLONE_CHILD_CLEARTID action if a
: core dump has been initiated.

The resulting patch from Roland (https://lkml.org/lkml/2006/10/26/269)
seems to have a larger scope than the original patch asked for.  It
seems that limitting the scope of the check to core dumping should work
for SIGSEGV issue describe above.

[Changelog partly based on Andreas' description]
Fixes: fec1d0115240 ("[PATCH] Disable CLONE_CHILD_CLEARTID for abnormal exit")
Link: http://lkml.kernel.org/r/1471968749-26173-1-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Tested-by: William Preston <wpreston@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@hack.frob.com>
Cc: Andreas Schwab <schwab@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMAINTAINERS: Vladimir has moved
Vladimir Davydov [Thu, 1 Sep 2016 23:15:09 +0000 (16:15 -0700)]
MAINTAINERS: Vladimir has moved

vdavydov@{parallels,virtuozzo}.com will bounce from now on.

Link: http://lkml.kernel.org/r/20160831180752.GB10353@esperanza
Signed-off-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>