0day-ci / linux

…ving to a new bio_vec

In __btrfs_lookup_bio_sums() we set the file offset value at the
beginning of every iteration of the while loop. This is incorrect since
the blocks mapped by the current bvec->bv_page might not yet have been
completely processed.

This commit fixes the issue by setting the file offset value when we
move to the next bvec of the bio.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…ksize aware

The file extent relocation code currently assumes blocksize to be same
as PAGE_CACHE_SIZE. This commit adds code to support subpage blocksize
scenario.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…clone range

After cloning the required extents, we truncate all the pages that map
the file range being cloned. In subpage-blocksize scenario, we could
have dirty blocks before and/or after the clone range in the
leading/trailing pages. Truncating these pages would lead to data
loss. Hence this commit forces such dirty blocks to be flushed to disk
before performing the clone operation.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

The function implementing the dedupe ioctl
i.e. btrfs_ioctl_file_extent_same(), returns with an error in
subpage-blocksize scenario. This was done due to the fact that Btrfs did
not have code to deal with block size < page size. This commit removes
this restriction since we now support "block size < page size".

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…from being unlocked more than once

extent_clear_unlock_delalloc() can unlock a page more than once as shown
below (assume 4k as the block size and 64k as the page size).

cow_file_range
  create 4k ordered extent corresponding to page offsets 0 - 4095
  extent_clear_unlock_delalloc corresponding to page offsets 0 - 4095
    unlock page
  create 4k ordered extent corresponding to page offsets 4096 - 8191
  extent_clear_unlock_delalloc corresponding to page offsets 4096 - 8191
    unlock page

To prevent such a scenario this commit passes "delalloc end" to
extent_clear_unlock_delalloc() to help decide whether the page can be unlocked
or not.

NOTE: Since extent_clear_unlock_delalloc() is used by compression code
as well, the commit passes ordered extent "end" as the value for the
argument corresponding to "delalloc end" for invocations made from
compression code path. This will be fixed by a future commit that gets
compression to work in subpage-blocksize scenario.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

This commit gets file defragmentation code to work in subpage-blocksize
scenario. It does this by keeping track of page offsets that mark block
boundaries and passing them as arguments to the functions that implement
the defragmentation logic.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

The patch "Btrfs: subpage-blocksize: Prevent writes to an extent buffer
when PG_writeback flag is set" requires btrfs_try_tree_write_lock() to
be a true try lock w.r.t to both spinning and blocking locks. During
2015's Vault Conference Btrfs meetup, Chris Mason had suggested that he
will write up a suitable locking function to be used when writing dirty
pages that map metadata blocks. Until we have a suitable locking
function available, this patch temporarily disables the commit
f82c458.

…writeback flag is set

In non-subpage-blocksize scenario, BTRFS_HEADER_FLAG_WRITTEN flag
prevents Btrfs code from writing into an extent buffer whose pages are
under writeback. This facility isn't sufficient for achieving the same
in subpage-blocksize scenario, since we have more than one extent buffer
mapped to a page.

Hence this patch adds a new flag (i.e. EXTENT_BUFFER_HEAD_WRITEBACK) and
corresponding code to track the writeback status of the page and to
prevent writes to any of the extent buffers mapped to the page while
writeback is going on.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

In case of subpage-blocksize, the file blocks to be punched may map only
part of a page. For file blocks inside such pages, we need to check for
the presence of BLK_STATE_UPTODATE flag.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

… ordered extent.

In subpage-blocksize scenario a page can have more than one block. So in
addition to PagePrivate2 flag, we would have to track the I/O status of
each block of a page to reliably mark the ordered extent as complete.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

In subpage-blocksize scenario, extent allocations for only some of the
dirty blocks of a page can succeed, while allocation for rest of the
blocks can fail. This patch allows I/O against such pages to be
submitted.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

… < PAGE_SIZE

This patch allows mounting filesystems with sectorsize smaller than the
PAGE_SIZE.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…o a page

For the subpage-blocksize scenario, this patch adds the ability to write
a single extent buffer to the disk.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…_SIZE

In the case of subpage-blocksize, this patch makes it possible to read
only a single metadata block from the disk instead of all the metadata
blocks that map into a page.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

In order to handle multiple extent buffers per page, first we need to create a
way to handle all the extent buffers that are attached to a page.

This patch creates a new data structure 'struct extent_buffer_head', and moves
fields that are common to all extent buffers from 'struct extent_buffer' to
'struct extent_buffer_head'

Also, this patch moves EXTENT_BUFFER_TREE_REF, EXTENT_BUFFER_DUMMY and
EXTENT_BUFFER_IN_TREE flags from extent_buffer->ebflags  to
extent_buffer_head->bflags.

Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…e locked page's range

find_delalloc_range indirectly depends on EXTENT_UPTODDATE to make sure that
the delalloc range returned intersects with the file range mapped by the
page. Since we now track "uptodate" state in a per-page
bitmap (i.e. in btrfs_page_private->bstate), this commit makes an explicit
check to make sure that the delalloc range starts from within the file range
mapped by the page.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

For the subpage-blocksize scenario, a page can contain multiple
blocks. In such cases, this patch handles writing data to files.

Also, When setting EXTENT_DELALLOC, we no longer set EXTENT_UPTODATE bit on
the extent_io_tree since uptodate status is being tracked by the bitmap
pointed to by page->private.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

For the subpage-blocksize scenario, a page can contain multiple
blocks. In such cases, this patch handles reading data from files.

To track the status of individual blocks of a page, this patch makes use
of a bitmap pointed to by the newly introduced per-page 'struct
btrfs_page_private'.

The per-page btrfs_page_private->io_lock plays the same role as
BH_Uptodate_Lock (see end_buffer_async_read()) i.e. without the io_lock
we may end up in the following situation,

NOTE: Assume 64k page size and 4k block size. Also assume that the first
12 blocks of the page are contiguous while the next 4 blocks are
contiguous. When reading the page we end up submitting two "logical
address space" bios. So end_bio_extent_readpage function is invoked
twice, once for each bio.

|-------------------------+-------------------------+-------------|
| Task A                  | Task B                  | Task C      |
|-------------------------+-------------------------+-------------|
| end_bio_extent_readpage |                         |             |
| process block 0         |                         |             |
| - clear BLK_STATE_IO    |                         |             |
| - page_read_complete    |                         |             |
| process block 1         |                         |             |
|                         |                         |             |
|                         |                         |             |
|                         | end_bio_extent_readpage |             |
|                         | process block 0         |             |
|                         | - clear BLK_STATE_IO    |             |
|                         | - page_read_complete    |             |
|                         | process block 1         |             |
|                         |                         |             |
| process block 11        | process block 3         |             |
| - clear BLK_STATE_IO    | - clear BLK_STATE_IO    |             |
| - page_read_complete    | - page_read_complete    |             |
|   - returns true        |   - returns true        |             |
|   - unlock_page()       |                         |             |
|                         |                         | lock_page() |
|                         |   - unlock_page()       |             |
|-------------------------+-------------------------+-------------|

We end up incorrectly unlocking the page twice and "Task C" ends up
working on an unlocked page. So private->io_lock makes sure that only
one of the tasks gets "true" as the return value when page_io_complete()
is invoked. As an optimization the patch gets the io_lock only when the
last block of the bio_vec is being processed.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

…scm/linux/kernel/git/acme/linux into perf/core

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

User visible changes:

- Print callchains asked for events requested via 'perf trace --event' too:
  (Arnaldo Carvalho de Melo)

  # trace -e nanosleep --call dwarf --event sched:sched_switch/call-graph=fp/ usleep 1
   0.346 (0.005 ms): usleep/24428 nanosleep(rqtp: 0x7fffa15a0540) ...
   0.346 (        ): sched:sched_switch:usleep:24428 [120] S ==> swapper/3:0 [120])
                                    __schedule+0xfe200402 ([kernel.kallsyms])
                                    schedule+0xfe200035 ([kernel.kallsyms])
                                    do_nanosleep+0xfe20006f ([kernel.kallsyms])
                                    hrtimer_nanosleep+0xfe2000dc ([kernel.kallsyms])
                                    sys_nanosleep+0xfe20007a ([kernel.kallsyms])
                                    do_syscall_64+0xfe200062 ([kernel.kallsyms])
                                    return_from_SYSCALL_64+0xfe200000 ([kernel.kallsyms])
                                    __nanosleep+0xffff005b8d602010 (/usr/lib64/libc-2.22.so)
   0.400 (0.059 ms): usleep/24428  ... [continued]: nanosleep()) = 0
                                    __nanosleep+0x10 (/usr/lib64/libc-2.22.so)
                                    usleep+0x34 (/usr/lib64/libc-2.22.so)
                                    main+0x1eb (/usr/bin/usleep)
                                    __libc_start_main+0xf0 (/usr/lib64/libc-2.22.so)
                                    _start+0x29 (/usr/bin/usleep)

- Allow requesting that some CPUs or PIDs be highlighted in 'perf sched map' (Jiri Olsa)

- Compact 'perf sched map' to show just CPUs with activity, improving the output
  in high core count systems (Jiri Olsa)

- Fix segfault with 'perf trace --no-syscalls -e syscall-names' by bailing out
  such request, doesn't make sense to ask for no syscalls and then specify which
  ones should be printed (Arnaldo Carvalho de Melo)

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

Doesn't make sense and was causing a segfault, fix it.

  # trace -e clone --no-syscalls --event sched:*exec firefox
  The -e option can't be used with --no-syscalls.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-ccrahezikdk2uebptzr1eyyi@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Those were converted to be evsel methods long ago, move the
source to where it belongs.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-vja8rjmkw3gd5ungaeyb5s2j@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Introducing --cpus option that will display only given cpus. Could be
used together with color-cpus option.

  $ perf sched map  --cpus 0,1
        *A0   309999.786924 secs A0 => rcu_sched:7
        *.    309999.786930 secs
    *B0  .    309999.786931 secs B0 => rcuos/2:25
     B0 *A0   309999.786947 secs

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-9-git-send-email-jolsa@kernel.org
[ Added entry to man page ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Adding --color-cpus option to display selected cpus with background
color (red by default).  It helps on navigating through the perf sched
map output.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-8-git-send-email-jolsa@kernel.org
[ Added entry to man page ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Adding --color-pids option to display selected pids in color (blue by
default). It helps on navigating through the 'perf sched map' output.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-7-git-send-email-jolsa@kernel.org
[ Added entry to man page ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

It will be used in following patch.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-6-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

As preparation for next patch.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-5-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Add compact map display that does not output the whole cpu matrix, only
cpus that got event.

  $ perf sched map --compact
    *A0   1082427.094098 secs A0 => perf:19404 (CPU 2)
     A0 *.    1082427.094127 secs .  => swapper:0 (CPU 1)
     A0  .  *B0   1082427.094174 secs B0 => rcuos/2:25 (CPU 3)
     A0  .  *.    1082427.094177 secs
    *C0  .   .    1082427.094187 secs C0 => migration/2:21
     C0 *A0  .    1082427.094193 secs
    *.   A0  .    1082427.094195 secs
    *D0  A0  .    1082427.094402 secs D0 => rngd:968
    *.   A0  .    1082427.094406 secs
     .  *E0  .    1082427.095221 secs E0 => kworker/1:1:5333
     .   E0 *F0   1082427.095227 secs F0 => xterm:3342

It helps to display sane output for small thread loads on big cpu
servers.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-4-git-send-email-jolsa@kernel.org
[ Add entry in 'perf sched' man page ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Adding cpu_map__has() to return bool of cpu presence in cpus map.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Adding thread_map__has() to return bool of pid presence in threads map.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460467771-26532-2-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

We already were able to ask for callchains for a specific event:

  # trace -e nanosleep --call dwarf --event sched:sched_switch/call-graph=fp/ usleep 1

This would enable tracing just the "nanosleep" syscall, with callchains
at syscall exit and would ask the kernel for frame pointer callchains to
be enabled for the "sched:sched_switch" tracepoint event, its just that
we were not resolving the callchain and printing it in 'perf trace', do
it:

  # trace -e nanosleep --call dwarf --event sched:sched_switch/call-graph=fp/ usleep 1
     0.425 ( 0.013 ms): usleep/6718 nanosleep(rqtp: 0x7ffcc1d16e20) ...
     0.425 (         ): sched:sched_switch:usleep:6718 [120] S ==> swapper/2:0 [120])
                                       __schedule+0xfe200402 ([kernel.kallsyms])
                                       schedule+0xfe200035 ([kernel.kallsyms])
                                       do_nanosleep+0xfe20006f ([kernel.kallsyms])
                                       hrtimer_nanosleep+0xfe2000dc ([kernel.kallsyms])
                                       sys_nanosleep+0xfe20007a ([kernel.kallsyms])
                                       do_syscall_64+0xfe200062 ([kernel.kallsyms])
                                       return_from_SYSCALL_64+0xfe200000 ([kernel.kallsyms])
                                       __nanosleep+0xffff008b8cbe2010 (/usr/lib64/libc-2.22.so)
     0.486 ( 0.073 ms): usleep/6718  ... [continued]: nanosleep()) = 0
                                       __nanosleep+0x10 (/usr/lib64/libc-2.22.so)
                                       usleep+0x34 (/usr/lib64/libc-2.22.so)
                                       main+0x1eb (/usr/bin/usleep)
                                       __libc_start_main+0xf0 (/usr/lib64/libc-2.22.so)
                                       _start+0x29 (/usr/bin/usleep)
  #

Pretty compact, huh? DWARF callchains for raw_syscalls:sys_exit + frame
pointer callchains for a tracepoint, if your hardware supports LBR, go
wild with /call-graph=lbr/, guess the next step is to lift this from
'perf script':

  -F, --fields <str>    comma separated output fields prepend with 'type:'. Valid types: hw,sw,trace,raw.
                        Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr,symoff,period,iregs,brstack,brstacksym,flags

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-2e7yiv5hqdm8jywlmfivvx2v@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

The new sanity check introduced by:

  2665784 ("perf/core: Verify we have a single perf_hw_context PMU")

... triggered on the AMD uncore driver.

Uncore PMUs are per node, they cannot have per-task counters. Fix it.

Reported-by: Borislav Petkov <bp@suse.de>
Reported-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@redhat.com
Cc: alexander.shishkin@linux.intel.com
Cc: eranian@google.com
Cc: jolsa@redhat.com
Cc: linux-tip-commits@vger.kernel.org
Cc: vincent.weaver@maine.edu
Link: http://lkml.kernel.org/r/20160404140208.GA3448@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

At the moment, initialization path is using test_cpu_cap(&boot_cpu_data),
to detect PT, which is just open coding boot_cpu_has(). Use the latter
instead.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Acked-by: Borislav Petkov <bp@suse.de>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: eranian@google.com
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1459953307-14372-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

The uprobe_xol_ops structures are never modified, so declare them as const.

Done with the help of Coccinelle.

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: kernel-janitors@vger.kernel.org
Link: http://lkml.kernel.org/r/1460200649-32526-1-git-send-email-Julia.Lawall@lip6.fr
Signed-off-by: Ingo Molnar <mingo@kernel.org>

…scm/linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements from Arnaldo Carvalho de Melo:

User visible changes:

- Automagically create a 'bpf-output' event, easing the setup of BPF
  C "scripts" that produce output via the perf ring buffer. Now it is
  just a matter of calling any perf tool, such as 'trace', with a C
  source file that references the __bpf_stdout__ output channel and
  that channel will be created and connected to the script:

  # trace -e nanosleep --event test_bpf_stdout.c usleep 1
    0.013 ( 0.013 ms): usleep/2818 nanosleep(rqtp: 0x7ffcead45f40                                        ) ...
    0.013 (         ): __bpf_stdout__:Raise a BPF event!..)
    0.015 (         ): perf_bpf_probe:func_begin:(ffffffff81112460))
    0.261 (         ): __bpf_stdout__:Raise a BPF event!..)
    0.262 (         ): perf_bpf_probe:func_end:(ffffffff81112460 <- ffffffff81003d92))
    0.264 ( 0.264 ms): usleep/2818  ... [continued]: nanosleep()) = 0
  #

  Further work is needed to reduce the number of lines in a perf bpf C source
  file, this being the part where we greatly reduce the command line setup (Wang Nan)

- 'perf trace' now supports callchains, with 'trace --call-graph dwarf' using
  libunwind, just like 'perf top', to ask the kernel for stack dumps for CFI
  processing. This reduces the overhead by asking just for userspace callchains
  and also only for the syscall exit tracepoint (raw_syscalls:sys_exit)
  (Milian Wolff, Arnaldo Carvalho de Melo)

  Try it with, for instance:

     # perf trace --call dwarf ping 127.0.0.1

  An excerpt of a system wide 'perf trace --call dwarf" session is at:

   https://fedorapeople.org/~acme/perf/perf-trace--call-graph-dwarf--all-cpus.txt

  You may need to bump the number of mmap pages, using -m/--mmap-pages,
  but on a Broadwell machine the defaults allowed system wide tracing to
  work without losing that many records, experiment with just some
  syscalls, like:

    # perf trace --call dwarf -e nanosleep,futex

  All the targets available for 'perf record', 'perf top' (--pid, --tid, --cpu,
  etc) should work. Also --duration may be interesting to try.

  To get filenames from in various syscalls pointer args (open, ettc), add this
  to the mix:

  # perf probe 'vfs_getname=getname_flags:72 pathname=filename:string'

  Making this work is next in line:

     # trace --call dwarf --ev sched:sched_switch/call-graph=fp/ usleep 1

  I.e. honouring per-tracepoint callchains in 'perf trace' in addition to
  in raw_syscalls:sys_exit.

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

…scm/linux/kernel/git/acme/linux into perf/core

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

User visible changes:

- Beautify more syscall arguments in 'perf trace', using the type column in
  tracepoint /format fields to attach, for instance, a pid_t resolver to the
  thread COMM, also attach a mode_t beautifier in the same fashion
  (Arnaldo Carvalho de Melo)

- Build the syscall table id <-> name resolver using the same .tbl file
  used in the kernel to generate headers, to avoid the delay in getting
  new syscalls supported in the audit-libs external dependency, done so
  far only for x86_64 (Arnaldo Carvalho de Melo)

- Improve the documentation of event specifications (Andi Kleen)

- Process update events in 'perf script', fixing up this use case:

    # perf stat -a -I 1000 -e cycles record | perf script -s script.py

- Shared object symbol adjustment fixes, fixing symbol resolution in
  Android (Wang Nan)

Infrastructure changes:

- Add dedicated unwind addr_space member into thread struct, to allow
  tools to use thread->priv, noticed while working on having callchains
  in 'perf trace' (Jiri Olsa)

Build fixes:

- Fix the build in Ubuntu 12.04 (Arnaldo Carvalho de Melo, Vinson Lee)

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