tools build: Add HOSTARCH Makefile variable

For tools that needs to be always compiled with the host headers.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-907q32k2nep6q670dkxypmu6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests kmod-path: Fix build on ubuntu:16.04-x-armhf

Cross building it on Ubuntu 16.04 to ARM ends up showing we get
the free() prototype by luck in other environments, fix it.

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>
Link: http://lkml.kernel.org/n/tip-0ktfgmmyhcfw8ondka2013f3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add AVX-512 instructions to the new instructions test

Previous patches added support for Intel's AVX-512 instructions to the
kernel and perf tools instruction decoders.

AVX-512 instructions are documented in Intel Architecture Instruction
Set Extensions Programming Reference (February 2016).

Add a representative set of instructions to perf's "new instructions"
test. e.g.

perf test "new instructions"

Or to view a particular instruction:

perf test -v "new instructions" 2>&1 | grep vbroadcasti64x4

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: X86 ML <x86@kernel.org>
Link: http://lkml.kernel.org/r/1469003437-32706-5-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add AVX-512 support to the instruction decoder used by Intel PT

Add support for Intel's AVX-512 instructions to perf tools instruction
decoder used by Intel PT.  The kernel's instruction decoder was updated in
a previous patch.

AVX-512 instructions are documented in Intel Architecture Instruction Set
Extensions Programming Reference (February 2016).

AVX-512 instructions are identified by a EVEX prefix which, for the purpose
of instruction decoding, can be treated as though it were a 4-byte VEX
prefix.

Existing instructions which can now accept an EVEX prefix need not be
further annotated in the op code map (x86-opcode-map.txt). In the case of
new instructions, the op code map is updated accordingly.

Also add associated Mask Instructions that are used to manipulate mask
registers used in AVX-512 instructions.

A representative set of instructions is added to the perf tools new
instructions test in a subsequent patch.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: X86 ML <x86@kernel.org>
Link: http://lkml.kernel.org/r/1469003437-32706-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

x86/insn: Add AVX-512 support to the instruction decoder

Add support for Intel's AVX-512 instructions to the instruction decoder.

AVX-512 instructions are documented in Intel Architecture Instruction
Set Extensions Programming Reference (February 2016).

AVX-512 instructions are identified by a EVEX prefix which, for the
purpose of instruction decoding, can be treated as though it were a
4-byte VEX prefix.

Existing instructions which can now accept an EVEX prefix need not be
further annotated in the op code map (x86-opcode-map.txt). In the case
of new instructions, the op code map is updated accordingly.

Also add associated Mask Instructions that are used to manipulate mask
registers used in AVX-512 instructions.

The 'perf tools' instruction decoder is updated in a subsequent patch.
And a representative set of instructions is added to the perf tools new
instructions test in a subsequent patch.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: X86 ML <x86@kernel.org>
Link: http://lkml.kernel.org/r/1469003437-32706-3-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

x86/insn: perf tools: Fix vcvtph2ps instruction decoding

vcvtph2ps does not have an immediate operand, so remove the erroneous
'Ib' from its opcode map entry. Add vcvtph2ps to the perf tools new
instructions test to verify it.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: X86 ML <x86@kernel.org>
Link: http://lkml.kernel.org/r/1469003437-32706-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'perf-core-for-mingo-20160718' 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:

 - Properly report when a function wildcard produces no matches in 'perf probe'
   (Masami Hiramatsu)

 - Balance opening and reading events in 'perf stat', which could cause
   it to get stuck trying to close invalid file descriptors (Mark Rutland)

Infrastructure changes:

 - Copy more headers from the kernel, this time for headers that
   were just including the contents of its kernel counterparts, should
   help resolving the problems with linux-next, where some uapi related
   patches seem to be breaking tools/object/ build. (Arnaldo Carvalho de Melo)

   Some more combing will be done, but at least it is possible to build
   perf out of tree, via a detached tarball (make help | grep perf),
   without including kernel files in its MANIFEST (Arnaldo Carvalho de Melo)

 - Fix smatch found errors that were not causing problems, but are
   mistakes nonetheless (Dan Carpenter)

 - Fix string vs. byte array resolving in the python script code (Jiri Olsa)

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

perf tests: Add is_printable_array test

Add automated test for is_printable_array function.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1468685480-18951-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Make is_printable_array global

It's used from 2 objects in perf, so it's better to keep just one copy.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1468685480-18951-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script python: Fix string vs byte array resolving

Jirka reported that python code returns all arrays as strings.  This
makes impossible to get all items for byte array tracepoint field
containing 0x00 value item.

Fixing this by scanning full length of the array and returning it as
PyByteArray object in case non printable byte is found.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Reported-and-Tested-by: Jiri Pirko <jiri@mellanox.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1468685480-18951-2-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Warn unmatched function filter correctly

Warn unmatched function filter correctly instead of warning
"symbol-loading error", since that can be a filter issue.

From the technical point of view, this adds a filter chech in map__load
and if there is a filter, it returns -2 (filter-out), instead of -1
(error), and perf-probe checks it and change message.

E.g. without this fix:

  # perf probe -F rt_sp*
  no symbols found in [kernel.kallsyms], maybe install a debug package?
  Failed to load symbols in kernel

With this fix:

  # perf probe -F rt_sp*
  no symbols passed the given filter.
  Failed to find symbols matched to "rt_sp*"
    Error: Failed to show functions.

Reported-and-Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146885835596.16106.2293540792775552481.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cpu_map: Add more helpers

In some cases it's necessry to figure out the map-local index of a given
Linux logical CPU ID. Add a new helper, cpu_map__idx, to acquire this.
As the logic is largely the same as the existing cpu_map__has, this is
rewritten in terms of the new helper.

At the same time, add the inverse operation, cpu_map__cpu, which yields
the logical CPU id for a map-local index. While this can be performed
manually, wrapping this in a helper can make code more legible.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1468577293-19667-3-git-send-email-mark.rutland@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf stat: Balance opening and reading events

In create_perf_stat_counter, when a target CPU has not been provided, we
call __perf_evsel__open with empty_cpu_map, and open a single FD per
thread. However, in read_counter we assume that we opened events for the
product of threads and CPUs described in the evsel's cpu_map.

Thus, if an evsel has a cpu_map with more than one entry, we will
attempt to access FDs that we didn't open. This could result in a number
of problems (e.g. blocking while reading from STDIN if the fd memory
happened to be initialised to zero).

This is problematic for systems were a logical CPU PMU covers some
arbitrary subset of CPUs. The cpu_map of any evsel for that PMU will be
initialised based on the cpumask exposed through sysfs, even if the user
requests per-thread events.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1468577293-19667-2-git-send-email-mark.rutland@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Copy linux/{hash,poison}.h and check for drift

We were also using this directly from the kernel sources, the two last
cases, fix it.

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>
Link: http://lkml.kernel.org/n/tip-7o14xvacqcjc5llc7gvjjyl8@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove include/linux/list.h from perf's MANIFEST

It hasn't been used since we made tools/ self sufficiente wrt list.h.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: d1b39d41ebec ("tools: Make list.h self-sufficient")
Link: http://lkml.kernel.org/n/tip-w20ueqlf22kh7ctjqo0zjpig@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Copy the bitops files accessed from the kernel and check for drift

copy some more kernel files accessed from tools/, check for drift.

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>
Link: http://lkml.kernel.org/n/tip-omz8xdyvvxgjiuqzwj6ecm6j@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Remove: kernel unistd*h files from perf's MANIFEST, not used

No need to copy it to a detached tarball as they aren't used anymore

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>
Link: http://lkml.kernel.org/n/tip-lopmaqi439ke10g1j9cxrxwt@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove tools/perf/util/include/linux/const.h

Not used anymore, remove one more file referencing kernel sources, i.e.
outside of tools/

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>
Link: http://lkml.kernel.org/n/tip-ykfjt3t8l0npxfwmekiwwyu6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove tools/perf/util/include/asm/byteorder.h

Not used anymore. This also stops include linux/swab.h directly
from the kernel sources, remove that reference from the MANIFEST.

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

perf tools: Add missing linux/compiler.h include to perf-sys.h

It uses the likely/unlikely macros, so need to include
<linux/compiler.h>.

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>
Link: http://lkml.kernel.org/n/tip-p0xrhgbkicsii9ohmhhprqpi@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf jit: Remove some no-op error handling

The 'info.e_machine' struct member is an uint16_t so 'm' is never less
than zero.  It looks like this was maybe left over code from earlier
versions so I've just removed it.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: kernel-janitors@vger.kernel.org
Link: http://lkml.kernel.org/r/20160715210836.GB19522@mwanda
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf jit: Add missing curly braces

It doesn't change the runtime behavior, but my static checker complains
that curly braces were intended.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: kernel-janitors@vger.kernel.org
Link: http://lkml.kernel.org/r/20160715210712.GA19522@mwanda
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'perf-core-for-mingo-20160715' 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:

 - Allow reading from a backward ring buffer (one setup via sys_perf_event_open()
   with perf_event_attr.write_backward = 1) (Wang Nan)

Infrastructure changes:

 - Fix the build on Android NDK r12b (initially just for ARM), that is now port
   of my perf-build container collection and will get tested prior to sending
   patches upstream (Arnaldo Carvalho de Melo)

 - Add correct header for IPv6 definitions

 - Fix bitsperlong.h fallout (Arnaldo Carvalho de Melo, Peter Zijlstra)

 - Use base 0 (auto) in filename__read_ull(), so that we can handle hex values too (Jiri Olsa)

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

objtool: Initialize variable to silence old compiler

gcc version 4.1.2 20080704 (Red Hat 4.1.2-55) barfs with:

    CC       /tmp/build/objtool/builtin-check.o
  cc1: warnings being treated as errors
  builtin-check.c: In function 'cmd_check':
  builtin-check.c:667: warning: 'prev_rela' may be used uninitialized in this function
  mv: cannot stat `/tmp/build/objtool/.builtin-check.o.tmp': No such file or directory
  make[1]: *** [/tmp/build/objtool/builtin-check.o] Error 1

Init it to NULL to silence it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-qolo31rl2ojlwj1lj9dhemyz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

objtool: Add -I$(srctree)/tools/arch/$(ARCH)/include/uapi

So that it can find asm/bitsperlong.h to get the __BITS_PER_LONG
definition.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-pr3pvskh65pey4po7t122z4j@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Add --tail-synthesize option

When working with overwritable ring buffer there's a inconvenience
problem: if perf dumps data after a long period after it starts,
non-sample events may lost, which makes following 'perf report' unable
to identify proc name and mmap layout. For example:

 # perf record -m 4 -e raw_syscalls:* -g --overwrite --switch-output \
        dd if=/dev/zero of=/dev/null

send SIGUSR2 after dd runs long enough. The resuling perf.data lost
correct comm and mmap events:

 # perf script -i perf.data.2016061522374354
 perf 24478 [004] 2581325.601789:  raw_syscalls:sys_exit: NR 0 = 512
 ^^^^
 Should be 'dd'
                   27b2e8 syscall_slow_exit_work+0xfe2000e3 (/lib/modules/4.6.0-rc3+/build/vmlinux)
                   203cc7 do_syscall_64+0xfe200117 (/lib/modules/4.6.0-rc3+/build/vmlinux)
                   b18d83 return_from_SYSCALL_64+0xfe200000 (/lib/modules/4.6.0-rc3+/build/vmlinux)
             7f47c417edf0 [unknown] ([unknown])
             ^^^^^^^^^^^^
             Fail to unwind

This patch provides a '--tail-synthesize' option, allows perf to collect
system status when finalizing output file. In resuling output file, the
non-sample events reflect system status when dumping data.

After this patch:
 # perf record -m 4 -e raw_syscalls:* -g --overwrite --switch-output --tail-synthesize \
        dd if=/dev/zero of=/dev/null

 # perf script -i perf.data.2016061600544998
 dd 27364 [004] 2583244.994464: raw_syscalls:sys_enter: NR 1 (1, ...
 ^^
 Correct comm
                   203a18 syscall_trace_enter_phase2+0xfe2001a8 ([kernel.kallsyms])
                   203aa5 syscall_trace_enter+0xfe200055 ([kernel.kallsyms])
                   203caa do_syscall_64+0xfe2000fa ([kernel.kallsyms])
                   b18d83 return_from_SYSCALL_64+0xfe200000 ([kernel.kallsyms])
                    d8e50 __GI___libc_write+0xffff01d9639f4010 (/tmp/oxygen_root-w00229757/lib64/libc-2.18.so)
                    ^^^^^
                    Correct unwind

This option doesn't aim to solve this problem completely. If a process
terminates before SIGUSR2, we still lost its COMM and MMAP events. For
example, we can't unwind correctly from the final perf.data we get from
the previous example, because when perf collects the final output file
(when we press C-c), 'dd' has been terminated so its '/proc/<pid>/mmap'
becomes empty.

However, this is a cheaper choice. To completely solve this problem we
need to continously output non-sample events. To satisify the
requirement of daemonization, we need to merge them periodically. It is
possible but requires much more code and cycles.

Automatically select --tail-synthesize when --overwrite is provided.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-16-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf session: Don't warn about out of order event if write_backward is used

If write_backward attribute is set, records are written into kernel
ring buffer from end to beginning, but read from beginning to end.
To avoid 'XX out of order events recorded' warning message (timestamps
of records is in reverse order when using write_backward), suppress the
warning message if write_backward is selected by at lease one event.

Result:

Before this patch:
  # perf record -m 1 -e raw_syscalls:sys_exit/overwrite/ \
                     -e raw_syscalls:sys_enter \
                     dd if=/dev/zero of=/dev/null count=300
  300+0 records in
  300+0 records out
  153600 bytes (154 kB) copied, 0.000601617 s, 255 MB/s
  [ perf record: Woken up 5 times to write data ]
  Warning:
  40 out of order events recorded.
  [ perf record: Captured and wrote 0.096 MB perf.data (696 samples) ]

After this patch:
  # perf record -m 1 -e raw_syscalls:sys_exit/overwrite/ \
                     -e raw_syscalls:sys_enter \
                     dd if=/dev/zero of=/dev/null count=300
  300+0 records in
  300+0 records out
  153600 bytes (154 kB) copied, 0.000644873 s, 238 MB/s
  [ perf record: Woken up 5 times to write data ]
  [ perf record: Captured and wrote 0.096 MB perf.data (696 samples) ]

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-15-git-send-email-wangnan0@huawei.com
Signed-off-by: He Kuang <hekuang@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Enable overwrite settings

This patch allows following config terms and option:

Globally setting events to overwrite;

  # perf record --overwrite ...

Set specific events to be overwrite or no-overwrite.

  # perf record --event cycles/overwrite/ ...
  # perf record --event cycles/no-overwrite/ ...

Add missing config terms and update the config term array size because
the longest string length has changed.

For overwritable events, it automatically selects attr.write_backward
since perf requires it to be backward for reading.

Test result:

  # perf record --overwrite -e syscalls:*enter_nanosleep* usleep 1
  [ perf record: Woken up 2 times to write data ]
  [ perf record: Captured and wrote 0.011 MB perf.data (1 samples) ]
  # perf evlist -v
  syscalls:sys_enter_nanosleep: type: 2, size: 112, config: 0x134, { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|CPU|PERIOD|RAW, disabled: 1, inherit: 1, mmap: 1, comm: 1, enable_on_exec: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, write_backward: 1
  # Tip: use 'perf evlist --trace-fields' to show fields for tracepoint events

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-14-git-send-email-wangnan0@huawei.com
Signed-off-by: He Kuang <hekuang@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Make {pause,resume} internal helpers

There's no user of these two function outside evlist.c. Remove them from
public namespace.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-13-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Read from overwritable ring buffer

Drive the evlist->bkw_mmap_state state machine during draining and when
SIGUSR2 is received. Read the backward ring buffer in record__mmap_read_all.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-12-git-send-email-wangnan0@huawei.com
Signed-off-by: He Kuang <hekuang@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Setup backward mmap state machine

Introduce a bkw_mmap_state state machine to evlist:

                     .________________(forbid)_____________.
                     |                                     V
 NOTREADY --(0)--> RUNNING --(1)--> DATA_PENDING --(2)--> EMPTY
                     ^  ^              |   ^               |
                     |  |__(forbid)____/   |___(forbid)___/|
                     |                                     |
                      \_________________(3)_______________/

 NOTREADY     : Backward ring buffers are not ready
 RUNNING      : Backward ring buffers are recording
 DATA_PENDING : We are required to collect data from backward ring buffers
 EMPTY        : We have collected data from backward ring buffers.

 (0): Setup backward ring buffer
 (1): Pause ring buffers for reading
 (2): Read from ring buffers
 (3): Resume ring buffers for recording

We can't avoid this complexity. Since we deliberately drop records from
overwritable ring buffer, there's no way for us to check remaining from
ring buffer itself (by checking head and old pointers). Therefore, we
need DATA_PENDING and EMPTY state to help us recording what we have done
to the ring buffer.

In record__mmap_read_evlist(), drive this state machine from DATA_PENDING
to EMPTY.

In perf_evlist__mmap_per_evsel(), drive this state machine from NOTREADY
to RUNNING when creating backward mmap.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-11-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Drop evlist->backward

Now there's no real user of evlist->backward. Drop it. We are going to
use evlist->backward_mmap as a container for backward ring buffer.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-10-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Map backward events to backward_mmap

In perf_evlist__mmap_per_evsel(), select backward_mmap for backward
events.  Utilize new perf_mmap APIs. Dynamically alloc backward_mmap.

Remove useless functions.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-9-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Introduce backward_mmap array for evlist

Add backward_mmap to evlist, free it together with normal mmap.

Improve perf_evlist__pick_pc(), search backward_mmap if evlist->mmap is
not available.

This patch doesn't alloc this array. It will be allocated conditionally
in the following commits.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-8-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Extract common code in mmap failure processing

In perf_evlist__mmap_per_cpu() and perf_evlist__mmap_per_thread(), in
case of mmap failure, successfully created maps should be cleared.

Current code uses two loops calling __perf_evlist__munmap() for each
function.

This patch extracts common code to perf_evlist__munmap_nofree() and use
previous introduced decoupled API perf_mmap__munmap(). Now
__perf_evlist__munmap() can be removed because of no user.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-7-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Record mmap cookie into fdarray private field

Insetad of saving a index into fdarray entries private field, save the
corresponding 'struct perf_mmap' pointer, and release them directly
using perf_mmap__put().

Following commits introduce multiple mmap arrays to evlist. Without this
patch, perf_evlist__munmap_filtered() is unable to retrive correct
'struct perf_mmap' pointer.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-6-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Decouple record__mmap_read() and evlist.

Perf evlist will have multiple mmap arrays. Update record__mmap_read():
it should read from 'struct perf_mmap' directly.

Also, make record__mmap_read() ready to read from backward ring buffer.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-5-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Update mmap related APIs and helpers

Currently, the evlist mmap related helpers and APIs accept evlist and
idx, and dereference 'struct perf_mmap' by evlist->mmap[idx]. This is
unnecessary, and force each evlist contains only one mmap array.

Following commits are going to introduce multiple mmap arrays to a
evlist.  This patch refators these APIs and helpers, introduces
functions accept perf_mmap pointer directly. New helpers and APIs are
decoupled with perf_evlist, and become perf_mmap functions (so they have
perf_mmap prefix).

Old functions are reimplemented with new functions. Some of them will be
removed in following commits.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-4-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib fd array: Allow associating a pointer cookie with each entry

Add a 'ptr' field to fdarray->priv array.

This feature will be used by following commits, which introduce
muiltiple 'struct perf_mmap' arrays for different types of mapping.

Because of this, during fdarray__filter(), a simple 'idx' is not enough.

Add a pointer cookie that allows to directly associate a 'struct
perf_mmap' pointer to an fdarray entry.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Simplify BITS_PER_LONG define

Do it using (__CHAR_BIT__ * __SIZEOF_LONG__), simpler, works everywhere,
reduces the complexity by ditching CONFIG_64BIT, that was being
synthesized from yet another set of defines, which proved fragile,
breaking the build on linux-next for no obvious reasons.

Committer Note:

Except on:

gcc version 4.1.2 20080704 (Red Hat 4.1.2-55)

Fallback to __WORDSIZE in that case...

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20160715072243.GP30154@twins.programming.kicks-ass.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Drop redundant evsel->overwrite indicator

evsel->overwrite indicator means an event should be put into
overwritable ring buffer. In current implementation, it equals to
evsel->attr.write_backward. To reduce compliexity, remove
evsel->overwrite, use evsel->attr.write_backward instead.

In addition, in __perf_evsel__open(), if kernel doesn't support
write_backward and user explicitly set it in evsel, don't fallback
like other missing feature, since it is meaningless to fall back to
a forward ring buffer in this case: we are unable to stably read
from an forward overwritable ring buffer.

Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nilay Vaish <nilayvaish@gmail.com>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468485287-33422-2-git-send-email-wangnan0@huawei.com
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib api fs: Use base 0 in filename__read_ull

By using 0 for base, the strtoull() detects the base automatically (see
'man strtoull').

ATM we have just one user of this function, the cpu__get_max_freq
function reading the "cpuinfo_max_freq" sysfs file. It should not get
affected by this change.

Committer note:

This change seems motivated by this discussion:

"[PATCH] [RFC V1]s390/perf: fix 'start' address of module's map"
http://lkml.kernel.org/r/20160711120155.GA29929@krava

I.e. this patches paves the way for filename__read_ull() to be used in a
S/390 related fix.

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>
Cc: Songshan Gong <gongss@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1468567797-27564-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Bail out at "--sort dcacheline" and cacheline_size not known

There are cases where further work would be needed to overcome the fact
that neither sysconf(_SC_LEVEL1_DCACHE_LINESIZE) nor
/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size are
available in some systems (Android, for instance), so bail out when such
a situation takes place.

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>
Link: http://lkml.kernel.org/n/tip-ho8d8g8mh0o2dri7ckcccafi@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Just pr_debug() about not being able to read cacheline_size

So far the cacheline_size is only useful for the "dcacheline" --sort
order, i.e. if that is not used, which is the norm, then the user
shouldn't care that he is running this, say, on an Android system where
sysconf(_SC_LEVEL1_DCACHE_LINESIZE) and the
/sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size sysfs file
isn't available.

An upcoming patch will emit an warning only for "--sort ...,dcacheline,...".

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Chris Phlipot <cphlipot0@gmail.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>
Link: http://lkml.kernel.org/n/tip-580cnkvftunyvt9n7unsholi@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Make "__always_inline" just "inline" on Android

As the gcc there is producing tons of:

  "warning: always_inline function might not be inlinable"

At least on android-ndk-r12/platforms/android-24/arch-arm, so, for the
time being, use this big hammer.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Chris Phlipot <cphlipot0@gmail.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>
Link: http://lkml.kernel.org/n/tip-97l3eg3fnk5shmo4rsyyvj2t@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Do not provide dup sched_getcpu() prototype on Android

The Bionic libc has this definition, so don't duplicate it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Chris Phlipot <cphlipot0@gmail.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>
Link: http://lkml.kernel.org/n/tip-rmd19832zkt07e4crdzyen9z@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Add correct header for ipv6 definitions

We need to include netinet/in.h to get the in6_addr struct definition, needed to
build it on the Android NDK:

  In file included from event-parse.c:36:0:
  /home/acme/android/android-ndk-r12/platforms/android-24/arch-arm/usr/include/netinet/ip6.h:82:18: error: field 'ip6_src' has incomplete type
    struct in6_addr ip6_src; /* source address */

And it is the canonical way of getting IPv6 definitions, as described,
for instance, in Linux's 'man ipv6'

Doing that uncovers another problem: this source file uses PRIu64 but
doesn't include it, depending on it being included by chance via the now
replaced header (netinet/ip6.h), fix it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Chris Phlipot <cphlipot0@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-tilr31n3yaba1whsd47qlwa3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'perf-core-for-mingo-20160713' 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:

- Finish merging initial SDT (Statically Defined Traces) support, see
  cset comments for details about how it all works (Masami Hiramatsu)

- Support attaching eBPF programs to tracepoints (Wang Nan)

Infrastructure changes:

- Fix up BITS_PER_LONG setting (Arnaldo Carvalho de Melo)

- Add fallback from ELF_C_READ_MMAP to ELF_C_READ in objtool, fixing
  the build in libelf implementations lacking that elf_begin() cmd,
  such as Alpine Linux's (Arnaldo Carvalho de Melo)

- Avoid checking code drift on busybox's diff in objtool (Arnaldo Carvalho de Melo)

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

perf test: Add a test case for SDT event

Add a basic test case for SDT event support.  This test scans an SDT
event in perftools and check whether the SDT event is correctly stored
into the buildid cache.

Here is an example:

  ----
  $ perf test sdt -v
  47: Test SDT event probing                                   :
  --- start ---
  test child forked, pid 20732
  Found 72 SDTs in /home/mhiramat/ksrc/linux/tools/perf/perf
  Writing cache: %sdt_perf:test_target=test_target
  Cache committed: 0
  symbol:test_target file:(null) line:0 offset:0 return:0 lazy:(null)
  test child finished with 0
  ---- end ----
  Test SDT event probing: Ok
  ----

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831796546.17065.1502584370844087537.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf build: Add sdt feature detection

This checks whether sys/sdt.h is available or not, which is required for
DTRACE_PROBE().

We can disable this feature by passing NO_SDT=1 when building.

This flag will be used for SDT test case and further SDT events in
perftools.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831795615.17065.17513820540591053933.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Support a special SDT probe format

Support a special SDT probe format which can omit the '%' prefix only if
the SDT group name starts with "sdt_". So, for example both of
"%sdt_libc:setjump" and "sdt_libc:setjump" are acceptable for perf probe
--add.

E.g. without this:

  # perf probe -a sdt_libc:setjmp
  Semantic error :There is non-digit char in line number.
  ...

With this:

  # perf probe -a sdt_libc:setjmp
  Added new event:
    sdt_libc:setjmp      (on %setjmp in /usr/lib64/libc-2.20.so)

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

   perf record -e sdt_libc:setjmp -aR sleep 1

Suggested-by: Brendan Gregg <brendan.d.gregg@gmail.com>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831794674.17065.13359473252168740430.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Support @BUILDID or @FILE suffix for SDT events

Support @BUILDID or @FILE suffix for SDT events. This allows perf to add
probes on SDTs/pre-cached events on given FILE or the file which has
given BUILDID (also, this complements BUILDID.)

For example, both gcc and libstdc++ has same SDTs as below.  If you
would like to add a probe on sdt_libstdcxx:catch on gcc, you can do as
below.

  ----
  # perf list sdt | tail -n 6
    sdt_libstdcxx:catch@/usr/bin/gcc(0cc207fc4b27)     [SDT event]
    sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49)
    sdt_libstdcxx:rethrow@/usr/bin/gcc(0cc207fc4b27)   [SDT event]
    sdt_libstdcxx:rethrow@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49)
    sdt_libstdcxx:throw@/usr/bin/gcc(0cc207fc4b27)     [SDT event]
    sdt_libstdcxx:throw@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49)
  # perf probe -a %sdt_libstdcxx:catch@0cc
  Added new event:
    sdt_libstdcxx:catch  (on %catch in /usr/bin/gcc)

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

   perf record -e sdt_libstdcxx:catch -aR sleep 1
  ----

Committer note:

Doing the full sequence of steps to get the results above:

With a clean build-id cache:

  [root@jouet ~]# rm -rf ~/.debug/
  [root@jouet ~]# perf list sdt

  List of pre-defined events (to be used in -e):

  [root@jouet ~]#

No events whatsoever, then, we can add all events in gcc to the build-id
cache, doing a --add + --dry-run:

  [root@jouet ~]# perf probe --dry-run --cache -x /usr/bin/gcc --add %sdt_libstdcxx:\*
  Added new events:
    sdt_libstdcxx:throw  (on %* in /usr/bin/gcc)
    sdt_libstdcxx:rethrow (on %* in /usr/bin/gcc)
    sdt_libstdcxx:catch  (on %* in /usr/bin/gcc)

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

perf record -e sdt_libstdcxx:catch -aR sleep 1

  [root@jouet ~]#

It really didn't add any events, it just cached them:

  [root@jouet ~]# perf probe -l
  [root@jouet ~]#

We can see that it was cached as:

  [root@jouet ~]# ls -la ~/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/
  total 976
  drwxr-xr-x. 2 root root   4096 Jul 13 21:47 .
  drwxr-xr-x. 3 root root   4096 Jul 13 21:47 ..
  -rwxr-xr-x. 4 root root 985912 Jun 22 18:52 elf
  -rw-r--r--. 1 root root    303 Jul 13 21:47 probes
  [root@jouet ~]# file ~/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/elf
  /root/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/elf: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2, stripped
  [root@jouet ~]# cat ~/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/probes
  %sdt_libstdcxx:throw=throw
  p:sdt_libstdcxx/throw /usr/bin/gcc:0x71ffd
  %sdt_libstdcxx:rethrow=rethrow
  p:sdt_libstdcxx/rethrow /usr/bin/gcc:0x720b8
  %sdt_libstdcxx:catch=catch
  p:sdt_libstdcxx/catch /usr/bin/gcc:0x7307f
  %sdt_libgcc:unwind=unwind
  p:sdt_libgcc/unwind /usr/bin/gcc:0x7eec0
  #sdt_libstdcxx:*=%*
  [root@jouet ~]#

Ok, now we can use 'perf probe' to refer to those cached entries as:

  Humm, nope, doing as above we end up with:

  [root@jouet ~]# perf probe -a %sdt_libstdcxx:catch
  Semantic error :* is bad for event name -it must follow C symbol-naming rule.
    Error: Failed to add events.
  [root@jouet ~]#

But it worked at some point, lets try not using --dry-run:

Resetting everything:

  # rm -rf ~/.debug/
  # perf probe -d *:*
  # perf probe -l
  # perf list sdt

    List of pre-defined events (to be used in -e):

  #

Ok, now it cached everything, even things we haven't asked it to
(sdt_libgcc:unwind):

  [root@jouet ~]# perf probe -x /usr/bin/gcc --add %sdt_libstdcxx:\*
  Added new events:
    sdt_libstdcxx:throw  (on %* in /usr/bin/gcc)
    sdt_libstdcxx:rethrow (on %* in /usr/bin/gcc)
    sdt_libstdcxx:catch  (on %* in /usr/bin/gcc)

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

perf record -e sdt_libstdcxx:catch -aR sleep 1

  [root@jouet ~]# perf list sdt

  List of pre-defined events (to be used in -e):

    sdt_libgcc:unwind                                  [SDT event]
    sdt_libstdcxx:catch                                [SDT event]
    sdt_libstdcxx:rethrow                              [SDT event]
    sdt_libstdcxx:throw                                [SDT event]
  [root@jouet ~]#

And we have the events in place:

  [root@jouet ~]# perf probe -l
    sdt_libstdcxx:catch  (on execute_cfa_program+1551@../../../libgcc/unwind-dw2.c in /usr/bin/gcc)
    sdt_libstdcxx:rethrow (on d_print_subexpr+280@libsupc++/cp-demangle.c in /usr/bin/gcc)
    sdt_libstdcxx:throw  (on d_print_subexpr+93@libsupc++/cp-demangle.c in /usr/bin/gcc)
  [root@jouet ~]#

And trying to use them at least has 'perf trace --event sdt*:*' working.

Then, if we try to add the ones in libstdc++:

  [root@jouet ~]# perf probe -x /usr/lib64/libstdc++.so.6 -a %sdt_libstdcxx:\*
  Error: event "catch" already exists.
   Hint: Remove existing event by 'perf probe -d'
         or force duplicates by 'perf probe -f'
         or set 'force=yes' in BPF source.
    Error: Failed to add events.
  [root@jouet ~]#

Doesn't work, dups, but at least this served to, unbeknownst to the user, add
the SDT probes in /usr/lib64/libstdc++.so.6!

  [root@jouet ~]# perf list sdt

  List of pre-defined events (to be used in -e):

    sdt_libgcc:unwind                                  [SDT event]
    sdt_libstdcxx:catch@/usr/bin/gcc(9a0730e2bcc6)     [SDT event]
    sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event]
    sdt_libstdcxx:rethrow@/usr/bin/gcc(9a0730e2bcc6)   [SDT event]
    sdt_libstdcxx:rethrow@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event]
    sdt_libstdcxx:throw@/usr/bin/gcc(9a0730e2bcc6)     [SDT event]
    sdt_libstdcxx:throw@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event]
  [root@jouet ~]#

Now we should be able to get to the original cset comment, if we remove all
SDTs events in place, not from the cache, from the kernel, where it was set up as:

  [root@jouet ~]# ls -la /sys/kernel/debug/tracing/events/sdt_libstdcxx/
  total 0
  drwxr-xr-x.  5 root root 0 Jul 13 22:00 .
  drwxr-xr-x. 80 root root 0 Jul 13 21:56 ..
  drwxr-xr-x.  2 root root 0 Jul 13 22:00 catch
  -rw-r--r--.  1 root root 0 Jul 13 22:00 enable
  -rw-r--r--.  1 root root 0 Jul 13 22:00 filter
  drwxr-xr-x.  2 root root 0 Jul 13 22:00 rethrow
  drwxr-xr-x.  2 root root 0 Jul 13 22:00 throw
  [root@jouet ~]#

  [root@jouet ~]# head -2 /sys/kernel/debug/tracing/events/sdt_libstdcxx/throw/format
  name: throw
  ID: 2059
  [root@jouet ~]#

Now to remove it:

  [root@jouet ~]# perf probe -d sdt_libstdc*:*
  Removed event: sdt_libstdcxx:catch
  Removed event: sdt_libstdcxx:rethrow
  Removed event: sdt_libstdcxx:throw
  [root@jouet ~]#

Which caused:

  [root@jouet ~]# ls -la /sys/kernel/debug/tracing/events/sdt_libstdcxx/
  ls: cannot access '/sys/kernel/debug/tracing/events/sdt_libstdcxx/': No such file or directory
  [root@jouet ~]#

Ok, now we can do:

  [root@jouet ~]# perf list sdt_libstdcxx:catch

  List of pre-defined events (to be used in -e):

    sdt_libstdcxx:catch@/usr/bin/gcc(9a0730e2bcc6)     [SDT event]
    sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event]
  [root@jouet ~]#

So, these are not really 'pre-defined events', i.e. we can't use them with
'perf record --event':

  [root@jouet ~]# perf record --event sdt_libstdcxx:catch*
  event syntax error: 'sdt_libstdcxx:catch*'
                       \___ unknown tracepoint

  Error: File /sys/kernel/debug/tracing/events/sdt_libstdcxx/catch* not found.
  Hint: Perhaps this kernel misses some CONFIG_ setting to enable this feature?.
<SNIP>
  [root@jouet ~]#

To have it really pre-defined we must use perf probe to get its definition from
the cache and set it up in the kernel, creating the tracepoint to _then_ use it
with 'perf record --event':

  [root@jouet ~]# perf probe -a sdt_libstdcxx:catch
  Semantic error :There is non-digit char in line number.
  <SNIP>

Oops, there is another gotcha here, we need that pesky '%' character:

  [root@jouet ~]# perf probe -a %sdt_libstdcxx:catch
  Added new events:
    sdt_libstdcxx:catch  (on %catch in /usr/bin/gcc)
    sdt_libstdcxx:catch_1 (on %catch in /usr/lib64/libstdc++.so.6.0.22)

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

perf record -e sdt_libstdcxx:catch_1 -aR sleep 1

  [root@jouet ~]#

But then we added _two_ events, one with the name we expected, the other one
with a _ added, when doing the analysis we need to pay attention to who maps to
who.

And here is where we get to the point of this patch, which is to be able to
disambiguate those definitions for 'catch' in the build-id cache, but first we need
remove those events we just added:

[root@jouet ~]# perf probe -d %sdt_libstdcxx:catch

Oops, that didn't remove anything, we need to _remove_ that % char in this case:

  [root@jouet ~]# perf probe -d sdt_libstdcxx:catch
  Removed event: sdt_libstdcxx:catch

And we need to remove the other event added, i.e. I forgot to add a * at the end:

  [root@jouet ~]# perf probe -d sdt_libstdcxx:catch*
  Removed event: sdt_libstdcxx:catch_1
  [root@jouet ~]#

Ok, disambiguating it using what is in this patch:

  [root@jouet ~]# perf list sdt_libstdcxx:catch

  List of pre-defined events (to be used in -e):

    sdt_libstdcxx:catch@/usr/bin/gcc(9a0730e2bcc6)     [SDT event]
    sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event]
  [root@jouet ~]#
  [root@jouet ~]# perf probe -a %sdt_libstdcxx:catch@9a07
  Added new event:
    sdt_libstdcxx:catch  (on %catch in /usr/bin/gcc)

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

perf record -e sdt_libstdcxx:catch -aR sleep 1

  [root@jouet ~]# perf probe -l
    sdt_libstdcxx:catch  (on execute_cfa_program+1551@../../../libgcc/unwind-dw2.c in /usr/bin/gcc)
  [root@jouet ~]#

Yeah, it works! But we need to try and simplify this :-)

Update: Some aspects of this simplification take place in the following
        patches.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831793746.17065.13065062753978236612.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf list: Show SDT and pre-cached events

Show SDT and pre-cached events by perf-list with "sdt". This also shows
the binary and build-id where the events are placed only when there are
same name events on different binaries.

e.g.:

  # perf list sdt

  List of pre-defined events (to be used in -e):

    sdt_libc:lll_futex_wake                            [SDT event]
    sdt_libc:lll_lock_wait_private                     [SDT event]
    sdt_libc:longjmp                                   [SDT event]
    sdt_libc:longjmp_target                            [SDT event]
  ...
    sdt_libstdcxx:rethrow@/usr/bin/gcc(0cc207fc4b27)   [SDT event]
    sdt_libstdcxx:rethrow@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49)
    sdt_libstdcxx:throw@/usr/bin/gcc(0cc207fc4b27)     [SDT event]
    sdt_libstdcxx:throw@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49)

The binary path and build-id are shown in below format;

  <GROUP>:<EVENT>@<PATH>(<BUILD-ID>)

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20160624090646.25421.44225.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Search SDT/cached event from all probe caches

Search SDT/cached event from all probe caches if user doesn't pass any
binary. With this, we don't have to specify target binary for SDT and
named cached events (which start with %).

E.g. without this, a target binary must be passed with -x.

  # perf probe -x /usr/lib64/libc-2.20.so -a %sdt_libc:\*

With this change, we don't need it anymore.

  # perf probe -a %sdt_libc:\*

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831792812.17065.2353705982669445313.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Allow wildcard for cached events

Allo glob wildcard for reusing cached/SDT events. E.g.

  # perf probe -x /usr/lib64/libc-2.20.so -a %sdt_libc:\*

This example adds probes for all SDT in libc.
Note that the SDTs must have been scanned by perf buildid-cache.

Committer note:

Using it to check what of those SDT probes would take place when doing
a cargo run (rust):

  # trace --no-sys --event sdt_libc:* cargo run
     0.000 sdt_libc:setjmp:(7f326b69c4d1))
    28.423 sdt_libc:setjmp:(7f4b0a5364d1))
    29.000 sdt_libc:setjmp:(7f4b0a5364d1))
    88.597 sdt_libc:setjmp:(7fc01fd414d1))
    89.220 sdt_libc:setjmp:(7fc01fd414d1))
    95.501 sdt_libc:setjmp:(7f326b69c4d1))
     Running `target/debug/hello_world`
    97.110 sdt_libc:setjmp:(7f95e09234d1))
  Hello, world!
  #

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831791813.17065.17846564230840594888.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe-cache: Add for_each_probe_cache_entry() wrapper

Add for_each_probe_cache_entry() wrapper macro for hiding list in
probe_cache.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831790386.17065.15082256697569419710.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Make --list show only available cached events

Make "perf probe --cache --list" show only available cached events by
checking build-id validity.

E.g. without this patch:
  ----
  $ ./perf probe --cache --add oldevent=cmd_probe
  $ make #(to update ./perf)
  $ ./perf probe --cache --add newevent=cmd_probe
  $ ./perf probe --cache --list
  /home/mhiramat/ksrc/linux/tools/perf/perf (061e90539bac69
  probe_perf:newevent=cmd_probe
  /home/mhiramat/ksrc/linux/tools/perf/perf (c2e44d614e33e1
  probe_perf:oldevent=cmd_probe
  ----
It shows both of old and new events but user can not use old one.
With this;
  ----
  $ ./perf probe --cache -l
  /home/mhiramat/ksrc/linux/tools/perf/perf (061e90539bac69
  probe_perf:newevent=cmd_probe
  ----

This shows only new events which are on the existing binary.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831789417.17065.17896487479879669610.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Accept %sdt and %cached event name

To improve usability, support %[PROVIDER:]SDTEVENT format to add new
probes on SDT and cached events.

e.g.
  ----
  # perf probe -x /lib/libc-2.17.so  %lll_lock_wait_private
  Added new event:
    sdt_libc:lll_lock_wait_private (on %lll_lock_wait_private in /usr/lib/libc-2.17.so)

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

          perf record -e sdt_libc:lll_lock_wait_private -aR sleep 1

  # perf probe -l | more
    sdt_libc:lll_lock_wait_private (on __lll_lock_wait_private+21 in /usr/lib/libc-2.17.so)
  ----

Note that this is not only for SDT events, but also normal
events with event-name.

e.g. define "myevent" on cache (-n doesn't add the real probe)
  ----
  # perf probe -x ./perf --cache -n --add 'myevent=dso__load $params'
  ----
  Reuse the "myevent" from cache as below.
  ----
  # perf probe -x ./perf %myevent
  ----

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831788372.17065.3645054540325909346.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Fix to show correct error message for $vars and $params

Fix to show correct error messages for $vars and $params because
those special variables requires debug information to find the
real variables or function parameters.

E.g. without this fix;
  ----
  # perf probe -x /lib64/libc-2.23.so getaddrinfo \$params
  Failed to write event: Invalid argument
  Please upgrade your kernel to at least 3.14 to have access to feature $params
    Error: Failed to add events.
  ----

Perf ends up with an error, but the message is not correct.  With this
fix, perf shows correct error message as below.

  ----
  # perf probe -x /lib64/libc-2.23.so getaddrinfo \$params
  The /usr/lib64/libc-2.23.so file has no debug information.
  Rebuild with -g, or install an appropriate debuginfo package.
    Error: Failed to add events.
  ----

Reported-and-Tested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/146831787438.17065.6152436996780110699.stgit@devbox
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf bpf: Support BPF program attach to tracepoints

To support 98b5c2c65c29 ("perf, bpf: allow bpf programs attach to
tracepoints"), this patch allows BPF scripts to select tracepoints in
their section name.

Example:

  # cat test_tracepoint.c
  /*********************************************/
  #include <uapi/linux/bpf.h>
  #define SEC(NAME) __attribute__((section(NAME), used))
  SEC("raw_syscalls:sys_enter")
  int func(void *ctx)
  {
  /*
   * /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/format:
   * ...
   * field:long id; offset:8; size:8; signed:1;
   * ...
   * ctx + 8 select 'id'
   */
  u64 id = *((u64 *)(ctx + 8));
  if (id == 1)
  return 1;
  return 0;
  }
  SEC("_write=sys_write")
  int _write(void *ctx)
  {
  return 1;
  }
  char _license[] SEC("license") = "GPL";
  int _version SEC("version") = LINUX_VERSION_CODE;
  /*********************************************/
  # perf record -e ./test_tracepoint.c  dd if=/dev/zero of=/dev/null count=5
  5+0 records in
  5+0 records out
  2560 bytes (2.6 kB) copied, 6.2281e-05 s, 41.1 MB/s
  [ perf record: Woken up 1 times to write data ]
  # perf script
         dd 13436 [005] 1596.490869: raw_syscalls:sys_enter: NR 1 (1, 178d000, 200, 7ffe82470d60, ffffffffffffe020, fffff
         dd 13436 [005] 1596.490871:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490873: raw_syscalls:sys_enter: NR 1 (1, 178d000, 200, ffffffffffffe000, ffffffffffffe020, f
         dd 13436 [005] 1596.490874:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490876: raw_syscalls:sys_enter: NR 1 (1, 178d000, 200, ffffffffffffe000, ffffffffffffe020, f
         dd 13436 [005] 1596.490876:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490878: raw_syscalls:sys_enter: NR 1 (1, 178d000, 200, ffffffffffffe000, ffffffffffffe020, f
         dd 13436 [005] 1596.490879:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490881: raw_syscalls:sys_enter: NR 1 (1, 178d000, 200, ffffffffffffe000, ffffffffffffe020, f
         dd 13436 [005] 1596.490882:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490900: raw_syscalls:sys_enter: NR 1 (2, 7ffe8246e640, 1f, 40acb8, 7f44bac74700, 7f44baa4fba
         dd 13436 [005] 1596.490901:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490917: raw_syscalls:sys_enter: NR 1 (2, 7ffe8246e640, 1a, fffffffa, 7f44bac74700, 7f44baa4f
         dd 13436 [005] 1596.490918:  perf_bpf_probe:_write: (ffffffff812351e0)
         dd 13436 [005] 1596.490932: raw_syscalls:sys_enter: NR 1 (2, 7ffe8246e640, 1a, fffffff9, 7f44bac74700, 7f44baa4f
         dd 13436 [005] 1596.490933:  perf_bpf_probe:_write: (ffffffff812351e0)

Committer note:

Further testing:

  # trace --no-sys --event /home/acme/bpf/tracepoint.c cat /etc/passwd > /dev/null
     0.000 raw_syscalls:sys_enter:NR 1 (1, 7f0490504000, c48, 7f0490503010, ffffffffffffffff, 0))
     0.006 perf_bpf_probe:_write:(ffffffff81241bc0))
  #

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468406646-21642-6-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf bpf: Rename bpf__foreach_tev() to bpf__foreach_event()

Following commit will allow BPF script attach to tracepoints.
bpf__foreach_tev() will iterate over all events, not only kprobes.
Rename it to bpf__foreach_event().

Since only group and event are used by caller, there's no need to pass
full 'struct probe_trace_event' to bpf_prog_iter_callback_t. Pass only
these two strings. After this patch bpf_prog_iter_callback_t natually
support tracepoints.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468406646-21642-5-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf event parser: Add const qualifier to evt_name and sys_name

Add missing 'const' qualifiers so following commits are able to create
tracepoints using const strings.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468406646-21642-4-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib bpf: Report error when kernel doesn't support program type

Now libbpf support tracepoint program type. Report meaningful error when kernel
version is less than 4.7.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468406646-21642-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib bpf: New API to adjust type of a BPF program

Add 4 new APIs to adjust and query the type of a BPF program.
Load program according to type set by caller. Default is set to
BPF_PROG_TYPE_KPROBE.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1468406646-21642-2-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

objtool: Avoid checking code drift on busybox's diff

That doesn't have -I to match lines.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-1zqv1h6okt70e2huokkdtf1u@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

objtool: Add fallback from ELF_C_READ_MMAP to ELF_C_READ

Not all libelf implementations have this "Please, ELF_C_READ, but use
mmap if possible" elf_begin() cmd, so provide a fallback to plain old
ELF_C_READ.

Case in point: Alpine Linux 3.4.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-1fctuknrawgoi5xqon4mu9dv@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Fix up BITS_PER_LONG setting

It was set based on CONFIG_64BIT, that is available only when using
Kconfig, which we're working towards but not to the point of having this
CONFIG variable set, so synthesize it from available compiler defined
defines, __SIZEOF_LONG__ or, lacking that, __WORDSIZE.

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>
Link: http://lkml.kernel.org/n/tip-og5fmkr17856lhupacihwxvb@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Work around BITS_PER_LONG related build failure in objtool

The objtool build fails with the recent changes to the bits-per-long
headers:

  tools/include/linux/bitops.h:12:0: error: "BITS_PER_LONG" redefined [-Werror]

Which got introduced by:

  bb9707077b4e tools: Copy the bitsperlong.h files from the kernel

Work it around for the time being.

Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-core-for-mingo-20160712' 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 demangling of symbols in programs written in the Rust language (David Tolnay)

- Add support for tracepoints in the python binding, including an example, that
  sets up and parses sched:sched_switch events, tools/perf/python/tracepoint.py
  (Jiri Olsa)

- Introduce --stdio-color to set up the color output mode selection in
  'annotate' and 'report', allowing emit color escape sequences when
  redirecting the output of these tools (Arnaldo Carvalho de Melo)

Infrastructure changes:

- Various tweaks to allow the 'perf trace' beautifiers to build without using
  kernel headers and in a wider range of Linux distributions/releases
  (Arnaldo Carvalho de Melo)

- Stop using kernel source files, instead copy what is needed and
  check when the original kernel source file gets modified, warning
  the developers about it. This helps in building the tool in older
  systems and even in recent ones, for just added kernel features
  for which ABI details (struct changes, defines, etc) still are not
  available on system headers (Arnaldo Carvalho de Melo)

- Be consistent in how to use strerror_r, adding a wrapper that makes sure that
  it returns a pointer to passed buffer, and using the XSI variant, that is
  available in more libc implementations (Arnaldo Carvalho de Melo)

- Avoid checking code drift on busibox's diff perf intel-pt-decoder, as it
  doesn't have the '-I' command line switch to check for regexps
  (Arnaldo Carvalho de Melo)

- Add missing headers in various places (Arnaldo Carvalho de Melo)

- Remove unneeded headers from various other places (Arnaldo Carvalho de Melo)

- Add feature detection for gelf_getnote(), disabling SDT support if not
  present (Arnaldo Carvalho de Melo)

- Fix oddities with GCC 5.3.0 by initializing some variables
  (Arnaldo Carvalho de Melo)

- With those changes in place perf now builds on Alpine Linux 3.4, in addition to
  on centos (5, 6, 7), debian (7, 8, experimental), fedora (21, 22, 23, 24, rawhide),
  mageia 5, opensuse (13.2, 42.1) and ubuntu (12.04.5, 14.04.4, 15.10, 16.04) and
  will be test build on those systems prior to future pull requests.

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

tools lib traceevent: Add filter on task CPU id

Add a 'CPU' special field to allow the filter in trace-cmd report to
filter on the task's CPU.

By adding a special field 'CPU' (all caps) the user can now filter out
tasks based on which CPU they are on. This is useful when filtering out
(or in) a bunch of threads.

  -F 'CPU == 0'

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/20160712093306.5b058103@gandalf.local.home
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Add tracepoint example

To show how to enable a tracepoint and access its fields.

Committer note:

Testing it:

  # ls -l /tmp/build/perf/python/perf.so
  -rwxrwxr-x. 1 acme acme 1563256 Jul 12 16:19 /tmp/build/perf/python/perf.so
  # export PYTHONPATH=/tmp/build/perf/python/
  # tools/perf/python/tracepoint.py 2> /dev/null | head -200 | tail -10
  time 76345337296548 prev_comm=swapper/0 prev_pid=0 prev_prio=120 prev_state=0x0 ==> next_comm=tracepoint.py- next_pid=18479 next_prio=120
  time 76345338520479 prev_comm=gnome-shelln-b prev_pid=2186 prev_prio=120 prev_state=0x1 ==> next_comm=swapper/1 next_pid=0 next_prio=120
  time 76345337309942 prev_comm=tracepoint.py- prev_pid=18479 prev_prio=120 prev_state=0x1 ==> next_comm=swapper/0 next_pid=0 next_prio=120
  time 76345337312302 prev_comm=swapper/0 prev_pid=0 prev_prio=120 prev_state=0x0 ==> next_comm=tracepoint.py- next_pid=18479 next_prio=120
  time 76345337324927 prev_comm=tracepoint.py- prev_pid=18479 prev_prio=120 prev_state=0x1 ==> next_comm=swapper/0 next_pid=0 next_prio=120
  time 76345337327115 prev_comm=swapper/0 prev_pid=0 prev_prio=120 prev_state=0x0 ==> next_comm=tracepoint.py- next_pid=18479 next_prio=120
  time 76345338621750 prev_comm=swapper/2 prev_pid=0 prev_prio=120 prev_state=0x0 ==> next_comm=rcuos/2 next_pid=29 next_prio=120
  time 76345338607922 prev_comm=swapper/3 prev_pid=0 prev_prio=120 prev_state=0x0 ==> next_comm=rcu_sched next_pid=7 next_prio=120
  time 76345337338817 prev_comm=tracepoint.py- prev_pid=18479 prev_prio=120 prev_state=0x1 ==> next_comm=swapper/0 next_pid=0 next_prio=120
  time 76345338627156 prev_comm=swapper/1 prev_pid=0 prev_prio=120 prev_state=0x0 ==> next_comm=head-terminal- next_pid=18480 next_prio=120
  #
  # strip /tmp/build/perf/python/perf.so
  # ls -l /tmp/build/perf/python/perf.so
  -rwxrwxr-x. 1 acme acme 319616 Jul 12 16:25 /tmp/build/perf/python/perf.so

Reported-and-Tested-by: Jiri Pirko <jiri@mellanox.com>
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/1468148882-10362-10-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Add support to resolve tracepoint fields

Adding tp_getattro callback for sample event. It resolves tracepoint
fields in runtime.

It's now possible to access tracepoint fields in normal fashion like
hardcoded ones (see the example in the next patch).

Reported-and-Tested-by: Jiri Pirko <jiri@mellanox.com>
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/1468148882-10362-9-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Add struct evsel into struct pyrf_event

To be able to find out event configuration info during sample parsing.

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/1468148882-10362-8-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Add perf.tracepoint method

To get id of the tracepoint from subsystem and name strings. The
interface is:

  id = perf.tracepoint(sys, name)

In case of error -1 is returned.

It will be used to get python tracepoint event's config value for
tracepoint event.

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/1468148882-10362-7-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Put perf.event objects into dictionary

Make perf.event object parts of the perf module dictionary so we can
address them by name.

The following objects/names are added:

  mmap_event
  lost_event
  comm_event
  task_event
  throttle_event
  task_event
  read_event
  sample_event
  switch_event

We can now use it in python script like:
  ...
  event = evlist.read_on_cpu(cpu)
  ...
  if not isinstance(event, perf.sample_event):

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/1468148882-10362-6-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Fix pyrf_evlist__read_on_cpu event consuming

We can't consume the event before parsing it. Under heavy load we could
get caught by kernel writer overwriting the event we're trying to parse.

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/1468148882-10362-5-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf python: Init perf_event_attr::size in perf.evsel constructor

Currently 0 is passed as perf_event_attr::size, which could block usage
of new features.

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/1468148882-10362-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Introduce trace_event__tp_format_id()

To get struct event_format object from tracepoint ID.  It will be used
in following patches.

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/1468148882-10362-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Make event2evsel public

It will be used outside of evlist.c object in folowing patches.

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/1468148882-10362-2-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf symbols: Add Rust demangling

Rust demangling is another step after bfd demangling. Add a diagnosis to
identify mangled Rust symbols based on the hash that the Rust mangler appends
as the last path component, as well as other characteristics.  Add a demangler
to reconstruct the original symbol.

Committer notes:

How I tested it:

Enabled COPR on Fedora 24 and then installed the 'rust-binary' package,
with it:

  $ cat src/main.rs
  fn main() {
      println!("Hello, world!");
  }
  $ cat Cargo.toml
  [package]

  name = "hello_world"
  version = "0.0.1"
  authors = [ "Arnaldo Carvalho de Melo <acme@kernel.org>" ]

  $ perf record cargo bench
   Compiling hello_world v0.0.1 (file:///home/acme/projects/hello_world)
     Running target/release/hello_world-d4b9dab4b2a47d75

  running 0 tests

  test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured

  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.096 MB perf.data (1457 samples) ]
  $

Before this patch:

  $ perf report --stdio --dsos librbml-e8edd0fd.so
  # dso: librbml-e8edd0fd.so
  #
  # Total Lost Samples: 0
  #
  # Samples: 1K of event 'cycles:u'
  # Event count (approx.): 979599126
  #
  # Overhead  Command  Symbol
  # ........  .......  .............................................................................................................
  #
       1.78%  rustc    [.] rbml::reader::maybe_get_doc::hb9d387df6024b15b
       1.50%  rustc    [.] _$LT$reader..DocsIterator$LT$$u27$a$GT$$u20$as$u20$std..iter..Iterator$GT$::next::hd9af9e60d79a35c8
       1.20%  rustc    [.] rbml::reader::doc_at::hc88107fba445af31
       0.46%  rustc    [.] _$LT$reader..TaggedDocsIterator$LT$$u27$a$GT$$u20$as$u20$std..iter..Iterator$GT$::next::h0cb40e696e4bb489
       0.35%  rustc    [.] rbml::reader::Decoder::_next_int::h66eef7825a398bc3
       0.29%  rustc    [.] rbml::reader::Decoder::_next_sub::h8e5266005580b836
       0.15%  rustc    [.] rbml::reader::get_doc::h094521c645459139
       0.14%  rustc    [.] _$LT$reader..Decoder$LT$$u27$doc$GT$$u20$as$u20$serialize..Decoder$GT$::read_u32::h0acea2fff9669327
       0.07%  rustc    [.] rbml::reader::Decoder::next_doc::h6714d469c9dfaf91
       0.07%  rustc    [.] _ZN4rbml6reader10doc_as_u6417h930b740aa94f1d3aE@plt
       0.06%  rustc    [.] _fini
  $

After:

  $ perf report --stdio --dsos librbml-e8edd0fd.so
  # dso: librbml-e8edd0fd.so
  #
  # Total Lost Samples: 0
  #
  # Samples: 1K of event 'cycles:u'
  # Event count (approx.): 979599126
  #
  # Overhead  Command  Symbol
  # ........  .......  .................................................................
  #
     1.78%  rustc    [.] rbml::reader::maybe_get_doc
     1.50%  rustc    [.] <reader::DocsIterator<'a> as std::iter::Iterator>::next
     1.20%  rustc    [.] rbml::reader::doc_at
     0.46%  rustc    [.] <reader::TaggedDocsIterator<'a> as std::iter::Iterator>::next
     0.35%  rustc    [.] rbml::reader::Decoder::_next_int
     0.29%  rustc    [.] rbml::reader::Decoder::_next_sub
     0.15%  rustc    [.] rbml::reader::get_doc
     0.14%  rustc    [.] <reader::Decoder<'doc> as serialize::Decoder>::read_u32
     0.07%  rustc    [.] rbml::reader::Decoder::next_doc
     0.07%  rustc    [.] _ZN4rbml6reader10doc_as_u6417h930b740aa94f1d3aE@plt
     0.06%  rustc    [.] _fini
  $

Signed-off-by: David Tolnay <dtolnay@gmail.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/5780B7FA.3030602@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add feature detection for gelf_getnote()

That is not present on some libelf implementations, such as the one used
in Alpine Linux: libelf-0.8.13.

This ends up disabling the SDT code, that relies on this function.

One alternative would be to provide an weak fallback implementation or
the open coded variant used by the buildid sysfs notes reading code.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-82lh22ybedy9b9lych8xj12g@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt-decoder: Avoid checking code drift on busibox's diff

That doesn't have -I to match lines.

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>
Link: http://lkml.kernel.org/n/tip-7nz9hnbk7a9p91ou927ye5yh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Don't add kernel directories to the header search path

We've decided not to access kernel source files because changes there
could break the tooling side, this is one more step in that direction.

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>
Link: http://lkml.kernel.org/n/tip-ar0hupkxl45h5hk09l2rprj3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add the tools/ stringify copy to the MANIFEST

So that we don't end up using the kernel one when building out of tree,
via a detached tarball.

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>
Fixes: 737ef7d32cb4 ("tools include: Copy linux/stringify.h from the kernel")
Link: http://lkml.kernel.org/n/tip-t8yn1d7y0magk889ymc8jlai@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Copy the bitsperlong.h files from the kernel

We use it in bitops/__ffs.h and bitops/atomic.h, that we also got from
the kernel, but were getting it from either newer systems that carry it
in /usr/include, or from the kernel sources, that we decided not to
touch from tools/ code. Fix it.

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>
Link: http://lkml.kernel.org/n/tip-lwqvgbuitjmrdpjmjp6zqnyx@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script python: Silence -Werror=maybe-uninitialized on gcc 5.3.0

Sounds like a compiler bug, but to silence it, initialize those
variables to NULL.

Noticed on:

Target: x86_64-alpine-linux-musl
Configured with: /home/buildozer/aports/main/gcc/src/gcc-5.3.0/configure
--prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info
--build=x86_64-alpine-linux-musl --host=x86_64-alpine-linux-musl
--target=x86_64-alpine-linux-musl --with-pkgversion='Alpine 5.3.0'
--enable-checking=release --disable-fixed-point --disable-libstdcxx-pch
--disable-multilib --disable-nls --disable-werror --disable-symvers
--enable-__cxa_atexit --enable-esp --enable-cloog-backend
--enable-languages=c,c++,objc,java,fortran,ada --disable-libssp
--disable-libmudflap --disable-libsanitizer --enable-shared
--enable-threads --enable-tls --with-system-zlib
Thread model: posix
gcc version 5.3.0 (Alpine 5.3.0)

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>
Link: http://lkml.kernel.org/n/tip-zyvsjvbl45o7hzcuz78wu2xi@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf symbols: Provide a GElf_Nhdr typedef

This one can be safely defined to be Elf64_Nhdr, as it is in elfutils's
libelf, but not on musl libc, as both Elf64_Nhdr and  Elf32_Nhdr have
the same layout.

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>
Link: http://lkml.kernel.org/n/tip-w8z8614l03lc8bip4ijbywbt@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace beauty seccomp: Remove seccomp.h include

All we need from it is already conditionally defined, and this header
file is not present in older systems, so ditch it.

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>
Link: http://lkml.kernel.org/n/tip-3jxpz9gwahk4e7ltqtnr1rjg@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace beauty futex_op: Add missing defines for older systems

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>
Link: http://lkml.kernel.org/n/tip-6qkuhv2mrcxmpy5sasc3c9tf@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Fallback to reading sysfs to get cacheline size

On systems where sysconf(_SC_LEVEL1_DCACHE_LINESIZE) is not available,
such as musl LIBC and Android's bionic libc.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Chris Phlipot <cphlipot0@gmail.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>
Link: http://lkml.kernel.org/n/tip-772obxzby758g7m2wmzcejxz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Copy the header files needed by perf tools

Those kernel files were being directly accessed, which we're not
allowing anymore to avoid that changes in the kernel side break tooling.

Warn if these copies drift from the original files.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Yarygin <yarygin@linux.vnet.ibm.com>
Cc: Christoffer Dall <christoffer.dall@linaro.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Eric Auger <eric.auger@linaro.org>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: Yunlong Song <yunlong.song@huawei.com>
Link: http://lkml.kernel.org/n/tip-mnopguymhnwzjhw3mowllvsy@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

kvm arm/arm64: Remove trailing whitespace from headers

Noticed while making a copy of these files to tools/ where those kernel
files were being directly accessed, which we're not allowing anymore to
avoid that changes in the kernel side break tooling.

Cc: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Eric Auger <eric.auger@linaro.org>
Cc: Marc Zyngier <marc.zyngier@arm.com>
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>
Link: http://lkml.kernel.org/n/tip-82thftcdhj2j5wt6ir4vuyhk@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Remove unused sys/ptrace.h include

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>
Link: http://lkml.kernel.org/n/tip-ogtjqc0hxm961djgiwboe2q7@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove unneeded magic.h include from util.h

Not used anymore, IIRC it was for useless PROC_FS_MAGIC procfs checks,
but those are long gone.

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: Steven Rostedt <rostedt@goodmis.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-v027did3kvj0vz7bofgzkw29@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Introduce weak alternative to sched_getcpu()

Which is just a wrapper for sys_getcpu and is not present in at least
musl libc.

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>
Link: http://lkml.kernel.org/n/tip-kblef7svmhr0g93kkx78envg@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Copy uapi/linux/hw_breakpoint.h from the kernel

To allow the build to complete on older systems, where those files are
either not uptodate, lacking some recent additions or not present at
all.

And check if the copy drifts from the kernel.

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>
Link: http://lkml.kernel.org/n/tip-3jz31pz4nw526uko5da9e7o3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test bpf: Use epoll_wait() instead of epoll_pwait()

The prototype for epoll_wait() is available in older distros, so use it
instead of epoll_pwait() (removing the last NULL arg, the sigmask,
makes it the same thing anyway) to avoid breaking the build.

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>
Link: http://lkml.kernel.org/n/tip-pwiwizloxt0jujy8em80qut3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools: Copy uapi/asm/perf_regs.h from the kernel

To allow the build to complete on older systems, where those files are
either not uptodate, lacking some recent additions or not present at
all.

And check if the copy drifts from the kernel.

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>
Link: http://lkml.kernel.org/n/tip-sxf7rpow2blsno5f7t6n0sqz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib bpf: Copy bpf.h and bpf_common.h from the kernel

To allow the build to complete on older systems, where those files are
either not uptodate, lacking some recent additions or not present at
all.

And check if the copy drifts from the kernel, as in this synthetic test:

    BUILD:   Doing 'make -j4' parallel build
  Warning: tools/include/linux/bpf.h differs from kernel
  Warning: tools/include/linux/bpf_common.h differs from kernel

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>
Link: http://lkml.kernel.org/n/tip-5plvi2gq4x469dcyybiu226q@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf bench: Copy kernel files needed to build mem{cpy,set} x86_64 benchmarks

We can't access kernel files directly from tools/, so copy the required
bits, and make sure that we detect when the original files, in the
kernel, gets modified.

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>
Link: http://lkml.kernel.org/n/tip-z7e76274ch5j4nugv048qacb@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>