perf evlist: Use unshare(CLONE_FS) in sb threads to let setns(CLONE_NEWNS) work

When we started using a thread to catch the PERF_RECORD_BPF_EVENT meta
data events to then ask the kernel for further info (BTF, etc) for BPF
programs shortly after they get loaded, we forgot to use
unshare(CLONE_FS) as was done in:

  868a832918f6 ("perf top: Support lookup of symbols in other mount namespaces.")

Do it so that we can enter the namespaces to read the build-ids at the
end of a 'perf record' session for the DSOs that had hits.

Before:

Starting a 'stress-ng --cpus 8' inside a container and then, outside the
container running:

  # perf record -a --namespaces sleep 5
  # perf buildid-list | grep stress-ng
  #

We would end up with a 'perf.data' file that had no entry in its
build-id table for the /usr/bin/stress-ng binary inside the container
that got tons of PERF_RECORD_SAMPLEs.

After:

  # perf buildid-list | grep stress-ng
  f2ed02c68341183a124b9b0f6e2e6c493c465b29 /usr/bin/stress-ng
  #

Then its just a matter of making sure that that binary debuginfo package
gets available in a place that 'perf report' will look at build-id keyed
ELF files, which, in my case, on a f30 notebook, was a matter of
installing the debuginfo file for the distro used in the container,
fedora 31:

  # rpm -ivh http://fedora.c3sl.ufpr.br/linux/development/31/Everything/x86_64/debug/tree/Packages/s/stress-ng-debuginfo-0.07.29-10.fc31.x86_64.rpm

Then, because perf currently looks for those debuginfo files (richer ELF
symtab) inside that namespace (look at the setns calls):

  openat(AT_FDCWD, "/proc/self/ns/mnt", O_RDONLY) = 137
  openat(AT_FDCWD, "/proc/13169/ns/mnt", O_RDONLY) = 139
  setns(139, CLONE_NEWNS)                 = 0
  stat("/usr/bin/stress-ng", {st_mode=S_IFREG|0755, st_size=3065416, ...}) = 0
  openat(AT_FDCWD, "/usr/bin/stress-ng", O_RDONLY) = 140
  fcntl(140, F_GETFD)                     = 0
  fstat(140, {st_mode=S_IFREG|0755, st_size=3065416, ...}) = 0
  mmap(NULL, 3065416, PROT_READ, MAP_PRIVATE, 140, 0) = 0x7ff2fdc5b000
  munmap(0x7ff2fdc5b000, 3065416)         = 0
  close(140)                              = 0
  stat("stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory)
  stat("/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory)
  stat("/usr/bin/.debug/stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory)
  stat("/usr/lib/debug/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory)
  stat("/root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29", 0x7fff45d711e0) = -1 ENOENT (No such file or directory)

To only then go back to the "host" namespace to look just in the users's
~/.debug cache:

  setns(137, CLONE_NEWNS)                 = 0
  chdir("/root")                          = 0
  close(137)                              = 0
  close(139)                              = 0
  stat("/root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf", 0x7fff45d732e0) = -1 ENOENT (No such file or directory)

It continues to fail to resolve symbols:

  # perf report | grep stress-ng | head -5
     9.50%  stress-ng-cpu    stress-ng    [.] 0x0000000000021ac1
     8.58%  stress-ng-cpu    stress-ng    [.] 0x0000000000021ab4
     8.51%  stress-ng-cpu    stress-ng    [.] 0x0000000000021489
     7.17%  stress-ng-cpu    stress-ng    [.] 0x00000000000219b6
     3.93%  stress-ng-cpu    stress-ng    [.] 0x0000000000021478
  #

To overcome that we use:

  # perf buildid-cache -v --add /usr/lib/debug/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug
  Adding f2ed02c68341183a124b9b0f6e2e6c493c465b29 /usr/lib/debug/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug: Ok
  #
  # ls -la /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf
  -rw-r--r--. 3 root root 2401184 Jul 27 07:03 /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf
  # file /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf
  /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter \004, BuildID[sha1]=f2ed02c68341183a124b9b0f6e2e6c493c465b29, for GNU/Linux 3.2.0, with debug_info, not stripped, too many notes (256)
  #

Now it finally works:

  # perf report | grep stress-ng | head -5
    23.59%  stress-ng-cpu    stress-ng    [.] ackermann
    23.33%  stress-ng-cpu    stress-ng    [.] is_prime
    17.36%  stress-ng-cpu    stress-ng    [.] stress_cpu_sieve
     6.08%  stress-ng-cpu    stress-ng    [.] stress_cpu_correlate
     3.55%  stress-ng-cpu    stress-ng    [.] queens_try
  #

I'll make sure that it looks for the build-id keyed files in both the
"host" namespace (the namespace the user running 'perf record' was a the
time of the recording) and in the container namespace, as it shouldn't
matter where a content based key lookup finds the ELF file to use in
resolving symbols, etc.

Reported-by: Karl Rister <krister@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Krister Johansen <kjlx@templeofstupid.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Stanislav Fomichev <sdf@google.com>
Cc: Thomas-Mich Richter <tmricht@linux.vnet.ibm.com>
Fixes: 657ee5531903 ("perf evlist: Introduce side band thread")
Link: https://lkml.kernel.org/n/tip-g79k0jz41adiaeuqud742t2l@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Move 'enum perf_user_event_type' to perf/event.h

So it's available for libperf's users.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-24-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Rename the PERF_RECORD_ structs to have a "perf" prefix

Even more, to have a "perf_record_" prefix, so that they match the
PERF_RECORD_ enum they map to.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-23-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add 'union perf_event' to perf/event.h

So it's available for libperf's users.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-22-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_COMPRESSED 'struct compressed_event' to perf/event.h

Move the PERF_RECORD_COMPRESSED event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-21-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_HEADER_FEATURE 'struct feature_event' to perf/event.h

Move the PERF_RECORD_HEADER_FEATURE event definition to libperf's
event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-20-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_TIME_CONV 'struct time_conv_event' to perf/event.h

Move the PERF_RECORD_TIME_CONV event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-19-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_STAT_ROUND 'struct stat_round_event' to perf/event.h

Move the PERF_RECORD_STAT_ROUND event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-18-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_STAT 'struct stat_event' to perf/event.h

Move the PERF_RECORD_STAT event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-17-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_STAT_CONFIG 'struct stat_config_event' to perf/event.h

Move the PERF_RECORD_STAT_CONFIG event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-16-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_THREAD_MAP 'struct thread_map_event' to perf/event.h

Move the PERF_RECORD_THREAD_MAP event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-15-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_SWITCH 'struct context_switch_event' to perf/event.h

Move the PERF_RECORD_SWITCH event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-14-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_ITRACE_START 'struct itrace_start_event' to perf/event.h

Move the PERF_RECORD_ITRACE_START event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-13-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_AUX 'struct aux_event' to perf/event.h

Move the PERF_RECORD_AUX event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-12-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_AUXTRACE_ERROR 'struct auxtrace_error_event' to perf/event.h

Move the PERF_RECORD_AUXTRACE_ERROR event definition to libperf's
event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-11-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_AUXTRACE 'struct auxtrace_event' to perf/event.h

Move the PERF_RECORD_AUXTRACE event definition to libperf's event.h.

Ipn order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-10-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_AUXTRACE_INFO 'struct auxtrace_info_event' to perf/event.h

Move the PERF_RECORD_AUXTRACE_INFO event definition to libperf's
event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-9-jolsa@kernel.org
[ Fix cs_etm__print_auxtrace_info() arg to be __u64 too to fix the CORESIGHT=1 build ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_ID_INDEX 'struct id_index_event' to perf/event.h

Move the PERF_RECORD_ID_INDEX event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Add the PRI_ld64 define, so we can use it in printf output.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-8-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_HEADER_BUILD_ID 'struct build_id_event' to perf/event.h

Move the PERF_RECORD_HEADER_BUILD_ID event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Adding the fix value for build_id variable, because it will never
change.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-7-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_HEADER_TRACING_DATA 'struct tracing_data_event' to perf/event.h

Move the PERF_RECORD_HEADER_TRACING_DATA event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-6-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_HEADER_EVENT_TYPE 'struct event_type_event' to perf/event.h

Move the PERF_RECORD_HEADER_EVENT_TYPE event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-5-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_EVENT_UPDATE 'struct event_update_event' to perf/event.h

Move the PERF_RECORD_EVENT_UPDATE event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-4-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_CPU_MAP 'struct cpu_map_event' to perf/event.h

Move the PERF_RECORD_CPU_MAP event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-3-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_HEADER_ATTR 'struct attr_event' to perf/event.h

Move the PERF_RECORD_HEADER_ATTR event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190828135717.7245-2-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf top: Fix event group with more than two events

The event group feature links relevant hist entries among events so that
they can be displayed together.  During the link process, each hist
entry in non-leader events is connected to a hist entry in the leader
event.  This is done in order of events specified in the command line so
it assumes that events are linked in the order.

But 'perf top' can break the assumption since it does the link process
multiple times.  For example, a hist entry can be in the third event
only at first so it's linked after the leader.  Some time later, second
event has a hist entry for it and it'll be linked after the entry of the
third event.

This makes the code compilicated to deal with such unordered entries.
This patch simply unlink all the entries after it's printed so that they
can assume the correct order after the repeated link process.  Also it'd
be easy to deal with decaying old entries IMHO.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190827231555.121411-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf top: Decay all events in the evlist

Currently perf top only decays entries in a selected evsel.  I don't
know whether it's intended (maybe due to performance reason?) but anyway
it might show incorrect output when event group is used since users will
see leader event is decayed but others are not.

This patch moves the decay code into perf_top__resort_hists() so that
stdio and TUI code shared the logic.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190827231555.121411-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf clang: Delete needless util-cxx.h header

It was put in place just to make sure the 'new' C++ operator wouldn't
clash with some argument name in util.h, but there is not anymore any
such argument and also the reason stated for util.h to be included there
was to get the __maybe_unused definition, that is in linux/compiler.h,
so use that instead and nuke util-cxx.h.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-1r5tvfnwiydjxhukgqs6bi11@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Remove needless util.h from evlist.h

There is no need for that util/util.h include there and, remove it,
pruning the include tree, fix the fallout by adding necessary headers to
places that were getting needed includes indirectly from evlist.h ->
util.h.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-s9f7uve8wvykr5itcm7m7d8q@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove needless util.h include from builtin.h

And fix up places where util.h is needed but was obtained indirectly via
builtin.h.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-a01ig3c4t76ye5wkqmtgk9qn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Warn that perf_event_paranoid can restrict kernel symbols

Warn that /proc/sys/kernel/perf_event_paranoid can also restrict kernel
symbols.

Signed-off-by: Igor Lubashev <ilubashe@akamai.com>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: James Morris <jmorris@namei.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1566869956-7154-6-git-send-email-ilubashe@akamai.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf symbols: Use CAP_SYSLOG with kptr_restrict checks

The kernel is using CAP_SYSLOG capability instead of uid==0 and euid==0
when checking kptr_restrict. Make perf do the same.

Also, the kernel is a more restrictive than "no restrictions" in case of
kptr_restrict==0, so add the same logic to perf.

Signed-off-by: Igor Lubashev <ilubashe@akamai.com>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: James Morris <jmorris@namei.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1566869956-7154-5-git-send-email-ilubashe@akamai.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Kernel profiling is disallowed only when perf_event_paranoid > 1

Perf was too restrictive about sysctl kernel.perf_event_paranoid. The
kernel only disallows profiling when perf_event_paranoid > 1. Make perf
do the same.

Committer testing:

For a non-root user:

  $ id
  uid=1000(acme) gid=1000(acme) groups=1000(acme),10(wheel) context=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
  $

Before:

We were restricting it to just userspace (:u suffix) even for a
workload started by the user:

  $ perf record sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.001 MB perf.data (8 samples) ]
  $ perf evlist
  cycles:u
  $ perf evlist -v
  cycles:u: size: 112, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|PERIOD, read_format: ID, disabled: 1, inherit: 1, exclude_kernel: 1, mmap: 1, comm: 1, freq: 1, enable_on_exec: 1, task: 1, precise_ip: 3, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1
  $ perf report --stdio
  # To display the perf.data header info, please use --header/--header-only options.
  #
  # Total Lost Samples: 0
  #
  # Samples: 8  of event 'cycles:u'
  # Event count (approx.): 1040396
  #
  # Overhead  Command  Shared Object     Symbol
  # ........  .......  ................  ......................
  #
      68.36%  sleep    libc-2.29.so      [.] _dl_addr
      27.33%  sleep    ld-2.29.so        [.] dl_main
       3.80%  sleep    ld-2.29.so        [.] _dl_setup_hash
  #
  # (Tip: Order by the overhead of source file name and line number: perf report -s srcline)
  #
  $
  $

After:

When the kernel allows profiling the kernel in that scenario:

  $ perf record sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.023 MB perf.data (11 samples) ]
  $ perf evlist
  cycles
  $ perf evlist -v
  cycles: size: 112, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|PERIOD, read_format: ID, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1, enable_on_exec: 1, task: 1, precise_ip: 3, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1
  $
  $ perf report --stdio
  # To display the perf.data header info, please use --header/--header-only options.
  #
  # Total Lost Samples: 0
  #
  # Samples: 11  of event 'cycles'
  # Event count (approx.): 1601964
  #
  # Overhead  Command  Shared Object     Symbol
  # ........  .......  ................  ..........................
  #
      28.14%  sleep    [kernel.vmlinux]  [k] __rb_erase_color
      27.21%  sleep    [kernel.vmlinux]  [k] unmap_page_range
      27.20%  sleep    ld-2.29.so        [.] __tunable_get_val
      15.24%  sleep    [kernel.vmlinux]  [k] thp_get_unmapped_area
       1.96%  perf     [kernel.vmlinux]  [k] perf_event_exec
       0.22%  perf     [kernel.vmlinux]  [k] native_sched_clock
       0.02%  perf     [kernel.vmlinux]  [k] intel_bts_enable_local
       0.00%  perf     [kernel.vmlinux]  [k] native_write_msr
  #
  # (Tip: Boolean options have negative forms, e.g.: perf report --no-children)
  #
  $

Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Igor Lubashev <ilubashe@akamai.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: James Morris <jmorris@namei.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1566869956-7154-4-git-send-email-ilubashe@akamai.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Use CAP_SYS_ADMIN with perf_event_paranoid checks

The kernel is using CAP_SYS_ADMIN instead of euid==0 to override
perf_event_paranoid check. Make perf do the same.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org> # coresight part
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: James Morris <jmorris@namei.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1566869956-7154-3-git-send-email-ilubashe@akamai.com
Signed-off-by: Igor Lubashev <ilubashe@akamai.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf event: Check ref_reloc_sym before using it

Check for ref_reloc_sym before using it instead of checking
symbol_conf.kptr_restrict and relying solely on that check.

Reported-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Igor Lubashev <ilubashe@akamai.com>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: James Morris <jmorris@namei.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1566869956-7154-2-git-send-email-ilubashe@akamai.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf arch powerpc: Sync powerpc syscall.tbl

Copy over powerpc syscall.tbl to grab changes from the below commits:

  commit cee3536d24a1 ("powerpc: Wire up clone3 syscall")
  commit 1a271a68e030 ("arch: mark syscall number 435 reserved for clone3")
  commit 7615d9e1780e ("arch: wire-up pidfd_open()")
  commit d8076bdb56af ("uapi: Wire up the mount API syscalls on non-x86 arches [ver #2]")
  commit 39036cd27273 ("arch: add pidfd and io_uring syscalls everywhere")
  commit 48166e6ea47d ("y2038: add 64-bit time_t syscalls to all 32-bit architectures")
  commit d33c577cccd0 ("y2038: rename old time and utime syscalls")
  commit 00bf25d693e7 ("y2038: use time32 syscall names on 32-bit")
  commit 8dabe7245bbc ("y2038: syscalls: rename y2038 compat syscalls")
  commit 0d6040d46817 ("arch: add split IPC system calls where needed")

Reported-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/20190827071458.19897-1-naveen.n.rao@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/x86/intel: Support PEBS output to PT

If PEBS declares ability to output its data to Intel PT stream, use the
aux_output attribute bit to enable PEBS data output to PT. This requires
a PT event to be present and scheduled in the same context. Unlike the
DS area, the kernel does not extract PEBS records from the PT stream to
generate corresponding records in the perf stream, because that would
require real time in-kernel PT decoding, which is not feasible. The PMI,
however, can still be used.

The output setting is per-CPU, so all PEBS events must be either writing
to PT or to the DS area, therefore, in case of conflict, the conflicting
event will fail to schedule, allowing the rotation logic to alternate
between the PEBS->PT and PEBS->DS events.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: kan.liang@linux.intel.com
Link: https://lkml.kernel.org/r/20190806084606.4021-3-alexander.shishkin@linux.intel.com

perf: Allow normal events to output AUX data

In some cases, ordinary (non-AUX) events can generate data for AUX events.
For example, PEBS events can come out as records in the Intel PT stream
instead of their usual DS records, if configured to do so.

One requirement for such events is to consistently schedule together, to
ensure that the data from the "AUX output" events isn't lost while their
corresponding AUX event is not scheduled. We use grouping to provide this
guarantee: an "AUX output" event can be added to a group where an AUX event
is a group leader, and provided that the former supports writing to the
latter.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: kan.liang@linux.intel.com
Link: https://lkml.kernel.org/r/20190806084606.4021-2-alexander.shishkin@linux.intel.com

Merge tag 'perf-core-for-mingo-5.4-20190826' of git://git./linux/kernel/git/acme/linux into perf/core

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

perf report:

  Andi Kleen:

  - Make --ns time sort key output column wide enough for nanoseconds.

perf script:

  Gustavo A. R. Silva:

  - Fix memory leaks in list_scripts()

perf tests:

  James Clark:

  - Fixes hang in zstd compression test by changing the source of random data.

perf trace:

  Arnaldo Carvalho de Melo:

  - augmented_raw_syscalls.c BPF helper improvements.

  Benjamin Peterson:

  - Fix off-by-one error in ioctl cmd->string table.

libperf:

  Jiri Olsa:

  - Move most PERF_RECORD_ structs to perf/event.h.

headers:

  Arnaldo Carvalho de Melo:

  - Move cacheline related routines to separate source files.

  - Move record_opts and other record declarations to separate files.

  - Explicitly add some more needed headers here and there.

  Souptick Joarder:

  - Remove some duplicate include directives.

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

perf evsel: Rename perf_missing_features::bpf_event to ::bpf

No need for that _event suffix, do just like all the other meta events
and do away with that.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Link: https://lkml.kernel.org/n/tip-bvc83f380dva83wlg52yd10t@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tool: Rename perf_tool::bpf_event to bpf

No need for that _event suffix, do just like all the other meta event
handlers and suppress that suffix.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Link: https://lkml.kernel.org/n/tip-03spzxtqafbabbbmnm7y4xfx@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Rename perf_event::bpf_event to perf_event::bpf

Just like all the other meta events, that extra _event suffix is just
redundant, ditch it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Link: https://lkml.kernel.org/n/tip-505qwpaizq1k0t6pk13v1ibd@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Rename perf_event::ksymbol_event to perf_event::ksymbol

Just like all the other meta events, that extra _event suffix is just
redundant, ditch it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Link: https://lkml.kernel.org/n/tip-0q8b2xnfs17q0g523oej75s0@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Rename the PERF_RECORD_ structs to have a "perf" suffix

Even more, to have a "perf_record_" prefix, so that they match the
PERF_RECORD_ enum they map to.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-qbabmcz2a0pkzt72liyuz3p8@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_SAMPLE 'struct sample_event' to perf/event.h

Move the PERF_RECORD_SAMPLE event definition to libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-13-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_BPF_EVENT 'struct bpf_event' to perf/event.h

Move the PERF_RECORD_BPF_EVENT event definition to libperf's event.h.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-12-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_KSYMBOL 'struct ksymbol_event' to perf/event.h

Move the PERF_RECORD_KSYMBOL event definition into libperf's event.h
header include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values
as stated in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add and use new PRI_lu64 and PRI_lx64 macros for that.  Use extra '_' to
ease up the reading and differentiate them from standard PRI*64 macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-11-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_THROTTLE 'struct throttle_event' to perf/event.h

Move the PERF_RECORD_THROTTLE event definition into libperf's event.h
header include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values as stated
in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add and use new PRI_lu64 and PRI_lx64 macros for that.  Use extra '_' to
ease up the reading and differentiate them from standard PRI*64 macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-10-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_READ 'struct read_event' to perf/event.h

Move the PERF_RECORD_READ event definition to libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values
as stated in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add and use new PRI_lu64 and PRI_lx64 macros for that.  Use extra '_' to
ease up the reading and differentiate them from standard PRI*64 macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-9-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_LOST_SAMPLES 'struct lost_samples_event' to perf/event.h

Move the PERF_RECORD_LOST_SAMPLES event definition into libperf's
event.h header include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values
as stated in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add and use new PRI_lu64 and PRI_lx64 macros for that.  Use extra '_' to
ease up the reading and differentiate them from standard PRI*64 macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-8-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_LOST 'struct lost_event' to perf/event.h

Move the lost_event event definition to libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values as stated
in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add and use new PRI_lu64 and PRI_lx64 macros for that.  Use extra '_' to
ease up the reading and differentiate them from standard PRI*64 macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-7-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_FORK 'struct fork_event' to perf/event.h

Move the fork_event event definition into libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values
as stated in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add and use new PRI_lu64 and PRI_lx64 macros for that.  Using extra '_'
to ease up the reading and differentiate them from standard PRI*64
macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-6-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_NAMESPACES 'struct namespaces_event' to perf/event.h

Move the namespaces_event event definition into libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-5-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_COMM 'struct comm_event' to perf/event.h

Moving comm_event event definition into libperf's event.h
header include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8'
types used events to their generic '__u*' versions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-4-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_MMAP2 'struct mmap2_event' to perf/event.h

Moving mmap2_event event definition into libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values
as stated in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Adding and using new PRI_lu64 and PRI_lx64 macros to be used for
that.  Using extra '_' to ease up the reading and differentiate
them from standard PRI*64 macros.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lkml.kernel.org/n/tip-ufs9ityr5w2xqwtd5w3p6dm4@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf: Add PERF_RECORD_MMAP 'struct mmap_event' to perf/event.h

Move the mmap_event event definition to libperf's event.h header
include.

In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used
events to their generic '__u*' versions.

Perf added 'u*' types mainly to ease up printing __u64 values as stated
in the linux/types.h comment:

  /*
   * We define u64 as uint64_t for every architecture
   * so that we can print it with "%"PRIx64 without getting warnings.
   *
   * typedef __u64 u64;
   * typedef __s64 s64;
   */

Add  and use new PRI_lu64 and PRI_lx64 macros for that.  Use extra '_'
to ease up reading and differentiate them from standard PRI*64 macros.

Committer notes:

Fixup the PRI_l[ux]64 macros on 32-bit arches, conditionally defining it
with that extra 'l' modifier only on arches where __u64 is long long,
leaving it aside on 32-bit arches.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190825181752.722-2-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script: Fix memory leaks in list_scripts()

In case memory resources for *buf* and *paths* were allocated, jump to
*out* and release them before return.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Gustavo A. R. Silva <gustavo@embeddedor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Addresses-Coverity-ID: 1444328 ("Resource leak")
Fixes: 6f3da20e151f ("perf report: Support builtin perf script in scripts menu")
Link: http://lkml.kernel.org/r/20190408162748.GA21008@embeddedor
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Fix --ns time sort key output

If the user specified --ns, the column to print the sort time stamp
wasn't wide enough to actually print the full nanoseconds.

Widen the time key column width when --ns is specified.

Before:

  % perf record -a sleep 1
  % perf report --sort time,overhead,symbol --stdio --ns
  ...
       2.39%  187851.10000  [k] smp_call_function_single   -      -
       1.53%  187851.10000  [k] intel_idle                 -      -
       0.59%  187851.10000  [.] __wcscmp_ifunc             -      -
       0.33%  187851.10000  [.] 0000000000000000           -      -
       0.28%  187851.10000  [k] cpuidle_enter_state        -      -

After:

  % perf report --sort time,overhead,symbol --stdio --ns
  ...
       2.39%  187851.100000000  [k] smp_call_function_single   -      -
       1.53%  187851.100000000  [k] intel_idle                 -      -
       0.59%  187851.100000000  [.] __wcscmp_ifunc             -      -
       0.33%  187851.100000000  [.] 0000000000000000           -      -
       0.28%  187851.100000000  [k] cpuidle_enter_state        -      -

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/20190823210338.12360-2-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Use timestamp__scnprintf_nsec() for time sort key

Use timestamp__scnprintf_nsec() to print nanoseconds for the time sort
key, instead of open coding.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/20190823210338.12360-1-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove duplicate headers

Removed headers which are included twice.

Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Reviewed-by: Mukesh Ojha <mojha@codeaurora.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1566663319-4283-1-git-send-email-jrdr.linux@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Reduce perf_event_output() boilerplate

Add a augmented__output() helper to reduce the boilerplate of sending
the augmented tracepoint to the PERF_EVENT_ARRAY BPF map associated with
the bpf-output event used to communicate with the userspace perf trace
tool.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-ln99gt0j4fv0kw0778h6vphm@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Introduce helper to get the scratch space

We need more than the BPF stack can give us to format the
raw_syscalls:sys_enter augmented tracepoint, so we use a PERCPU_ARRAY
map for that, use a helper to shorten the sequence to access that area.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Postpone tmp map lookup to after pid_filter

No sense in doing that lookup before figuring out if it will be used,
i.e. if the pid is being filtered that tmp space lookup will be useless.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-o74yggieorucfg4j74tb6rta@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Rename augmented_filename to augmented_arg

Because it is not used only for strings, we already use it for sockaddr
structs and will use it for all other types.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-w9nkt3tvmyn5i4qnwng3ap1k@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace beauty ioctl: Fix off-by-one error in cmd->string table

While tracing a program that calls isatty(3), I noticed that strace
reported TCGETS for the request argument of the underlying ioctl(2)
syscall while perf trace reported TCSETS. strace is corrrect. The bug in
perf was due to the tty ioctl beauty table starting at 0x5400 rather
than 0x5401.

Committer testing:

  Using augmented_raw_syscalls.o and settings to make 'perf trace'
  use strace formatting, i.e. with this in ~/.perfconfig

  # cat ~/.perfconfig
  [trace]
add_events = /home/acme/git/linux/tools/perf/examples/bpf/augmented_raw_syscalls.c
show_zeros = yes
show_duration = no
no_inherit = yes
show_timestamp = no
show_arg_names = no
args_alignment = 40
show_prefix = yes

  # strace -e ioctl stty > /dev/null
  ioctl(0, TCGETS, {B38400 opost isig icanon echo ...}) = 0
  ioctl(1, TIOCGWINSZ, 0x7fff8a9b0860)    = -1 ENOTTY (Inappropriate ioctl for device)
  ioctl(1, TCGETS, 0x7fff8a9b0540)        = -1 ENOTTY (Inappropriate ioctl for device)
  +++ exited with 0 +++
  #

Before:

  # perf trace -e ioctl stty > /dev/null
  ioctl(0, TCSETS, 0x7fff2cf79f20)        = 0
  ioctl(1, TIOCSWINSZ, 0x7fff2cf79f40)    = -1 ENOTTY (Inappropriate ioctl for device)
  ioctl(1, TCSETS, 0x7fff2cf79c20)        = -1 ENOTTY (Inappropriate ioctl for device)
  #

After:

  # perf trace -e ioctl stty > /dev/null
  ioctl(0, TCGETS, 0x7ffed0763920)        = 0
  ioctl(1, TIOCGWINSZ, 0x7ffed0763940)    = -1 ENOTTY (Inappropriate ioctl for device)
  ioctl(1, TCGETS, 0x7ffed0763620)        = -1 ENOTTY (Inappropriate ioctl for device)
  #

Signed-off-by: Benjamin Peterson <benjamin@python.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Fixes: 1cc47f2d46206d67285aea0ca7e8450af571da13 ("perf trace beauty ioctl: Improve 'cmd' beautifier")
Link: http://lkml.kernel.org/r/20190823033625.18814-1-benjamin@python.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Fixes hang in zstd compression test by changing the source of random data

Running 'perf test' with zstd compression linked will hang at the test
'Zstd perf.data compression/decompression' because /dev/random blocks
reads until there is enough entropy. This means that the test will
appear to never complete unless the mouse is continually moved while
running it.

Signed-off-by: James Clark <james.clark@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Jeremy Linton <jeremy.linton@arm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/3d8cc701-df4e-f949-1715-5118b530e990@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf sort: Remove needless headers from sort.h, provide fwd struct decls

Reducing the includes hell a bit more, speeding up the build and
avoiding needless rebuilds when just one of those files gets updated.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-u63el2vqsovsmnhebx1rcixo@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf srcline: Add missing srcline.h header to files needing its defs

When srcline was introduced it wrongly added the include to util/sort.h,
even with that header not needing the definitions it provides, fix it by
adding it to the places that need it as a pre patch to remove srcline.h
from sort.h.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-shuebppedtye8hrgxk15qe3x@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cacheline: Move cacheline related routines to separate files

To disentangle util/sort.h a bit more.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-6kbf2cauas06rbqp15pyter5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Move record_opts and other record decls out of perf.h

And into a separate util/record.h, to better isolate things and make
sure that those who use record_opts and the other moved declarations
are explicitly including the necessary header.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-31q8mei1qkh74qvkl9nwidfq@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf stat: Remove needless headers from stat.h

Just a forward declaration for 'struct timespec' is needed, ditch the
rest.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-6shdqw801oqe7ax6r307k27r@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cpumap: No need to include perf.h, ditch it

From a quick look this was never needed and just polluted the build,
needlessly making things including cpumap.h to be rebuild if perf.h or
anything it includes gets changed.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-x10p8slllqkn3fc3bntjx3n0@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/x86/intel/pt: Get rid of reverse lookup table for ToPA

In order to quickly find a ToPA entry by its page offset in the buffer,
we're using a reverse lookup table. The problem with it is that it's a
large array of mostly similar pointers, especially so now that we're
using high order allocations from the page allocator. Because its size
is limited to whatever is the maximum for kmalloc(), it places a limit
on the number of ToPA entries per buffer, and therefore, on the total
buffer size, which otherwise doesn't have to be there.

Replace the reverse lookup table with a simple runtime lookup. With the
high order AUX allocations in place, the runtime penalty of such a lookup
is much smaller and in cases where all entries in a ToPA table are of
the same size, the complexity is O(1).

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20190821124727.73310-7-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/pt: Free up space in a ToPA descriptor

Currently, we're storing physical address of a ToPA table in its
descriptor, which is completely unnecessary. Since the descriptor
and the table itself share the same page, reducing the descriptor
size leaves more space for the table.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20190821124727.73310-6-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/pt: Split ToPA metadata and page layout

PT uses page sized ToPA tables, where the ToPA table resides at the bottom
and its driver-specific metadata taking up a few words at the top of the
page. The split is currently calculated manually and needs to be redone
every time a field is added to or removed from the metadata structure.
Also, the 32-bit version can be made smaller.

By splitting the table and metadata into separate structures, we are making
the compiler figure out the division of the page.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20190821124727.73310-5-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/pt: Use pointer arithmetics instead in ToPA entry calculation

Currently, pt_buffer_reset_offsets() calculates the current ToPA entry by
casting pointers to addresses and performing ungainly subtractions and
divisions instead of a simpler pointer arithmetic, which would be perfectly
applicable in that case. Fix that.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20190821124727.73310-4-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/pt: Use helpers to obtain ToPA entry size

There are a few places in the PT driver that need to obtain the size of
a ToPA entry, some of them for the current ToPA entry in the buffer.
Use helpers for those, to make the lines shorter and more readable.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20190821124727.73310-3-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/pt: Clean up ToPA allocation path

Some of the allocation parameters are passed as function arguments,
while the CPU number for per-cpu allocation is passed via the buffer
object. There's no reason for this.

Pass the CPU as a function argument instead.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/20190821124727.73310-2-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'v5.3-rc6' into perf/core, to pick up fixes

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

Linux 5.3-rc6

Merge tag 'auxdisplay-for-linus-v5.3-rc7' of git://github.com/ojeda/linux

Pull auxdisplay cleanup from Miguel Ojeda:
 "Make ht16k33_fb_fix and ht16k33_fb_var constant (Nishka Dasgupta)"

* tag 'auxdisplay-for-linus-v5.3-rc7' of git://github.com/ojeda/linux:
  auxdisplay: ht16k33: Make ht16k33_fb_fix and ht16k33_fb_var constant

Merge tag 'for-linus-5.3-rc6' of git://git./linux/kernel/git/rw/uml

Pull UML fix from Richard Weinberger:
 "Fix time travel mode"

* tag 'for-linus-5.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml:
  um: fix time travel mode

Merge tag 'for-linus-5.3-rc6' of git://git./linux/kernel/git/rw/ubifs

Pull UBIFS and JFFS2 fixes from Richard Weinberger:
 "UBIFS:
   - Don't block too long in writeback_inodes_sb()
   - Fix for a possible overrun of the log head
   - Fix double unlock in orphan_delete()

  JFFS2:
   - Remove C++ style from UAPI header and unbreak picky toolchains"

* tag 'for-linus-5.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/ubifs:
  ubifs: Limit the number of pages in shrink_liability
  ubifs: Correctly initialize c->min_log_bytes
  ubifs: Fix double unlock around orphan_delete()
  jffs2: Remove C++ style comments from uapi header

Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:
 "A few fixes for x86:

   - Fix a boot regression caused by the recent bootparam sanitizing
     change, which escaped the attention of all people who reviewed that
     code.

   - Address a boot problem on machines with broken E820 tables caused
     by an underflow which ended up placing the trampoline start at
     physical address 0.

   - Handle machines which do not advertise a legacy timer of any form,
     but need calibration of the local APIC timer gracefully by making
     the calibration routine independent from the tick interrupt. Marked
     for stable as well as there seems to be quite some new laptops
     rolled out which expose this.

   - Clear the RDRAND CPUID bit on AMD family 15h and 16h CPUs which are
     affected by broken firmware which does not initialize RDRAND
     correctly after resume. Add a command line parameter to override
     this for machine which either do not use suspend/resume or have a
     fixed BIOS. Unfortunately there is no way to detect this on boot,
     so the only safe decision is to turn it off by default.

   - Prevent RFLAGS from being clobbers in CALL_NOSPEC on 32bit which
     caused fast KVM instruction emulation to break.

   - Explain the Intel CPU model naming convention so that the repeating
     discussions come to an end"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/retpoline: Don't clobber RFLAGS during CALL_NOSPEC on i386
  x86/boot: Fix boot regression caused by bootparam sanitizing
  x86/CPU/AMD: Clear RDRAND CPUID bit on AMD family 15h/16h
  x86/boot/compressed/64: Fix boot on machines with broken E820 table
  x86/apic: Handle missing global clockevent gracefully
  x86/cpu: Explain Intel model naming convention

Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timekeeping fix from Thomas Gleixner:
 "A single fix for a regression caused by the generic VDSO
  implementation where a math overflow causes CLOCK_BOOTTIME to become a
  random number generator"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  timekeeping/vsyscall: Prevent math overflow in BOOTTIME update

Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Thomas Gleixner:
 "Handle the worker management in situations where a task is scheduled
  out on a PI lock contention correctly and schedule a new worker if
  possible"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/core: Schedule new worker even if PI-blocked

Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Thomas Gleixner:
 "Two small fixes for kprobes and perf:

   - Prevent a deadlock in kprobe_optimizer() causes by reverse lock
     ordering

   - Fix a comment typo"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  kprobes: Fix potential deadlock in kprobe_optimizer()
  perf/x86: Fix typo in comment

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

Pull irq fix from Thomas Gleixner:
 "A single fix for a imbalanced kobject operation in the irq decriptor
  code which was unearthed by the new warnings in the kobject code"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Properly pair kobject_del() with kobject_add()

Merge branch 'akpm' (patches from Andrew)

Mergr misc fixes from Andrew Morton:
 "11 fixes"

Mostly VM fixes, one psi polling fix, and one parisc build fix.

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm/kasan: fix false positive invalid-free reports with CONFIG_KASAN_SW_TAGS=y
  mm/zsmalloc.c: fix race condition in zs_destroy_pool
  mm/zsmalloc.c: migration can leave pages in ZS_EMPTY indefinitely
  mm, page_owner: handle THP splits correctly
  userfaultfd_release: always remove uffd flags and clear vm_userfaultfd_ctx
  psi: get poll_work to run when calling poll syscall next time
  mm: memcontrol: flush percpu vmevents before releasing memcg
  mm: memcontrol: flush percpu vmstats before releasing memcg
  parisc: fix compilation errrors
  mm, page_alloc: move_freepages should not examine struct page of reserved memory
  mm/z3fold.c: fix race between migration and destruction

Merge tag 'dma-mapping-5.3-5' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fixes from Christoph Hellwig:
 "Two fixes for regressions in this merge window:

   - select the Kconfig symbols for the noncoherent dma arch helpers on
     arm if swiotlb is selected, not just for LPAE to not break then Xen
     build, that uses swiotlb indirectly through swiotlb-xen

   - fix the page allocator fallback in dma_alloc_contiguous if the CMA
     allocation fails"

* tag 'dma-mapping-5.3-5' of git://git.infradead.org/users/hch/dma-mapping:
  dma-direct: fix zone selection after an unaddressable CMA allocation
  arm: select the dma-noncoherent symbols for all swiotlb builds

mm/kasan: fix false positive invalid-free reports with CONFIG_KASAN_SW_TAGS=y

The code like this:

ptr = kmalloc(size, GFP_KERNEL);
page = virt_to_page(ptr);
offset = offset_in_page(ptr);
kfree(page_address(page) + offset);

may produce false-positive invalid-free reports on the kernel with
CONFIG_KASAN_SW_TAGS=y.

In the example above we lose the original tag assigned to 'ptr', so
kfree() gets the pointer with 0xFF tag.  In kfree() we check that 0xFF
tag is different from the tag in shadow hence print false report.

Instead of just comparing tags, do the following:

1) Check that shadow doesn't contain KASAN_TAG_INVALID.  Otherwise it's
   double-free and it doesn't matter what tag the pointer have.

2) If pointer tag is different from 0xFF, make sure that tag in the
   shadow is the same as in the pointer.

Link: http://lkml.kernel.org/r/20190819172540.19581-1-aryabinin@virtuozzo.com
Fixes: 7f94ffbc4c6a ("kasan: add hooks implementation for tag-based mode")
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reported-by: Walter Wu <walter-zh.wu@mediatek.com>
Reported-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc.c: fix race condition in zs_destroy_pool

In zs_destroy_pool() we call flush_work(&pool->free_work).  However, we
have no guarantee that migration isn't happening in the background at
that time.

Since migration can't directly free pages, it relies on free_work being
scheduled to free the pages.  But there's nothing preventing an
in-progress migrate from queuing the work *after*
zs_unregister_migration() has called flush_work().  Which would mean
pages still pointing at the inode when we free it.

Since we know at destroy time all objects should be free, no new
migrations can come in (since zs_page_isolate() fails for fully-free
zspages).  This means it is sufficient to track a "# isolated zspages"
count by class, and have the destroy logic ensure all such pages have
drained before proceeding.  Keeping that state under the class spinlock
keeps the logic straightforward.

In this case a memory leak could lead to an eventual crash if compaction
hits the leaked page.  This crash would only occur if people are
changing their zswap backend at runtime (which eventually starts
destruction).

Link: http://lkml.kernel.org/r/20190809181751.219326-2-henryburns@google.com
Fixes: 48b4800a1c6a ("zsmalloc: page migration support")
Signed-off-by: Henry Burns <henryburns@google.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Henry Burns <henrywolfeburns@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Jonathan Adams <jwadams@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc.c: migration can leave pages in ZS_EMPTY indefinitely

In zs_page_migrate() we call putback_zspage() after we have finished
migrating all pages in this zspage.  However, the return value is
ignored.  If a zs_free() races in between zs_page_isolate() and
zs_page_migrate(), freeing the last object in the zspage,
putback_zspage() will leave the page in ZS_EMPTY for potentially an
unbounded amount of time.

To fix this, we need to do the same thing as zs_page_putback() does:
schedule free_work to occur.

To avoid duplicated code, move the sequence to a new
putback_zspage_deferred() function which both zs_page_migrate() and
zs_page_putback() call.

Link: http://lkml.kernel.org/r/20190809181751.219326-1-henryburns@google.com
Fixes: 48b4800a1c6a ("zsmalloc: page migration support")
Signed-off-by: Henry Burns <henryburns@google.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Henry Burns <henrywolfeburns@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Jonathan Adams <jwadams@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_owner: handle THP splits correctly

THP splitting path is missing the split_page_owner() call that
split_page() has.

As a result, split THP pages are wrongly reported in the page_owner file
as order-9 pages.  Furthermore when the former head page is freed, the
remaining former tail pages are not listed in the page_owner file at
all.  This patch fixes that by adding the split_page_owner() call into
__split_huge_page().

Link: http://lkml.kernel.org/r/20190820131828.22684-2-vbabka@suse.cz
Fixes: a9627bc5e34e ("mm/page_owner: introduce split_page_owner and replace manual handling")
Reported-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

userfaultfd_release: always remove uffd flags and clear vm_userfaultfd_ctx

userfaultfd_release() should clear vm_flags/vm_userfaultfd_ctx even if
mm->core_state != NULL.

Otherwise a page fault can see userfaultfd_missing() == T and use an
already freed userfaultfd_ctx.

Link: http://lkml.kernel.org/r/20190820160237.GB4983@redhat.com
Fixes: 04f5866e41fb ("coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping")
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reported-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Tested-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Jann Horn <jannh@google.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

psi: get poll_work to run when calling poll syscall next time

Only when calling the poll syscall the first time can user receive
POLLPRI correctly.  After that, user always fails to acquire the event
signal.

Reproduce case:
 1. Get the monitor code in Documentation/accounting/psi.txt
 2. Run it, and wait for the event triggered.
 3. Kill and restart the process.

The question is why we can end up with poll_scheduled = 1 but the work
not running (which would reset it to 0).  And the answer is because the
scheduling side sees group->poll_kworker under RCU protection and then
schedules it, but here we cancel the work and destroy the worker.  The
cancel needs to pair with resetting the poll_scheduled flag.

Link: http://lkml.kernel.org/r/1566357985-97781-1-git-send-email-joseph.qi@linux.alibaba.com
Signed-off-by: Jason Xing <kerneljasonxing@linux.alibaba.com>
Signed-off-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: Caspar Zhang <caspar@linux.alibaba.com>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: memcontrol: flush percpu vmevents before releasing memcg

Similar to vmstats, percpu caching of local vmevents leads to an
accumulation of errors on non-leaf levels.  This happens because some
leftovers may remain in percpu caches, so that they are never propagated
up by the cgroup tree and just disappear into nonexistence with on
releasing of the memory cgroup.

To fix this issue let's accumulate and propagate percpu vmevents values
before releasing the memory cgroup similar to what we're doing with
vmstats.

Since on cpu hotplug we do flush percpu vmstats anyway, we can iterate
only over online cpus.

Link: http://lkml.kernel.org/r/20190819202338.363363-4-guro@fb.com
Fixes: 42a300353577 ("mm: memcontrol: fix recursive statistics correctness & scalabilty")
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: memcontrol: flush percpu vmstats before releasing memcg

Percpu caching of local vmstats with the conditional propagation by the
cgroup tree leads to an accumulation of errors on non-leaf levels.

Let's imagine two nested memory cgroups A and A/B.  Say, a process
belonging to A/B allocates 100 pagecache pages on the CPU 0.  The percpu
cache will spill 3 times, so that 32*3=96 pages will be accounted to A/B
and A atomic vmstat counters, 4 pages will remain in the percpu cache.

Imagine A/B is nearby memory.max, so that every following allocation
triggers a direct reclaim on the local CPU.  Say, each such attempt will
free 16 pages on a new cpu.  That means every percpu cache will have -16
pages, except the first one, which will have 4 - 16 = -12.  A/B and A
atomic counters will not be touched at all.

Now a user removes A/B.  All percpu caches are freed and corresponding
vmstat numbers are forgotten.  A has 96 pages more than expected.

As memory cgroups are created and destroyed, errors do accumulate.  Even
1-2 pages differences can accumulate into large numbers.

To fix this issue let's accumulate and propagate percpu vmstat values
before releasing the memory cgroup.  At this point these numbers are
stable and cannot be changed.

Since on cpu hotplug we do flush percpu vmstats anyway, we can iterate
only over online cpus.

Link: http://lkml.kernel.org/r/20190819202338.363363-2-guro@fb.com
Fixes: 42a300353577 ("mm: memcontrol: fix recursive statistics correctness & scalabilty")
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

parisc: fix compilation errrors

Commit 0cfaee2af3a0 ("include/asm-generic/5level-fixup.h: fix variable
'p4d' set but not used") converted a few functions from macros to static
inline, which causes parisc to complain,

  In file included from include/asm-generic/4level-fixup.h:38:0,
                   from arch/parisc/include/asm/pgtable.h:5,
                   from arch/parisc/include/asm/io.h:6,
                   from include/linux/io.h:13,
                   from sound/core/memory.c:9:
  include/asm-generic/5level-fixup.h:14:18: error: unknown type name 'pgd_t'; did you mean 'pid_t'?
   #define p4d_t    pgd_t
                    ^
  include/asm-generic/5level-fixup.h:24:28: note: in expansion of macro 'p4d_t'
   static inline int p4d_none(p4d_t p4d)
                              ^~~~~

It is because "4level-fixup.h" is included before "asm/page.h" where
"pgd_t" is defined.

Link: http://lkml.kernel.org/r/20190815205305.1382-1-cai@lca.pw
Fixes: 0cfaee2af3a0 ("include/asm-generic/5level-fixup.h: fix variable 'p4d' set but not used")
Signed-off-by: Qian Cai <cai@lca.pw>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: move_freepages should not examine struct page of reserved memory

After commit 907ec5fca3dc ("mm: zero remaining unavailable struct
pages"), struct page of reserved memory is zeroed.  This causes
page->flags to be 0 and fixes issues related to reading
/proc/kpageflags, for example, of reserved memory.

The VM_BUG_ON() in move_freepages_block(), however, assumes that
page_zone() is meaningful even for reserved memory.  That assumption is
no longer true after the aforementioned commit.

There's no reason why move_freepages_block() should be testing the
legitimacy of page_zone() for reserved memory; its scope is limited only
to pages on the zone's freelist.

Note that pfn_valid() can be true for reserved memory: there is a
backing struct page.  The check for page_to_nid(page) is also buggy but
reserved memory normally only appears on node 0 so the zeroing doesn't
affect this.

Move the debug checks to after verifying PageBuddy is true.  This
isolates the scope of the checks to only be for buddy pages which are on
the zone's freelist which move_freepages_block() is operating on.  In
this case, an incorrect node or zone is a bug worthy of being warned
about (and the examination of struct page is acceptable bcause this
memory is not reserved).

Why does move_freepages_block() gets called on reserved memory? It's
simply math after finding a valid free page from the per-zone free area
to use as fallback.  We find the beginning and end of the pageblock of
the valid page and that can bring us into memory that was reserved per
the e820.  pfn_valid() is still true (it's backed by a struct page), but
since it's zero'd we shouldn't make any inferences here about comparing
its node or zone.  The current node check just happens to succeed most
of the time by luck because reserved memory typically appears on node 0.

The fix here is to validate that we actually have buddy pages before
testing if there's any type of zone or node strangeness going on.

We noticed it almost immediately after bringing 907ec5fca3dc in on
CONFIG_DEBUG_VM builds.  It depends on finding specific free pages in
the per-zone free area where the math in move_freepages() will bring the
start or end pfn into reserved memory and wanting to claim that entire
pageblock as a new migratetype.  So the path will be rare, require
CONFIG_DEBUG_VM, and require fallback to a different migratetype.

Some struct pages were already zeroed from reserve pages before
907ec5fca3c so it theoretically could trigger before this commit.  I
think it's rare enough under a config option that most people don't run
that others may not have noticed.  I wouldn't argue against a stable tag
and the backport should be easy enough, but probably wouldn't single out
a commit that this is fixing.

Mel said:

: The overhead of the debugging check is higher with this patch although
: it'll only affect debug builds and the path is not particularly hot.
: If this was a concern, I think it would be reasonable to simply remove
: the debugging check as the zone boundaries are checked in
: move_freepages_block and we never expect a zone/node to be smaller than
: a pageblock and stuck in the middle of another zone.

Link: http://lkml.kernel.org/r/alpine.DEB.2.21.1908122036560.10779@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pavel Tatashin <pavel.tatashin@microsoft.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/z3fold.c: fix race between migration and destruction

In z3fold_destroy_pool() we call destroy_workqueue(&pool->compact_wq).
However, we have no guarantee that migration isn't happening in the
background at that time.

Migration directly calls queue_work_on(pool->compact_wq), if destruction
wins that race we are using a destroyed workqueue.

Link: http://lkml.kernel.org/r/20190809213828.202833-1-henryburns@google.com
Signed-off-by: Henry Burns <henryburns@google.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Jonathan Adams <jwadams@google.com>
Cc: Henry Burns <henrywolfeburns@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>