Merge tag 'perf-core-for-mingo-4.16-20171206' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core

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

- Improve build messages for files needed by Intel-PT, originally copied
  from the kernel sources, that drifted from its original (Adrian Hunter)

- Allow computing 'perf stat' style metrics in 'perf script' (Andi Kleen)

- Fix 'perf report -D' output for user metadata events (Arnaldo Carvalho de Melo)

- Add feature test for pthread_barrier_t availability (Arnaldo Carvalho de Melo)

- Allow again using x86's asm.h when building for the 'bpf' clang target,
  making some 'perf test' LLVM/BPF entries work again (Arnaldo Carvalho de Melo)

- Use cpumaps in 'perf bench futex', eliminating some code duplication
  (Davidlohr Bueso)

- Improve PMU infrastructure to support amp64's ThunderX2 implementation
  defined core events (Ganapatrao Kulkarni)

- Add hint about how to add USDT probes for Node.js (Hansuk Hong)

- s/390 needs -fPIC to be incrementally linked or linked to shared
  libraries (Hendrik Brueckner)

- Use pthread_barrier to synch 'perf bench futex wake-parallel' waker
  threads (James Yang)

- Fix up build in hardened environments, such as Fedora 27 (Jiri Olsa)

- Add a tip about cacheline events in 'perf c2c' (Sangwon Hong)

- Set browser mode right before setup_browser(), because we may have
  errors printed before that, which were getting lost (Seokho Song)

- s390x doesn't support PERF_TYPE_BREAKPOINT, so disable 'perf test'
  cases 19 and 20 on s390x, that tests that feature (Thomas Richter)

- Fix unnecessary memory allocation for s390x 'perf annotate' objdump
  parsing, which could lead to thousands of needless entries in the
  instruction handling array (Thomas Richter)

- Fix objdump comment parsing for Intel mov dissassembly (Thomas Richter)

- Clarify usage of 'overwrite' and 'backward' in the evlist/mmap code,
  removing the 'overwrite' parameter from several functions as it was
  always used it as 'false' (Wang Nan)

- Fix 'perf record' backward recording, it wasn't doing what was
  expected: overwriting records when the ring buffer gets full (Wang Nan)

- Use more flexible pattern matching for CPU identification for perf
  vendor event's mapfile.csv, removing the need for a new perf binary
  for a sligthly different chip revision that shares the same set of
  counters (William Cohen)

Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>

1 files changed, 5 insertions, 3 deletions

diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c
index 22ea7936d92f..facad1e279a8 100644
--- a/tools/perf/util/annotate.c
+++ b/tools/perf/util/annotate.c
@@ -322,6 +322,8 @@ static int comment__symbol(char *raw, char *comment, u64 *addrp, char **namep)
 		return 0;
 
 	*addrp = strtoull(comment, &endptr, 16);
+	if (endptr == comment)
+		return 0;
 	name = strchr(endptr, '<');
 	if (name == NULL)
 		return -1;
@@ -435,8 +437,8 @@ static int mov__parse(struct arch *arch, struct ins_operands *ops, struct map *m
 		return 0;
 
 	comment = ltrim(comment);
-	comment__symbol(ops->source.raw, comment, &ops->source.addr, &ops->source.name);
-	comment__symbol(ops->target.raw, comment, &ops->target.addr, &ops->target.name);
+	comment__symbol(ops->source.raw, comment + 1, &ops->source.addr, &ops->source.name);
+	comment__symbol(ops->target.raw, comment + 1, &ops->target.addr, &ops->target.name);
 
 	return 0;
 
@@ -480,7 +482,7 @@ static int dec__parse(struct arch *arch __maybe_unused, struct ins_operands *ops
 		return 0;
 
 	comment = ltrim(comment);
-	comment__symbol(ops->target.raw, comment, &ops->target.addr, &ops->target.name);
+	comment__symbol(ops->target.raw, comment + 1, &ops->target.addr, &ops->target.name);
 
 	return 0;
 }