perf lock contention: Trim backtrace by skipping traceiter functions
authorAnne Macedo <retpolanne@posteo.net>
Tue, 19 Mar 2024 14:36:26 +0000 (14:36 +0000)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Thu, 21 Mar 2024 23:44:35 +0000 (20:44 -0300)
The 'perf lock contention' program currently shows the caller of the locks
as __traceiter_contention_begin+0x??. This caller can be ignored, as it is
from the traceiter itself. Instead, it should show the real callers for
the locks.

When fiddling with the --stack-skip parameter, the actual callers for
the locks start to show up. However, just ignore the
__traceiter_contention_begin and the __traceiter_contention_end symbols
so the actual callers will show up.

Before this patch is applied:

sudo perf lock con -a -b -- sleep 3
 contended   total wait     max wait     avg wait         type   caller

         8      2.33 s       2.28 s     291.18 ms     rwlock:W   __traceiter_contention_begin+0x44
         4      2.33 s       2.28 s     582.35 ms     rwlock:W   __traceiter_contention_begin+0x44
         7    140.30 ms     46.77 ms     20.04 ms     rwlock:W   __traceiter_contention_begin+0x44
         2     63.35 ms     33.76 ms     31.68 ms        mutex   trace_contention_begin+0x84
         2     46.74 ms     46.73 ms     23.37 ms     rwlock:W   __traceiter_contention_begin+0x44
         1     13.54 us     13.54 us     13.54 us        mutex   trace_contention_begin+0x84
         1      3.67 us      3.67 us      3.67 us      rwsem:R   __traceiter_contention_begin+0x44

Before this patch is applied - using --stack-skip 5

sudo perf lock con --stack-skip 5 -a -b -- sleep 3
 contended   total wait     max wait     avg wait         type   caller

         2      2.24 s       2.24 s       1.12 s      rwlock:W   do_epoll_wait+0x5a0
         4      1.65 s     824.21 ms    412.08 ms     rwlock:W   do_exit+0x338
         2    824.35 ms    824.29 ms    412.17 ms     spinlock   get_signal+0x108
         2    824.14 ms    824.14 ms    412.07 ms     rwlock:W   release_task+0x68
         1     25.22 ms     25.22 ms     25.22 ms        mutex   cgroup_kn_lock_live+0x58
         1     24.71 us     24.71 us     24.71 us     spinlock   do_exit+0x44
         1     22.04 us     22.04 us     22.04 us      rwsem:R   lock_mm_and_find_vma+0xb0

After this patch is applied:

sudo ./perf lock con -a -b -- sleep 3
 contended   total wait     max wait     avg wait         type   caller

         4      4.13 s       2.07 s       1.03 s      rwlock:W   release_task+0x68
         2      2.07 s       2.07 s       1.03 s      rwlock:R   mm_update_next_owner+0x50
         2      2.07 s       2.07 s       1.03 s      rwlock:W   do_exit+0x338
         1     41.56 ms     41.56 ms     41.56 ms        mutex   cgroup_kn_lock_live+0x58
         2     36.12 us     18.83 us     18.06 us     rwlock:W   do_exit+0x338

Signed-off-by: Anne Macedo <retpolanne@posteo.net>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240319143629.3422590-1-retpolanne@posteo.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/util/machine.c
tools/perf/util/machine.h

index 527517db31821ffa3d80c349c7433dcdc4f7cdfe..5eb9044bc223c36a9e5b90263e134824d17b47c5 100644 (file)
@@ -3266,6 +3266,17 @@ bool machine__is_lock_function(struct machine *machine, u64 addr)
 
                sym = machine__find_kernel_symbol_by_name(machine, "__lock_text_end", &kmap);
                machine->lock.text_end = map__unmap_ip(kmap, sym->start);
+
+               sym = machine__find_kernel_symbol_by_name(machine, "__traceiter_contention_begin", &kmap);
+               if (sym) {
+                       machine->traceiter.text_start = map__unmap_ip(kmap, sym->start);
+                       machine->traceiter.text_end = map__unmap_ip(kmap, sym->end);
+               }
+               sym = machine__find_kernel_symbol_by_name(machine, "trace_contention_begin", &kmap);
+               if (sym) {
+                       machine->trace.text_start = map__unmap_ip(kmap, sym->start);
+                       machine->trace.text_end = map__unmap_ip(kmap, sym->end);
+               }
        }
 
        /* failed to get kernel symbols */
@@ -3280,5 +3291,18 @@ bool machine__is_lock_function(struct machine *machine, u64 addr)
        if (machine->lock.text_start <= addr && addr < machine->lock.text_end)
                return true;
 
+       /* traceiter functions currently don't have their own section
+        * but we consider them lock functions
+        */
+       if (machine->traceiter.text_start != 0) {
+               if (machine->traceiter.text_start <= addr && addr < machine->traceiter.text_end)
+                       return true;
+       }
+
+       if (machine->trace.text_start != 0) {
+               if (machine->trace.text_start <= addr && addr < machine->trace.text_end)
+                       return true;
+       }
+
        return false;
 }
index e28c787616fe4ea3cd6b0ed30aece17cfd0b92e3..4312f6db6de0a042cbd30ae3d19cdf6f73d887e5 100644 (file)
@@ -49,7 +49,7 @@ struct machine {
        struct {
                u64       text_start;
                u64       text_end;
-       } sched, lock;
+       } sched, lock, traceiter, trace;
        pid_t             *current_tid;
        size_t            current_tid_sz;
        union { /* Tool specific area */