]>
Commit | Line | Data |
---|---|---|
7264e8a9 SL |
1 | From ff9acdd435edf2a89c5c9ec652b792dbe8df0a1f Mon Sep 17 00:00:00 2001 |
2 | From: Stephane Eranian <eranian@google.com> | |
3 | Date: Wed, 6 Mar 2019 11:50:48 -0800 | |
4 | Subject: perf/x86/intel: Fix handling of wakeup_events for multi-entry PEBS | |
5 | ||
6 | [ Upstream commit 583feb08e7f7ac9d533b446882eb3a54737a6dbb ] | |
7 | ||
8 | When an event is programmed with attr.wakeup_events=N (N>0), it means | |
9 | the caller is interested in getting a user level notification after | |
10 | N samples have been recorded in the kernel sampling buffer. | |
11 | ||
12 | With precise events on Intel processors, the kernel uses PEBS. | |
13 | The kernel tries minimize sampling overhead by verifying | |
14 | if the event configuration is compatible with multi-entry PEBS mode. | |
15 | If so, the kernel is notified only when the buffer has reached its threshold. | |
16 | Other PEBS operates in single-entry mode, the kenrel is notified for each | |
17 | PEBS sample. | |
18 | ||
19 | The problem is that the current implementation look at frequency | |
20 | mode and event sample_type but ignores the wakeup_events field. Thus, | |
21 | it may not be possible to receive a notification after each precise event. | |
22 | ||
23 | This patch fixes this problem by disabling multi-entry PEBS if wakeup_events | |
24 | is non-zero. | |
25 | ||
26 | Signed-off-by: Stephane Eranian <eranian@google.com> | |
27 | Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> | |
28 | Reviewed-by: Andi Kleen <ak@linux.intel.com> | |
29 | Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> | |
30 | Cc: Arnaldo Carvalho de Melo <acme@redhat.com> | |
31 | Cc: Jiri Olsa <jolsa@redhat.com> | |
32 | Cc: Linus Torvalds <torvalds@linux-foundation.org> | |
33 | Cc: Peter Zijlstra <peterz@infradead.org> | |
34 | Cc: Thomas Gleixner <tglx@linutronix.de> | |
35 | Cc: Vince Weaver <vincent.weaver@maine.edu> | |
36 | Cc: kan.liang@intel.com | |
37 | Link: https://lkml.kernel.org/r/20190306195048.189514-1-eranian@google.com | |
38 | Signed-off-by: Ingo Molnar <mingo@kernel.org> | |
39 | Signed-off-by: Sasha Levin <sashal@kernel.org> | |
40 | --- | |
41 | arch/x86/kernel/cpu/perf_event_intel.c | 2 +- | |
42 | 1 file changed, 1 insertion(+), 1 deletion(-) | |
43 | ||
44 | diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c | |
45 | index 7b79c80ce029a..325ed90511cff 100644 | |
46 | --- a/arch/x86/kernel/cpu/perf_event_intel.c | |
47 | +++ b/arch/x86/kernel/cpu/perf_event_intel.c | |
48 | @@ -2513,7 +2513,7 @@ static int intel_pmu_hw_config(struct perf_event *event) | |
49 | return ret; | |
50 | ||
51 | if (event->attr.precise_ip) { | |
52 | - if (!event->attr.freq) { | |
53 | + if (!(event->attr.freq || event->attr.wakeup_events)) { | |
54 | event->hw.flags |= PERF_X86_EVENT_AUTO_RELOAD; | |
55 | if (!(event->attr.sample_type & | |
56 | ~intel_pmu_free_running_flags(event))) | |
57 | -- | |
58 | 2.20.1 | |
59 |