lkml.org 
[lkml]   [2023]   [Jun]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v1] perf pmus: Add placeholder core PMU
From
Date
Hello,

On 2023/6/28 2:28, Ian Rogers wrote:
> If loading a core PMU fails, legacy hardware/cache events may segv due
> to there being no PMU. Create a placeholder empty PMU for this
> case. This was discussed in:
> https://lore.kernel.org/lkml/20230614151625.2077-1-yangjihong1@huawei.com/
>
> Reported-by: Yang Jihong <yangjihong1@huawei.com>
> Signed-off-by: Ian Rogers <irogers@google.com>
> ---
> tools/perf/util/pmu.c | 25 +++++++++++++++++++++++++
> tools/perf/util/pmu.h | 1 +
> tools/perf/util/pmus.c | 7 ++++++-
> 3 files changed, 32 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
> index 8d5ecd4ff1a9..7f984a7f16ca 100644
> --- a/tools/perf/util/pmu.c
> +++ b/tools/perf/util/pmu.c
> @@ -928,6 +928,31 @@ struct perf_pmu *perf_pmu__lookup(struct list_head *pmus, int dirfd, const char
> return NULL;
> }
>
> +/* Creates the PMU when sysfs scanning fails. */
> +struct perf_pmu *perf_pmu__create_placeholder_core_pmu(struct list_head *core_pmus)
> +{
> + struct perf_pmu *pmu = zalloc(sizeof(*pmu));
> +
> + if (!pmu)
> + return NULL;
> +
> + pmu->name = strdup("cpu");
> + if (!pmu->name) {
> + free(pmu);
> + return NULL;
> + }
> +
> + pmu->is_core = true;
> + pmu->type = PERF_TYPE_RAW;
> + pmu->cpus = cpu_map__online();
> +
> + INIT_LIST_HEAD(&pmu->format);
> + INIT_LIST_HEAD(&pmu->aliases);
> + INIT_LIST_HEAD(&pmu->caps);
> + list_add_tail(&pmu->list, core_pmus);
> + return pmu;
> +}
> +
> void perf_pmu__warn_invalid_formats(struct perf_pmu *pmu)
> {
> struct perf_pmu_format *format;
> diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
> index 8807a624e918..203b92860e3c 100644
> --- a/tools/perf/util/pmu.h
> +++ b/tools/perf/util/pmu.h
> @@ -286,6 +286,7 @@ int perf_pmu__event_source_devices_fd(void);
> int perf_pmu__pathname_fd(int dirfd, const char *pmu_name, const char *filename, int flags);
>
> struct perf_pmu *perf_pmu__lookup(struct list_head *pmus, int dirfd, const char *lookup_name);
> +struct perf_pmu *perf_pmu__create_placeholder_core_pmu(struct list_head *core_pmus);
> void perf_pmu__delete(struct perf_pmu *pmu);
>
> #endif /* __PMU_H */
> diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
> index 0866dee3fc62..3cd9de42139e 100644
> --- a/tools/perf/util/pmus.c
> +++ b/tools/perf/util/pmus.c
> @@ -153,7 +153,12 @@ static void pmu_read_sysfs(bool core_only)
>
> closedir(dir);
> if (core_only) {
> - read_sysfs_core_pmus = true;
> + if (!list_empty(&core_pmus))
> + read_sysfs_core_pmus = true;
> + else {
> + if (perf_pmu__create_placeholder_core_pmu(&core_pmus))
> + read_sysfs_core_pmus = true;
> + }
> } else {
> read_sysfs_core_pmus = true;
> read_sysfs_all_pmus = true;
>
Great. This patch solves the problem, the test results are shown in [1].

Tested-by: Yang Jihong <yangjihong1@huawei.com>

[1]:
# perf --debug verbose=2 record true
<SNIP>
------------------------------------------------------------
perf_event_attr:
size 136
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD
read_format ID|LOST
disabled 1
inherit 1
mmap 1
comm 1
freq 1
enable_on_exec 1
task 1
precise_ip 3
sample_id_all 1
mmap2 1
comm_exec 1
ksymbol 1
bpf_event 1
------------------------------------------------------------
sys_perf_event_open: pid 703 cpu 0 group_fd -1 flags 0x8
sys_perf_event_open failed, error -2
decreasing precise_ip by one (2)
------------------------------------------------------------
perf_event_attr:
size 136
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD
read_format ID|LOST
disabled 1
inherit 1
mmap 1
comm 1
freq 1
enable_on_exec 1
task 1
precise_ip 2
sample_id_all 1
mmap2 1
comm_exec 1
ksymbol 1
bpf_event 1
------------------------------------------------------------
sys_perf_event_open: pid 703 cpu 0 group_fd -1 flags 0x8
sys_perf_event_open failed, error -2
decreasing precise_ip by one (1)
------------------------------------------------------------
perf_event_attr:
size 136
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD
read_format ID|LOST
disabled 1
inherit 1
mmap 1
comm 1
freq 1
enable_on_exec 1
task 1
precise_ip 1
sample_id_all 1
mmap2 1
comm_exec 1
ksymbol 1
bpf_event 1
------------------------------------------------------------
sys_perf_event_open: pid 703 cpu 0 group_fd -1 flags 0x8
sys_perf_event_open failed, error -2
decreasing precise_ip by one (0)
------------------------------------------------------------
perf_event_attr:
size 136
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD
read_format ID|LOST
disabled 1
inherit 1
mmap 1
comm 1
freq 1
enable_on_exec 1
task 1
sample_id_all 1
mmap2 1
comm_exec 1
ksymbol 1
bpf_event 1
------------------------------------------------------------
sys_perf_event_open: pid 703 cpu 0 group_fd -1 flags 0x8
sys_perf_event_open failed, error -2
Warning:
The cycles event is not supported, trying to fall back to cpu-clock-ticks
------------------------------------------------------------
perf_event_attr:
type 1
size 136
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD
read_format ID|LOST
disabled 1
inherit 1
mmap 1
comm 1
freq 1
enable_on_exec 1
task 1
precise_ip 3
sample_id_all 1
mmap2 1
comm_exec 1
ksymbol 1
bpf_event 1
------------------------------------------------------------
sys_perf_event_open: pid 703 cpu 0 group_fd -1 flags 0x8 = 5
sys_perf_event_open: pid 703 cpu 1 group_fd -1 flags 0x8 = 6
sys_perf_event_open: pid 703 cpu 2 group_fd -1 flags 0x8 = 7
sys_perf_event_open: pid 703 cpu 3 group_fd -1 flags 0x8 = 9
sys_perf_event_open: pid 703 cpu 4 group_fd -1 flags 0x8 = 10
sys_perf_event_open: pid 703 cpu 5 group_fd -1 flags 0x8 = 11
sys_perf_event_open: pid 703 cpu 6 group_fd -1 flags 0x8 = 12
sys_perf_event_open: pid 703 cpu 7 group_fd -1 flags 0x8 = 13
<SNIP>
[ perf record: Captured and wrote 0.006 MB perf.data (116 samples) ]

Thanks,
Yang

\
 
 \ /
  Last update: 2023-06-28 05:02    [W:0.048 / U:0.088 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site