* [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group
@ 2021-06-09 4:45 Jin Yao
2021-06-09 20:15 ` Jiri Olsa
2021-12-08 6:50 ` Ian Rogers
0 siblings, 2 replies; 4+ messages in thread
From: Jin Yao @ 2021-06-09 4:45 UTC (permalink / raw)
To: acme, jolsa, peterz, mingo, alexander.shishkin
Cc: Linux-kernel, ak, kan.liang, yao.jin, Jin Yao
A group mixed with hybrid event and global event is allowed. For example,
group leader is 'intel_pt//' and the group member is 'cpu_atom/cycles/'.
e.g.
perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u'
The challenge is their available cpus are not fully matched. For example,
'intel_pt//' is available on CPU0-CPU23, but 'cpu_atom/cycles/' is
available on CPU16-CPU23.
When getting the group id for group member, we must be very careful.
Because the cpu for 'intel_pt//' is not equal to the cpu for
'cpu_atom/cycles/'. Actually the cpu here is the index of evsel->core.cpus,
not the real CPU ID.
e.g. cpu0 for 'intel_pt//' is CPU0, but cpu0 for 'cpu_atom/cycles/' is CPU16.
Before:
# perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u' -vv uname
...
------------------------------------------------------------
perf_event_attr:
type 10
size 128
config 0xe601
{ sample_period, sample_freq } 1
sample_type IP|TID|TIME|CPU|IDENTIFIER
read_format ID
disabled 1
inherit 1
exclude_kernel 1
exclude_hv 1
enable_on_exec 1
sample_id_all 1
exclude_guest 1
------------------------------------------------------------
sys_perf_event_open: pid 4084 cpu 0 group_fd -1 flags 0x8 = 5
sys_perf_event_open: pid 4084 cpu 1 group_fd -1 flags 0x8 = 6
sys_perf_event_open: pid 4084 cpu 2 group_fd -1 flags 0x8 = 7
sys_perf_event_open: pid 4084 cpu 3 group_fd -1 flags 0x8 = 9
sys_perf_event_open: pid 4084 cpu 4 group_fd -1 flags 0x8 = 10
sys_perf_event_open: pid 4084 cpu 5 group_fd -1 flags 0x8 = 11
sys_perf_event_open: pid 4084 cpu 6 group_fd -1 flags 0x8 = 12
sys_perf_event_open: pid 4084 cpu 7 group_fd -1 flags 0x8 = 13
sys_perf_event_open: pid 4084 cpu 8 group_fd -1 flags 0x8 = 14
sys_perf_event_open: pid 4084 cpu 9 group_fd -1 flags 0x8 = 15
sys_perf_event_open: pid 4084 cpu 10 group_fd -1 flags 0x8 = 16
sys_perf_event_open: pid 4084 cpu 11 group_fd -1 flags 0x8 = 17
sys_perf_event_open: pid 4084 cpu 12 group_fd -1 flags 0x8 = 18
sys_perf_event_open: pid 4084 cpu 13 group_fd -1 flags 0x8 = 19
sys_perf_event_open: pid 4084 cpu 14 group_fd -1 flags 0x8 = 20
sys_perf_event_open: pid 4084 cpu 15 group_fd -1 flags 0x8 = 21
sys_perf_event_open: pid 4084 cpu 16 group_fd -1 flags 0x8 = 22
sys_perf_event_open: pid 4084 cpu 17 group_fd -1 flags 0x8 = 23
sys_perf_event_open: pid 4084 cpu 18 group_fd -1 flags 0x8 = 24
sys_perf_event_open: pid 4084 cpu 19 group_fd -1 flags 0x8 = 25
sys_perf_event_open: pid 4084 cpu 20 group_fd -1 flags 0x8 = 26
sys_perf_event_open: pid 4084 cpu 21 group_fd -1 flags 0x8 = 27
sys_perf_event_open: pid 4084 cpu 22 group_fd -1 flags 0x8 = 28
sys_perf_event_open: pid 4084 cpu 23 group_fd -1 flags 0x8 = 29
------------------------------------------------------------
perf_event_attr:
size 128
config 0x800000000
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD|IDENTIFIER|AUX
read_format ID
inherit 1
exclude_kernel 1
exclude_hv 1
freq 1
sample_id_all 1
exclude_guest 1
aux_sample_size 4096
------------------------------------------------------------
sys_perf_event_open: pid 4084 cpu 16 group_fd 5 flags 0x8
sys_perf_event_open failed, error -22
The group_fd 5 is not correct. It should be 22 (the fd of
'intel_pt' on CPU16).
After:
# perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u' -vv uname
...
------------------------------------------------------------
perf_event_attr:
type 10
size 128
config 0xe601
{ sample_period, sample_freq } 1
sample_type IP|TID|TIME|CPU|IDENTIFIER
read_format ID
disabled 1
inherit 1
exclude_kernel 1
exclude_hv 1
enable_on_exec 1
sample_id_all 1
exclude_guest 1
------------------------------------------------------------
sys_perf_event_open: pid 5162 cpu 0 group_fd -1 flags 0x8 = 5
sys_perf_event_open: pid 5162 cpu 1 group_fd -1 flags 0x8 = 6
sys_perf_event_open: pid 5162 cpu 2 group_fd -1 flags 0x8 = 7
sys_perf_event_open: pid 5162 cpu 3 group_fd -1 flags 0x8 = 9
sys_perf_event_open: pid 5162 cpu 4 group_fd -1 flags 0x8 = 10
sys_perf_event_open: pid 5162 cpu 5 group_fd -1 flags 0x8 = 11
sys_perf_event_open: pid 5162 cpu 6 group_fd -1 flags 0x8 = 12
sys_perf_event_open: pid 5162 cpu 7 group_fd -1 flags 0x8 = 13
sys_perf_event_open: pid 5162 cpu 8 group_fd -1 flags 0x8 = 14
sys_perf_event_open: pid 5162 cpu 9 group_fd -1 flags 0x8 = 15
sys_perf_event_open: pid 5162 cpu 10 group_fd -1 flags 0x8 = 16
sys_perf_event_open: pid 5162 cpu 11 group_fd -1 flags 0x8 = 17
sys_perf_event_open: pid 5162 cpu 12 group_fd -1 flags 0x8 = 18
sys_perf_event_open: pid 5162 cpu 13 group_fd -1 flags 0x8 = 19
sys_perf_event_open: pid 5162 cpu 14 group_fd -1 flags 0x8 = 20
sys_perf_event_open: pid 5162 cpu 15 group_fd -1 flags 0x8 = 21
sys_perf_event_open: pid 5162 cpu 16 group_fd -1 flags 0x8 = 22
sys_perf_event_open: pid 5162 cpu 17 group_fd -1 flags 0x8 = 23
sys_perf_event_open: pid 5162 cpu 18 group_fd -1 flags 0x8 = 24
sys_perf_event_open: pid 5162 cpu 19 group_fd -1 flags 0x8 = 25
sys_perf_event_open: pid 5162 cpu 20 group_fd -1 flags 0x8 = 26
sys_perf_event_open: pid 5162 cpu 21 group_fd -1 flags 0x8 = 27
sys_perf_event_open: pid 5162 cpu 22 group_fd -1 flags 0x8 = 28
sys_perf_event_open: pid 5162 cpu 23 group_fd -1 flags 0x8 = 29
------------------------------------------------------------
perf_event_attr:
size 128
config 0x800000000
{ sample_period, sample_freq } 4000
sample_type IP|TID|TIME|PERIOD|IDENTIFIER|AUX
read_format ID
inherit 1
exclude_kernel 1
exclude_hv 1
freq 1
sample_id_all 1
exclude_guest 1
aux_sample_size 4096
------------------------------------------------------------
sys_perf_event_open: pid 5162 cpu 16 group_fd 22 flags 0x8 = 30
sys_perf_event_open: pid 5162 cpu 17 group_fd 23 flags 0x8 = 31
sys_perf_event_open: pid 5162 cpu 18 group_fd 24 flags 0x8 = 32
sys_perf_event_open: pid 5162 cpu 19 group_fd 25 flags 0x8 = 33
sys_perf_event_open: pid 5162 cpu 20 group_fd 26 flags 0x8 = 34
sys_perf_event_open: pid 5162 cpu 21 group_fd 27 flags 0x8 = 35
sys_perf_event_open: pid 5162 cpu 22 group_fd 28 flags 0x8 = 36
sys_perf_event_open: pid 5162 cpu 23 group_fd 29 flags 0x8 = 37
------------------------------------------------------------
...
Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
---
tools/perf/util/evsel.c | 25 +++++++++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 4a3cd1b5bb33..f81ac6962aec 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -1581,6 +1581,27 @@ int __evsel__read_on_cpu(struct evsel *evsel, int cpu, int thread, bool scale)
return 0;
}
+static int evsel__match_other_cpu(struct evsel *evsel, struct evsel *other,
+ int cpu)
+{
+ int cpuid;
+
+ cpuid = perf_cpu_map__cpu(evsel->core.cpus, cpu);
+ return perf_cpu_map__idx(other->core.cpus, cpuid);
+}
+
+static int evsel__hybrid_group_cpu(struct evsel *evsel, int cpu)
+{
+ struct evsel *leader = evsel->leader;
+
+ if ((evsel__is_hybrid(evsel) && !evsel__is_hybrid(leader)) ||
+ (!evsel__is_hybrid(evsel) && evsel__is_hybrid(leader))) {
+ return evsel__match_other_cpu(evsel, leader, cpu);
+ }
+
+ return cpu;
+}
+
static int get_group_fd(struct evsel *evsel, int cpu, int thread)
{
struct evsel *leader = evsel->leader;
@@ -1595,6 +1616,10 @@ static int get_group_fd(struct evsel *evsel, int cpu, int thread)
*/
BUG_ON(!leader->core.fd);
+ cpu = evsel__hybrid_group_cpu(evsel, cpu);
+ if (cpu == -1)
+ return -1;
+
fd = FD(leader, cpu, thread);
BUG_ON(fd == -1);
--
2.17.1
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group
2021-06-09 4:45 [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group Jin Yao
@ 2021-06-09 20:15 ` Jiri Olsa
2021-06-10 16:43 ` Arnaldo Carvalho de Melo
2021-12-08 6:50 ` Ian Rogers
1 sibling, 1 reply; 4+ messages in thread
From: Jiri Olsa @ 2021-06-09 20:15 UTC (permalink / raw)
To: Jin Yao
Cc: acme, jolsa, peterz, mingo, alexander.shishkin, Linux-kernel, ak,
kan.liang, yao.jin
On Wed, Jun 09, 2021 at 12:45:55PM +0800, Jin Yao wrote:
> A group mixed with hybrid event and global event is allowed. For example,
> group leader is 'intel_pt//' and the group member is 'cpu_atom/cycles/'.
>
> e.g.
> perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u'
>
> The challenge is their available cpus are not fully matched. For example,
> 'intel_pt//' is available on CPU0-CPU23, but 'cpu_atom/cycles/' is
> available on CPU16-CPU23.
>
> When getting the group id for group member, we must be very careful.
> Because the cpu for 'intel_pt//' is not equal to the cpu for
> 'cpu_atom/cycles/'. Actually the cpu here is the index of evsel->core.cpus,
> not the real CPU ID.
>
> e.g. cpu0 for 'intel_pt//' is CPU0, but cpu0 for 'cpu_atom/cycles/' is CPU16.
>
> Before:
>
> # perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u' -vv uname
> ...
> ------------------------------------------------------------
> perf_event_attr:
> type 10
> size 128
> config 0xe601
> { sample_period, sample_freq } 1
> sample_type IP|TID|TIME|CPU|IDENTIFIER
> read_format ID
> disabled 1
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> enable_on_exec 1
> sample_id_all 1
> exclude_guest 1
> ------------------------------------------------------------
> sys_perf_event_open: pid 4084 cpu 0 group_fd -1 flags 0x8 = 5
> sys_perf_event_open: pid 4084 cpu 1 group_fd -1 flags 0x8 = 6
> sys_perf_event_open: pid 4084 cpu 2 group_fd -1 flags 0x8 = 7
> sys_perf_event_open: pid 4084 cpu 3 group_fd -1 flags 0x8 = 9
> sys_perf_event_open: pid 4084 cpu 4 group_fd -1 flags 0x8 = 10
> sys_perf_event_open: pid 4084 cpu 5 group_fd -1 flags 0x8 = 11
> sys_perf_event_open: pid 4084 cpu 6 group_fd -1 flags 0x8 = 12
> sys_perf_event_open: pid 4084 cpu 7 group_fd -1 flags 0x8 = 13
> sys_perf_event_open: pid 4084 cpu 8 group_fd -1 flags 0x8 = 14
> sys_perf_event_open: pid 4084 cpu 9 group_fd -1 flags 0x8 = 15
> sys_perf_event_open: pid 4084 cpu 10 group_fd -1 flags 0x8 = 16
> sys_perf_event_open: pid 4084 cpu 11 group_fd -1 flags 0x8 = 17
> sys_perf_event_open: pid 4084 cpu 12 group_fd -1 flags 0x8 = 18
> sys_perf_event_open: pid 4084 cpu 13 group_fd -1 flags 0x8 = 19
> sys_perf_event_open: pid 4084 cpu 14 group_fd -1 flags 0x8 = 20
> sys_perf_event_open: pid 4084 cpu 15 group_fd -1 flags 0x8 = 21
> sys_perf_event_open: pid 4084 cpu 16 group_fd -1 flags 0x8 = 22
> sys_perf_event_open: pid 4084 cpu 17 group_fd -1 flags 0x8 = 23
> sys_perf_event_open: pid 4084 cpu 18 group_fd -1 flags 0x8 = 24
> sys_perf_event_open: pid 4084 cpu 19 group_fd -1 flags 0x8 = 25
> sys_perf_event_open: pid 4084 cpu 20 group_fd -1 flags 0x8 = 26
> sys_perf_event_open: pid 4084 cpu 21 group_fd -1 flags 0x8 = 27
> sys_perf_event_open: pid 4084 cpu 22 group_fd -1 flags 0x8 = 28
> sys_perf_event_open: pid 4084 cpu 23 group_fd -1 flags 0x8 = 29
> ------------------------------------------------------------
> perf_event_attr:
> size 128
> config 0x800000000
> { sample_period, sample_freq } 4000
> sample_type IP|TID|TIME|PERIOD|IDENTIFIER|AUX
> read_format ID
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> freq 1
> sample_id_all 1
> exclude_guest 1
> aux_sample_size 4096
> ------------------------------------------------------------
> sys_perf_event_open: pid 4084 cpu 16 group_fd 5 flags 0x8
> sys_perf_event_open failed, error -22
>
> The group_fd 5 is not correct. It should be 22 (the fd of
> 'intel_pt' on CPU16).
>
> After:
>
> # perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u' -vv uname
> ...
> ------------------------------------------------------------
> perf_event_attr:
> type 10
> size 128
> config 0xe601
> { sample_period, sample_freq } 1
> sample_type IP|TID|TIME|CPU|IDENTIFIER
> read_format ID
> disabled 1
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> enable_on_exec 1
> sample_id_all 1
> exclude_guest 1
> ------------------------------------------------------------
> sys_perf_event_open: pid 5162 cpu 0 group_fd -1 flags 0x8 = 5
> sys_perf_event_open: pid 5162 cpu 1 group_fd -1 flags 0x8 = 6
> sys_perf_event_open: pid 5162 cpu 2 group_fd -1 flags 0x8 = 7
> sys_perf_event_open: pid 5162 cpu 3 group_fd -1 flags 0x8 = 9
> sys_perf_event_open: pid 5162 cpu 4 group_fd -1 flags 0x8 = 10
> sys_perf_event_open: pid 5162 cpu 5 group_fd -1 flags 0x8 = 11
> sys_perf_event_open: pid 5162 cpu 6 group_fd -1 flags 0x8 = 12
> sys_perf_event_open: pid 5162 cpu 7 group_fd -1 flags 0x8 = 13
> sys_perf_event_open: pid 5162 cpu 8 group_fd -1 flags 0x8 = 14
> sys_perf_event_open: pid 5162 cpu 9 group_fd -1 flags 0x8 = 15
> sys_perf_event_open: pid 5162 cpu 10 group_fd -1 flags 0x8 = 16
> sys_perf_event_open: pid 5162 cpu 11 group_fd -1 flags 0x8 = 17
> sys_perf_event_open: pid 5162 cpu 12 group_fd -1 flags 0x8 = 18
> sys_perf_event_open: pid 5162 cpu 13 group_fd -1 flags 0x8 = 19
> sys_perf_event_open: pid 5162 cpu 14 group_fd -1 flags 0x8 = 20
> sys_perf_event_open: pid 5162 cpu 15 group_fd -1 flags 0x8 = 21
> sys_perf_event_open: pid 5162 cpu 16 group_fd -1 flags 0x8 = 22
> sys_perf_event_open: pid 5162 cpu 17 group_fd -1 flags 0x8 = 23
> sys_perf_event_open: pid 5162 cpu 18 group_fd -1 flags 0x8 = 24
> sys_perf_event_open: pid 5162 cpu 19 group_fd -1 flags 0x8 = 25
> sys_perf_event_open: pid 5162 cpu 20 group_fd -1 flags 0x8 = 26
> sys_perf_event_open: pid 5162 cpu 21 group_fd -1 flags 0x8 = 27
> sys_perf_event_open: pid 5162 cpu 22 group_fd -1 flags 0x8 = 28
> sys_perf_event_open: pid 5162 cpu 23 group_fd -1 flags 0x8 = 29
> ------------------------------------------------------------
> perf_event_attr:
> size 128
> config 0x800000000
> { sample_period, sample_freq } 4000
> sample_type IP|TID|TIME|PERIOD|IDENTIFIER|AUX
> read_format ID
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> freq 1
> sample_id_all 1
> exclude_guest 1
> aux_sample_size 4096
> ------------------------------------------------------------
> sys_perf_event_open: pid 5162 cpu 16 group_fd 22 flags 0x8 = 30
> sys_perf_event_open: pid 5162 cpu 17 group_fd 23 flags 0x8 = 31
> sys_perf_event_open: pid 5162 cpu 18 group_fd 24 flags 0x8 = 32
> sys_perf_event_open: pid 5162 cpu 19 group_fd 25 flags 0x8 = 33
> sys_perf_event_open: pid 5162 cpu 20 group_fd 26 flags 0x8 = 34
> sys_perf_event_open: pid 5162 cpu 21 group_fd 27 flags 0x8 = 35
> sys_perf_event_open: pid 5162 cpu 22 group_fd 28 flags 0x8 = 36
> sys_perf_event_open: pid 5162 cpu 23 group_fd 29 flags 0x8 = 37
> ------------------------------------------------------------
> ...
>
> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
looks good to me
Acked-by: Jiri Olsa <jolsa@kernel.org>
thanks,
jirka
> ---
> tools/perf/util/evsel.c | 25 +++++++++++++++++++++++++
> 1 file changed, 25 insertions(+)
>
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 4a3cd1b5bb33..f81ac6962aec 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -1581,6 +1581,27 @@ int __evsel__read_on_cpu(struct evsel *evsel, int cpu, int thread, bool scale)
> return 0;
> }
>
> +static int evsel__match_other_cpu(struct evsel *evsel, struct evsel *other,
> + int cpu)
> +{
> + int cpuid;
> +
> + cpuid = perf_cpu_map__cpu(evsel->core.cpus, cpu);
> + return perf_cpu_map__idx(other->core.cpus, cpuid);
> +}
> +
> +static int evsel__hybrid_group_cpu(struct evsel *evsel, int cpu)
> +{
> + struct evsel *leader = evsel->leader;
> +
> + if ((evsel__is_hybrid(evsel) && !evsel__is_hybrid(leader)) ||
> + (!evsel__is_hybrid(evsel) && evsel__is_hybrid(leader))) {
> + return evsel__match_other_cpu(evsel, leader, cpu);
> + }
> +
> + return cpu;
> +}
> +
> static int get_group_fd(struct evsel *evsel, int cpu, int thread)
> {
> struct evsel *leader = evsel->leader;
> @@ -1595,6 +1616,10 @@ static int get_group_fd(struct evsel *evsel, int cpu, int thread)
> */
> BUG_ON(!leader->core.fd);
>
> + cpu = evsel__hybrid_group_cpu(evsel, cpu);
> + if (cpu == -1)
> + return -1;
> +
> fd = FD(leader, cpu, thread);
> BUG_ON(fd == -1);
>
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group
2021-06-09 20:15 ` Jiri Olsa
@ 2021-06-10 16:43 ` Arnaldo Carvalho de Melo
0 siblings, 0 replies; 4+ messages in thread
From: Arnaldo Carvalho de Melo @ 2021-06-10 16:43 UTC (permalink / raw)
To: Jiri Olsa
Cc: Jin Yao, jolsa, peterz, mingo, alexander.shishkin, Linux-kernel,
ak, kan.liang, yao.jin
Em Wed, Jun 09, 2021 at 10:15:29PM +0200, Jiri Olsa escreveu:
> On Wed, Jun 09, 2021 at 12:45:55PM +0800, Jin Yao wrote:
> > A group mixed with hybrid event and global event is allowed. For example,
> > group leader is 'intel_pt//' and the group member is 'cpu_atom/cycles/'.
> >
> > e.g.
> > perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u'
> >
> > The challenge is their available cpus are not fully matched. For example,
> > 'intel_pt//' is available on CPU0-CPU23, but 'cpu_atom/cycles/' is
> > available on CPU16-CPU23.
> >
> > When getting the group id for group member, we must be very careful.
> > Because the cpu for 'intel_pt//' is not equal to the cpu for
> > 'cpu_atom/cycles/'. Actually the cpu here is the index of evsel->core.cpus,
> > not the real CPU ID.
<SNIP>
> > Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
> looks good to me
> Acked-by: Jiri Olsa <jolsa@kernel.org>
Thanks, applied.
- Arnaldo
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group
2021-06-09 4:45 [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group Jin Yao
2021-06-09 20:15 ` Jiri Olsa
@ 2021-12-08 6:50 ` Ian Rogers
1 sibling, 0 replies; 4+ messages in thread
From: Ian Rogers @ 2021-12-08 6:50 UTC (permalink / raw)
To: Jin Yao, acme, jolsa
Cc: peterz, mingo, alexander.shishkin, linux-kernel, ak, kan.liang, yao.jin
On Tue, Jun 8, 2021 at 9:48 PM Jin Yao <yao.jin@linux.intel.com> wrote:
>
> A group mixed with hybrid event and global event is allowed. For example,
> group leader is 'intel_pt//' and the group member is 'cpu_atom/cycles/'.
>
> e.g.
> perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u'
>
> The challenge is their available cpus are not fully matched. For example,
> 'intel_pt//' is available on CPU0-CPU23, but 'cpu_atom/cycles/' is
> available on CPU16-CPU23.
>
> When getting the group id for group member, we must be very careful.
> Because the cpu for 'intel_pt//' is not equal to the cpu for
> 'cpu_atom/cycles/'. Actually the cpu here is the index of evsel->core.cpus,
> not the real CPU ID.
>
> e.g. cpu0 for 'intel_pt//' is CPU0, but cpu0 for 'cpu_atom/cycles/' is CPU16.
>
> Before:
>
> # perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u' -vv uname
> ...
> ------------------------------------------------------------
> perf_event_attr:
> type 10
> size 128
> config 0xe601
> { sample_period, sample_freq } 1
> sample_type IP|TID|TIME|CPU|IDENTIFIER
> read_format ID
> disabled 1
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> enable_on_exec 1
> sample_id_all 1
> exclude_guest 1
> ------------------------------------------------------------
> sys_perf_event_open: pid 4084 cpu 0 group_fd -1 flags 0x8 = 5
> sys_perf_event_open: pid 4084 cpu 1 group_fd -1 flags 0x8 = 6
> sys_perf_event_open: pid 4084 cpu 2 group_fd -1 flags 0x8 = 7
> sys_perf_event_open: pid 4084 cpu 3 group_fd -1 flags 0x8 = 9
> sys_perf_event_open: pid 4084 cpu 4 group_fd -1 flags 0x8 = 10
> sys_perf_event_open: pid 4084 cpu 5 group_fd -1 flags 0x8 = 11
> sys_perf_event_open: pid 4084 cpu 6 group_fd -1 flags 0x8 = 12
> sys_perf_event_open: pid 4084 cpu 7 group_fd -1 flags 0x8 = 13
> sys_perf_event_open: pid 4084 cpu 8 group_fd -1 flags 0x8 = 14
> sys_perf_event_open: pid 4084 cpu 9 group_fd -1 flags 0x8 = 15
> sys_perf_event_open: pid 4084 cpu 10 group_fd -1 flags 0x8 = 16
> sys_perf_event_open: pid 4084 cpu 11 group_fd -1 flags 0x8 = 17
> sys_perf_event_open: pid 4084 cpu 12 group_fd -1 flags 0x8 = 18
> sys_perf_event_open: pid 4084 cpu 13 group_fd -1 flags 0x8 = 19
> sys_perf_event_open: pid 4084 cpu 14 group_fd -1 flags 0x8 = 20
> sys_perf_event_open: pid 4084 cpu 15 group_fd -1 flags 0x8 = 21
> sys_perf_event_open: pid 4084 cpu 16 group_fd -1 flags 0x8 = 22
> sys_perf_event_open: pid 4084 cpu 17 group_fd -1 flags 0x8 = 23
> sys_perf_event_open: pid 4084 cpu 18 group_fd -1 flags 0x8 = 24
> sys_perf_event_open: pid 4084 cpu 19 group_fd -1 flags 0x8 = 25
> sys_perf_event_open: pid 4084 cpu 20 group_fd -1 flags 0x8 = 26
> sys_perf_event_open: pid 4084 cpu 21 group_fd -1 flags 0x8 = 27
> sys_perf_event_open: pid 4084 cpu 22 group_fd -1 flags 0x8 = 28
> sys_perf_event_open: pid 4084 cpu 23 group_fd -1 flags 0x8 = 29
> ------------------------------------------------------------
> perf_event_attr:
> size 128
> config 0x800000000
> { sample_period, sample_freq } 4000
> sample_type IP|TID|TIME|PERIOD|IDENTIFIER|AUX
> read_format ID
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> freq 1
> sample_id_all 1
> exclude_guest 1
> aux_sample_size 4096
> ------------------------------------------------------------
> sys_perf_event_open: pid 4084 cpu 16 group_fd 5 flags 0x8
> sys_perf_event_open failed, error -22
>
> The group_fd 5 is not correct. It should be 22 (the fd of
> 'intel_pt' on CPU16).
>
> After:
>
> # perf record --aux-sample -e '{intel_pt//,cpu_atom/cycles/}:u' -vv uname
> ...
> ------------------------------------------------------------
> perf_event_attr:
> type 10
> size 128
> config 0xe601
> { sample_period, sample_freq } 1
> sample_type IP|TID|TIME|CPU|IDENTIFIER
> read_format ID
> disabled 1
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> enable_on_exec 1
> sample_id_all 1
> exclude_guest 1
> ------------------------------------------------------------
> sys_perf_event_open: pid 5162 cpu 0 group_fd -1 flags 0x8 = 5
> sys_perf_event_open: pid 5162 cpu 1 group_fd -1 flags 0x8 = 6
> sys_perf_event_open: pid 5162 cpu 2 group_fd -1 flags 0x8 = 7
> sys_perf_event_open: pid 5162 cpu 3 group_fd -1 flags 0x8 = 9
> sys_perf_event_open: pid 5162 cpu 4 group_fd -1 flags 0x8 = 10
> sys_perf_event_open: pid 5162 cpu 5 group_fd -1 flags 0x8 = 11
> sys_perf_event_open: pid 5162 cpu 6 group_fd -1 flags 0x8 = 12
> sys_perf_event_open: pid 5162 cpu 7 group_fd -1 flags 0x8 = 13
> sys_perf_event_open: pid 5162 cpu 8 group_fd -1 flags 0x8 = 14
> sys_perf_event_open: pid 5162 cpu 9 group_fd -1 flags 0x8 = 15
> sys_perf_event_open: pid 5162 cpu 10 group_fd -1 flags 0x8 = 16
> sys_perf_event_open: pid 5162 cpu 11 group_fd -1 flags 0x8 = 17
> sys_perf_event_open: pid 5162 cpu 12 group_fd -1 flags 0x8 = 18
> sys_perf_event_open: pid 5162 cpu 13 group_fd -1 flags 0x8 = 19
> sys_perf_event_open: pid 5162 cpu 14 group_fd -1 flags 0x8 = 20
> sys_perf_event_open: pid 5162 cpu 15 group_fd -1 flags 0x8 = 21
> sys_perf_event_open: pid 5162 cpu 16 group_fd -1 flags 0x8 = 22
> sys_perf_event_open: pid 5162 cpu 17 group_fd -1 flags 0x8 = 23
> sys_perf_event_open: pid 5162 cpu 18 group_fd -1 flags 0x8 = 24
> sys_perf_event_open: pid 5162 cpu 19 group_fd -1 flags 0x8 = 25
> sys_perf_event_open: pid 5162 cpu 20 group_fd -1 flags 0x8 = 26
> sys_perf_event_open: pid 5162 cpu 21 group_fd -1 flags 0x8 = 27
> sys_perf_event_open: pid 5162 cpu 22 group_fd -1 flags 0x8 = 28
> sys_perf_event_open: pid 5162 cpu 23 group_fd -1 flags 0x8 = 29
> ------------------------------------------------------------
> perf_event_attr:
> size 128
> config 0x800000000
> { sample_period, sample_freq } 4000
> sample_type IP|TID|TIME|PERIOD|IDENTIFIER|AUX
> read_format ID
> inherit 1
> exclude_kernel 1
> exclude_hv 1
> freq 1
> sample_id_all 1
> exclude_guest 1
> aux_sample_size 4096
> ------------------------------------------------------------
> sys_perf_event_open: pid 5162 cpu 16 group_fd 22 flags 0x8 = 30
> sys_perf_event_open: pid 5162 cpu 17 group_fd 23 flags 0x8 = 31
> sys_perf_event_open: pid 5162 cpu 18 group_fd 24 flags 0x8 = 32
> sys_perf_event_open: pid 5162 cpu 19 group_fd 25 flags 0x8 = 33
> sys_perf_event_open: pid 5162 cpu 20 group_fd 26 flags 0x8 = 34
> sys_perf_event_open: pid 5162 cpu 21 group_fd 27 flags 0x8 = 35
> sys_perf_event_open: pid 5162 cpu 22 group_fd 28 flags 0x8 = 36
> sys_perf_event_open: pid 5162 cpu 23 group_fd 29 flags 0x8 = 37
> ------------------------------------------------------------
> ...
>
> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
> ---
> tools/perf/util/evsel.c | 25 +++++++++++++++++++++++++
> 1 file changed, 25 insertions(+)
>
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 4a3cd1b5bb33..f81ac6962aec 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -1581,6 +1581,27 @@ int __evsel__read_on_cpu(struct evsel *evsel, int cpu, int thread, bool scale)
> return 0;
> }
>
> +static int evsel__match_other_cpu(struct evsel *evsel, struct evsel *other,
> + int cpu)
> +{
> + int cpuid;
> +
> + cpuid = perf_cpu_map__cpu(evsel->core.cpus, cpu);
I came across this as part of:
https://lore.kernel.org/lkml/20211208024607.1784932-1-irogers@google.com/
The assumption that evsel's cpu maps are indexable by cpus looks wrong
- it wouldn't work for a SkylakeX's:
$ cat /sys/devices/uncore_imc_0/cpumask
0,18
The cpu here is a real CPU and not a cpu map index, so I'm not sure
what the cpu map translation is trying to do - if this is a full cpu
map then cpuid will equal cpu. Perhaps the code should be:
BUG_ON(!perf_cpu_map__has(evsel->core.cpus, cpu));
return perf_cpu_map__idx(other->core.cpus, cpu);
Thanks,
Ian
> + return perf_cpu_map__idx(other->core.cpus, cpuid);
> +}
> +
> +static int evsel__hybrid_group_cpu(struct evsel *evsel, int cpu)
> +{
> + struct evsel *leader = evsel->leader;
> +
> + if ((evsel__is_hybrid(evsel) && !evsel__is_hybrid(leader)) ||
> + (!evsel__is_hybrid(evsel) && evsel__is_hybrid(leader))) {
> + return evsel__match_other_cpu(evsel, leader, cpu);
> + }
> +
> + return cpu;
> +}
> +
> static int get_group_fd(struct evsel *evsel, int cpu, int thread)
> {
> struct evsel *leader = evsel->leader;
> @@ -1595,6 +1616,10 @@ static int get_group_fd(struct evsel *evsel, int cpu, int thread)
> */
> BUG_ON(!leader->core.fd);
>
> + cpu = evsel__hybrid_group_cpu(evsel, cpu);
> + if (cpu == -1)
> + return -1;
> +
> fd = FD(leader, cpu, thread);
> BUG_ON(fd == -1);
>
> --
> 2.17.1
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2021-12-08 6:50 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-09 4:45 [PATCH v2] perf evsel: Adjust hybrid event and global event mixed group Jin Yao
2021-06-09 20:15 ` Jiri Olsa
2021-06-10 16:43 ` Arnaldo Carvalho de Melo
2021-12-08 6:50 ` Ian Rogers
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).