2022-09-22 03:47:24

by Namhyung Kim

Subject: [PATCH] perf stat: Get a perf cgroup more portably in BPF

The perf_event_cgrp_id can be different on other configurations.
To be more portable as CO-RE, it needs to get the cgroup subsys id
using the bpf_core_enum_value() helper.

Suggested-by: Ian Rogers <[email protected]>
Signed-off-by: Namhyung Kim <[email protected]>
---
tools/perf/util/bpf_skel/bperf_cgroup.bpf.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c b/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
index c72f8ad96f75..35a8e38d3ee3 100644
--- a/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
+++ b/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
@@ -57,8 +57,10 @@ static inline int get_cgroup_v1_idx(__u32 *cgrps, int size)
__u32 *elem;
int level;
int cnt;
+ int perf_id;

- cgrp = BPF_CORE_READ(p, cgroups, subsys[perf_event_cgrp_id], cgroup);
+ perf_id = bpf_core_enum_value(enum cgroup_subsys_id, perf_event_cgrp_id);
+ cgrp = BPF_CORE_READ(p, cgroups, subsys[perf_id], cgroup);
level = BPF_CORE_READ(cgrp, level);

for (cnt = 0; i < MAX_LEVELS; i++) {
--
2.37.3.968.ga6b4b080e4-goog

2022-09-22 04:03:36

by Ian Rogers

Subject: Re: [PATCH] perf stat: Get a perf cgroup more portably in BPF

On Wed, Sep 21, 2022 at 8:33 PM Namhyung Kim <[email protected]> wrote:
>
> The perf_event_cgrp_id can be different on other configurations.
> To be more portable as CO-RE, it needs to get the cgroup subsys id
> using the bpf_core_enum_value() helper.
>
> Suggested-by: Ian Rogers <[email protected]>
> Signed-off-by: Namhyung Kim <[email protected]>
> ---
> tools/perf/util/bpf_skel/bperf_cgroup.bpf.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c b/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
> index c72f8ad96f75..35a8e38d3ee3 100644
> --- a/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
> +++ b/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
> @@ -57,8 +57,10 @@ static inline int get_cgroup_v1_idx(__u32 *cgrps, int size)
> __u32 *elem;
> int level;
> int cnt;
> + int perf_id;
>
> - cgrp = BPF_CORE_READ(p, cgroups, subsys[perf_event_cgrp_id], cgroup);
> + perf_id = bpf_core_enum_value(enum cgroup_subsys_id, perf_event_cgrp_id);
> + cgrp = BPF_CORE_READ(p, cgroups, subsys[perf_id], cgroup);

LGTM, do we need the same fix in off cpu?
https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/util/bpf_skel/off_cpu.bpf.c?h=perf/core#n123

Thanks,
Ian

> level = BPF_CORE_READ(cgrp, level);
>
> for (cnt = 0; i < MAX_LEVELS; i++) {
> --
> 2.37.3.968.ga6b4b080e4-goog
>

2022-09-22 05:12:34

by Namhyung Kim

Subject: Re: [PATCH] perf stat: Get a perf cgroup more portably in BPF

Hi Ian,

On Wed, Sep 21, 2022 at 8:54 PM Ian Rogers <[email protected]> wrote:
>
> On Wed, Sep 21, 2022 at 8:33 PM Namhyung Kim <[email protected]> wrote:
> >
> > The perf_event_cgrp_id can be different on other configurations.
> > To be more portable as CO-RE, it needs to get the cgroup subsys id
> > using the bpf_core_enum_value() helper.
> >
> > Suggested-by: Ian Rogers <[email protected]>
> > Signed-off-by: Namhyung Kim <[email protected]>
> > ---
> > tools/perf/util/bpf_skel/bperf_cgroup.bpf.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c b/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
> > index c72f8ad96f75..35a8e38d3ee3 100644
> > --- a/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
> > +++ b/tools/perf/util/bpf_skel/bperf_cgroup.bpf.c
> > @@ -57,8 +57,10 @@ static inline int get_cgroup_v1_idx(__u32 *cgrps, int size)
> > __u32 *elem;
> > int level;
> > int cnt;
> > + int perf_id;
> >
> > - cgrp = BPF_CORE_READ(p, cgroups, subsys[perf_event_cgrp_id], cgroup);
> > + perf_id = bpf_core_enum_value(enum cgroup_subsys_id, perf_event_cgrp_id);
> > + cgrp = BPF_CORE_READ(p, cgroups, subsys[perf_id], cgroup);
>
> LGTM, do we need the same fix in off cpu?
> https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/util/bpf_skel/off_cpu.bpf.c?h=perf/core#n123

Right, will send v2.

Thanks,
Namhyung