2023-12-07 16:36:00

by Kyle Huey

Subject: [PATCH v2 2/3] perf/bpf: Allow a bpf program to suppress all sample side effects.

Returning zero from a bpf program attached to a perf event already
suppresses any data output. By clearing pending_kill, returning zero from a
bpf program will effectively pretend the sample never happened for all
userspace purposes.

Signed-off-by: Kyle Huey <[email protected]>
---
kernel/events/core.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 19fddfc27a4a..6cda05a4969d 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -10421,8 +10421,10 @@ static void bpf_overflow_handler(struct perf_event *event,
rcu_read_unlock();
out:
__this_cpu_dec(bpf_prog_active);
- if (!ret)
+ if (!ret) {
+ event->pending_kill = 0;
return;
+ }

event->orig_overflow_handler(event, data, regs);
}
--
2.34.1

2023-12-07 19:13:29

by Andrii Nakryiko

Subject: Re: [PATCH v2 2/3] perf/bpf: Allow a bpf program to suppress all sample side effects.

On Thu, Dec 7, 2023 at 8:35 AM Kyle Huey <[email protected]> wrote:
>
> Returning zero from a bpf program attached to a perf event already
> suppresses any data output. By clearing pending_kill, returning zero from a
> bpf program will effectively pretend the sample never happened for all
> userspace purposes.
>
> Signed-off-by: Kyle Huey <[email protected]>
> ---
> kernel/events/core.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>

LGTM

Acked-by: Andrii Nakryiko <[email protected]>

> diff --git a/kernel/events/core.c b/kernel/events/core.c
> index 19fddfc27a4a..6cda05a4969d 100644
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -10421,8 +10421,10 @@ static void bpf_overflow_handler(struct perf_event *event,
> rcu_read_unlock();
> out:
> __this_cpu_dec(bpf_prog_active);
> - if (!ret)
> + if (!ret) {
> + event->pending_kill = 0;
> return;
> + }
>
> event->orig_overflow_handler(event, data, regs);
> }
> --
> 2.34.1
>