Re: [PATCH v1 2/8] tracing/ftrace: guard syscall probe with preempt_notrace

[Andrii Nakryiko]

On Fri, Oct 4, 2024 at 7:53 AM Mathieu Desnoyers
<mathieu.desnoyers@xxxxxxxxxxxx> wrote:
>
> On 2024-10-04 16:52, Steven Rostedt wrote:
> > On Fri, 4 Oct 2024 10:19:36 -0400
> > Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx> wrote:
> >
> >> The eBPF people want to leverage this. When I last discussed this with
> >> eBPF maintainers, they were open to adapt eBPF after this infrastructure
> >> series is merged. Based on this eBPF attempt from 2022:
> >>
> >> https://lore.kernel.org/lkml/c323bce9-a04e-b1c3-580a-783fde259d60@xxxxxx/
> >
> > Sorry, I wasn't part of that discussion.
> >
> >>
> >> The sframe code is just getting in shape (2024), but is far from being ready.
> >>
> >> Everyone appears to be waiting for this infrastructure work to go in
> >> before they can build on top. Once this infrastructure is available,
> >> multiple groups can start working on introducing use of this into their
> >> own code in parallel.
> >>
> >> Four years into this effort, and this is the first time we're told we need
> >> to adapt in-tree tracers to handle the page faults before this can go in.
> >>
> >> Could you please stop moving the goal posts ?
> >
> > I don't think I'm moving the goal posts. I was mentioning to show an
> > in-tree user. If BPF wants this, I'm all for it. The only thing I saw was a
> > generalization in the cover letter about perf, bpf and ftrace using
> > faultible tracepoints. I just wanted to see a path for that happening.
>
> AFAIU eBPF folks are very eager to start making use of this, so we won't
> have to wait long.

I already gave my ack on BPF parts of this patch set, but I'll
elaborate a bit more here for the record. There seems to be two things
that's been discussed.

First, preempt_disable() vs migrate_disable(). We only need the
latter, but the former just preserves current behavior and I think
it's fine, we can follow up with BPF-specific bits later to optimize
and clean this up further. No big deal.

Second, whether BPF can utilize sleepable (faultable) tracepoints
right now with these changes. No, we need a bit more work (again, in
BPF specific parts) to allow faultable tracepoint attachment for BPF
programs. But it's a bit nuanced piece of code to get everything
right, and it's best done by someone more familiar with BPF internals.
So I wouldn't expect Mathieu to do this either.

So, tl;dr, I think patches are fine as-is (from BPF perspective), and
we'd like to see them applied and get to bpf-next for further
development on top of that.

>
> Thanks,
>
> Mathieu
>
> --
> Mathieu Desnoyers
> EfficiOS Inc.
> https://www.efficios.com
>