Re: [PATCH] ring-buffer: Do not trigger WARN_ON() due to a commit_overrun

[Masami Hiramatsu (Google) <mhiramat@kernel.org>]

On Tue, 27 May 2025 16:11:30 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:

> On Tue, 27 May 2025 12:11:40 -0400
> Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > But there just happens to be one scenario where this can legitimately
> > happen. That is on a commit_overrun. A commit overrun is when an interrupt
> > preempts an event being written to the buffer and then the interrupt adds
> > so many new events that it fills and wraps the buffer back to the commit.
> > Any new events would then be dropped and be reported as "missed_events".
> 
> I'll probably update the commit log, but the way I triggered this was to run:
> 
>  # perf record -o perf-test.dat -a -- trace-cmd record --nosplice  -e all -p function hackbench 50

Hmm, so this runs 3 commands, hackbench, which is traced by trace-cmd, which
is traced by perf.

> 
> Which causes perf to trigger a bunch of interrupts while trace-cmd enables
> function tracing and all events. This is on a debug kernel that has
> lockdep, KASAN and interrupt and preemption disabling events enabled.

Ah, that is the full-set of the interrupt and tracing :)

Thanks,

> 
> Basically, this causes a lot to be traced in an interrupt. Enough to fill
> 1.4 megs of the tracing buffer with events in interrupts before a single
> event could be recorded.
> 
> I've never triggered this when those extreme conditions were not there.
> 
> -- Steve


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>