On Tue, Aug 11, 2026 at 09:48:57AM -0400, Steven Rostedt wrote:
> On Tue, 11 Aug 2026 10:09:27 +0100
> Vincent Donnefort <[email protected]> wrote:
> 
> > > > This is a pre-existing issue, but if an oversized event causes
> > > > trace_remote_iter_print_event() to return -EOVERFLOW, will this break
> > > > the loop without calling trace_remote_iter_move()?
> > > > 
> > > > Since the event is left unconsumed in the buffer, could this lead to an
> > > > uninterruptible infinite loop (hard lockup) when reading the trace 
> > > > pipe?  
> > > 
> > > I'll let you answer the above.
> > > 
> > > -- Steve  
> > 
> > I don't believe this can realistically happen, trace_seq buffer is 8K and 
> > remote
> > events are really small. 
> 
> They are small until they are not ;-)
> 
> Is there anything that forces their size to be small?
> 
> -- Steve

I don't see anything preventing the declaration of a massive print function at
the moment...

... and neither the registration of an event with a size bigger than a page.

Is it something we should consider? I have ignored it (in both trace_remote and
simple_ring_buffer) as it seems very much unlikely anyone would do that.

-- 
Vincent

Reply via email to