Re: [PATCH 1/2] mmiotrace: fix buffer overrun detection

[Frédéric Weisbecker]

2008/11/24 Pekka Paalanen <pq@xxxxxx>:
> On Mon, 24 Nov 2008 21:30:24 +0100
> "Frédéric Weisbecker" <fweisbec@xxxxxxxxx> wrote:
>
>> 2008/11/24 Pekka Paalanen <pq@xxxxxx>:
>> >
>> > Oh right, the busylooping bug when echoing into trace_marker,
>> > I already forgot about that, since I couldn't reproduce it.
>> > I guess there's nothing to do, until someone finds a way to
>> > reproduce it.
>>
>>
>> Hi Pekka.
>>
>> Are you talking about this old bug that broke the pipe?
>> If so, you fixed it yourself partially with a patch that corrected
>> tracing_read_pipe.
>> Another origin was the none tracer that set tracer_enabled to 0.
>> But i has been fixed with the nop tracer.
>>
>> Perhaps you are talking of something else?
>
> Yeah, I'm talking about the shell spinning 90% in kernel and 10% in
> user, I tried to explain it in the part you left out from the quote.
> That is, doing 'echo foo > trace_marker' makes the shell/echo spin.
> And this doesn't seem to be deterministic, or the exact triggering
> sequence is too complex. I didn't notice any other adverse effects
> than the shell getting stuck.
>
> Maybe a corner case related to having only a single page in the
> ring buffer... nah, more likely something unexpected under
> trace_marker callbacks. Returning zero instead of error on
> write(2) maybe... something that makes 'echo' retry again and
> again, never succeeding or really failing.
>
> --
> Pekka Paalanen
> http://www.iki.fi/pq/
>


Perhaps that:

int trace_vprintk(unsigned long ip, const char *fmt, va_list args)
{
     ......
     if (tracing_disabled)
		return 0;
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/