Re: Trace event testing time - 3s to 41s

[Steven Rostedt]

On Sun, 2009-09-13 at 11:15 -0400, Parag Warudkar wrote:
> Somewhere between -rc9 and yesterday's git, my desktop went from using
> 3s worth of trace event testing time to 41s !
> 
> It looks like we not only added a whole bunch of tracing points but
> also are doing the tests twice post -rc9 - the addition being testing
> with function tracer.
> 
> grep "Testing event" dmesg.31rc9 |wc -l
> 100
> 
> grep "Testing event" dmesg.31git |wc -l
> 1172

This was due to the fact that we test every syscall now.

> 
> While one could get away with "don't enable that config option then"
> response, for testing kernels daily this sounds like it is going to be
> very annoying as it adds a considerable boot lag.
> 
> So I was thinking could we move this testing to a kthread with idle
> priority or something - may be disabling tracing until those tests are
> finished, if that sort of thing is necessary?
> Or does it need to happen synchronous to booting?

I don't have time at the moment, but I would not mind moving the syscall
testing to a thread. Heck, that way we can actually test syscalls
too ;-)

-- Steve


--
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/