[linux-audio-dev] Re: [announce] [patch] Voluntary Kernel Preemption Patch

Ingo Molnar mingo at elte.hu
Sun Jul 11 08:03:31 EDT 2004


* Andrew Morton <akpm at osdl.org> wrote:

> Oh I can buy the make-the-bugs-less-probable practical argument, but
> sheesh.  If you insist on going this way we can stick the patch in
> after 2.7 has forked.  I spose.  The patch will actually slow the rate
> of improvement of the kernel :(

it has already improved the kernel by already detecting at least 2 bugs
that were not found via CONFIG_PREEMPT or other means (the cfq-iosched
one and the proc_pid_flush() one) - less then 1 week after i wrote the
patch.

It also already resulted in a number of independent improvements: e.g.
the lock-break changes it includes, and all the work on the debug
infrastructure.

So i fail to see how it would slow the rate of improvement of the
kernel. It's simply the middle step towards CONFIG_PREEMPT - a feature
that several major distribution have not enabled so far. It's a more
opt-in approach instead of CONFIG_PREEMPT's opt-out approach, that's
all.

(that having said it's not at all clear that voluntary-preempt would get
into Fedora. If it causes stability problems it will be zapped. But at
least the problems it triggers we could trigger _even on the current UP
kernel_, and that's the difference to preempt - CONFIG_PREEMPT triggers
_new_ problems on UP.)

also, frankly, at minimum the latency side of CONFIG_PREEMPT has been
underdeveloped lately. Nobody seemed to care about real latencies that
happen. Just do a 'du /; du /; sync' on a box with enough RAM and listen
to audio. Or trigger some swapping. (these particular latencies are
fixed in the voluntary-preempt patchset.)

	Ingo


More information about the linux-audio-dev mailing list