cvs commit: src/sys/kern uipc_usrreq.c
rwatson at freebsd.org
Fri Jun 11 02:55:03 GMT 2004
On Thu, 10 Jun 2004, Alfred Perlstein wrote:
> * Robert Watson <rwatson at freebsd.org> [040610 19:18] wrote:
> > Agreed. If the caller doesn't mind us dropping the unp_mtx, the first
> > logical thing to try is to replace unp_gcing with an sx lock (or the
> > like), and only acquire unp_mtx when calling into code that needs it.
> That's not really needed. The idea is just to avoid initiating another
> gc scan while one is in progress.
It would almost be tempting to do that cleanup in a separate worker
(perhaps a callout), and just launch the cleanup using a cv when
necessary. That would get all the work out of a context that probably
includes socket activity, etc.
> I'll look at it if I find the time/motivation.
Sounds good to me :-). I'll move ahead with merging other stack locking
bits in the mean time.
Robert N M Watson FreeBSD Core Team, TrustedBSD Projects
robert at fledge.watson.org Senior Research Scientist, McAfee Research
More information about the cvs-all