Date: Thu, 2 Nov 2006 12:31:52 -0500 From: John Baldwin <jhb@freebsd.org> To: Scott Long <scottl@samsco.org> Cc: Perforce Change Reviews <perforce@freebsd.org>, Scott Long <scottl@freebsd.org>, Hans Petter Selasky <hselasky@c2i.net> Subject: Re: PERFORCE change 108878 for review Message-ID: <200611021231.53607.jhb@freebsd.org> In-Reply-To: <454A2679.4030609@samsco.org> References: <200611010112.kA11C1Jt066210@repoman.freebsd.org> <200611021151.19396.jhb@freebsd.org> <454A2679.4030609@samsco.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday 02 November 2006 12:10, Scott Long wrote: > John Baldwin wrote: > > On Thursday 02 November 2006 06:22, Hans Petter Selasky wrote: > > > >>On Wednesday 01 November 2006 16:47, John Baldwin wrote: > >> > >>>On Tuesday 31 October 2006 20:12, Scott Long wrote: > >>> > >>>>http://perforce.freebsd.org/chv.cgi?CH=108878 > >>>> > >>>>Change 108878 by scottl@scottl-x64 on 2006/11/01 01:11:30 > >>>> > >>>> For some wonderful reason, you cannot pass &Giant to msleep. Work > >>>> around that in a crude fashion. Also add some more assertions. > >>> > >>>Ah, yes, that would be most unhappy. I guess mostly the idea is that Giant > >>>should be rather implicit and explicitly using Giant for an object lock is > >>>discouraged. I'm not sure that is what you are doing though. I think > >>>maybe you are borrowing Giant that's already held? > >> > >>I use this patch: > >> > >>/* preliminary fix for a bug in msleep on FreeBSD, > >> * which cannot sleep with Giant: > >> */ > >>#define msleep(i,m,p,w,t) msleep(i,(((m) == &Giant) ? NULL : (m)),p,w,t) > >> > >>Really this issue should be fixed. It happens just because GIANT_DROP is done > >>too early in "msleep()". > > > > > > Giant is special in msleep() and friends to make sure it is first in the > > lock order, but unlock doesn't matter for lock order, and actually, the > > current order is less intuitive. I think it's the way it is now because we > > inherited it from BSD/OS. Also in theory old code under Giant should be > > using tsleep() and not msleep() anyway. It actually won't hurt to move > > DROP_GIANT later though. > > > > How about this: > > > > This won't work for what I'm using it for. It's not a big deal, though. ? It should make msleep(&Giant) work just the same as msleep(&foo). In this case if Giant is only singly locked, DROP_GIANT just won't do anything. -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200611021231.53607.jhb>