Date: Wed, 13 Apr 2005 22:39:24 +0200 From: Antoine Brodin <antoine.brodin@laposte.net> To: Alfred Perlstein <alfred@freebsd.org> Cc: mdodd@freebsd.org Subject: Re: cvs commit: src/share/man/man4 unix.4 src/sys/sys socket.h un.h unpcb.h src/sys/kern uipc_usrreq.c Message-ID: <20050413223924.75ac1047.antoine.brodin@laposte.net> In-Reply-To: <20050413195312.GD60345@elvis.mu.org> References: <200504130001.j3D01kuD081602@repoman.freebsd.org> <20050413184246.19212b84.antoine.brodin@laposte.net> <20050413124956.M3762@sasami.jurai.net> <20050413195312.GD60345@elvis.mu.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Alfred Perlstein <alfred@freebsd.org> wrote: > * Matthew N. Dodd <mdodd@FreeBSD.ORG> [050413 10:04] wrote: > > On Wed, 13 Apr 2005, Antoine Brodin wrote: > > >I'm not sure it's safe to sooptcopyout/sooptcopyin while holding the > > >unp mutex in uipc_ctloutput. It can fault. > > > > I was concerned about this but saw that other code holds locks across > > sooptcopy*() calls. > > > > If there is a deadlock problem it may be time to add a per unpcb lock. > > This should be tripping an assertion, copyout/copyin is not ok while > holding a mutex. Does anyone want to add the logic to sooptcopy*? A simple WITNESS_WARN(WARN_GIANTOK|WARN_SLEEPOK, NULL, "Sleeping in sofoo"); should do the trick. Cheers, Antoine
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050413223924.75ac1047.antoine.brodin>