Date: Tue, 24 Jun 2003 23:18:15 -0700 (PDT) From: Nate Lawson <nate@root.org> To: Andrea Campi <andrea@webcom.it> Cc: cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/dev/pccbb pccbb.c pccbbvar.h Message-ID: <20030624231534.K71390@root.org> In-Reply-To: <20030625001613.GA636@webcom.it> References: <200306120337.h5C3bSmB075154@repoman.freebsd.org> <20030625001613.GA636@webcom.it>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 25 Jun 2003, Andrea Campi wrote:
> On Wed, Jun 11, 2003 at 08:37:28PM -0700, Warner Losh wrote:
> > Make cbb interrupts MPSAFE:
> > o Register ISR INTR_MPSAFE.
> ...
> > Revision Changes Path
> > 1.76 +31 -20 src/sys/dev/pccbb/pccbb.c
> > 1.19 +3 -2 src/sys/dev/pccbb/pccbbvar.h
>
> This commit causes my ep0 to stop working after some traffic is exchanged;
> the symptoms are full buffers (e.g. dhclient: send_packet: No buffer space available)
>
> I traced the issue to this commit by dichotomic search; I also verified reverting to
> 1.78 and applying this patch is enough to get it to work again (I'm not actually sure
> all three hunks are needed).
>
> Could you look into this issue?
Um, did you cvsup to 1.78? I found an error in the original commit where
it locks Giant twice (instead of unlocking) which would give exactly these
symptoms.
> --- pccbb.c Thu Jun 12 05:37:28 2003
> +++ pccbb.c.fixed Wed Jun 25 02:05:28 2003
> @@ -1087,11 +1086,7 @@
> }
> if (sc->flags & CBB_CARD_OK) {
> STAILQ_FOREACH(ih, &sc->intr_handlers, entries) {
> - if ((ih->flags & INTR_MPSAFE) != 0)
> - mtx_lock(&Giant);
> (*ih->intr)(ih->arg);
> - if ((ih->flags & INTR_MPSAFE) != 0)
> - mtx_lock(&Giant);
> }
> }
> }
By your patch above, it looks like you were not using 1.78.
-Nate
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030624231534.K71390>