Date: Tue, 21 Aug 2001 07:20:01 -0700 (PDT) From: Andre Albsmeier <andre.albsmeier@mchp.siemens.de> To: freebsd-bugs@FreeBSD.org Subject: Re: kern/29845: 4.4-PRERELEASE crashes under heavy net I/O Message-ID: <200108211420.f7LEK1F60355@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR kern/29845; it has been noted by GNATS. From: Andre Albsmeier <andre.albsmeier@mchp.siemens.de> To: Ian Dowse <iedowse@maths.tcd.ie> Cc: Andre Albsmeier <andre.albsmeier@mchp.siemens.de>, Warner Losh <imp@harmony.village.org>, freebsd-gnats-submit@FreeBSD.org Subject: Re: kern/29845: 4.4-PRERELEASE crashes under heavy net I/O Date: Tue, 21 Aug 2001 16:17:49 +0200 On Tue, 21-Aug-2001 at 14:42:54 +0100, Ian Dowse wrote: > > [ Andre tried adding the kludge from if_sl.c that merges the tty and > net interrupt masks - not as a solution - but just to determine if > these crashes are caused by an spl problem. It seems they are. ] > > In message <20010821143627.A26964@curry.mchp.siemens.de>, Andre Albsmeier write > s: > >On Tue, 21-Aug-2001 at 11:52:57 +0100, Ian Dowse wrote: > >> The fact that it is only pccard cards that have problems really > >> suggests a problem there, but the crashes are so random that it > >> has to be some kind of spl problem. Previously the cards got their > >> own IRQ, so they would set it up with the right interrupt mask. > >> Now all pccard interrupt handlers are called from the pcic one, > >> so I don't think splimp() is blocking these interrupts. > > > >Yes, that seems to do it! I have my dd from /dev/zero to /dev/null via > >rsh running in both directions for about 5 minutes now... No problems > >so far. > > So at the moment, when the network code calls splimp(), it does > not block NIC interrupts that come in via the pccard code. That > certainly explains all the odd crashes. > > I'm not sure how to solve this problem properly, but it seems that > pcic_pci_setup_intr() needs to call bus_generic_setup_intr() to > properly update the interrupt masks. I assume there is a reason > for not just using bus_generic_setup_intr() as the pcic_pci > bus_setup_intr method? > > Thanks for trying out that kludge Andre! Hopefully there's enough > information now to get it fixed properly. Well I was only whining about the problem, you fixed it (or at least isolated it) :-) Anyway, I am looking forward to testing other suggestions. It seems that I have an environment that triggers the problem easily. Thanks, -Andre To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200108211420.f7LEK1F60355>