Date: Tue, 30 Sep 2008 08:54:10 +0200 From: Marius Strobl <marius@alchemy.franken.de> To: Sean Thomas Caron <scaron@umich.edu> Cc: freebsd-sparc64@freebsd.org Subject: Re: Kernel panic in 7.0-RELEASE with fatm driver Message-ID: <20080930065410.GA87677@alchemy.franken.de> In-Reply-To: <20080929201947.86701vp3bs8vcatc@web.mail.umich.edu> References: <20080929201947.86701vp3bs8vcatc@web.mail.umich.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
--+QahgC5+KEYLbs62 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline On Mon, Sep 29, 2008 at 08:19:47PM -0400, Sean Thomas Caron wrote: > Hi folks, > > I've been running FreeBSD 6.x on a number of Ultrasparc based systems > using the Cranor NATM driver with a bunch of Fore PCA-200E cards. It > generally worked but as the machines got hit with higher and higher > network load I would get kernel panics at fairly random intervals on > "sbdrop" and "sbflush". Sometimes I would get two per day, sometimes I > wouldn't get any for two weeks. > > So after a while at this I decided to try 7.0-RELEASE out hoping > whatever bug was causing these random kernel panics might have gotten > fixed. > > Test system is Sun Fire v120, 550 MHz, 1 GB RAM, PCA-200e installed in > PCI slot. > > I pruned a bunch of stuff from GENERIC and added the usual Cranor NATM > stuff that I was using in 6.x. In particular - > > # ATM > > options NATM > device atm > device fatm > device utopia > > Kernel does have multiprocessing enabled (as GENERIC did) even though > it's a uniprocessor machine. If you're compiling a custom kernel anyway you should leave out options SMP on a UP machine for performance reasons. > > I built the new kernel then reboot and it panics immediately after > touching the ATM card - > > fatm0: <FORE PCA200E> mem 0x200000-0x3fffff at device 5.0 on pci2 > panic: bus_dma_tag_create: parent DMA tag NULL > cpuid = 0 > Uptime: 1s > Automatic reboot in 15 seconds - press a key on the console to abort > --> Press a key on the console to reboot, > --> or switch off the system now. > > I'm hoping maybe someone has seen this before and has a patch laying > around somewhere, or if nothing else, it might get passed along to the > fatm maintainer eventually? Please give the attached patch a try. > > If anyone wants to see any further information e.g. full kernel > configuration file, I will be happy to pass along. I can also procure > a backtrace if someone can tell me how to hardcode a dump path into > the kernel; I can't find a procedure documented anywhere. > FYI, you can specifiy the path (defaulting to /var/crash) for the dump via "dumpdir" in /etc/rc.conf. Marius --+QahgC5+KEYLbs62 Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="fatm_bus_get_dma_tag.diff" Index: if_fatm.c =================================================================== --- if_fatm.c (revision 183423) +++ if_fatm.c (working copy) @@ -2880,7 +2880,7 @@ fatm_attach(device_t dev) * restrictions for the controller (and PCI bus) and is never used * to do anything. */ - if (bus_dma_tag_create(NULL, 1, 0, + if (bus_dma_tag_create(bus_get_dma_tag(dev), 1, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, BUS_SPACE_MAXSIZE_32BIT, MAXDMASEGS, BUS_SPACE_MAXSIZE_32BIT, 0, NULL, NULL, --+QahgC5+KEYLbs62--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20080930065410.GA87677>