Date: Mon, 25 Sep 2000 17:12:40 +0100 From: Steve Roome <steve@sse0691.bri.hp.com> To: kafeend <jimmychonga@pop1.crosswinds.net> Cc: hardware@freebsd.org Subject: Re: UDMA write error Message-ID: <20000925171240.F8111@moose.bri.hp.com> In-Reply-To: <3.0.6.32.20000923164403.00795b10@pop.crosswinds.net>; from jimmychonga@pop1.crosswinds.net on Sat, Sep 23, 2000 at 04:44:03PM -0700 References: <3.0.6.32.20000923164403.00795b10@pop.crosswinds.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Sep 23, 2000 at 04:44:03PM -0700, kafeend wrote: > I came across your answer to Nader's problem. And i have the same it > seems like all VIA chipsets have this problem :). I wanted to know > how to make it not start up in ata mode. the output of sysctl > hw.atamodes is "hw.atamodes:pio,pio,"...but it still gives me that > error the UDMA write error on boot. How can i fix it so that this > won't happen anymore. Also i don't have an /etc/sysctl.conf file > :(. Thanks kafeend Firstly you really need to get your mail client to wrap lines at 72 chars or thereabouts.. I saw another post, possibly from you recently on -stable and I almost replied but it was a right mess because it wasn't wrapped the way I (and a lot of others) like. Secondly, I've cc:d this to freebsd-hardware because it seems relevant, and hopefully someone will either find my comments useful and perhaps even get back to me with corrections or explanations for where I've got things wrong. Finally, in response, the problem, UDMA devices dying horribly... I have a 82c686 as well, (mine is the onboard one on an FIC SD-11, which is not really the best of motherboards, but does the job and works OKAY being one of the earliest boards to support athlons). I also have a 34Gb IBM disk, UDMA66 (and apparenly UDMA 100) compatible. Originally I had my disk setup in UDMA 66 mode, and it went all pear shaped and died with lots of bad blocks while writing or reading #<some block numbers>. The disk actually then reported itself as dead to Windows as well, and I got as far as sending it back to the manufacturors to get a replacement. However, it turns out I was one of the first to get bit by this problem, and all that had happened is that the IDE driver had toasted the disk due to some sort of corruption. I finally wiped the disk with a util from IBM for ... wiping your disks! The same disk now works, and I have moved it so that it is better cooled and rechecked all the cabling. I think that your problem, and a lot of other problems people have been having with this are due to UDMA going wrong because a disk has overheated, I think *guessing* that writing to a device with UDMA which is just in the process of dying due to overheating can cause problems with the blocks on the disk. Not just in the drivers.. Note that after my problem I couldn't even get the IBM DOS disk utils to properly identify and check the disk until after a complete low level wipe. Interestingly my problem occurred about 10GB into the FreeBSD partition of the disk, yet it rendered the Windows partition corrupt as well. Now, as a solution you can turn off UDMA, (I've put it here because I think you should read the bit above anyway!) but I don't beleive there is as much wrong with the driver as people are making out, and I'm not going to advocate that solution as it's probably solvable (soluble?!) by simply checking the cabling and cooling of the disk. Others will tell you to switch to PIO mode, but since moving my disk I've hammered it hard! I've got an Athlon 500, and after two or three succesive make -j<x> buildworlds I've come to the conclusion that although it looked like UDMA driver crapness it was just UDMA driver stressed hardware better than I'd installed it! Since then I've also repartitioned the disk much more cleanly and I move some very large files about and do TV captures etc. The only other possible culprit for data corruption might be using the UDMA driver to write more than 2Gb to a Windows (FAT32) filesystem. I've remember something about this, I think it might even cause a panic if you try that, but I'mm not willing to test it any further because I don't fancy wiping my disk again. I'll leave that as an excersice for someone on the hardware list who feels like having a bad week =) In summary though, you probably do actually have bad blocks on this disk now, I know that after my problem I did have blocks marked as bad on the disk. It killed two filesystems (about 15Gb apart!) and I had to wipe the disk to get rid of them. But I've never actually changed to PIO mode, and I don't think I or you need to! Of course, as always, your mileage may vary, Steve To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hardware" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000925171240.F8111>