Re: Unusual filesystem/memory corruption
Michael Perry wrote:
> Quoting John Toon on Sun, Aug 12, 2001 at 12:04:20AM +0100:
> > On 11 Aug 2001 16:37:58 -0500, Mike Brownlow wrote:
> > > * hdparm options: -m16 -d1
> > > * Machine is behind firewall, and has few services for external
> > > access
> > >
> > > I suspect hardware failure caused it, but there are still a few software
> > > unknowns. I'm starting to lean on corruption due to using -m16 for
> > > hdparm. Any other suggestions appreciated...
> > Ouch. Sounds bad.
> > In general, you should _not_ use hdparm, for the simple reason that if
> > you build a custom kernel with all the necessary options, it will
> > default to optimum performance anyway (you can set "Use DMA by default"
> > under the block device options).
> > I'm running a 2.4.5 XFS patched custom kernel that I built, and DMA/32
> > bit disc access etc. is fully operational by default since I selected
> > the appropriate kernel options. With modern kernels, hdparm is totally
> > unnecessary and potentially dangerous (except for performing performance
> > tests on your drive of course!).
> > What chipset does your mainboard use?
> > John.
> I had a similar thing occur on a system running a VIA/southbridge
> chipset on 2.2.18. this a K7T style motherboard with a KT133 chipset.
> I had been using it with a Western Digital 6g drive and the system was
> so corrupt and unusable it could not find mounted file systems, the
> /etc/fstab file corrupt, and on... I was also using hdparm.
Yep. Sounds like mine. Expect I was using 2.4.7 as well as hdparm.
Oops. The drive was a maxtor though. I don't think the IBM drives had
any problems though.
> I re-installed and went to a 2.4.5 kernel with the via/southbridge ide
> driver. The system now runs happily with twin maxtor 40g ide drives on
> the 2.4.7 kernel with the via ide driver compiled in.
Okay, I think using hdparm is what did me in. For now I just
got a 40g scsi to work on while I examine the trashed drives.
Thanks for the suggestions!
Mike Brownlow ><> http://www.wsmake.org/~mike/
1024D/8AA6EAFD 3861 96B3 EEA2 285C BE23 F706 3E1E EBB2 8AA6 EAFD