On Saturday 26 July 2008 07:41, Ron Johnson wrote: > On 07/25/08 20:21, Andrew Reid wrote: > [snip] > > > - For certain nVidia chipsets in combination with large XFS > > file systems, you need to boot with "iommu=soft" in order > > to avoid (infrequent) random filesystem errors. See the > > Debian "etch" release notes for details. > > I don't see this mentioned in the i386 notes. Does it only affect > AMD64? And, specifically, AMD chips? > http://www.mail-archive.com/[EMAIL PROTECTED]/msg97076.html
That's the one -- you're right that it's apparently only a 64-bit problem, I hadn't appreciated that, since all my servers are 64-bit at this point. I found it here: http://www.debian.org/releases/stable/amd64/release-notes/ch-information.en.html#s-nvidia-iommu ... and it's the same one. > Also, has it been fixed in later kernels? The bugzilla page for kernel bug <A HREF="http://bugzilla.kernel.org/show_bug.cgi?id=7768";>#7768</A> says it's closed and resolved by code_fix, so apparently has been fixed, although one would have to dig deeper to say in which version the fix first appeared. > Lastly, what, in this instance, is "large"? I was having this problem intermittently with a cron-triggered rsync backup on a 4.2TB software RAID6 array hosting an xfs filesystem that was 80% full. The problem seemed to appear around the time the filesystem crossed the 80% threshold, so "large" might mean 3.2 TB or so. Booting with "iommu=soft" made the problem go away, with no evident loss of server performance. -- A. -- Andrew Reid / [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
