RE: 1.3.x and disk corruptions

David Eagles (eaglesd@cmpsf.com.au)
Tue, 24 Oct 1995 10:27:10 +-1000


I have tried to run several versions of 1.3 kernels, and so far each one =

has caused disk corruption of the sort below; these came straight from=20
the system logs.

I recently purchased a new PCI Pentium/100 with dual 850MB EIDE drives =
and a Triton-based M.B. I noticed I was having the same problems as =
soon as I installed v1.3.33 or above (although the time between each =
upgrade was small, so I can't be 100% sure) which I suspect is due to =
problems with the new drivers. I'd suggest going back to 1.3.31 or =
1.3.32 as I've done and see how things go then.

For those working on the drivers, the files in question were definately =
in the /usr/src/linux tree (but that's probably due to the fact that =
most of the I/O I was doing was to that area) and only seemed to show =
corruption during a compile. If I cleaned out the src/linux tree, =
untarred the distribution and looked at the files, they all seemed ok, =
but then when I compiled 75% of the time it would bomb out with errors =
such as truncated file, etc. Looking at the files/directories in =
question showed many files had been corrupted and showed file sizes in =
the hundreds of MB - they weren't actually this big though, as df =
reported the correct available/used space.

Anyway, more food for thought.

My system is:
PCI Triton M.B.
2 x 850MB EIDE HDD using onboard EIDE controller
SMC Elite 16 Ethernet Card
16 MB Memory

Kernel was 1.3.33 (or higher) with _NO_ PPP, SLIP or anything compiled =
in (compiled as modules however, but not installed). Just ELF, PCI,
Network (with Dummy driver) and new IDE driver (no SCSI).