Simon Matter wrote:
>
> I was disturbing this list with my IDE UDMA corruption problem several
> times now and I have another question about that.
>
> When I put heavy load on more than one IDE UDMA disk simultaneously, I
> get filesystem corruption. This happens on several IDE / Motherboard
> combinations. I have tried different kernels and until
> 2.4.3-SGI_XFS_1.0.1_PR3 it was all the same.
>
> Now, could it be that we have to blame egcs-2.91.66 for this? That's why
> I post to the XFS list because we are the people compiling the RH
> kernels with egcs-2.91.66 instead of gcc-2.96-85. As I read on this list
> XFS kernels can not be compiled with 2.96-85 because it breaks
> something, could it be that egcs-2.91.66 breaks something else in the
> RedHat Kernels. Why then does RedHat use 2.96-85 for the kernel?
I personally have heard more about problems with people trying 2.96 than
2.91.66. From the kernel list, it seems unlikely that kgcc is at fault
(that doesn't mean it can't be at fault).
>
> BTW unfortunately I don't have the failing systems available for more
> test. I'm just trying to find what went wrong.
>
> Simon
Probably someone has mentioned this already, I'm curious about what to
use to check for bad blocks? Since fsck.xfs just returns, and I'm not
sure if the blockmap in the man page for xfs_repair refers to the same
thing as checking for bad blocks and marking them, what is the
specifically recommended way to check for bad blocks on an XFS
partition? Is the filesystem itself able to work around bad blocks
(especially if the drive is degrading and new bad blocks are appearing)?
If so, does it log a message to let the admin know of a degrading drive?
D. Stimits, stimits@xxxxxxxxxx
|