RE: File System Errors (Additional Data)
>
> I'm beginning to suspect that Roland's fixes to the ext2fs server (to
> solve the problems people were experiencing with the new
> e2fsprogs 1.15) are to blame. The
>
> hdxx: bas access
>
> messages are probably caused by the fact that the ext2fs server tries
> to access a nonexisting block, either because of bad filesystem
> metadata, or because it doesn't interpret the data in the right way.
Is it possible that these are the result of ext2fs not understanding
the partitioning scheme in the case of logical drives?
> Roland might fix the bug eventually, but the best thing to do would be
> for someone who actually is experiencing the problems to try to find
> out what's wrong. Either by debugging the filesystem or by carefully
> inspecting the code and comparing it with an older version that
> doesn't have the change associated with the 1999-08-23 ChangeLog
> entry.
>
Earlier today I diffed a 19990725 version of the hurd (just the hurd,
not gnumach, libc, etc.) against 19990922.
It appears that several large changes were made to ext2fs:
1. It was updated to Linux 2.3.14
2. balloc.c was modified to allow freeing of blocks across group
boundaries.
I also took a quick look through the Linux Kernel archives, but did
not see any messages regarding errors with the 2.3.14-series ext2fs
support.
What would be the most helpful course of action to take? I intend
to generate a debug-version of the Hurd and see if I can isolate
a case where the corruption occurs.
-Brent
Reply to: