Bug#367125: knfsd / bad inode number on ext3

["Peter Kruse" <pjodrr@gmail.com>]

Hello,

thanks for your info.  That means it is a bug in ext3.  It seems
to ignore the errors=continue mount option, although
according to the manpage it should support all the
options that ext2 does.
Unfortunately I'm not able to reproduce this exact behaviour, however
I am able to produce a filesystem error and make ext3 remount
the filesystem read-only, while ext2 continues on an error.
Here is what happens with ext2:

May 26 15:46:25 nfstest1 kernel: attempt to access beyond end of device
May 26 15:46:25 nfstest1 kernel: drbd1: rw=0, want=146538512, limit=146538496
May 26 15:46:25 nfstest1 kernel: EXT2-fs error (device drbd1):
read_inode_bitmap: Cannot read inode bitmap - block_group = 559,
inode_bitmap = 18317313

And this is with ext3:

May 26 13:31:37 nfstest1 kernel: attempt to access beyond end of device
May 26 13:31:37 nfstest1 kernel: drbd1: rw=0, want=146538512, limit=146538496
May 26 13:31:37 nfstest1 kernel: EXT3-fs error (device drbd1):
read_inode_bitmap: Cannot read inode bitmap - block_group = 559,
inode_bitmap = 18317313
May 26 13:31:37 nfstest1 kernel: Aborting journal on device drbd1.
May 26 13:31:37 nfstest1 kernel: EXT3-fs error (device drbd1) in
ext3_new_inode: IO failure
May 26 13:31:37 nfstest1 kernel: EXT3-fs error (device drbd1) in
ext3_create: IO failure
May 26 13:31:37 nfstest1 kernel: ext3_abort called.
May 26 13:31:37 nfstest1 kernel: EXT3-fs error (device drbd1):
ext3_journal_start_sb:Detected aborted journal
May 26 13:31:37 nfstest1 kernel: Remounting filesystem read-only
May 26 13:31:37 nfstest1 kernel: __journal_remove_journal_head:
freeing b_committed_data

I would consider this a bug, maybe it's only a bug in the manpage ;)

 Peter