This actually started out as tracking down oopses from unmounting full
LVM snapshots... this is part of the way there. Mounting the (readonly)
lvm snapshot device with quota flags that differed from the source filesystem
would cause an I/O error (since a transaction was needed on the ro device),
shutting down the filesystem.
Still need to track down the oops after the shutdown, but at least now the
shutdown won't happen - the fs won't mount if the flags don't match (and
you get a helpful error message in the logs).
I tested this with overflowed snapshots (from the original problem) and
things seem to work, except when unloading the module one of the xfs cache
zones can't be freed, we leak something somewhere.
This fix does address a problem by itself, though - now on to the next part. :)
Date: Sun Mar 3 14:34:15 PST 2002
Workarea: stout.americas.sgi.com:/localhome/eric/2.4.x-xfs/workarea-alwaysclean
The following file(s) were checked into:
bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs
Modid: 2.4.x-xfs:slinx:113180a
linux/fs/xfs/xfs_mount.c - 1.271
- Don't allow quota flag change on readonly device
|