I get EXT2-fs panic: ext2_write_inode: unable to read i-node block
followed by an inode number and a block number, which vary from time
to time.
OK, so there are two problems. The first is one that we know about
already --- we need to have better error recovery when the device driver
signals an error from a read.
The other problem is that the SCSI subsystem reported a disk error when
a SCSI bus reset happens, instead of retrying the request. Does that
sound right?
- Ted