Hi!
Accessing a 4.3G FAT32 partition of an external SCSI disk (IBM ddrs 4.3G)
heavily, I get both a corrupted filesystem and a nuked partition table.
syslog says:
[normal operation]
kernel: sym53c810a-0-<1,0>: phase change 2-3 6@0000a045 resid=2.
kernel: sym53c810a-0-<1,0>: phase change 2-7 10@03328e45 resid=4.
[2 seconds delay]
kernel: Filesystem panic (dev 08:05).
kernel: File without EOF
kernel: File system has been set read-only
kernel: FAT: fat_truncate called though fs is read-only, uhh...
kernel: Filesystem panic (dev 08:05).
kernel: File without EOF
kernel: FAT: fat_truncate called though fs is read-only, u hh...
kernel: Filesystem panic (dev 08:05).
kernel: FAT error
kernel: Directory 134587845: bad FAT
[...]
This happens with 2.0.35 and 2.2.6, too.
So what is my problem? The filesystem or the SCSI system? Are there known
Bugs?
$ cat /proc/scsi/sym53c8xx/0
General information:
Chip sym53c810a, device id 0x1, revision id 0x23
IO port address 0xe400, IRQ number 11
Using memory mapped IO at virtual address 0xc4826000
Synchronous period factor 25, max commands per lun 32
$
SCSI termination is correct. The SCSI controller shares an interrupt with a
Riva128 Display adapter.
I tried to find an easy way to reproduce this. The mostly reproducable way
seems to be:
rm -rf on a tree like 1/2/3/4/5/8 of empty files
4 dd if=/dev/zero of=file.$i started with a delay of roughly 10 seconds
I'm not sure if the rm -rf is really neccessary - it was still running from
one of the previos tests (creating 100,000 dirs each with 10 empty files).
While performing one short test with ext2 nothing got corrupted.
Thanx
Rainer
--
KeyID=58341901 fingerprint=A5 57 04 B3 69 88 A1 FB 78 1D B5 64 E0 BF 72 EB
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
