OK, more info:
# mkraid /dev/md0
handling MD device /dev/md0
analyzing super-block
disk 0: /dev/sdb1, 4257193kB, raid superblock at 4257088kB
disk 1: /dev/sdc1, 4257193kB, raid superblock at 4257088kB
disk 2: /dev/sdd1, 4192933kB, raid superblock at 4192832kB
/dev/md0: Invalid argument
leaves this in the logs:
Mar 22 16:32:28 raidtest kernel: bind<sdb1,1>
Mar 22 16:32:28 raidtest kernel: no chunksize?
Mar 22 16:32:28 raidtest kernel: bind<sdc1,2>
Mar 22 16:32:28 raidtest kernel: no chunksize?
Mar 22 16:32:28 raidtest kernel: bind<sdd1,3>
Mar 22 16:32:28 raidtest kernel: no chunksize?
Mar 22 16:32:28 raidtest kernel: sdd1's event counter: 00000000
Mar 22 16:32:28 raidtest kernel: sdc1's event counter: 00000000
Mar 22 16:32:28 raidtest kernel: sdb1's event counter: 00000000
Mar 22 16:32:28 raidtest kernel: md: bug in file md.c, line 1536
Now, md.c is one of the things modified by the patch, so I looked at it a
bit:
* chunk-size has to be a power of 2 and multiples of PAGE_SIZE
*/
if ( (1 << ffz(~chunk_size)) != chunk_size) {
MD_BUG();
return -EINVAL;
and figured, contrary to documentation, that it actually needs a chunk
size defined.
(http://www.ostenfeld.dk/~jakob/Software-RAID.HOWTO/Software-RAID.HOWTO-3.html#ss3.2,
which I thought was the latest info, does not include chunk_size in the
example for a linear device.)
Adding a chunk_size to /etc/raidtab fixed my problem.
Having the persistent-superblock defined lets it be seen on boot-up, which
is nice! I can just stick an entry for /dev/md0 in /etc/fstab.
Thanks to all who helped!
-Ross
