A few days ago, I wrote to the list about a problem I was having with
making a large filesystem on a HP NetRAID (really an AMI MegaRaid). I also
wrote to the developers of the driver (names in the megaraid.c file) and
they responded with a fix that worked. I'm including it here, with the
hopes that it can help somebody else.
First is the response from the developer, then my original description.
Kevin
Xi Chen wrote:
>This is a known problem with SCSI upper layer timeout value in
>/usr/src/linux/drivers/scsi/sd.c. The increased timeout value is needed
>because the RAID controller will allow hundreds of commands to be queued at
>once, but the disks can only accept 64 at a time so some have to wait in an
>intermediate queue. If this value is not changed, false scsi timeouts and
>aborts may occur from the upper scsi layer under a heavy load.
>
>So, make following change in sd.c file: changing
>#define SD_TIMEOUT (15 * HZ)
>to:
>#define SD_TIMEOUT (30 * HZ)
I originally wrote:
>I'm trying to make a 20 gig file system on a HP NetServer. It has an "HP
>Net Raid" card, which I'm driving with the AMI MegaRAID driver. I've used
>kernel 2.2.10, unpatched, and added AMI's megaraid.c version 1.01 manually.
> This morning, I tried again with Alan's proposed 2.0.11pre2 patch (which
>has the megaraid.c 1.01 with it). I still had the same problem.
>
>I can fdisk and create the partition just fine (my RAID has 6 18 gig
>drives, 2 hot swap, 4 in a RAID 5, exposing about 52.1 gigs of space). But
>when I mkfs the 20 gig partition, it:
>
>Writing inode tables: up to 2501, then says "done".
>Site there after writing "done" for 2:30.
>Says "Writing superblocks & filesystem accounting info".
>Sits there for about 2 minutes.
>Aborts with "aborting command due to timeout". Prints the string:
>(10) 20 03 25 c5 ff 00 00 02 00
>Adds a second "aborting command due to timeout" with the string:
>(10) 20 03 28 86 5D 00 00 02 00
>
>Then it adds "Abort: 1500 1500".
>
>Finally, "Blocked mailbox on exit" appears, and repeats, over and over
>again, scrolling up the screen about 2 per second...the only way to recover
>is a hard reset.
>
>I'm going to try some other things: mkfs with a smaller number of inodes,
>etc, but I figured I'd pose my problem to the gurus and see if anybody's
>seen anything like this.
>
>The same RAID drive has a 1 gig and a 5 gig partition on it, created and
>formatted just fine.
>
>Any help? Any suggestions?
>
>Thanks...
>
>Kevin
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
