2.0.27: SCSI hassles

Chris Evans (chris.evans@lady-margaret-hall.oxford.ac.uk)
Tue, 10 Dec 1996 14:42:21 +0000 (GMT)


Just installing a new 2Gb SCSI drive in our Linux box here. A few points...

1) I'm running badblocks -w on the new partition. Even though dropped to
single user mode, the performance is abysmal. I'm reported ~27Mb of
buffers in use...(!)

2) I had BAD problems running the aic7xxx.c driver with a few tweaked
options. Note that said options aren't tagged as experimental. The
options I enabled were TAGGED QUEUING, SCB_PAGING, and I also raised
CMNDS_PER_LUN from 4 (I think) to 8.

These changes always worked perfectly with just one SCSI drive installed.
However, the moment I touched the 2nd drive with badblocks -w /dev/sdb,
the machine was unimpressed, and gave me a kernel panic (skb_free
maybe...? not logged to disk of course). I then got the usual glut of
reset attempts, flushing/discarding of SCB's ad nauseum. This was 2.0.26.
I rebooted onto 2.0.27 with the aic7xxx.c #define changes reversed, and
so far a badblocks -w /dev/sdb1 is going happily.

Just thought someone should know.


PS Re: This 3c59x stuff, I'm now running 0.29-all and will report on its
stability. Looking at the one-liner typo bugfix over 0.28c-all, I would
not be surprised if problems disappear.

>> PGP key available on request <<
>>Presence in pub available on request<<