[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Fwd: linux-image-2.6-17-mac tries

> > I have no idea what kind of machine this happened on. If it's not Sonic or
> > IIfx, locking or interrupt reentry seems most likely. The 5380 driver of
> > old mostly relies on interrupts being masked on entry, and the big kernel
> > lock held.
> I hadn't looked at it that closely. However, that sounds like it's asking
> for trouble. The BKL has been pretty heavily deprecated in 2.6. There
> have been a ton of changes to irq handling as well.

I am aware of that.

> > IIRC the SCSI driver for Atari still worked in 2.6.13, with the limited
> > testing I gave it. I've patched 2.6.17 to compile but didn't have a chance
> > to test it yet.
> I'm not sure if it is still the case, but in the copy of the tree I
> have handy, atari_scsi uses a local copy of NCR5380.c rather than

Sure, but once upon a time, that local copy was derived from NCR5380.c and
the business end of the logic didn't change.

> the shared copy being used by mac_scsi. The sun3_scsi driver appears
> to have a local copy as well. I wouldn't trust the atari driver working
> as being much of an indication because of this.

I'm not even sure the Atari version still works in 2.6.17.

> > SCSI-1 disks have been reported to fail before... Either way, I'll test
> > the related Atari driver as soon as possible.
> I have to admit to lacking information with some of the SCSI specific
> issues. In any case, it would certainly be good to test the Atari code.

I have no idea what in particular the SCSI-1 disks do that the driver
doesn't like, except for not handling disconnects properly. It might even
be about select with or without attention signaling that goes wrong. IIRC
mixing SCSI-1 and SCSI-2 with the 5380 is a no-no. The details are
extremely fuzzy in my recollection, though.


Reply to: