old ISA DMA bug in 2.6.12?

old ISA DMA bug in 2.6.12?

[Bob Tracy]

I was enjoying yet another session of beating my head against the wall
trying to do useful things with old hardware :-), and managed to cause a
kernel panic by simply trying to mount a cdrom in the context of a DSL-N
installation.

The SCSI host adapter is an Adaptec AHA-1542B, and when I try to mount a
cdrom, I manage to run afoul of the BAD_DMA() check in aha1542.c: the
buffer returned is not in the lower 16 MB of memory.

The same 2.6.12 kernel + hardware combination works fine as long as I
confine my I/O to the hard disk that's also attached to the AHA-1542B.

-- 
-----------------------------------------------------------------------
Bob Tracy                   WTO + WIPO = DMCA? http://www.anti-dmca.org
rct@frus.com
-----------------------------------------------------------------------

Re: old ISA DMA bug in 2.6.12?

[Robert Hancock]

Bob Tracy wrote:
> I was enjoying yet another session of beating my head against the wall
> trying to do useful things with old hardware :-), and managed to cause a
> kernel panic by simply trying to mount a cdrom in the context of a DSL-N
> installation.
> 
> The SCSI host adapter is an Adaptec AHA-1542B, and when I try to mount a
> cdrom, I manage to run afoul of the BAD_DMA() check in aha1542.c: the
> buffer returned is not in the lower 16 MB of memory.
> 
> The same 2.6.12 kernel + hardware combination works fine as long as I
> confine my I/O to the hard disk that's also attached to the AHA-1542B.

Looks like the aha1542 driver doesn't set the DMA mask, so the kernel 
will default to thinking it can do 32-bit DMA when it should be 24-bit.

-- 
Robert Hancock      Saskatoon, SK, Canada
To email, remove "nospam" from hancockr@nospamshaw.ca
Home Page: http://www.roberthancock.com/