Re: [PATCH] 2.6.xx: sata_mv: another critical fix

[Jeff Garzik <jeff@garzik.org>]

Linus Torvalds wrote:
> Maybe back-porting any critical sata_mv fixes to 2.6.16.x is appropriate, 
> considering that I don't think RH or SuSE will necessarily want to pull 
> the whole thing.

Agreed -- in theory -- but I'll let Mark Lord or somebody else do that, 
if they are motivated.  sata_mv is labelled "HIGHLY EXPERIMENTAL" for a 
reason :)  Its a known bug report generator, with bugs far outpacing 
fixes at present.

Key patches to backport or test, in priority order, would be:

[libata sata_mv] do not enable PCI MSI by default
[libata sata_mv] add 6042 support, fix 60xx/50xx EDMA configuration
[PATCH] Add missing FUA write to sata_mv dma command list

After that, you're at the mercy of not-yet-worked-around hardware bugs, 
and (as Mark Lord appears to be finding) some driver bugs as well.

sata_mv development slowed to a trickle for a long time, after the 
original developer disappeared, and I didn't have time to dig deep into 
the hardware details.  Mark Lord recently started picking up the pieces, 
so things are looking much better already.  I have another patch from 
Mark to forward to you and 2.6.16.x (stable@), coming to you today.

	Jeff