Re: Radeondrm w/RV370 / X300 SE on 7.0_BETA -- working?

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Tue, Oct 07, 2014 at 03:45:12PM +0200, Havard Eidnes wrote:
> >>Though it would be nice if the kernel didn't simply reset the
> >>machine when the twa driver is used :)
> >>
> >>Hmm, I have another machine which runs netbsd-5 which appears to
> >>have the same controller type and which works fine:
> >>
> >>mt# pcictl /dev/pci2 list
> >>002:00:0: 3ware 9650-series RAID (RAID mass storage, revision 0x01)
> >>mt# pcictl /dev/pci2 list -n
> >>002:00:0: 0x100413c1 (0x01040001)
> >>mt# 
> >
> > OK, there look to have been some twa(4) driver changes since netbsd-5,
> > has the driver probed correctly on the working machine?
> 
> On the machine I'm currently testing with: no, not that I can
> think of.  I'm a little confused why ahcisata_pci.c's match()
> routing returns 3 while twa.c's match() routine returns 1 --
> isn't this supposed to be a "priority value" where the highest
> one wins?

yes

> 
> How does other folks' 3ware controllers probe with twa, then?
> Only because ahcisata doesn't match their controllers?

I guess so (that they don't identify as PCI_SUBCLASS_MASS_STORAGE_RAID);
I looked at the ahci match function and it's this way for several years.
Anyway I don't think it's correct for ahci_pci_match() to read
AHCI_GHC and use the result before we're sure it's an AHCI controller.

But for the time being, changing twa_match to return 4 is the right way to do.
Other RAID controller drivers should probably return 4 or higher too.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--