Re: Everyone's SRM woes

["Christopher C. Chimelis" <chris debian org>]

On Sun, 11 Mar 2001 ppetrakis@alphalinux.org wrote:

> Maybe instead of a small list of adapter cards that alpha/linux uses
> want/need to use.
> How about relaxing the the scope of which SRM decides to support a given
> type of card,
> for example symbios based cards. I never understood why SRM needed to
> KNOW the PCI vendor
> ID of evert card it wanted to support. Like symbios support, why doesnt
> a fireport 40
> show up? It's symbios based, works fine in linux. Same chipset as some
> of the older
> "supported" scsi cards though it doesnt show up. Now users who want to
> get the "SRM supported"
> Ultra whatever symbios xxxxx SCSI card have no choice but to go to
> intraserver and pay their
> prices. 

Easier said than done.  The drivers in SRM do need to recognise the PCI
vendor IDs of SCSI boards in order to support them, but that's not unique
to SRM.  Even Linux drivers do this.  Granted, it should be very easy to
add vendor IDs for cards that use the same chipsets as the other supported
adapters, but it would still need to be done.

> SRM could supply a generic pk driver , gpk if you will. That would try
> and start any host
> adapter of chipset class X. You get the idea. Adaptec support shouldnt
> be a problem at all
> to implement. I know Compaq SRM supports the 2940 in DP264 and API's SRM
> can support that, the
> 2940U2 ,and the 160. So it's not like the driver support has been
> written it just hasnt been
> implemented across ALL Alpha platforms. 

Are you referring to trying to starting everything of a certain PCI device
class or a certain chipset class?  If it's the latter, there's often no
way to know what chipset is on a board via PCI probes.  If the former,
this definitely wouldn't work in most cases since all SCSI controllers
share the same PCI class type (0100).

I think the hardest part about doing anything with SCSI BIOS is that, from
what I've seen, there's no common way to initialise those cards like there
is for vid cards (int10).

C