RE: AIC79xx driver question

RE: AIC79xx driver question

[Cress, Andrew R]

Justin,

I was looking at the code for aic79xx, and it appears that the channel is
hard-coded to 'A' in a number of places, rather than using SCB_GET_CHANNEL()
(e.g.: calls to ahd_reset_channel in aic79xx_core.c).  Was this
intentional?  Is there only one channel per aic79xx host? 

Andy 

-----Original Message-----
[...]

Updated aic7xxx and aic79xx drivers for 2.4 and 2.5 can be found here:

http://people.FreeBSD.org/~gibbs/linux/tarballs

Notable changes:

o Both drivers support Domain Validation
o Memory mapped I/O is now a config option
o New memory mapped I/O register test that will hopefully
  weed out broken VIA chipsets
o The reboot notifier hook has been disabled.  The driver wants
  to shut itself down prior to reboot, but the reboot notifier is
  now called too early in 2.5.X for this to be effective.  It looks
  like it might be possible for the driver to grow a device shutdown
  routine, but it was not obvious with a quick look at this new feature
  how to ensure that the [sd, sr, etc.] driver's shutdown routines
  would be called prior to the aic7xxx driver's routine.  Some guidance
  in this area would be appreciated.
o Tag depth changes are now communicated to the midlayer.
 
--
Justin

RE: AIC79xx driver question

[Justin T. Gibbs]

> Justin,
> 
> I was looking at the code for aic79xx, and it appears that the channel is
> hard-coded to 'A' in a number of places, rather than using
> SCB_GET_CHANNEL() (e.g.: calls to ahd_reset_channel in aic79xx_core.c).
> Was this intentional?  Is there only one channel per aic79xx host? 

Yes.  There is only one channel.  Much of the code was ported from
the aic7xxx driver which, due to the aic7770's twin/narrow capability,
must have a concept of "channel".  At some point, the channel references
may be completely removed from the code, but they are harmless.

--
Justin

Re: AIC79xx driver question

[Jeff Nguyen]

Hello Justin,

The latest aic79xx source code (v1.1.0) found on Adaptec Web site
does not compile under 2.4.20 kernel. The makefile references to
an object file, aic7xxx_reg_print.o, which does not exist at all. 

Is there a patch that I can run to fix this error?

Jeff