Onboard SCSI controller support on Ultra 2

[Pyun YongHyeon]

On Thu, Feb 17, 2005 at 11:06:03PM +0100, Marius Strobl wrote:
 > On Thu, Feb 17, 2005 at 11:48:34AM +0900, Pyun YongHyeon wrote:
 > > On Wed, Feb 16, 2005 at 02:02:08PM +0100, Marius Strobl wrote:
 > > [...]
 > >  > 
 > >  > > 
 > >  > > > > The prob i did find was with QFE card support; they all appear as
 > >  > > > > hme, have the correct individual MAC addresses--- but little else,
 > >  > > > > and errors galore.
 > >  > > 
 > >  > > The QFE actually *is* a quad HME card, that's why FreeBSD uses the
 > >  > > same driver for it.  What kind of problems do you have with it?
 > >  > > 
 > >  > > I've only used PCI-based QFEs so far, but never encountered any
 > >  > > problems with them -- they even make a fine Quad card in IA32 machines
 > >  > > now that FreeBSD supports them outside of OpenFirmware systems.
 > >  > > 
 > >  > 
 > >  > AFAIK the SBus QFE cards only have a single interrupt for all four
 > >  > HMEs and the FreeBSD hme(4) doesn't take this into account.
 > >  > 
 > > AFAIK HME interrupt handler is called with softc argument. So I guess
 > > there is no problem since it can correctly find an instance of softc
 > > and related information.
 > 
 > Depends on how it actually works. I can't remember were I read about it
 > but my impression was that it doesn't work like a shared IRQ in the
 > usual sense.
 > Anyway, there were reports in the past that SBus QFE don't work properly
 > and act like not getting interrupts.
 > 

Then it seems the SBus QFE has single interrupt source. If this is
the case hme(4) should be taught to look into all instances of QFE
softc. Adding a code to inspect all softc(maybe be seperate
interrupt handler wrapper that calls real interrupt handler for each
port) would be simple but I guess it also needs special attach/detach
routine for QFE and hme(4) should know a global softc where it starts
the traverse.

-- 
Regards,
Pyun YongHyeon
http://www.kr.freebsd.org/~yongari	|	yongari at freebsd.org