Eines schoenen Tages schrieb Izumi Tsutsui:
> kiyohara@kk.iij4u.or.jp wrote:
> 
> > I guess that the version of GT is old perhaps.  ;-<
> 
> According to Linux-MIPS source:
> http://www.linux-mips.org/cvsweb/linux/arch/mips/pci/fixup-cobalt.c?rev=1.11
> it always set timeout and retry values to maximum.
> How about the attached patch? I've also put compiled -current kernel:
> http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-cobalt-GENERIC-gtfix-20050614.gz

Unfortunatly it also hangs hard in the same place :( 


> But does the problem really depend on GT version?
> I doubt that "the STOP line disconnected from Galileo to VIA on PCI"
> is GT revision specific, but I guess it Cobalt model specific.

Yes, probably. Unfortunalty I do not own a Raq/Cube 2, so i cannot test the
kernel there.


> I've been thinking that we should have some variable which
> indicates cobalt models (like "cobalt_board_id" on Linux)
> and it could be fetched from VT82C586 register space,
> but I have not tried to implement it yet.

Hm. If there would be something to indicate wich cobalt model the machine
is, one could also disable the DMA modes for the IDE deviced. Raq1 will also
crash, if one tries to use these modes. The usual restore cd kernel for
version 2 will not boot because of that issue.

Best Regards,
Oliver
-- 
--------------------------------------------------------
And remember: "To Infinity And Far Beyond ... Somehow?!"

Hi! I'm a .signature virus! Copy me in your ~/.signature
to help me spread!             <- Save this lifeform ;-)