Re: kbuild 2.5.26 - arch/alpha

[Jay Estabrook <Jay.Estabrook@compaq.com>]

On Tue, Jul 23, 2002 at 09:08:26PM +0200, Sven Koch wrote:
> 
> I am using Stock 2.4.19-rc2 with the following simple patch on an xl-300
> with milo:
> (without it, it breaks while initalizing the ncr-scsi-controller)
> 
> --- linux/arch/alpha/kernel/core_cia.c.orig	Sun Oct 21 19:30:58 2001
> +++ linux/arch/alpha/kernel/core_cia.c	Fri Jul 19 16:11:46 2002
> @@ -382,10 +382,10 @@
>  	for (i = 0; i < CIA_BROKEN_TBIA_SIZE / sizeof(unsigned long); ++i)
>  		ppte[i] = pte;
> 
> -	*(vip)CIA_IOC_PCI_W1_BASE = CIA_BROKEN_TBIA_BASE | 3;
> -	*(vip)CIA_IOC_PCI_W1_MASK = (CIA_BROKEN_TBIA_SIZE*1024 - 1)
> +	*(vip)CIA_IOC_PCI_W3_BASE = CIA_BROKEN_TBIA_BASE | 3;
> +	*(vip)CIA_IOC_PCI_W3_MASK = (CIA_BROKEN_TBIA_SIZE*1024 - 1)
>  				    & 0xfff00000;
> -	*(vip)CIA_IOC_PCI_T1_BASE = virt_to_phys(ppte) >> 2;
> +	*(vip)CIA_IOC_PCI_T3_BASE = virt_to_phys(ppte) >> 2;
>  }
> 
>  static void __init

Yes, this will help on XLT-300.

What this patch does is revert the code to an older version which uses
PCI DMA window #3 for the scatter/gather operations, thus avoiding the
use of window #1 for that operation.

On older machines like the XLT-300 and EB164, their core logic, CIA
Rev 1, appears to have a bug in that PCI DMA windows #1 and #2 cannot
be used for scatter/gather. Boxes based on CIA rev 2, and PYXIS, do
not have the same problem.

The patches I attached to an earlier posting essentially do this for
the rev 1 CIA machines, but leave active the code for dual-address
cycle (DAC) support for the CIA rev 2  and PYXIS based machines.

 --Jay++

-----------------------------------------------------------------------------
Jay A Estabrook                            HPTC - LINUX support
Hewlett-Packard Company - MRO1-2/K15       (508) 467-2080
200 Forest Street, Marlboro MA 01752       Jay.Estabrook@hp.com
-----------------------------------------------------------------------------