Re: UP2000, setup for SRMconsole, and the sym53c8xx driver

[G. Hugh Song]

Thank you for the reply, Gerard.

On Tue, 27 Jun 2000, [ISO-8859-1] Gérard Roudier wrote:

> On Tue, 27 Jun 2000, G. Hugh Song wrote:
> 
> > If you ever compiled a Linux kernel on UP2000 for this type of
> > SCSI controller, would you please send me .config file under
> > /usr/src/linux?
> 
> I donnot have in mind the name of the core logic used by the UP2000 but it
> should be a tsunami (for 21264 processor).

That is "dp264".
Mr. Alan Cox.  Please add "UP2000" and "UP1000" in the list for dp264.
I myself was quite puzzled because of this.

> For this one, I would recommend you to put the board on BUS #0 (HOSE #0)
> if not yet. Linux/Alpha tampers physical addresses in order not to lose
> the HOSE number and tampering seems to differ between linux-2.2.X and
> linux-2.3/4.X.

Unfortunately, I am not that knowledgeable enough to understand BUS #0.
So, I am gonna try further without tampering anything.  As I report below.
I succeeded obtaining a working kernel out of Kernel 2.2.15.  However,
Kernel-2.2.16 (and after) with the same .config as 2.2.15/.config is not
detected by the SRM console with the following SRM-console error message:

-------------------------------------
failed to send Read to dka0.0.0.8.1
failed to send Read to dka0.0.0.8.1
failed to send Read to dka0.0.0.8.1
failed to send Read to dka0.0.0.8.1
ext2_blkno: error on iblk read
failed to send Read to dka0.0.0.8.1
failed to send Read to dka0.0.0.8.1
failed to send Read to dka0.0.0.8.1

[ and it goes on forever ]
----------------------------------------

I'd like to think that this behavior could be related to the BUS number
problem.  So, I will try the new driver which you suggested below.

> Btw, I actually donnot want to know how these address tamperings work (or
> fail), but when HOSE number is zero, physical/bus addresses passed to the
> driver should match the reality.
> 
> If (when?) the board will be on hose #0 and if failure is still there, you
> may want to try latest driver versions (if not yet :) ):
> 
> ftp://ftp.tux.org/pub/roudier/drivers/linux/experimental/sym53c8xx-1.7-pre2.tar.gz
> 
> It contains full driver files you want to copy to
>     /usr/linux/drivers/scsi/.
> 
> Regards,
>   Gérard.
> 
> 
> > I am having a trouble with both ncr53c8xx and sym53c8xx
> > for ITI-6101U2W using 53c895.
> > The error message I am getting is as follows:
> > 
> > ========================================================
> > sym53c895-0 on-chip RAM at 0x109002000
> > CACHE TEST FAILED: script execution failed
> > start=fffced08, pc=fffced08, end=fffced28
> > CACHE INCORRECTLY CONFIGURED
> > sym53c895-0: giving up
> > 
> > =======================================================
> > I was able to locate the place in scsi/sym53c8xx.c.
> > But I don't have any further idea what to do.

The above error appears with the following choise:
CONFIG_ALPHA_DP264=y
CONFIG_PCI=y
CONFIG_ALPHA_EV6=y
CONFIG_ALPHA_TSUNAMI=y
CONFIG_ALPHA_SRM=y
CONFIG_ALPHA_SRM_SETUP=y
CONFIG_SMP=y
CONFIG_PCI_QUIRKS=y
CONFIG_PCI_OPTIMIZE=y

I solved the above problem.  The trouble comes from the PCI setup for the
kernel for dp264.  One is supposed to select "dp264" and to choose "Yes"
to "CONFIG_PCI" and "CONFIG_ALPHA_SRM".  However, if I choose "Yes" to
"CONFIG_ALPHA_SRM_SETUP", then the above error appears and fails.  
There is not much relevant information about this trouble in the Help 
dialog for "CONFIG_ALPHA_SRM_SETUP".
This is so subtle that I had real hard time to figure out.  
So the working setup is as follows:

CONFIG_ALPHA_DP264=y
CONFIG_PCI=y
CONFIG_ALPHA_EV6=y
CONFIG_ALPHA_TSUNAMI=y
CONFIG_ALPHA_SRM=y
# CONFIG_ALPHA_SRM_SETUP is not set
CONFIG_SMP=y
CONFIG_PCI_QUIRKS=y
# CONFIG_PCI_OPTIMIZE is not set

The above trouble is also related to the  error 
"dp264 device interrup NOT YET IMPLEMENTED" in
arch/alpha/kernel/sys_dp264.c.
This happens with the following set of choices:
CONFIG_ALPHA_DP264=y
CONFIG_PCI=y
CONFIG_ALPHA_EV6=y
CONFIG_ALPHA_TSUNAMI=y
# CONFIG_ALPHA_SRM is not set
CONFIG_SMP=y
CONFIG_PCI_QUIRKS=y
# CONFIG_PCI_OPTIMIZE is not set

Please, Mr. Alan Cox, would you please revise the Help dialog for the 
above item?  Just a one-line information about this would help a lot 
of people using UP1000 or UP2000 with the SRM console.

> > 
> > Regards,
> > 
> > G. Hugh Song
> > 

Thank you all,

G. Hugh Song