[parisc-linux] Re: Problem installing .9 iso image on C240

[parisc-linux] Re: Problem installing .9 iso image on C240

[Richard Hirst]

On Mon, Jul 09, 2001 at 09:05:36PM -0700, Charles Boone wrote:
> I got the .9.1 release, it only wants to boot the
> 32bit version.  so I  tried running the install both
> with compatibility and without.
....
....
>      YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI                 
>                          
> PSW: 00000000000011001111111100001110                 
>                          
> r0-3     00000000 10335810 101fc9ec ffffffff          
> r4-7     1fe02060 0000000e 00000000 10336010          
> r8-11    0000004e 0000004d 0004000e 102c0f14
> r12-15   00000000 ffffffff 000000fd f0100000     
> r16-19   10300580 f000020c f0000204 0000ff40     
> r20-23   1fe02060 0000ff40 101fc9b0 0000000e     
> r24-27   0000ff40 f2000064 c1b21ec0 1028a010     
> r28-31   1fe03aa0 000001c0 10300a00 1010d2c8
> sr0-3    00000000 000006e2 00000000 00000000
> sr4-7    00000000 00000000 00000000 00000000
>  					     
> IASQ: 00000000 00000000 IAOQ: 101fc2e8 101fc9ec
>  IIR: 081a025c    ISR: 9227ffc8  IOR: 00000064
>  ORIG_R28: 00000000

IAOQ = _gsc_readl + 8
called from dino_in32()
I guess it was dino_in32(0xc1b21ec0), which looks odd.

Don't know why it should fail like that in the middle of the
install.  Is it repeatable?


Richard

[parisc-linux] Re: Problem installing .9 iso image on C240

[Charles Boone]

Richard,

Yes, I can repeat it everytime I try to install, same
as with the .9 version.  I do have one question, why
would it not try to install the 64bit version? I was
pretty sure that a C240 is a 64bit system.

I am assuming that this is not a problem reading the
cdrom, I think that I can try another cdrom.

Thanks,
Charles
--- Richard Hirst <rhirst@linuxcare.com> wrote:
> On Mon, Jul 09, 2001 at 09:05:36PM -0700, Charles
> Boone wrote:
> > I got the .9.1 release, it only wants to boot the
> > 32bit version.  so I  tried running the install
> both
> > with compatibility and without.
> ....
> ....
> >      YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI             
>    
> >                          
> > PSW: 00000000000011001111111100001110             
>    
> >                          
> > r0-3     00000000 10335810 101fc9ec ffffffff      
>    
> > r4-7     1fe02060 0000000e 00000000 10336010      
>    
> > r8-11    0000004e 0000004d 0004000e 102c0f14
> > r12-15   00000000 ffffffff 000000fd f0100000     
> > r16-19   10300580 f000020c f0000204 0000ff40     
> > r20-23   1fe02060 0000ff40 101fc9b0 0000000e     
> > r24-27   0000ff40 f2000064 c1b21ec0 1028a010     
> > r28-31   1fe03aa0 000001c0 10300a00 1010d2c8
> > sr0-3    00000000 000006e2 00000000 00000000
> > sr4-7    00000000 00000000 00000000 00000000
> >  					     
> > IASQ: 00000000 00000000 IAOQ: 101fc2e8 101fc9ec
> >  IIR: 081a025c    ISR: 9227ffc8  IOR: 00000064
> >  ORIG_R28: 00000000
> 
> IAOQ = _gsc_readl + 8
> called from dino_in32()
> I guess it was dino_in32(0xc1b21ec0), which looks
> odd.
> 
> Don't know why it should fail like that in the
> middle of the
> install.  Is it repeatable?
> 
> 
> Richard
> 
> 


__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
http://personal.mail.yahoo.com/

[parisc-linux] Re: Problem installing .9 iso image on C240

[Richard Hirst]

On Tue, Jul 10, 2001 at 07:23:53AM -0700, Charles Boone wrote:
> Richard,
> 
> Yes, I can repeat it everytime I try to install, same
> as with the .9 version.  I do have one question, why
> would it not try to install the 64bit version? I was
> pretty sure that a C240 is a 64bit system.

I think C240 will run 32 or 64 bit kernels, but 64 bit kernel isn't
working on it yet.

> I am assuming that this is not a problem reading the
> cdrom, I think that I can try another cdrom.

I agree, it doesn't sound like a corrupt CD to me.  More likely
some kernel problem.  I'll try another install on my c360, but it
worked last time I tried.

Richard

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Grant Grundler]

Charles Boone wrote:
...
> I do have one question, why
> would it not try to install the 64bit version? I was
> pretty sure that a C240 is a 64bit system.

PDC on the C240 (and similar C-models of that generation) only
support PDC calls in "narrow" (ie 32-bit) mode. Sounds like the decision
to install 32 vs 64-bit depends on the PDC capabilities bit.
Can anyone offhand confirm/deny?

BTW, though the C240 has a PA2.0 CPU (ergo 64-bit capable), I'm not sure
rbrad has been able to resolve all the issues when trying to run 64-bit
kernels on his C200. He did resolve many of them - including support
for 32-bit PDC calls from 64-bit kernels.

grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Richard Hirst]

On Tue, Jul 10, 2001 at 11:45:15PM -0600, Grant Grundler wrote:
> Charles Boone wrote:
> ...
> > I do have one question, why
> > would it not try to install the 64bit version? I was
> > pretty sure that a C240 is a 64bit system.
> 
> PDC on the C240 (and similar C-models of that generation) only
> support PDC calls in "narrow" (ie 32-bit) mode. Sounds like the decision
> to install 32 vs 64-bit depends on the PDC capabilities bit.
> Can anyone offhand confirm/deny?

Hmm, our installer installs either a 32 bit kernel or a 64 bit one
which does wide PDC calls.  So, I guess we will always install a
32 bit kernel on Cxx.  Installing a 64 bit kernel with narrow PDC
calls later would be an option (once it works).  Do we leave building
that kernel as an exercise for the user, or or do we have one packaged
in the archive?

Richard

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Paul Bame]

= > > I do have one question, why
= > > would it not try to install the 64bit version? I was
= > > pretty sure that a C240 is a 64bit system.

As others said, it's a 64-bit-capable CPU with 32-bit-only firmware and
only 32-bit kernels work there right now.  I don't know that a 64-bit
kernel buys you much on a C240 -- am I mistaken?

= > PDC on the C240 (and similar C-models of that generation) only
= > support PDC calls in "narrow" (ie 32-bit) mode. Sounds like the decision
= > to install 32 vs 64-bit depends on the PDC capabilities bit.
= > Can anyone offhand confirm/deny?

Yes, palo asks the firmware whether one or both of 32&64-bit OSs will
run on the box.  Cxxx says 32 bit only.  If both are possible, palo
chooses 32-bit for now.

= a 64 bit kernel with narrow PDC
= calls later would be an option (once it works).  Do we leave building
= that kernel as an exercise for the user, or or do we have one packaged
= in the archive?

When it works we can add a config file for building a suitably flavored
debian kernel-image package.

	-P

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Grant Grundler]

Paul Bame wrote:
> I don't know that a 64-bit kernel buys you much on a C240 -- am I mistaken?

I think not.

I thought about making the same comment since in general wide-mode
just runs 2-5% slower. But some apps could take advantage of the wider
words for substantial perf gains...but since we don't support 64-bit apps
(yet) and only a few isolated places in the kernel would benefit (eg
cksum'ing code), I don't see 64-bit kernel a "must have" for C240.

> = > PDC on the C240 (and similar C-models of that generation) only
> = > support PDC calls in "narrow" (ie 32-bit) mode. Sounds like the decision
> = > to install 32 vs 64-bit depends on the PDC capabilities bit.
> = > Can anyone offhand confirm/deny?
> 
> Yes, palo asks the firmware whether one or both of 32&64-bit OSs will
> run on the box.  Cxxx says 32 bit only.  If both are possible, palo
> chooses 32-bit for now.

Ok - thanks!
And I think that's fine in general. Someone with a specific
need for 64-bit ops can build/install a 64-bit kernel after the fact.
Someday we might have a real 64-bit userspace and it might be interesting
to get pickier about install support for PA2.0 CPU's.

grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Grant Grundler]

Richard Hirst wrote:
> Installing a 64 bit kernel with narrow PDC
> calls later would be an option (once it works).  Do we leave building
> that kernel as an exercise for the user, or or do we have one packaged
> in the archive?

I was under the impression the "CONFIG_PDC_NARROW" option would only
"activate" itself in one boxes that needed it. Ie it does grow the kernel
and AFAIK is really only useful for Cxxx models.
Ryan, do I have it right?

I am pretty sure K/D/R-class boxes have a different source base for PDC
which "properly" support wide-mode PDC calls.

thanks,
grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Ryan Bradetich]

On Wed, Jul 11, 2001 at 11:23:34PM -0600, Grant Grundler wrote:
> Richard Hirst wrote:
> > Installing a 64 bit kernel with narrow PDC
> > calls later would be an option (once it works).  Do we leave building
> > that kernel as an exercise for the user, or or do we have one packaged
> > in the archive?
> 
> I was under the impression the "CONFIG_PDC_NARROW" option would only
> "activate" itself in one boxes that needed it. Ie it does grow the kernel
> and AFAIK is really only useful for Cxxx models.
> Ryan, do I have it right?

Not quite.  Let me try to clarify (This is the help text for the 
CONFIG_PDC_NARROW option)....

CONFIG_PDC_NARROW
  Saying Y here will allow developers with a C180, C200, C240, C360,
  J200, J210, and/or a J2240 to test 64-bit kernels by providing 
  a wrapper for the 32-bit PDC calls.  Since the machines which 
  which require this option do not support over 4G of RAM, this
  option is targeted for developers of these machines wishing to
  test changes on both 32-bit and 64-bit configurations.

  If unsure, say N.

At one time I had run-time detection to detect if the machine was 
booted into wide mode it could detect if the pdc wrappers were needed.
After some discussion, we decided to make it a compiler option since
it added overhead to every pdc call and would rarely be used.

If the help text for this option needs to be clarified, please send
me your comments and I'll update the help text. :)

Thanks,

- Ryan

> I am pretty sure K/D/R-class boxes have a different source base for PDC
> which "properly" support wide-mode PDC calls.
> 
> thanks,
> grant
> 
> Grant Grundler
> parisc-linux {PCI|IOMMU|SMP} hacker
> +1.408.447.7253
> 
> 
> _______________________________________________
> parisc-linux mailing list
> parisc-linux@lists.parisc-linux.org
> http://lists.parisc-linux.org/cgi-bin/mailman/listinfo/parisc-linux
> 

-- 

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Grant Grundler]

Richard Hirst wrote:
> IAOQ = _gsc_readl + 8
> called from dino_in32()
> I guess it was dino_in32(0xc1b21ec0), which looks odd.

It's definitely wrong. dino_in32() pokes dino to generate IO Port
transactions and Dino lives in 0xF... address range. This suggests the
pci_dev->sys_data is garbage (perhaps not initialized?) or points to garbage.
Perhaps for Dino's built-in PS/2 or RS232 port?

Sorry, I didn't follow this thread earlier. Does console output hint
which PCI device was involved (if any)?

> Don't know why it should fail like that in the middle of the
> install.  Is it repeatable?

If it's a problem with devices on-board Dino, I'd expect the problem to
be reproducible but with different symptoms depending on what's in pci_dev.

grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Richard Hirst]

On Tue, Jul 10, 2001 at 11:36:25PM -0600, Grant Grundler wrote:
> Richard Hirst wrote:
> > IAOQ = _gsc_readl + 8
> > called from dino_in32()
> > I guess it was dino_in32(0xc1b21ec0), which looks odd.
> 
> It's definitely wrong. dino_in32() pokes dino to generate IO Port
> transactions and Dino lives in 0xF... address range. This suggests the
> pci_dev->sys_data is garbage (perhaps not initialized?) or points to garbage.
> Perhaps for Dino's built-in PS/2 or RS232 port?
> 
> Sorry, I didn't follow this thread earlier. Does console output hint
> which PCI device was involved (if any)?

Not obviously, and it failed in the middle of the install, so presumably
whatever device was involved had been accessed successfully earlier.

Richard

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Charles Boone]

One piece of info that I did not include,  which I
thought might be relative to the ps/2 issue.  I am not
sure it is, but I do not have a keyboard or mouse
connected on the c240.  I was pretty sure that it
would not be a problem, since it would be just like
having a dumb terminal on it.


--- Richard Hirst <rhirst@linuxcare.com> wrote:
> On Tue, Jul 10, 2001 at 11:36:25PM -0600, Grant
> Grundler wrote:
> > Richard Hirst wrote:
> > > IAOQ = _gsc_readl + 8
> > > called from dino_in32()
> > > I guess it was dino_in32(0xc1b21ec0), which
> looks odd.
> > 
> > It's definitely wrong. dino_in32() pokes dino to
> generate IO Port
> > transactions and Dino lives in 0xF... address
> range. This suggests the
> > pci_dev->sys_data is garbage (perhaps not
> initialized?) or points to garbage.
> > Perhaps for Dino's built-in PS/2 or RS232 port?
> > 
> > Sorry, I didn't follow this thread earlier. Does
> console output hint
> > which PCI device was involved (if any)?
> 
> Not obviously, and it failed in the middle of the
> install, so presumably
> whatever device was involved had been accessed
> successfully earlier.
> 
> Richard
> 


__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
http://personal.mail.yahoo.com/

Re: [parisc-linux] Re: Problem installing .9 iso image on C240

[Richard Hirst]

On Thu, Jul 12, 2001 at 07:42:25AM -0700, Charles Boone wrote:
> One piece of info that I did not include,  which I
> thought might be relative to the ps/2 issue.  I am not
> sure it is, but I do not have a keyboard or mouse
> connected on the c240.  I was pretty sure that it
> would not be a problem, since it would be just like
> having a dumb terminal on it.

I agree, that is unlikely to be a problem.

Richard