CP0_STATUS interrupt mask patch

CP0_STATUS interrupt mask patch

[Harald Koerfgen]

Hi Gang,

two weeks ago I asked for comments on a patch which is essential for Linux
on DECstations but may trigger undiscovered bugs on other machines.

Well, I haven't got any feedback :-(

On the other hand, nobody said "don't do this!" :-)

If nobody stops me I am going to commit the patch I posted two weeks ago
and see what happens B)
---
Regards,
Harald

P.S.: Yes, I *do* have my asbestos suit within reach.

Re: CP0_STATUS interrupt mask patch

[Ralf Baechle]

On Fri, Apr 30, 1999 at 10:55:22AM +0200, Harald Koerfgen wrote:

> two weeks ago I asked for comments on a patch which is essential for Linux
> on DECstations but may trigger undiscovered bugs on other machines.
> 
> Well, I haven't got any feedback :-(
> 
> On the other hand, nobody said "don't do this!" :-)
> 
> If nobody stops me I am going to commit the patch I posted two weeks ago
> and see what happens B)
>
> P.S.: Yes, I *do* have my asbestos suit within reach.

Not required, Real Penguins Stay Cool (TM).

  Ralf

Re: CP0_STATUS interrupt mask patch

[Charles Lepple]

On Fri, 30 Apr 1999, Harald Koerfgen wrote:

> on DECstations but may trigger undiscovered bugs on other machines.

I just pulled down the CVS kernel with the patch, and it seems that it
causes a 'keyboard timeout[2]' to be printed on the console after the SCSI
bus is probed. I undid the patch manually (ie, the rest of the kernel is
still from the cvs update) and it boots fine. Anyone else seen this?

--Charles Lepple
System Administrator, Virginia Tech EE Workstation Labs
clepple@foo.tho.org || http://foo.tho.org/charles/

Re: CP0_STATUS interrupt mask patch

[Charles Lepple]

Charles Lepple wrote:
> I just pulled down the CVS kernel with the patch, and it seems that it
> causes a 'keyboard timeout[2]' to be printed on the console after the SCSI

Omitted a critical detail -- it hangs hard-core after this message.
Evidently the power button was initialized by this point, because it
doesn't turn off immediately, and it doesn't start the LED blinking
indicating an impending power-off.

System summary: 
Indy R5000
gcc version egcs-2.90.27 980315 (egcs-1.0.2 release)
binutils-2.8.1

-- 
Charles Lepple
System Administrator, Virginia Tech EE Workstation Labs
clepple@ee.vt.edu || http://www.foo.tho.org/charles/

Re: CP0_STATUS interrupt mask patch

[Ralf Baechle]

On Wed, May 05, 1999 at 01:44:23AM +0000, Charles Lepple wrote:

> Charles Lepple wrote:
> > I just pulled down the CVS kernel with the patch, and it seems that it
> > causes a 'keyboard timeout[2]' to be printed on the console after the SCSI
> 
> Omitted a critical detail -- it hangs hard-core after this message.
> Evidently the power button was initialized by this point, because it
> doesn't turn off immediately, and it doesn't start the LED blinking
> indicating an impending power-off.

The power button is handled during interrupts so when the interrupt
handling is fsck'ed the button won't work, too.

I'll look at it.

  Ralf

Re: CP0_STATUS interrupt mask patch

[Ulf Carlsson]

On Wed, May 05, 1999 at 05:40:22PM +0200, Ralf Baechle wrote:
> On Wed, May 05, 1999 at 01:44:23AM +0000, Charles Lepple wrote:
> 
> > Charles Lepple wrote:
> > > I just pulled down the CVS kernel with the patch, and it seems that it
> > > causes a 'keyboard timeout[2]' to be printed on the console after the SCSI
> > 
> > Omitted a critical detail -- it hangs hard-core after this message.
> > Evidently the power button was initialized by this point, because it
> > doesn't turn off immediately, and it doesn't start the LED blinking
> > indicating an impending power-off.
> 
> The power button is handled during interrupts so when the interrupt
> handling is fsck'ed the button won't work, too.
> 
> I'll look at it.

I'm having this problem as well..

- Ulf

Re: CP0_STATUS interrupt mask patch

[Ulf Carlsson]

On Wed, May 05, 1999 at 05:40:22PM +0200, Ralf Baechle wrote:
> On Wed, May 05, 1999 at 01:44:23AM +0000, Charles Lepple wrote:
> 
> > Charles Lepple wrote:
> > > I just pulled down the CVS kernel with the patch, and it seems that it
> > > causes a 'keyboard timeout[2]' to be printed on the console after the SCSI
> > 
> > Omitted a critical detail -- it hangs hard-core after this message.
> > Evidently the power button was initialized by this point, because it
> > doesn't turn off immediately, and it doesn't start the LED blinking
> > indicating an impending power-off.
> 
> The power button is handled during interrupts so when the interrupt
> handling is fsck'ed the button won't work, too.
> 
> I'll look at it.

Ok, you don't have to look at it I think since I just smashed the bug.

I found that indy_sc_wipe corrupted the interrupt mask, and that it relayed on
that it was restored by other functions to a correct state again.

My R4600SC boots fine now with Harald's new patches.

- Ulf