Re: Oops while shutdown on serial console

Re: Oops while shutdown on serial console

[Sven Hartge]

Um 11:11 Uhr am 10.10.05 schrieb David S. Miller:

>> The Debian variant of 2.6.12 works fine and does shutdown correct. I have 
>> yet to check, if they include any patch to solve this issue or if only 
>> 2.6.13 and higher are affected.
 
> I believe they have a hack workaround that checks for a NULL tty in the 
> wakeup routine, it's the wrong fix.

Is this Oops specific to all sparc64 systems or does it just happen on 
sbus ones? The same question applies to the bug I sent the patch earlier, 
because I wonder why nobody has noticed before and 2.6.13 even got 
released while uncompilable on sparc64.

I am asking, because I have some E150 and E250, which I would like to 
convert from SunOS 2.6 to Linux, but I am dependant on proper serial 
console support.

Maybe I am just (again) the "lucky" guy, who ownes some obscure systems 
which trigger strange bugs. (Wouldn't be the first time for me, as I also 
own a PowerstackII, which was broken from 2.5.30 to 2.6.10 and nearly 
nobody was able to reproduce my problems.)

Grüße,
Sven

-- 
Sven Hartge -- professioneller Unix-Geek
Meine Gedanken im Netz: http://sven.formvision.de/blog/

Re: Oops while shutdown on serial console

[David S. Miller]

From: Sven Hartge <hartge@ds9.argh.org>
Date: Mon, 10 Oct 2005 20:44:42 +0200 (CEST)

> Is this Oops specific to all sparc64 systems or does it just happen on 
> sbus ones?

Specific to all sparc systems using serial console, it's some
problem in either the UART layer or the sparc serial drivers.

> The same question applies to the bug I sent the patch earlier, 

Specific to sparc64 systems using iptables and having discontiguous
physical cpu numbers.

> I am asking, because I have some E150 and E250, which I would like to 
> convert from SunOS 2.6 to Linux, but I am dependant on proper serial 
> console support.

I'll try to fix the serial problem this week.

Re: Oops while shutdown on serial console

[Sven Hartge]

Um 13:11 Uhr am 10.10.05 schrieb David S. Miller:

>> The same question applies to the bug I sent the patch earlier, 
> Specific to sparc64 systems using iptables and having discontiguous
> physical cpu numbers.

Huh? I think you mixed this one up. I meant the one-line fix which 
included asm/io.h into irq.c for readb()/writeb(). 
 
>> I am asking, because I have some E150 and E250, which I would like to 
>> convert from SunOS 2.6 to Linux, but I am dependant on proper serial 
>> console support.
 
> I'll try to fix the serial problem this week.

Thank you. Without proper serial console support I'd have a hard time 
defending myself against those NetBSD guys in my department ;)

Grüße,
Sven.

-- 
Sven Hartge -- professioneller Unix-Geek
Meine Gedanken im Netz: http://sven.formvision.de/blog/

Re: Oops while shutdown on serial console

[David S. Miller]

From: Sven Hartge <hartge@ds9.argh.org>
Date: Mon, 10 Oct 2005 22:26:17 +0200 (CEST)

> Um 13:11 Uhr am 10.10.05 schrieb David S. Miller:
> 
> >> The same question applies to the bug I sent the patch earlier, 
> > Specific to sparc64 systems using iptables and having discontiguous
> > physical cpu numbers.
> 
> Huh? I think you mixed this one up. I meant the one-line fix which 
> included asm/io.h into irq.c for readb()/writeb(). 

Most configurations usually get the asm/io.h file indirectly
by other includes, which is probably why others have not hit
it before.

Re: Oops while shutdown on serial console

[Sven Hartge]

Um 13:38 Uhr am 10.10.05 schrieb David S. Miller:
> Sven Hartge wrote:
>> Um 13:11 Uhr am 10.10.05 schrieb David S. Miller:

>>>> The same question applies to the bug I sent the patch earlier, 
>>> Specific to sparc64 systems using iptables and having discontiguous
>>> physical cpu numbers.

>> Huh? I think you mixed this one up. I meant the one-line fix which 
>> included asm/io.h into irq.c for readb()/writeb(). 

> Most configurations usually get the asm/io.h file indirectly
> by other includes, which is probably why others have not hit
> it before.

Most systems today are ones with PCI, so they all pull in pci_impl.h, 
which includes asm/io.h. As I am using a SBUS-Sparc, I hit the bug. Like I 
wrote, just me and my bizarre old computers again ;)

Thanks again for your quick answer.

Grüße,
Sven.

-- 
Sven Hartge -- professioneller Unix-Geek
Meine Gedanken im Netz: http://sven.formvision.de/blog/