git-current: slub breaks s2ram with fglrx...

git-current: slub breaks s2ram with fglrx...

[Soeren Sonnenburg]

OK,

I've found the second root for my 

'2.6.22-rc regression: s2ram fails to suspend + fails to resume w/ Xorg':

The first one was just the wrong coretemp patch (already fixed by Jean).

The second one happens only with Xorg/fglrx loaded and slub enabled, as
I've found after a useless git bisect session, I've traced down the
other s2ram does not work problem to be caused by slub ... FYI: I am now
on git-current which should be rc5:

slab & fglrx works
slub & console works
slub & X11+fglrx creates hangs on suspend (black screen - no further
idea ...)

It should be noted that withouth the proprietary  fglrx.ko module loaded
the machine just reboots on s2ram (though IIRC without fglrx.ko loaded
it worked with 2.6.19 or so...)

Is this a kernel thing or yet another bug in fglrx ?

Soeren 
-- 
Sometimes, there's a moment as you're waking, when you become aware of
the real world around you, but you're still dreaming.

Re: git-current: slub breaks s2ram with fglrx...

[Rafael J. Wysocki]

On Sunday, 17 June 2007 11:49, Soeren Sonnenburg wrote:
> OK,
> 
> I've found the second root for my 
> 
> '2.6.22-rc regression: s2ram fails to suspend + fails to resume w/ Xorg':
> 
> The first one was just the wrong coretemp patch (already fixed by Jean).
> 
> The second one happens only with Xorg/fglrx loaded and slub enabled, as
> I've found after a useless git bisect session, I've traced down the
> other s2ram does not work problem to be caused by slub ... FYI: I am now
> on git-current which should be rc5:
> 
> slab & fglrx works
> slub & console works
> slub & X11+fglrx creates hangs on suspend (black screen - no further
> idea ...)
> 
> It should be noted that withouth the proprietary  fglrx.ko module loaded
> the machine just reboots on s2ram (though IIRC without fglrx.ko loaded
> it worked with 2.6.19 or so...)
> 
> Is this a kernel thing or yet another bug in fglrx ?

I have no idea.

I guess we should let Christoph know about it (CC added), but I'm afraid
we won't be able to figure out what's wrong without access to the driver's
source code ...

Greetings,
Rafael


-- 
"Premature optimization is the root of all evil." - Donald Knuth

Re: git-current: slub breaks s2ram with fglrx...

[Soeren Sonnenburg]

On Sun, 2007-06-17 at 13:04 +0200, Rafael J. Wysocki wrote:
> On Sunday, 17 June 2007 11:49, Soeren Sonnenburg wrote:
> > OK,
[...]
> > slab & fglrx works
> > slub & console works
> > slub & X11+fglrx creates hangs on suspend (black screen - no further
> > idea ...)
> > 
> > It should be noted that withouth the proprietary  fglrx.ko module loaded
> > the machine just reboots on s2ram (though IIRC without fglrx.ko loaded
> > it worked with 2.6.19 or so...)
> > 
> > Is this a kernel thing or yet another bug in fglrx ?
> 
> I have no idea.
> 
> I guess we should let Christoph know about it (CC added), but I'm afraid
> we won't be able to figure out what's wrong without access to the driver's
> source code ...

I have one question: What if I don't load fglrx.ko but still use the
proprietary binary driver for xorg. Shouldn't s2ram work with that ? And
if not is this considered a kernel bug or still a ati-binary driver
bug ?

Soeren
-- 
Sometimes, there's a moment as you're waking, when you become aware of
the real world around you, but you're still dreaming.

Re: git-current: slub breaks s2ram with fglrx...

[Michal Piotrowski]

Hi Soeren,

On 17/06/07, Soeren Sonnenburg <kernel@nn7.de> wrote:
> On Sun, 2007-06-17 at 13:04 +0200, Rafael J. Wysocki wrote:
> > On Sunday, 17 June 2007 11:49, Soeren Sonnenburg wrote:
> > > OK,
> [...]
> > > slab & fglrx works
> > > slub & console works
> > > slub & X11+fglrx creates hangs on suspend (black screen - no further
> > > idea ...)
> > >
> > > It should be noted that withouth the proprietary  fglrx.ko module loaded
> > > the machine just reboots on s2ram (though IIRC without fglrx.ko loaded
> > > it worked with 2.6.19 or so...)
> > >
> > > Is this a kernel thing or yet another bug in fglrx ?
> >
> > I have no idea.
> >
> > I guess we should let Christoph know about it (CC added), but I'm afraid
> > we won't be able to figure out what's wrong without access to the driver's
> > source code ...
>
> I have one question: What if I don't load fglrx.ko but still use the
> proprietary binary driver for xorg.

AFAIK it will not work.

> Shouldn't s2ram work with that ? And
> if not is this considered a kernel bug or still a ati-binary driver
> bug ?
>
> Soeren
> --
> Sometimes, there's a moment as you're waking, when you become aware of
> the real world around you, but you're still dreaming.
>

Regards,
Michal

-- 
LOG
http://www.stardust.webpages.pl/log/

Re: git-current: slub breaks s2ram with fglrx...

[Soeren Sonnenburg]

On Sun, 2007-06-17 at 17:33 +0200, Michal Piotrowski wrote:
> Hi Soeren,

Hi Michael,

[...]
> > I have one question: What if I don't load fglrx.ko but still use the
> > proprietary binary driver for xorg.
> 
> AFAIK it will not work.

well I *know* that X works without fglrx loaded (no 3d acceleration/ no
xv extensions though). I would be happily using the no-fglrx setup if
s2ram worked (and is more stable than with fglrx loaded). But currently
it just reboots on resume, so ...

> Shouldn't s2ram work with that ? And
> if not is this considered a kernel bug or still a ati-binary driver
> bug ?

Soeren
-- 
Sometimes, there's a moment as you're waking, when you become aware of
the real world around you, but you're still dreaming.

Re: git-current: slub breaks s2ram with fglrx...

[Christoph Lameter]

Could you try to boot with "slub_debug" on the kernel command line? That 
may help you to diagnose the problem.