Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Allan Duncan]

This is a reissue of an earlier report, and I've since done some more digging.

Up to kernel 2.4.20-pre2 there was no problem, agpgart et al ran fine etc,
but from 2.4.20-pre4 onwards when Xwindows starts to load these modules
I am instantly thrown back to a booting machine.
The same kernels on a VIA MVP3 chipset box with a Matrox G200 are fine.

I have ascertained that any attempt to use agpgart triggers it.

Rather than clutter up the list with lots of log files, I've made a web page at
http://www.users.bigpond.com/allan.d/bug/Matrox.html
with all the info I could gather.

Any suggestions of how to improve the error messages around the failure point
are welcome.  Nothing is written into dmesg at the time of failure.

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Simon Fowler]

[-- Attachment #1: Type: text/plain, Size: 996 bytes --]

On Thu, Sep 12, 2002 at 11:56:20AM +1000, Allan Duncan wrote:
> This is a reissue of an earlier report, and I've since done some more digging.
> 
> Up to kernel 2.4.20-pre2 there was no problem, agpgart et al ran fine etc,
> but from 2.4.20-pre4 onwards when Xwindows starts to load these modules
> I am instantly thrown back to a booting machine.
> The same kernels on a VIA MVP3 chipset box with a Matrox G200 are fine.
> 
> I have ascertained that any attempt to use agpgart triggers it.
> 
I've seen the same thing with 2.4.20-pre5-ac1, and I'm just building
pre5-aa2 to see if there's any difference. This is with an AMD 751
system (an Asus K7M), an original Radeon, and the DRI CVS code. I
couldn't get any log messages out of it, though.

Simon

-- 
PGP public key Id 0x144A991C, or ftp://bg77.anu.edu.au/pub/himi/himi.asc
(crappy) Homepage: http://bg77.anu.edu.au
doe #237 (see http://www.lemuria.org/DeCSS) 
My DeCSS mirror: ftp://bg77.anu.edu.au/pub/mirrors/css/ 

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Nuitari]

On Thu, 12 Sep 2002, Allan Duncan wrote:

> This is a reissue of an earlier report, and I've since done some more digging.
> 
> Up to kernel 2.4.20-pre2 there was no problem, agpgart et al ran fine etc,
> but from 2.4.20-pre4 onwards when Xwindows starts to load these modules
> I am instantly thrown back to a booting machine.
> The same kernels on a VIA MVP3 chipset box with a Matrox G200 are fine.
> 
> I have ascertained that any attempt to use agpgart triggers it.
> 
> Rather than clutter up the list with lots of log files, I've made a web page at
> http://www.users.bigpond.com/allan.d/bug/Matrox.html
> with all the info I could gather.
> 
> Any suggestions of how to improve the error messages around the failure point
> are welcome.  Nothing is written into dmesg at the time of failure.
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 


Might be good to check with a previous bug report I posted if there are 
any similarities. My problem is that the kernel panics when opengl is used 
when I have some highmem available.

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Allan Duncan]

Allan Duncan wrote:
> 
> This is a reissue of an earlier report, and I've since done some more digging.
> 
> Up to kernel 2.4.20-pre2 there was no problem, agpgart et al ran fine etc,
> but from 2.4.20-pre4 onwards when Xwindows starts to load these modules
> I am instantly thrown back to a booting machine.
> The same kernels on a VIA MVP3 chipset box with a Matrox G200 are fine.
> 
> I have ascertained that any attempt to use agpgart triggers it.
> 
> Rather than clutter up the list with lots of log files, I've made a web page at
> http://www.users.bigpond.com/allan.d/bug/Matrox.html
> with all the info I could gather.
> 
> Any suggestions of how to improve the error messages around the failure point
> are welcome.  Nothing is written into dmesg at the time of failure.

Uh, the logs for -pre5 with AGP_VIA enabled are NOW referred to (instead of
duplicating the -pre5 w/o VIA).

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Andi Kleen]

Allan Duncan <allan.d@bigpond.com> writes:
> 
> Any suggestions of how to improve the error messages around the failure point
> are welcome.  Nothing is written into dmesg at the time of failure.

You're booting with mem=nopentium right ? It should go away when you turn
that off. I'm working on a fix. You can safely turn it off for now, the 
old problems that it worked around are fixed.

-Andi

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Simon Fowler]

[-- Attachment #1: Type: text/plain, Size: 817 bytes --]

On Thu, Sep 12, 2002 at 11:50:19AM +0200, Andi Kleen wrote:
> Allan Duncan <allan.d@bigpond.com> writes:
> > 
> > Any suggestions of how to improve the error messages around the failure point
> > are welcome.  Nothing is written into dmesg at the time of failure.
> 
> You're booting with mem=nopentium right ? It should go away when you turn
> that off. I'm working on a fix. You can safely turn it off for now, the 
> old problems that it worked around are fixed.
> 
The problem goes away without mem=nopentium - I've just booted into
2.4.20-pre5aa2 and fired up X.

Simon

-- 
PGP public key Id 0x144A991C, or ftp://bg77.anu.edu.au/pub/himi/himi.asc
(crappy) Homepage: http://bg77.anu.edu.au
doe #237 (see http://www.lemuria.org/DeCSS) 
My DeCSS mirror: ftp://bg77.anu.edu.au/pub/mirrors/css/ 

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Allan Duncan]

Simon Fowler wrote:
> 
> On Thu, Sep 12, 2002 at 11:50:19AM +0200, Andi Kleen wrote:
> > Allan Duncan <allan.d@bigpond.com> writes:
> > >
> > > Any suggestions of how to improve the error messages around the failure point
> > > are welcome.  Nothing is written into dmesg at the time of failure.
> >
> > You're booting with mem=nopentium right ? It should go away when you turn
> > that off. I'm working on a fix. You can safely turn it off for now, the
> > old problems that it worked around are fixed.
> >
> The problem goes away without mem=nopentium - I've just booted into
> 2.4.20-pre5aa2 and fired up X.

Not in my case, at least for 2.4.20-pre4.

At which kernels does the nopentium become obsolete?  Alan Cox mentioned some
confusion about this.  Obviously the latest ones, but does this extend as far
back as 2.4.19?

In order to close in on what changes are triggering this, I found the patch for
sched.c for -pre3 and ran that, and find that -pre3  is fine with or without
nopentium, so that narrows it to what was altered pre3 to pre4.

There was nothing obvious in Marcelo's log of changes, so I will trawl through
the diffs themselves tonight.

At the same time, I noticed that there seems to a fair bit of touchy
behaviour of AGP out there, so maybe what is proving fatal to me is the same as
the cause of flaky for others.

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Simon Fowler]

[-- Attachment #1: Type: text/plain, Size: 1381 bytes --]

On Fri, Sep 13, 2002 at 08:54:10AM +1000, Allan Duncan wrote:
> Not in my case, at least for 2.4.20-pre4.
> 
2.4.19 works for me with nopentium, 2.4.20-pre5 fails with nopentium
and works without it - I can't add anything beyond that.

> At which kernels does the nopentium become obsolete?  Alan Cox mentioned some
> confusion about this.  Obviously the latest ones, but does this extend as far
> back as 2.4.19?
> 
> In order to close in on what changes are triggering this, I found the patch for
> sched.c for -pre3 and ran that, and find that -pre3  is fine with or without
> nopentium, so that narrows it to what was altered pre3 to pre4.
> 
> There was nothing obvious in Marcelo's log of changes, so I will trawl through
> the diffs themselves tonight.
> 
> At the same time, I noticed that there seems to a fair bit of touchy
> behaviour of AGP out there, so maybe what is proving fatal to me is the same as
> the cause of flaky for others.

AGP/DRI has been flaky for me in all sorts of ways, but then I've been
using the DRI CVS code which does lots of strange things, so I can't
pin the problems on AGP . . .

Simon

-- 
PGP public key Id 0x144A991C, or ftp://bg77.anu.edu.au/pub/himi/himi.asc
(crappy) Homepage: http://bg77.anu.edu.au
doe #237 (see http://www.lemuria.org/DeCSS) 
My DeCSS mirror: ftp://bg77.anu.edu.au/pub/mirrors/css/ 

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Linux 2.4.20-pre4 & ff. blows away Xwindows with Matrox G400 and agpgart

[Allan Duncan]

Allan Duncan wrote:
> 
> Simon Fowler wrote:
> >
> > On Thu, Sep 12, 2002 at 11:50:19AM +0200, Andi Kleen wrote:
> > > Allan Duncan <allan.d@bigpond.com> writes:
> > > >
> > > > Any suggestions of how to improve the error messages around the failure point
> > > > are welcome.  Nothing is written into dmesg at the time of failure.
> > >
> > > You're booting with mem=nopentium right ? It should go away when you turn
> > > that off. I'm working on a fix. You can safely turn it off for now, the
> > > old problems that it worked around are fixed.
> > >
> > The problem goes away without mem=nopentium - I've just booted into
> > 2.4.20-pre5aa2 and fired up X.
> 
> Not in my case, at least for 2.4.20-pre4.

Errm.  Turns out I ran the same test twice, both with nopentium.

So, in summary, nopentium doesn't break 2.4.20-pre3 and earlier,
breaks 2.4.20-pre4 and later.

Thankyou all.