Linux-2.4.20: bug with radeonfb

Linux-2.4.20: bug with radeonfb

[José Luis Tallón]

Environment:	gcc-2.95.4, latest dev tools from Debian "Sarge"( testing )
Kernel:	2.4.20 + patch-2.4.20-ac1 from ftp.kernel.org
Machine:	i386 (intel P4-M/1700 ) / 512MB DDR266, chipset i845MP
Graphics subsystem:	ATI Radeon M7, 64MB DDR
			Samsung LTN-1050 flatpanel, 1400x1050 physical resolution (DVI connection)


First of all, i must congratulate all kernel developers for this wonderful 
work-of-art which Linux is and will continue to be :) Thanks for making 
this world a little better!

On to the point:

"Go's":	
	i845MP chipset supported ( didn't previously work ). ICH3 & co. detected 
and used. Some minor details left, though :)
	UDMA100 enabled and working

"No Go's":
	Radeon M7 AGP:
	- Framebuffer:	unreadable output ( did work _perfectly_ with Linux-2.4.19 ).
	Looks like there's a bug somewhere in the rendering code( kernel 
autodetects and configures a 175x65 framebuffer, as previously):
output seems to be "misplaced" -- might be a little assumption when 
calculating line offsets.
Text lines, although bent to the right( 60º angle or so ), look _extremely 
long_ (might be a wrong appreciation)
"Tux" logo is unrecogniceable too :(

	- DRM 4.1( v20020828 ) works nicely with XFree 4.2.1
	- Xvid extension works

dmesg gives this ( double-checked with 2.4.19 ): relevant sections are 
identical to 2.4.19
---
Pentium 4 Mobility - stepping 04
[...]
ref_clk=2700, ref_div=12, xclk=16600 from BIOS
Samsung LTN-1050P1-L02 flatpanel
DFP 1400x1050 from BIOS
Radeon M7 LW DDR SGRAM 64MB
colour framebuffer 175x65
---

Feel free to ask for whatever config detail i might have forgotten to add.
I'm not sure if i will have time enough to contribute (part of) a patch -- 
gotta leave some time for lunch and sleep ;) Wouldn't mind to help, however.


Please CC any relevant messages to me. Sorry for not subscribing, but i 
already receive 300+ e-mails daily :-|

TIA
Regards,
	J.L.

--
All my other computers run Linux.
This is an split-personality machine ;) 

Re: Linux-2.4.20: bug with radeonfb

[Adrian Bunk]

On Tue, Dec 03, 2002 at 01:52:09AM +0100, José Luis Tallón wrote:

> Environment:	gcc-2.95.4, latest dev tools from Debian "Sarge"( testing )
> Kernel:	2.4.20 + patch-2.4.20-ac1 from ftp.kernel.org
>...
> "No Go's":
> 	Radeon M7 AGP:
> 	- Framebuffer:	unreadable output ( did work _perfectly_ with 
> 	Linux-2.4.19 ).
> 	Looks like there's a bug somewhere in the rendering code( kernel 
> autodetects and configures a 175x65 framebuffer, as previously):
> output seems to be "misplaced" -- might be a little assumption when 
> calculating line offsets.
> Text lines, although bent to the right( 60º angle or so ), look _extremely 
> long_ (might be a wrong appreciation)
> "Tux" logo is unrecogniceable too :(
> 
> 	- DRM 4.1( v20020828 ) works nicely with XFree 4.2.1
> 	- Xvid extension works
> 
> dmesg gives this ( double-checked with 2.4.19 ): relevant sections are 
> identical to 2.4.19
> ---
> Pentium 4 Mobility - stepping 04
> [...]
> ref_clk=2700, ref_div=12, xclk=16600 from BIOS
> Samsung LTN-1050P1-L02 flatpanel
> DFP 1400x1050 from BIOS
> Radeon M7 LW DDR SGRAM 64MB
> colour framebuffer 175x65
>...

Does it work if you revert the patch against drivers/video/radeonfb.c 
that is in -ac?

Does it work in plain 2.4.20 (without the -ac patch)?

> TIA
> Regards,
> 	J.L.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Kernel for Pentium 4 hyperthreading?

[Scott Robert Ladd]

Okay, I'm going bald even faster than usual.

I've just received a new computer based on a 2.8 GHz Pentium 4 with
hyper-threading enabled. Yes, HT is enabled in the BIOS; yes, /proc/cpuinfo
shows the 'ht' flag; yes, I've compiled 2.4.20 (stock) with SMP and ACPI
enabled.

No, it doesn't work. cat /proc/cpuinfo reports a single CPU.

I've also tried a 2.5.51 kernel -- and it, indeed, does find "both"
processors, listing them in cpuinfo as siblings. Looking at the boot logs,
2.5.51 seems to work just fien with my CPU.

For many reasons, I'd prefer to be running the 2.4.20 kernel (if nothing
else, I'm having trouble getting loadable modules -- the nVidia drivers for
one -- to work on 2.5.51.)

Can 2.4.20 handle a Pentium 4 (not Xeon, mind you) with HT? What could I be
missing in my kernel build?

What is especially frustrating is that the factory-installed Windows XP had
no trouble at all using the HT-enable P4 (until I sent WinXP to the great
bit-bucket in the sky).

Thanks in advance.

..Scott

Re: Kernel for Pentium 4 hyperthreading?

[Dave Jones]

On Sat, Dec 14, 2002 at 08:11:45PM -0500, Scott Robert Ladd wrote:

 > I've just received a new computer based on a 2.8 GHz Pentium 4 with
 > hyper-threading enabled. Yes, HT is enabled in the BIOS; yes, /proc/cpuinfo
 > shows the 'ht' flag; yes, I've compiled 2.4.20 (stock) with SMP and ACPI
 > enabled.
 > No, it doesn't work. cat /proc/cpuinfo reports a single CPU.

Note that just because /proc/cpuinfo shows 'ht' does not mean you can
use it in hyperthreaded mode. To do that, you also have to have >1
sibling in the physical package. Non-Xeon type P4's don't have the
extra sibling, so don't function as a hyperthreaded CPU.

 > Can 2.4.20 handle a Pentium 4 (not Xeon, mind you) with HT? What could I be
 > missing in my kernel build?

It's more a case of whats missing in your CPU package 8-)
 
		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

RE: Kernel for Pentium 4 hyperthreading?

[Scott Robert Ladd]

Hi,

What I have is, indeed, a hyperthread-enabled Pentium 4. They aren't common;
I obtained this one direct from Intel through their Early Access Program.
The proof in the pudding is that both Windows XP and Linux 2.5.51 recognize
it as having "two" processors. The motherboard is an Intel Maryville2 (i850E
chipset), with an option to enable/disable HT on the first BIOS set-up
screen.

> Note that just because /proc/cpuinfo shows 'ht' does not mean you can
> use it in hyperthreaded mode. To do that, you also have to have >1
> sibling in the physical package. Non-Xeon type P4's don't have the
> extra sibling, so don't function as a hyperthreaded CPU.

Mine does, and so will any 3.06 or 3.6 GHz Pentium 4.

As it is, I'm past the worst of my troubles (knock on wood!) We'll see what
happens in the coming days the machine gets stressed. It looks stable with
2.5.51 -- so I guess I'm now a Linux kernel beta tester... ;)

Thanks much.

..Scott

--
Scott Robert Ladd
Coyote Gulch Productions,  http://www.coyotegulch.com
No ads -- just very free (and somewhat unusual) code.

Re: Kernel for Pentium 4 hyperthreading?

[Vergoz Michael (SYSDOOR)]

Hi,

It's possible to enable HT on any pentium 4, you just have to patch the bios
:P


> Hi,
>
> What I have is, indeed, a hyperthread-enabled Pentium 4. They aren't
common;
> I obtained this one direct from Intel through their Early Access Program.
> The proof in the pudding is that both Windows XP and Linux 2.5.51
recognize
> it as having "two" processors. The motherboard is an Intel Maryville2
(i850E
> chipset), with an option to enable/disable HT on the first BIOS set-up
> screen.
>
> > Note that just because /proc/cpuinfo shows 'ht' does not mean you can
> > use it in hyperthreaded mode. To do that, you also have to have >1
> > sibling in the physical package. Non-Xeon type P4's don't have the
> > extra sibling, so don't function as a hyperthreaded CPU.
>
> Mine does, and so will any 3.06 or 3.6 GHz Pentium 4.
>
> As it is, I'm past the worst of my troubles (knock on wood!) We'll see
what
> happens in the coming days the machine gets stressed. It looks stable with
> 2.5.51 -- so I guess I'm now a Linux kernel beta tester... ;)
>
> Thanks much.
>
> ..Scott
>
> --
> Scott Robert Ladd
> Coyote Gulch Productions,  http://www.coyotegulch.com
> No ads -- just very free (and somewhat unusual) code.
>
> -
> 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/
>

Re: Kernel for Pentium 4 hyperthreading?

[Dave Jones]

On Sun, Dec 15, 2002 at 04:47:39PM +0100, Vergoz Michael (SYSDOOR) wrote:
 > Hi,
 > 
 > It's possible to enable HT on any pentium 4, you just have to patch the bios
 > :P

A rumour perpetuated by many, and achieved by none to my knowledge.

		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: Kernel for Pentium 4 hyperthreading?

[Bill Davidsen]

On Sun, 15 Dec 2002, Dave Jones wrote:

> On Sun, Dec 15, 2002 at 04:47:39PM +0100, Vergoz Michael (SYSDOOR) wrote:
>  > Hi,
>  > 
>  > It's possible to enable HT on any pentium 4, you just have to patch the bios
>  > :P
> 
> A rumour perpetuated by many, and achieved by none to my knowledge.

If W2k can tell if I enable HT or not, I assume it's really on.
Particularly if it then tells me I have more than two CPUs, my license
doesn't cover that, and it doesn't love me anymore.

Actually, 2.5.recent can get this right using APCI and will also work.
Guess the rumor is true, although I would say "many P4" rather than "all
P4" because I believe Intel disables one sibling at the bonding step
(based on information elsewhere) for some CPUs.

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: Kernel for Pentium 4 hyperthreading?

[Dave Jones]

On Sun, Dec 15, 2002 at 10:07:43AM -0500, Scott Robert Ladd wrote:
 > Hi,
 > 
 > What I have is, indeed, a hyperthread-enabled Pentium 4. They aren't common;
 > I obtained this one direct from Intel through their Early Access Program.

Ah, apologies. Yes. In this case, you win. I bit the same problem you
had btw with this box in 2.4. You need an updated BIOS. Contact Intel.

		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

RE: Kernel for Pentium 4 hyperthreading?

[Scott Robert Ladd]

Dave Jones wrote:
> Ah, apologies. Yes. In this case, you win. I bit the same problem you
> had btw with this box in 2.4. You need an updated BIOS. Contact Intel.

I'll ask Intel if there's a BIOS update. Computers are almost as bad as
games now; the first thing you need to do before using them is patch!

What evokes my curiosity is that the 2.5.51 kernel detects and correctly
uses the processor siblings, while 2.4.20 does not. Given that 2.5.51 is
running quite well, I think I'll just stay on the bleeding edge of Linux for
a while.

..Scott

--
Scott Robert Ladd
Coyote Gulch Productions,  http://www.coyotegulch.com
No ads -- just very free (and somewhat unusual) code.

Re: Kernel for Pentium 4 hyperthreading?

[Dave Jones]

On Sun, Dec 15, 2002 at 12:40:59PM -0500, Scott Robert Ladd wrote:
 > Dave Jones wrote:
 > > Ah, apologies. Yes. In this case, you win. I bit the same problem you
 > > had btw with this box in 2.4. You need an updated BIOS. Contact Intel.
 > 
 > I'll ask Intel if there's a BIOS update. Computers are almost as bad as
 > games now; the first thing you need to do before using them is patch!
 > 
 > What evokes my curiosity is that the 2.5.51 kernel detects and correctly
 > uses the processor siblings, while 2.4.20 does not. Given that 2.5.51 is
 > running quite well, I think I'll just stay on the bleeding edge of Linux for
 > a while.

I think the problem was a missing MP table in the factory-shipped BIOS.
2.5 used ACPI to enumerate the siblings, whereas the 2.4 ACPI is a
little out of date in that department.
At least that was my random guess when I hit that problem.

		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk

Re: Linux-2.4.20: bug with radeonfb - solved

[José Luis Tallón]

At 01:56 15/12/2002 +0100, you wrote:
>On Tue, Dec 03, 2002 at 01:52:09AM +0100, José Luis Tallón wrote:
>
> > Environment:  gcc-2.95.4, latest dev tools from Debian "Sarge"( testing )
> > Kernel:       2.4.20 + patch-2.4.20-ac1 from ftp.kernel.org
> >...
> > "No Go's":
> >       Radeon M7 AGP:
> >       - Framebuffer:  unreadable output ( did work _perfectly_ with
> >       Linux-2.4.19 ).
> >       Looks like there's a bug somewhere in the rendering code( kernel
> > autodetects and configures a 175x65 framebuffer, as previously):
> > output seems to be "misplaced" -- might be a little assumption when
> > calculating line offsets.
> > Text lines, although bent to the right( 60º angle or so ), look _extremely
> > long_ (might be a wrong appreciation)
> > "Tux" logo is unrecogniceable too :(
> >
> >       - DRM 4.1( v20020828 ) works nicely with XFree 4.2.1
> >       - Xvid extension works
> >
> > dmesg gives this ( double-checked with 2.4.19 ): relevant sections are
> > identical to 2.4.19
> > ---
> > Pentium 4 Mobility - stepping 04
> > [...]
> > ref_clk=2700, ref_div=12, xclk=16600 from BIOS
> > Samsung LTN-1050P1-L02 flatpanel
> > DFP 1400x1050 from BIOS
> > Radeon M7 LW DDR SGRAM 64MB
> > colour framebuffer 175x65
> >...
>
>Does it work if you revert the patch against drivers/video/radeonfb.c
>that is in -ac?
>
>Does it work in plain 2.4.20 (without the -ac patch)?

<sweating and blushing>
Works perfect with 2.4.20 vanilla :-|
         ( I had previously trusted Alan's patches quite blindly )
- Chipset and DMA working


Will test with -ac2 and 21-pre1 and report back, so that we can isolate the 
bug ( quite a nasty one, I can assure you )

Thanks again.

         J.L.

Re: Linux-2.4.20: bug with radeonfb - solved

[Adrian Bunk]

On Sun, Dec 15, 2002 at 12:55:06PM +0100, José Luis Tallón wrote:

>...
> >Does it work if you revert the patch against drivers/video/radeonfb.c
> >that is in -ac?
> >
> >Does it work in plain 2.4.20 (without the -ac patch)?
> 
> <sweating and blushing>
> Works perfect with 2.4.20 vanilla :-|
>         ( I had previously trusted Alan's patches quite blindly )
> - Chipset and DMA working
> 
> 
> Will test with -ac2 and 21-pre1 and report back, so that we can isolate the 
> bug ( quite a nasty one, I can assure you )

If -ac2 fails, could you try to to revert the -ac change to
drivers/video/radeonfb.c and check whether this cures your problem?

> Thanks again.
> 
>         J.L.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed