Re: Linux 2.6.3 still doesn't boot on UltraSparc I

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Ben Collins]

> > > > Try this, keep the new head.S and newer SILO, and edit head.S so that
> > > > "HdrS version" is 0x202 instead of 0x300. See if that boots for you
> > > > (would confirm my suspicion).
> > > 
> > > It boots correctly this way. What can we do to help now ?
> > 
> > Can you enable CONFIG_DEBUG_BOOTMEM and boot with "linux -p"?
> 
> Here is the output :
> 
> boot: linux -p
> Allocated 8 Megs of memory at 0x4000000 for kernel
> Kernel doesn't support loading to high memory, relocating done.
> Loaded kernel version 2.6.3
> 
> PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
> Linux version 2.6.3 (root@ensilinx8) (gcc version 3.3.3 20040125 (prerelease) (Debian)) #4 SMP [...]
> ARCH: SUN4U
> PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
> PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
> PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
> PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
> [...]
> Looping, had to stop it using Break+A.
> 
> The same kernel boots without -p.

Oooh...that's interesting. It looks like it actually gets all the way
through head.S. The PROMLIB line is from the call to prom_init() in
head.S just before it finally calls start_kernel(), and the the "ARCH:
SUN4U" is in setup_arch(), which is well after head.S has jumped to the
main kernel.

Why it's looping with prom_init(), I'm not sure. Maybe the prom cif
handler and stack aren't getting setup properly. Dave, any ideas?

-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/
WatchGuard - http://www.watchguard.com/

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[David S. Miller]

On Tue, 24 Feb 2004 09:42:45 -0500
Ben Collins <bcollins@debian.org> wrote:

> Why it's looping with prom_init(), I'm not sure. Maybe the prom cif
> handler and stack aren't getting setup properly. Dave, any ideas?

Nothing jumps to my mind, it might even be just looping inside the
OBP itself, ie. in the prom_printf() call.

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Lucas Nussbaum]

On Wed, Feb 25, 2004 at 12:28:53AM -0800, "David S. Miller" <davem@redhat.com> wrote:
> On Tue, 24 Feb 2004 09:42:45 -0500
> Ben Collins <bcollins@debian.org> wrote:
> 
> > Why it's looping with prom_init(), I'm not sure. Maybe the prom cif
> > handler and stack aren't getting setup properly. Dave, any ideas?
> 
> Nothing jumps to my mind, it might even be just looping inside the
> OBP itself, ie. in the prom_printf() call.

Ok. Now, what's next ? I understand fixing a problem with some very old
Sparcs isn't one of your priorities. Do you want to drop support for
them in 2.6 ? Or there is still something you/we could do to try to fix
this ?

The box is a quadriprocessor with 1.6GB RAM, so we really are interested
in using it : it isn't _that_ slow.

Lucas

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Ben Collins]

On Thu, Feb 26, 2004 at 10:12:05AM +0100, Lucas Nussbaum wrote:
> On Wed, Feb 25, 2004 at 12:28:53AM -0800, "David S. Miller" <davem@redhat.com> wrote:
> > On Tue, 24 Feb 2004 09:42:45 -0500
> > Ben Collins <bcollins@debian.org> wrote:
> > 
> > > Why it's looping with prom_init(), I'm not sure. Maybe the prom cif
> > > handler and stack aren't getting setup properly. Dave, any ideas?
> > 
> > Nothing jumps to my mind, it might even be just looping inside the
> > OBP itself, ie. in the prom_printf() call.
> 
> Ok. Now, what's next ? I understand fixing a problem with some very old
> Sparcs isn't one of your priorities. Do you want to drop support for
> them in 2.6 ? Or there is still something you/we could do to try to fix
> this ?
> 
> The box is a quadriprocessor with 1.6GB RAM, so we really are interested
> in using it : it isn't _that_ slow.

Well, in the meantime, you can downgrade to silo prior to 1.4.0. Check
http://www.sparc-boot.org/pub/silo/old/

Get one of the silo-loaders tarballs, and unpack it. Put the second.b in
/boot and run "silo -f".

-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/
WatchGuard - http://www.watchguard.com/

2.6.3-rc2 fails to boot with "Illegal Instruction" on sparc64

[Lucas Nussbaum]

[ Please Cc me as I'm not subscribed to the list ]

Hi,

Linux 2.6.3-rc2 fails to boot on sparc64. 2.6.2 doesn't boot neither.

The output is :
Allocated 8 Megs of memory at 0x4000000 for kernel

Loaded kernel version 2.6.3    <--- SILO message
Illegal Instruction
{0} ok 

2.6.1 boots without any problem. 
silo v. 1.4.4

We used gcc 3.3 to compile the kernel. Same issue with gcc 3.2.
2.6.1 compiled with exactly the same build environment does boot.

The box is a Sun Entreprise 4000.
/proc/cpuinfo:
cpu             : TI UltraSparc I   (SpitFire)
fpu             : UltraSparc I integrated FPU
promlib         : Version 3 Revision 2
prom            : 3.2.4
type            : sun4u
ncpus probed    : 4
ncpus active    : 4
Cpu0Bogo        : 333.41
Cpu0ClkTck      : 0000000009f437c0
Cpu1Bogo        : 333.41
Cpu1ClkTck      : 0000000009f437c0
Cpu4Bogo        : 333.41
Cpu4ClkTck      : 0000000009f437c0
Cpu5Bogo        : 333.41
Cpu5ClkTck      : 0000000009f437c0
MMU Type        : Spitfire
State:
CPU0:           online
CPU1:           online
CPU4:           online
CPU5:           online

Any clues ?
Any other useful information I could provide ?
-- 
Lucas Nussbaum
Club GNU/Linux
ENSIMAG - Departement Telecommunications

Linux 2.6.3 still doesn't boot on UltraSparc I

[Lucas Nussbaum]

Hi,

As we reported in [1], Linux 2.6.(2|3) doesnt boot on sparc64. We traced the
problem back to some changes in arch/sparc64/kernel/head.S (2.6.3 does
boot with 2.6.1's head.S).

[1] http://marc.theaimsgroup.com/?l=linux-kernel&m=107643106916202&w=2

output :
Allocated 8 Megs of memory at 0x4000000 for kernel
Loaded kernel version 2.6.3    <--- SILO message
Illegal Instruction
{0} ok 

After a discussion with Ben Collins, it seems that only UltraSparc I are
affected. Could somebody check the new head.S's assembly code ? We are a
bit short here on sparc64 ASM...

Please tell us how we can help.

Thank you,
-- 
Lucas Nussbaum
Club GNU/Linux
ENSIMAG - Departement Telecommunications

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Ben Collins]

On Mon, Feb 23, 2004 at 02:32:13PM +0100, Lucas Nussbaum wrote:
> Hi,
> 
> As we reported in [1], Linux 2.6.(2|3) doesnt boot on sparc64. We traced the
> problem back to some changes in arch/sparc64/kernel/head.S (2.6.3 does
> boot with 2.6.1's head.S).
> 
> [1] http://marc.theaimsgroup.com/?l=linux-kernel&m=107643106916202&w=2
> 
> output :
> Allocated 8 Megs of memory at 0x4000000 for kernel
> Loaded kernel version 2.6.3    <--- SILO message
> Illegal Instruction
> {0} ok 
> 
> After a discussion with Ben Collins, it seems that only UltraSparc I are
> affected. Could somebody check the new head.S's assembly code ? We are a
> bit short here on sparc64 ASM...

So backing down to silo < 1.4.0 works with the new head.S on UltraSPARC
I? I don't think the problem itself is in head.S. Changing that would
force silo 1.4.x to not keep the kernel in high memory (would copy it
back down to 0x4000 where the old silo would have put it). I am guessing
the problem is elsewhere.

Try this, keep the new head.S and newer SILO, and edit head.S so that
"HdrS version" is 0x202 instead of 0x300. See if that boots for you
(would confirm my suspicion).

-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/
WatchGuard - http://www.watchguard.com/

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Lucas Nussbaum]

On Mon, Feb 23, 2004 at 08:25:02AM -0500, Ben Collins <bcollins@debian.org> wrote:
> So backing down to silo < 1.4.0 works with the new head.S on UltraSPARC
> I? I don't think the problem itself is in head.S. Changing that would
> force silo 1.4.x to not keep the kernel in high memory (would copy it
> back down to 0x4000 where the old silo would have put it). I am guessing
> the problem is elsewhere.
> 
> Try this, keep the new head.S and newer SILO, and edit head.S so that
> "HdrS version" is 0x202 instead of 0x300. See if that boots for you
> (would confirm my suspicion).

It boots correctly this way. What can we do to help now ?

Lucas

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Ben Collins]

On Mon, Feb 23, 2004 at 04:06:10PM +0100, Lucas Nussbaum wrote:
> On Mon, Feb 23, 2004 at 08:25:02AM -0500, Ben Collins <bcollins@debian.org> wrote:
> > So backing down to silo < 1.4.0 works with the new head.S on UltraSPARC
> > I? I don't think the problem itself is in head.S. Changing that would
> > force silo 1.4.x to not keep the kernel in high memory (would copy it
> > back down to 0x4000 where the old silo would have put it). I am guessing
> > the problem is elsewhere.
> > 
> > Try this, keep the new head.S and newer SILO, and edit head.S so that
> > "HdrS version" is 0x202 instead of 0x300. See if that boots for you
> > (would confirm my suspicion).
> 
> It boots correctly this way. What can we do to help now ?

Can you enable CONFIG_DEBUG_BOOTMEM and boot with "linux -p"?

-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/
WatchGuard - http://www.watchguard.com/

Re: Linux 2.6.3 still doesn't boot on UltraSparc I

[Lucas Nussbaum]

On Mon, Feb 23, 2004 at 10:45:49AM -0500, Ben Collins <bcollins@debian.org> wrote:
> On Mon, Feb 23, 2004 at 04:06:10PM +0100, Lucas Nussbaum wrote:
> > On Mon, Feb 23, 2004 at 08:25:02AM -0500, Ben Collins <bcollins@debian.org> wrote:
> > > So backing down to silo < 1.4.0 works with the new head.S on UltraSPARC
> > > I? I don't think the problem itself is in head.S. Changing that would
> > > force silo 1.4.x to not keep the kernel in high memory (would copy it
> > > back down to 0x4000 where the old silo would have put it). I am guessing
> > > the problem is elsewhere.
> > > 
> > > Try this, keep the new head.S and newer SILO, and edit head.S so that
> > > "HdrS version" is 0x202 instead of 0x300. See if that boots for you
> > > (would confirm my suspicion).
> > 
> > It boots correctly this way. What can we do to help now ?
> 
> Can you enable CONFIG_DEBUG_BOOTMEM and boot with "linux -p"?

Here is the output :

boot: linux -p
Allocated 8 Megs of memory at 0x4000000 for kernel
Kernel doesn't support loading to high memory, relocating done.
Loaded kernel version 2.6.3

PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
Linux version 2.6.3 (root@ensilinx8) (gcc version 3.3.3 20040125 (prerelease) (Debian)) #4 SMP [...]
ARCH: SUN4U
PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
PROMLIB: Sun IEEE Boot Prom 3.2.4 1996/05/30 11:17
[...]
Looping, had to stop it using Break+A.

The same kernel boots without -p.

Lucas