2.4 kernels broken on Takara?

2.4 kernels broken on Takara?

[Jay Maynard]

I'm trying to install Gentoo on a Takara (PICMG 21164A/500). The only
installer I've found that will run is the Debian 3.0r2 CD, which has a
2.2.20 kernel on it. That's not new enough for Gentoo, which needs a 2.4 to
install. I've tried installing a Debian system, then a 2.4 kernel on top of
that - but 2.4.18 and 2.4.24 prebuilt generic kernels, as well as a 2.4.24
custom kernel built specifically for the Takara, all blow up with a kernel
paging error during startup, right after the PCI bridges are detected. The
2.2.20 kernel iinstalled by Debian runs fine.

What do I need to do to find this problem and fix it?

Re: 2.4 kernels broken on Takara?

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 921 bytes --]

On Sun, 2004-02-08 10:39:09 -0600, Jay Maynard <jmaynard@conmicro.cx>
wrote in message <20040208103909.C25099@thebrain.conmicro.cx>:
> install. I've tried installing a Debian system, then a 2.4 kernel on top of
> that - but 2.4.18 and 2.4.24 prebuilt generic kernels, as well as a 2.4.24
> custom kernel built specifically for the Takara, all blow up with a kernel
> paging error during startup, right after the PCI bridges are detected. The
> 2.2.20 kernel iinstalled by Debian runs fine.

I remember there were problems with the PCI bridging code. Maybe you'd
try to give 2.6.x a shot...

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
   ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.4 kernels broken on Takara?

[Jay Maynard]

On Sun, Feb 08, 2004 at 10:14:55PM +0100, Jan-Benedict Glaw wrote:
> On Sun, 2004-02-08 10:39:09 -0600, Jay Maynard <jmaynard@conmicro.cx>
> wrote in message <20040208103909.C25099@thebrain.conmicro.cx>:
> > install. I've tried installing a Debian system, then a 2.4 kernel on top of
> > that - but 2.4.18 and 2.4.24 prebuilt generic kernels, as well as a 2.4.24
> > custom kernel built specifically for the Takara, all blow up with a kernel
> > paging error during startup, right after the PCI bridges are detected. The
> > 2.2.20 kernel iinstalled by Debian runs fine.
> I remember there were problems with the PCI bridging code. Maybe you'd
> try to give 2.6.x a shot...

No luck, although I got a little better look at the error messages. I've
retyped them here:

PCI: Bus 2, bridge: 0000:00:14.0
  IO window: 8000-8fff
  MEM window: 02400000-037fffff
  PREFETCH window: 02200000-022fffff
PCI: Setting latency timer of device 0000:00:13.0 to 64
Unable to handle kernel paging request at virtual address 0000000000000040
swapper(1): Oops 0
pc = [<fffffc00004eecf0>]  ra = [<fffffc00004eecc8>]  ps = 0000    Not tainted
[register dump here]
Trace:fffffc0000310124 fffffc00003151e8
Code: 261dfff1 401f0000 a4410038 2c810078 20210078 22107a7c <a0420040>
488100c4
Kernel panic: Attempted to kill init!

This is with the generic Debian 2.6.0 kernel. Now what?

Re: 2.4 kernels broken on Takara?

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 2565 bytes --]

On Sun, 2004-02-08 17:53:12 -0600, Jay Maynard <jmaynard@conmicro.cx>
wrote in message <20040208175312.A26586@thebrain.conmicro.cx>:
> On Sun, Feb 08, 2004 at 10:14:55PM +0100, Jan-Benedict Glaw wrote:
> > On Sun, 2004-02-08 10:39:09 -0600, Jay Maynard <jmaynard@conmicro.cx>
> > wrote in message <20040208103909.C25099@thebrain.conmicro.cx>:
> > > install. I've tried installing a Debian system, then a 2.4 kernel on top of
> > > that - but 2.4.18 and 2.4.24 prebuilt generic kernels, as well as a 2.4.24
> > > custom kernel built specifically for the Takara, all blow up with a kernel
> > > paging error during startup, right after the PCI bridges are detected. The
> > > 2.2.20 kernel iinstalled by Debian runs fine.
> > I remember there were problems with the PCI bridging code. Maybe you'd
> > try to give 2.6.x a shot...
> 
> No luck, although I got a little better look at the error messages. I've
> retyped them here:
> 
> PCI: Bus 2, bridge: 0000:00:14.0
>   IO window: 8000-8fff
>   MEM window: 02400000-037fffff
>   PREFETCH window: 02200000-022fffff
> PCI: Setting latency timer of device 0000:00:13.0 to 64
> Unable to handle kernel paging request at virtual address 0000000000000040

0x40 means that there (most probably) was an access to some struct by a
pointer. If this is a "normal" struct containing ints and pointers, it's
most probably access to the 9th structure element.

> swapper(1): Oops 0
> pc = [<fffffc00004eecf0>]  ra = [<fffffc00004eecc8>]  ps = 0000    Not tainted
> [register dump here]

This is the interesting part. Get the value of "pc" and those from the
stack printout which start with "fffffc" and decode them with the
matching System.map file (you'll find that in /boot). The values printed
on screen are typically a bit larger than those in the System.map file,
so search for the largest address line in that file which is still
smaller than what you've found on screen.

> Trace:fffffc0000310124 fffffc00003151e8
> Code: 261dfff1 401f0000 a4410038 2c810078 20210078 22107a7c <a0420040>
> 488100c4
> Kernel panic: Attempted to kill init!
> 
> This is with the generic Debian 2.6.0 kernel. Now what?

Decode the oops. That'll exactly tell where it happened.

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
   ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.4 kernels broken on Takara?

[Thomas Steudten]

Hi

Which compiler (gcc) do you use? Try gcc 3.3.2. But i think
this is a problem in the bridging code for the takara.
You should check the program counter (PC/ IP) against
your System.map to find the function in which the
trap occurs.
However some older kernel from debian, rh or suse should
work at least - i hope.


Jay Maynard wrote:

>>>install. I've tried installing a Debian system, then a 2.4 kernel on top of
>>>that - but 2.4.18 and 2.4.24 prebuilt generic kernels, as well as a 2.4.24
>>>custom kernel built specifically for the Takara, all blow up with a kernel
>>>paging error during startup, right after the PCI bridges are detected. The
>>>2.2.20 kernel iinstalled by Debian runs fine.
>>
>>I remember there were problems with the PCI bridging code. Maybe you'd
>>try to give 2.6.x a shot...
> 
> 
> No luck, although I got a little better look at the error messages. I've
> retyped them here:
> 
> PCI: Bus 2, bridge: 0000:00:14.0
>   IO window: 8000-8fff
>   MEM window: 02400000-037fffff
>   PREFETCH window: 02200000-022fffff
> PCI: Setting latency timer of device 0000:00:13.0 to 64
> Unable to handle kernel paging request at virtual address 0000000000000040
> swapper(1): Oops 0
> pc = [<fffffc00004eecf0>]  ra = [<fffffc00004eecc8>]  ps = 0000    Not tainted
> [register dump here]
> Trace:fffffc0000310124 fffffc00003151e8
> Code: 261dfff1 401f0000 a4410038 2c810078 20210078 22107a7c <a0420040>
> 488100c4
> Kernel panic: Attempted to kill init!

-- 
Tom

LINUX user since kernel 0.99.x 1994.
RPM Alpha packages at http://alpha.steudten.com/packages
Want to know what S.u.S.E 1995 cdrom-set contains?

Re: 2.4 kernels broken on Takara?

[Jay Maynard]

On Mon, Feb 09, 2004 at 10:02:59AM +0100, Jan-Benedict Glaw wrote:
> > pc = [<fffffc00004eecf0>]  ra = [<fffffc00004eecc8>]  ps = 0000    Not tainted
           ^-- takara_swizzle + 0x74

> > Trace:fffffc0000310124 fffffc00003151e8
          ^-- init + 0x24  ^-- kernel_thread + 0x28

> Decode the oops. That'll exactly tell where it happened.

Okkay, so takara_swizzle looks like it's hiccuping. Based on this, I suspect
it's the same failure in the 2.4.x kernels I've tried...I'll verify that a
little later.

Now, for a little source code digging...

Re: 2.4 kernels broken on Takara?

[Jay Maynard]

On Mon, Feb 09, 2004 at 07:57:17AM -0600, Jay Maynard wrote:
> Okkay, so takara_swizzle looks like it's hiccuping. Based on this, I suspect
> it's the same failure in the 2.4.x kernels I've tried...I'll verify that a
> little later.

Yes, it's the same failure in the 2.4 kernels.

> Now, for a little source code digging...

It appears that the problem is in this line of code:

        unsigned int busslot = PCI_SLOT(dev->bus->self->devfn);

The pointer dev->bus->self is zero. I added the following line to the
beginning of takara_swizzle:

        if (!dev->bus->self)
                printk(KERN_CRIT "takara_swizzle: null dev->bus->self "
                                        "pointer, bus number %d!\n",
                                        dev->bus->number);

and got the following at boot time:

 Linux version 2.6.0 (root@romy) (gcc version 3.3.2 (Debian)) #1 Fri Feb 9 17:40:17 CST 2024
Booting on Takara using machine vector Takara from SRM
Major Options: EV56 LEGACY_START VERBOSE_MCHECK MAGIC_SYSRQ
Command line: ro root=/dev/sdb2  console=ttyS0
memcluster 0, usage 1, start        0, end      159
memcluster 1, usage 0, start      159, end    32651
memcluster 2, usage 1, start    32651, end    32768
freeing pages 159:384
freeing pages 849:32651
reserving pages 849:850
pci: cia revision 2
On node 0 totalpages: 32651
  DMA zone: 32651 pages, LIFO batch:7
  Normal zone: 0 pages, LIFO batch:1
  HighMem zone: 0 pages, LIFO batch:1
Building zonelist for node : 0
Kernel command line: ro root=/dev/sdb2 console=ttyS0
PID hash table entries: 1024 (order 10: 16384 bytes)
Using epoch = 1980
Console: colour VGA+ 80x25
Memory: 253232k/261208k available (2115k kernel code, 6392k reserved, 374k data, 376k init)
Calibrating delay loop... 994.44 BogoMIPS
Dentry cache hash table entries: 32768 (order: 5, 262144 bytes)
Inode-cache hash table entries: 16384 (order: 4, 131072 bytes)
Mount-cache hash table entries: 512 (order: 0, 8192 bytes)
POSIX conformance testing by UNIFIX
NET: Registered protocol family 16
pci: passed tb register update test
pci: passed sg loopback i/o read test
pci: passed tbia test
pci: passed pte write cache snoop test
pci: failed valid tag invalid pte reload test (mcheck; workaround available)
pci: passed pci machine check test
pci: enabling save/restore of SRM state
PCI: Bus 1, bridge: 0000:00:13.0
  IO window: disabled.
  MEM window: disabled.
  PREFETCH window: disabled.
PCI: Bus 2, bridge: 0000:00:14.0
  IO window: 8000-8fff
  MEM window: 02200000-022fffff
  PREFETCH window: 02300000-023fffff
PCI: Setting latency timer of device 0000:00:13.0 to 64
takara_swizzle: null dev->bus->self pointer, bus number 0!
Unable to handle kernel paging request at virtual address 0000000000000040
swapper(1): Oops 0
pc = [<fffffc000059bf50>]  ra = [<fffffc000059bf34>]  ps = 0000    Not tainted
v0 = 000000000513b280  t0 = 0000000000000010  t1 = fffffc00005f8680
t2 = 0000000000000000  t3 = 0000000000000000  t4 = ffffffff00000000
t5 = 0000000000000001  t6 = fffffffffffe07f2  t7 = fffffc000015c000
s0 = fffffc000fef9000  s1 = 0000000000000001  s2 = fffffc000015fe18
s3 = 0000000000000007  s4 = fffffc0000538680  s5 = fffffc0000538680
s6 = 0000000000000008
a0 = fffffc858000a000  a1 = 00000000000007f2  a2 = ffffffffffffffff
a3 = 000000000000000a  a4 = fffffffffffffffb  a5 = 0000000000000000
t8 = 0000000000000000  t9 = fffffc000060a060  t10= 0000000000000000
t11= 000000000000000a  pv = fffffc00003f95e0  at = 0000000000000000
gp = fffffc0000658680  sp = fffffc000015fdc8
Trace:fffffc0000310144 fffffc0000313268
Code: a4290020  23bdc74c  43e00000  a4810038  28610078  203f0010 <a0440040> 43e303a3
Kernel panic: Attempted to kill init!

The addresses map to:
   fffffc0000310144 = init+0x24
   fffffc0000313268 = kernel_thread+0x28
   fffffc000059bf50 = takara_swizzle+0x80

Any guesses as to why that pointer's null when it shouldn't be?