Booting 2.5.67-ac1 makes the kernel panic

Booting 2.5.67-ac1 makes the kernel panic

[Christian Staudenmayer]

hello,

(sorry for posting this again, but the kind solution given
by Patrick Mansfield turn out to be no solution at all.
Thanks anyways :)

i recently compiled 2.5.67-ac1 on my machine which uses the
aic7xxx driver for an old adaptec 2940 scsi controller.
When booting, the kernel panics after loading the scsi driver
with the following message:

Process swapper (pid: 1, threadinfo=dbf8e000 task=dbf8c040)
Stack: c15259a8 c15259a8 00000001 00000002 dbf8fe48 c022cd1a c15259a8 000000ff
       dbf8fe3c dbf8fe40
Call Trace:
 [<c022cd1a>] ide_xlate_1024+0x106/0x18c
 [<c01654aa>] handle_ide_mess+0x14e/0x1e8
 [<c016557f>] msdos_partition+0x3b/0x3ec
 [<c01171fa>] release_console_sem+0x32/0x94
 [<c0117165>] printk+0xfd/0x114
 [<c016491d>] check_partition+0xa1/0xec
 [<c0164c13>] register_disk+0xa7/0x144
 [<c0164c28>] register_disk+0xbc/0x144
 [<c0218301>] add_disk+0x35/0x44
 [<c021829c>] exact_match+0x0/0xc
 [<c02182a8>] exact_lock+0x0/0x24
 [<c0273589>] sd_attach+0x1d/0x244
 [<c023d9c0>] scsi_register_device+0xb8/0xe4
 [<c010502c>] init+0x0/0x144
 [<c0105049>] init+0x1d/0x144
 [<c010502c>] init+0x0/0x144
 [<c01079c1>] kernel_thread_helper+0x5/0xc

Code: 8b 40 38 be 93 74 31 c0 ff d0 89 c1 c7 03 3f 00 00 00 85 ff
 <0>Kernel panic: Attempted to kill init!

Note: this does not happen with 2.4.20, 2.4.21-pre7, 2.4.21-pre7-ac1 or 2.5.67-bk8
It does, however happen with 2.5.67-ac2, but the error message is some lines longer
and some of the addresses have changed.

I'd be really grateful for any insight on this problem.

Greetings, Chris

__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com

Re: Booting 2.5.67-ac1 makes the kernel panic

[Andries Brouwer]

On Fri, Apr 18, 2003 at 03:34:23PM -0700, Christian Staudenmayer wrote:

> (sorry for posting this again, but the kind solution given
> by Patrick Mansfield turn out to be no solution at all.
> Thanks anyways :)
> 
> i recently compiled 2.5.67-ac1 on my machine which uses the
> aic7xxx driver for an old adaptec 2940 scsi controller.
> When booting, the kernel panics after loading the scsi driver
> with the following message:
> 
> Process swapper (pid: 1, threadinfo=dbf8e000 task=dbf8c040)
> Stack: c15259a8 c15259a8 00000001 00000002 dbf8fe48 c022cd1a c15259a8 000000ff
>        dbf8fe3c dbf8fe40
> Call Trace:
>  [<c022cd1a>] ide_xlate_1024+0x106/0x18c
>  [<c01654aa>] handle_ide_mess+0x14e/0x1e8

You have a SCSI disk, read its partitions, look whether
special disk manager or geometry translations are required
and crash.

Why? Of course Alan should have removed this ide_mess :-),
but I suppose Al is the culprit that broke ide_xlate_1024
converting it to bdev.

The code

int ide_xlate_1024 (struct block_device *bdev, int xparm, int ptheads, const char *msg)
{
        ide_drive_t *drive = bdev->bd_disk->private_data;

is just broken - there is no guarantee that bdev is an IDE disk,
and casting some private pointer to ide_drive_t and then
accessing fields must be unhealthy.

Try replacing the body of ide_xlate_1024 by just

	return 0;

Andries

Re: Booting 2.5.67-ac1 makes the kernel panic

[Christian Staudenmayer]

Hello,

for future reference, i can verify that the method described by Andries
Brouwer below worked fine, i can now boot 2.5.67-ac2 on my machine
using the aic7xxx driver for an old adaptec 2940 scsi controller.

Many thanks Andries!

Greetings, Christian Staudenmayer

--- Andries Brouwer <aebr@win.tue.nl> wrote:
> On Fri, Apr 18, 2003 at 03:34:23PM -0700, Christian Staudenmayer wrote:
> 
> > (sorry for posting this again, but the kind solution given
> > by Patrick Mansfield turn out to be no solution at all.
> > Thanks anyways :)
> > 
> > i recently compiled 2.5.67-ac1 on my machine which uses the
> > aic7xxx driver for an old adaptec 2940 scsi controller.
> > When booting, the kernel panics after loading the scsi driver
> > with the following message:
> > 
> > Process swapper (pid: 1, threadinfo=dbf8e000 task=dbf8c040)
> > Stack: c15259a8 c15259a8 00000001 00000002 dbf8fe48 c022cd1a c15259a8 000000ff
> >        dbf8fe3c dbf8fe40
> > Call Trace:
> >  [<c022cd1a>] ide_xlate_1024+0x106/0x18c
> >  [<c01654aa>] handle_ide_mess+0x14e/0x1e8
> 
> You have a SCSI disk, read its partitions, look whether
> special disk manager or geometry translations are required
> and crash.
> 
> Why? Of course Alan should have removed this ide_mess :-),
> but I suppose Al is the culprit that broke ide_xlate_1024
> converting it to bdev.
> 
> The code
> 
> int ide_xlate_1024 (struct block_device *bdev, int xparm, int ptheads, const char *msg)
> {
>         ide_drive_t *drive = bdev->bd_disk->private_data;
> 
> is just broken - there is no guarantee that bdev is an IDE disk,
> and casting some private pointer to ide_drive_t and then
> accessing fields must be unhealthy.
> 
> Try replacing the body of ide_xlate_1024 by just
> 
> 	return 0;
> 
> Andries
> 


__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com