sbp2 firewire failure on kernels 2.6.1-1.61, 2.6.2-1.81

[Dave Jones]

On Wed, 2004-02-18 at 08:02, Michel Alexandre Salim wrote:


> 
> the problem is probably not specific to ieee1394, actually, since I get
> similar errors connecting the drive to USB. Not to mention the drive
> geometry appears different too (under 2.4 fdisk reports one extra
> cylinder for USB2 if I recall correctly, and I had to take care not to
> allocate it to a partition).
> 
> rmmod-ing sbp2 after those errors results consistently in segfaults,
> here's the dump:

I've been going through every module we ship finding out if stuff
rmmod's cleanly. The amount that doesn't is truly terrifying.

> ieee1394: sbp2: Logged out of SBP-2 device
> Unable to handle kernel paging request at virtual address 6b6b6c1f
>  printing eip:
> f096d7c5
> *pde = 00000000
> Oops: 0000 [#1]
> CPU:    0
> EIP:    0060:[<f096d7c5>]    Not tainted
> EFLAGS: 00010246
> EIP is at scsi_device_put+0x5/0x40 [scsi_mod]
> eax: 6b6b6b6b   ebx: ecf5e89c   ecx: 00000000   edx: ecf5eddc
> esi: ef0297d8   edi: ef08675c   ebp: ef0298fc   esp: d7083f04
> ds: 007b   es: 007b   ss: 0068
> Process rmmod (pid: 20048, threadinfo=d7082000 task=d7f448a0)
> Stack: ef0297d8 f090e890 00000008 ef0297d8 ef08675c ef08675c 00000000
> f090deb8
>        ee6a7970 f091388c f091388c c0253086 f09138f0 f09138f0 c02530a8
>        f0adc6cc
>        f0adc680 c0253321 f0913894 f091388c 00000000 c02536f4 f0913a00
>        c0363920
> Call Trace:
>  [<f090e890>] sbp2_remove_device+0x180/0x190 [sbp2]
>  [<f090deb8>] sbp2_remove+0x38/0x50 [sbp2]
>  [<c0253086>] device_release_driver+0x56/0x60
>  [<c02530a8>] driver_detach+0x18/0x30
>  [<c0253321>] bus_remove_driver+0x51/0x90
>  [<c02536f4>] driver_unregister+0x14/0x3a
>  [<f0910aaa>] sbp2_module_exit+0xa/0x14 [sbp2]
>  [<c0145c3b>] sys_delete_module+0x11b/0x190
>  [<c0164b00>] unmap_vma+0x30/0xa0
>  [<c016513d>] do_munmap+0x1ad/0x280
>  [<c0177018>] __fput+0x98/0xf0
>  [<c010c49f>] syscall_call+0x7/0xb
>  
> Code: 8b 80 b4 00 00 00 8b 00 85 c0 74 0b ff 88 00 01 00 00 83 38
> ----
> 
> Any idea? Should I bugzilla this, and perhaps forward it to linux1394
> maintainers?

Let the 1394 folks know. If you bugzilla it, we'll likely just point
them at it anyway.  If they give you any nonsense about "its a vendor
kernel, we only support mainline", Fedora is 100% identical to upstream
in regard to firewire.

> Going to try a newer kernel first

Yes, please do.

	Dave