Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Paul » Tue, 15 Mar 2005 16:10:08


Hi;

Here is what I did:

# modprobe ide-scsi
# cd /proc/ide/hdd (this is a dvdrw drive)
# cat driver
# echo ide-scsi > driver
# cat driver
# echo ide-cdrom > driver

The shell is killed and Oops.

Machine flakey and half alive at this point. Reboot with Alt-sysrq.
The same thing works with 2.6.10, without Oops.

Here is the oops:
.config follows; let me know if any more information is required.

Paul
XXXX@XXXXX.COM

Unable to handle kernel NULL pointer dereference at virtual address 000001e8
printing eip:
d89f6338
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP
Modules linked in: ide_scsi usbcore serpent cryptoloop loop sg sr_mod sd_mod scsi_mod
CPU: 0
EIP: 0060:[pg0+407831352/1068467200] Not tainted VLI
EIP: 0060:[<d89f6338>] Not tainted VLI
EFLAGS: 00010002 (2.6.11)
EIP is at idescsi_queue+0x108/0x3d0 [ide_scsi]
eax: 00000000 ebx: d594b9a0 ecx: d594b9a0 edx: c04e46a8
esi: d129609e edi: d887c720 ebp: d1296040 esp: d0547cec
ds: 007b es: 007b ss: 0068
Process bash (pid: 8278, threadinfo=d0547000 task=d1f460a0)
Stack: d1296104 d887c6a7 d55ab0b0 d594b9a0 d1296040 d12960b4 c03ca896 d1296094
d1ed0540 000001dc c04e46a8 d7f05e00 d887c720 d1296040 00000286 d7f05e00
00000000 d887c555 d25fe800 d1296040 d25fe800 d7f05e00 d7cebcac d88817a3
Call Trace:
[pg0+406283943/1068467200] scsi_init_cmd_from_req+0xc7/0x140 [scsi_mod]
[<d887c6a7>] scsi_init_cmd_from_req+0xc7/0x140 [scsi_mod]
[_spin_lock+22/144] _spin_lock+0x16/0x90
[<c03ca896>] _spin_lock+0x16/0x90
[pg0+406284064/1068467200] scsi_done+0x0/0x20 [scsi_mod]
[<d887c720>] scsi_done+0x0/0x20 [scsi_mod]
[pg0+406283605/1068467200] scsi_dispatch_cmd+0x125/0x1b0 [scsi_mod]
[<d887c555>] scsi_dispatch_cmd+0x125/0x1b0 [scsi_mod]
[pg0+406304675/1068467200] scsi_request_fn+0x1d3/0x350 [scsi_mod]
[<d88817a3>] scsi_request_fn+0x1d3/0x350 [scsi_mod]
[blk_insert_request+149/176] blk_insert_request+0x95/0xb0
[<c02bd355>] blk_insert_request+0x95/0xb0
[pg0+406300158/1068467200] scsi_insert_special_req+0x2e/0x40 [scsi_mod]
[<d88805fe>] scsi_insert_special_req+0x2e/0x40 [scsi_mod]
[pg0+406300723/1068467200] scsi_wait_req+0x63/0xa0 [scsi_mod]
[<d8880833>] scsi_wait_req+0x63/0xa0 [scsi_mod]
[pg0+406300496/1068467200] scsi_wait_done+0x0/0x80 [scsi_mod]
[<d8880750>] scsi_wait_done+0x0/0x80 [scsi_mod]
[pg0+406188547/1068467200] sr_do_ioctl+0x73/0x250 [sr_mod]
[<d8865203>] sr_do_ioctl+0x73/0x250 [sr_mod]
[pg0+406187844/1068467200] sr_packet+0x14/0x30 [sr_mod]
[<d8864f44>] sr_packet+0x14/0x30 [sr_mod]
[cdrom_get_disc_info+83/160] cdrom_get_disc_info+0x53/0xa0
[<c02f6a73>] cdrom_get_disc_info+0x53/0xa0
[cdrom_mrw_exit+18/96] cdrom_mrw_exit+0x12/0x60
[<c02f2de2>] cdrom_mrw_exit+0x12/0x60
[_spin_lock+22/144] _spin_lock+0x16/0x90
[<c03ca896>] _spin_lock+0x16/0x90
[unregister_cdrom+140/192] unregister_cdrom+0x8c/0xc0
[<c02f2a3c>] unregister_cdrom+0x8c/0xc0
[pg0+406187935/1068467200] sr_kref_release+0x3f/0x70 [sr_mod]
[<d8864f9f>] sr_kref_release+0x3f/0x70 [sr_mod]
[pg0+406187872/1068467200] sr_kref_release+0x0/0x70 [sr_mod]
[<d8864f60>] sr_kref_release+0x0/0x70 [sr_mod]
[kref_put+53/160] kref_put+0x35/0xa0
[<c027d975>] kref_put+0x35/0xa0
[dput+30/416] dput+0x1e/0x1a0
[<c01661ce&
 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Alan Co » Wed, 16 Mar 2005 02:50:24


This has always been unsafe. Its something I suggested was removed a
long time ago because the locking for it is unfixable.

Alan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/

 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Bartlomiej » Wed, 16 Mar 2005 17:30:12


Locking is fixed in ide-dev-2.6 tree
(at the moment seem to be dropped from -mm?).

[ the FIXME comment left in ide-proc.c is also applicable to
modprobe/rmmod case as the issue is not /proc/.../driver specific ]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/
 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Alan Co » Wed, 16 Mar 2005 22:30:13


Excellent - I'm looking forward to dropping the -ac IDE locking patches

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/
 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Bartlomiej » Wed, 16 Mar 2005 23:00:23


There is still one thing TODO:

* fixing device drivers to refcount driver specific /proc/ide/ entries
(infrastructure is in-place now)

so don't drop your locking patches yet. :-)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/
 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Bartlomiej » Sun, 20 Mar 2005 01:50:12


Please see http://www.yqcomputer.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/
 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Paul » Sun, 20 Mar 2005 03:40:11

Bartlomiej Zolnierkiewicz < XXXX@XXXXX.COM >, on Fri Mar 18, 2005 [05:34:06 PM] said:


Hi;

What is the nature of the 'ide-dev-2.6 tree'? Are there broken
out patches available I can test vs. 2.6.11 or -mm? How do the 'ide fixes'
in -ac intersect with ide-dev? I am also curious if these patches could
have any effect on the pktcdvd problems I have reported.[*]
Thanks for the feedback.

Paul
XXXX@XXXXX.COM

* http://www.yqcomputer.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/
 
 
 

Repeatable IDE Oops for 2.6.11 (ide-scsi vs ide-cdrom)

Post by Bartlomiej » Tue, 22 Mar 2005 17:30:21


it is BK tree which is pulled into -mm, no patches against vanilla kernels (yet)


locking 'ide fixes' in -ac are another approach to the same problem


This issue looks like pktcdvdv/udf specific thing so these patches won't help.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to XXXX@XXXXX.COM
More majordomo info at http://www.yqcomputer.com/
Please read the FAQ at http://www.yqcomputer.com/