Re: swap on dmcrypt crashes machine

[Christophe Saout]

Am Dienstag, den 25.01.2005, 18:28 +0100 schrieb Michael Buesch:

> I set up swap on an encrypted dmcrypt device.
> While stressing swap usage with make -j200 in the
> kernel tree, the machine crashes:
> 
> Adding 1461872k swap on /dev/mapper/swap.  Priority:-2 extents:1
> ------------[ cut here ]------------
> kernel BUG at drivers/block/ll_rw_blk.c:2193!
> invalid operand: 0000 [#1]
> SMP 
> Modules linked in: ipv6 ohci_hcd tuner tvaudio msp3400 bttv video_buf firmware_class btcx_risc nvidia ehci_hcd uhci_hcd usbcore intel_agp agpgart evdev
> CPU:    0
> EIP:    0060:[<c0218b8a>]    Tainted: P      VLI
> EFLAGS: 00210002   (2.6.10-ck4-defaultidle) 
> EIP is at __blk_put_request+0x87/0x91

This seems to be this BUG:

BUG_ON(!list_empty(&req->queuelist));

in drivers/block/ll_rw_blk.c. I don't know how this could be non-empty
since scsi_end_request calls

if (blk_rq_tagged(req))
         blk_queue_end_tag(q, req);

just before calling end_that_request_last inside the same spinlock.

I don't know anything about tagged request handling but I don't think
this is the fault of the dm-crypt driver, perhaps it's stressing the
queue in some unexpected way? I have no idea what's going on here.

Attachment: signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil