Re: todays git: WARNING: at drivers/ata/libata-sff.c:1017 ata_sff_hsm_move+0x45e/0x750()

[Sergei Shtylyov <sshtylyov@ru.mvista.com>]

Hello.

Christian Borntraeger wrote:

>>> I did some printk debugging. 
>>> __atapi_pio_bytes is called with bytes=18. That means with 32bit transfers
>>> there is a 2 byte trailing transfer, which triggers this WARN_ON.
>>>       
>> Ouch.. that means all that someone issued a multiple segment data
>> transfer with a 2 byte aligned chunk in the middle.
>>
>> That is a bit of a suprise because we set blk_queue_dma_alignment to 3 so
>> the block layer should never hand us an sg list in that form. That means
>> we've either got a block layer bug or someone somewhere is generating
>> invalid sg lists internally to libata.
>>
>> Can you dump qc->cdb[] and see what command is getting issued in the
>> broken form ?
>>     
>
> I instrumented __atapi_pio_bytes to print some variables. This is the dump of
> the first two:
>
> [    1.484730] count: 36 sg->length: 36 qc->cursg_ofs: 0 bytes: 36
> [    1.485709] qc->cdb[]:12000000240000000000000000000000
> That works fine.
>   

   That's the typical INQUIRY command with 36-byte data length.

> [    1.499843] scsi 1:0:0:0: CD-ROM            HL-DT-ST DVDRAM GSA-U10N  1.05 PQ: 0 ANSI: 5
> [    1.508389] count: 18 sg->length: 96 qc->cursg_ofs: 0 bytes: 18
> [    1.509369] qc->cdb[]:03000000600000000000000000000000
> That one is broken.
>   

   That's the typical REQUEST SENSE command with 18-byte data length.

MBR, Sergei