libata: big endian bug in VPD page 89 (ATA Information)

libata: big endian bug in VPD page 89 (ATA Information)

[Douglas Gilbert]

In drivers/ata/libata-scsi.c in function ata_scsiop_inq_89() there is
this line, just before the return:
        memcpy(&rbuf[60], &args->id[0], 512);

args->id[0] is the first u16 word of an array from the ATA IDENTIFY
DEVICE response while rbuf is an array of u8 that will become the
response to a SCSI INQUIRY(VPD=89h). Given the definition of VPD
page 89h:
    byte 60+0:  ATA IDENTIFY DEVICE data word 0 bits 7:0
    byte 60+1:  ATA IDENTIFY DEVICE data word 0 bits 15:8
    byte 60+2:  ATA IDENTIFY DEVICE data word 1 bits 7:0
    ........

then that memcpy is just fine and dandy on a little endian machine.
On a big endian machine, not so much.

Would this call after the memcpy fix things?
     swap_buf_le16((u16 *)(rbuf + 60), ATA_ID_WORDS);

That function (in libata-core.c) only swaps bytes in 16 bit words
on big endian machines.

Doug Gilbert

Re: libata: big endian bug in VPD page 89 (ATA Information)

[Hannes Reinecke]

On 6/14/21 3:28 AM, Douglas Gilbert wrote:
> In drivers/ata/libata-scsi.c in function ata_scsiop_inq_89() there is
> this line, just before the return:
>        memcpy(&rbuf[60], &args->id[0], 512);
> 
> args->id[0] is the first u16 word of an array from the ATA IDENTIFY
> DEVICE response while rbuf is an array of u8 that will become the
> response to a SCSI INQUIRY(VPD=89h). Given the definition of VPD
> page 89h:
>    byte 60+0:  ATA IDENTIFY DEVICE data word 0 bits 7:0
>    byte 60+1:  ATA IDENTIFY DEVICE data word 0 bits 15:8
>    byte 60+2:  ATA IDENTIFY DEVICE data word 1 bits 7:0
>    ........
> 
> then that memcpy is just fine and dandy on a little endian machine.
> On a big endian machine, not so much.
> 
> Would this call after the memcpy fix things?
>     swap_buf_le16((u16 *)(rbuf + 60), ATA_ID_WORDS);
> 
> That function (in libata-core.c) only swaps bytes in 16 bit words
> on big endian machines.
> 
It might. But probably no-one ever ran libata code on big-endian machines.
They are becoming rare these days; I wouldn't know where to look.
So if you had a chance to run it please give it a go.
Truth to be told, I won't be surprised if there would be more issues 
lurking in the libata code.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@suse.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer