On 10/29/2014 10:07 AM, Paolo Bonzini wrote:
> On 10/29/2014 08:53 AM, Hannes Reinecke wrote:
>> All scsi functions take a scsi device as argument, which has
>> a LUN assigned to it. So we can get rid of specifying the 'lun'
>> as separate argument.
>
> ... except if you are sending a command to a non-existent LUN, in which
> case scsi_req_new detects this and redirects to a special SCSIReqOps
> instance:
>
> ... else if (lun != d->lun ||
> buf[0] == REPORT_LUNS ||
> (buf[0] == REQUEST_SENSE && d->sense_len)) {
> ops = &reqops_target_command;
> } else {
> ops = NULL;
> }
>
> Yes, it's ugly. If you have a better idea, I'm all ears. Passing a
> NULL SCSIDevice is hard though.
>
But still can happen even with the current code.
Look at eg hw/scsi/esp.c:
current_lun = scsi_device_find(&s->bus, 0, s->current_dev->id, lun);
s->current_req = scsi_req_new(current_lun, 0, buf, s);
datalen = scsi_req_enqueue(s->current_req);
If scsi_device_find() returns NULL we're done for.
>From what I've seeing we have two ways out of here:
- Make scsi_req_new accept NULL SCSIDevice pointer
- Make scsi_device_find() return a dummy device
with type TYPE_NO_LUN.
The latter is more appealing, as then we wouldn't need to
update the parsing logic later on.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
h...@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
