Re: [Qemu-devel] Re: [PATCH 1/2] [scsi-bus]: Add PR-OUT and PR-IN case for SCSIRequest xfer and xfer_mode setup

[Kevin Wolf <kwolf@redhat.com>]

Am 16.06.2010 15:09, schrieb Nicholas A. Bellinger:
> On Wed, 2010-06-16 at 06:03 -0700, Nicholas A. Bellinger wrote:
>> On Wed, 2010-06-16 at 14:13 +0200, Kevin Wolf wrote:
>>> Am 04.06.2010 16:06, schrieb Kevin Wolf:
>>>> Am 31.05.2010 03:43, schrieb Nicholas A. Bellinger:
>>>>> From: Nicholas Bellinger <nab@linux-iscsi.org>
>>>>>
>>>>> This patch updates hw/scsi-bus.c to add PERSISTENT_RESERVE_OUT and PERSISTENT_RESERVE_IN
>>>>> case in scsi_req_length() to extra the incoming buffer length into SCSIRequest->cmd.xfer,
>>>>> and adds a second PERSISTENT_RESERVE_OUT case in scsi_req_xfer_mode() in order to properly
>>>>> set SCSI_XFER_TO_DEV for WRITE data.
>>>>>
>>>>> Tested with Linux KVM guests and Megasas 8708EM2 HBA emulation and TCM_Loop target ports.
>>>>>
>>>>> Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>
>>>>> ---
>>>>>  hw/scsi-bus.c |    5 +++++
>>>>>  1 files changed, 5 insertions(+), 0 deletions(-)
>>>>>
>>>>> diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
>>>>> index b8e4b71..75ec74e 100644
>>>>> --- a/hw/scsi-bus.c
>>>>> +++ b/hw/scsi-bus.c
>>>>> @@ -325,6 +325,10 @@ static int scsi_req_length(SCSIRequest *req, uint8_t *cmd)
>>>>>      case INQUIRY:
>>>>>          req->cmd.xfer = cmd[4] | (cmd[3] << 8);
>>>>>          break;
>>>>> +    case PERSISTENT_RESERVE_OUT:
>>>>> +    case PERSISTENT_RESERVE_IN:
>>>>> +        req->cmd.xfer = cmd[8] | (cmd[7] << 8);
>>>>
>>>> Maybe I'm missing something, but isn't exactly the same value set in the
>>>> switch block above? (for cmd[0] >> 5 == 2)
>>>
>>> Nicholas? This isn't applied yet because I'm waiting for your answer.
>>>
>>> Is there a reason why it makes sense to do it explicitly here instead
>>> using the generic code a few lines above? I think the same applied to
>>> patch 2/2.
>>
>> Hi Kevin,
>>
>> I just tested this again and you are correct, the reassignment of
>> req->cmd.xfer for PR and Maintence CDBs is unnecessary in
>> scsi_req_length().  I will go ahead and drop part this from my tree.
>>
>> Please let me know if you would like me to resend the patch series.
>>
> 
> Actually, I should mention that I have only tested this with TYPE_DISK
> so far.  I think we do still need the Maintenance CDBs length
> reassignment for MMC and TYPE_ROM from the second patch:
> 
> diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
> index 75ec74e..7d80405 100644
> --- a/hw/scsi-bus.c
> +++ b/hw/scsi-bus.c
> @@ -329,6 +329,17 @@ static int scsi_req_length(SCSIRequest *req, uint8_t *cmd)
>      case PERSISTENT_RESERVE_IN:
>          req->cmd.xfer = cmd[8] | (cmd[7] << 8);
>          break;
> +    case MAINTENANCE_OUT:
> +    case MAINTENANCE_IN:
> +        if (req->dev->type != TYPE_ROM) {
> +            /* Used for MI_REPORT_TARGET_PGS, MO_SET_TARGET_PGS et al. */
> +            req->cmd.xfer = cmd[9] | (cmd[8] << 8) |
> +                            (cmd[7] << 16) | (cmd[6] << 24);
> +        } else {
> +            /* GPCMD_REPORT_KEY and GPCMD_SEND_KEY from multi media commands */
> +            req->cmd.xfer = cmd[9] | (cmd[8] << 8);
> +        }
> +        break;
>      }
> 
> Do you have a problem with leaving this reassignment in for TYPE_ROM..?

In a place where it's different from what was set above, I'm not against
it. I just wanted to reduce some code duplication.

Kevin