Re: [RFC PATCH] SCSI: Correct UA being ignored when notifying of media-changed

[Mike Christie <michaelc@cs.wisc.edu>]

On 08/03/2011 05:36 PM, James Bottomley wrote:
> On Wed, 2011-08-03 at 17:31 -0500, Mike Christie wrote:
>> On 08/03/2011 05:22 PM, James Bottomley wrote:
>>> On Wed, 2011-08-03 at 09:39 -0400, Rob Evers wrote:
>>>> Unit attention commands are getting ignored when a dvd is
>>>> replaced, resulting in errors when accessing the new dvd.
>>>>
>>>> This problem occurred when a system is booted in rescue
>>>> mode via dvd and the dvd is then replaced with another dvd.
>>>> Subsequent commands to a filesystem on the dvd, such as
>>>> 'fsck' and 'restore' report errors.
>>>>
>>>> Solution suggested by Mike Christie.
>>>>
>>>> Signed-off-by: Rob Evers <revers@redhat.com>
>>>> ---
>>>>  drivers/scsi/scsi_error.c |    4 ++++
>>>>  1 files changed, 4 insertions(+), 0 deletions(-)
>>>>
>>>> diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
>>>> index a4b9cdb..725cc11 100644
>>>> --- a/drivers/scsi/scsi_error.c
>>>> +++ b/drivers/scsi/scsi_error.c
>>>> @@ -294,6 +294,10 @@ static int scsi_check_sense(struct scsi_cmnd *scmd)
>>>>  		 */
>>>>  		if (scmd->device->expecting_cc_ua) {
>>>>  			scmd->device->expecting_cc_ua = 0;
>>>> +				if (sdev->removable) {
>>>> +					sdev->changed = 1;
>>>> +					return SUCCESS;
>>>> +				}
>>>
>>> Are you sure about this?  We only set expecting_cc_ua if the device was
>>> reset in error handling ... and there we definitely *don't* want to
>>> trigger a medium change.
>>>
>>
>> What is happening is that we have a quirky device, and when you switch
>> CD/DVDs the first command times out, so the scsi error handler runs.
>> That is why the above code was modified. Before the patch the scsi layer
>> would just retry the IO, and sr would not see the sense and so it would
>> not detect the media changed. So the above code was meant to work like
>> UA code in scsi_io_completion.
> 
> The SCSI standard requires all SCSI devices to return UA after a reset
> (that's the reason for the code).  So, if I put this patch in, every
> removable device would be flagged as having changed media if we go into
> an error handling situation and reset it.  I don't think that's a good
> idea to fix a single non-standard CD.
> 

Yeah, ok. Agree with you.