Re: scsi: scsi_debug: make timeout faults by set delay to maximum value

[JiangJianJun <jiangjianjun3@huawei.com>]

> The concern which I mentioned before was that we would have competing 
> and potentially conflicting methods to trigger and handle errors.
>
> This specifically is a change from Bart to fix my code. However I added 
> the change to check the abort result.
> 
> I think previously scsi_debug_abort() would just always report success, 
> even if scsi_debug_abort_cmnd() possibly fails, right?
> 
> Now for a fake timeout, there seems to be a problem in 
> scsi_debug_stop_cmnd() that we abort depending on the deferred type, but 
> this is not set for any fake timeout. Or - more specifically - it is not 
> reset for when the scmd is recycled. I think that we should just allow 
> the abort to be successful for that case.

I think your design is rational.

I think the method of simulating timeout is rational because discard-command is
equivalent to the queuecommand executing but returning a success status, which
is clearly not what this interface expects. It is also within expectation that
such an unrational approach might conflict with subsequent changes.

So, I tend to change the discard-command to setting a very much long delay for
schedule.