Re: [PATCH] scsi: sd: add runtime pm to open / release

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Wed, 2020-07-29 at 14:25 -0400, Alan Stern wrote:
> On Wed, Jul 29, 2020 at 06:43:48PM +0200, Martin Kepplinger wrote:
[...]
> > > > --- a/drivers/scsi/scsi_error.c
> > > > +++ b/drivers/scsi/scsi_error.c
> > > > @@ -554,16 +554,8 @@ int scsi_check_sense(struct scsi_cmnd
> > > > *scmd)
> > > >                  * so that we can deal with it there.
> > > >                  */
> > > >                 if (scmd->device->expecting_cc_ua) {
> > > > -                       /*
> > > > -                        * Because some device does not queue
> > > > unit
> > > > -                        * attentions correctly, we carefully
> > > > check
> > > > -                        * additional sense code and qualifier
> > > > so as
> > > > -                        * not to squash media change unit
> > > > attention.
> > > > -                        */
> > > > -                       if (sshdr.asc != 0x28 || sshdr.ascq !=
> > > > 0x00)
> > > > {
> > > > -                               scmd->device->expecting_cc_ua =
> > > > 0;
> > > > -                               return NEEDS_RETRY;
> > > > -                       }
> > > > +                       scmd->device->expecting_cc_ua = 0;
> > > > +                       return NEEDS_RETRY;
> > > 
> > > Well, yes, but you can't do this because it would lose us media
> > > change events in the non-suspend/resume case which we really
> > > don't want.  That's why I was suggesting a new flag.
> > > 
> > > James
> > 
> > also if I set expecting_cc_ua in resume() only, like I did?
> 
> That wouldn't make any difference.  The information sent by your
> card reader has sshdr.asc == 0x28 and sshdr.ascq == 0x00 (you can see
> it in the log).  So because of the code here in scsi_check_sense(),
> which you can't change, the Unit Attention sent by the card reader
> would not be  retried even if you do set the flag in resume().

But if we had a new flag, like expecting_media_change, you could set
that in resume and we could condition the above test in the code on it
and reset it and do a retry if it gets set.  I'm not saying we have to
do it this way, but it sounds like we have to do something in the
kernel, so I think the flag will become necessary but there might be a
bit of policy based dance around how it gets set in the kernel (to
avoid losing accurate media change events).

James