Re: [PATCH v5 09/10] scsi: scsi_debug: Map the region written by WRITE ATOMIC (16)

[Niklas Cassel <cassel@kernel.org>]

Hello John,

On Mon, Sep 21, 2026 at 04:50:31PM +0100, John Garry wrote:
> On 9/21/26 16:40, Niklas Cassel wrote:
> > @@ -6246,7 +6247,16 @@ static int resp_atomic_write(struct scsi_cmnd *scp,
> >		}
> >	}
> > +	if (lbp)
> > +		sdeb_meta_write_lock(sip);
> > +
> >	ret = do_device_access(sip, scp, 0, lba, len, 0, true, true);
> > +	if (unlikely(lbp))
>
> Is this supposed to be called if ret <= 0?
>
> > +		map_region(sip, lba, len);
> > +
> > +	if (lbp)

It does not have to be skipped: SBC-6 4.7.4.6.2 allows a device to map a
deallocated LBA at any time, and resp_write_dt0() and resp_write_scat()
also call map_region() without looking at the result, so I will keep it
unconditional.

Looking at this, I did however find another bug in resp_atomic_write().
While scsi_debug does fail a short WRITE ATOMIC (16) with DID_ERROR,
a short write will currently modify the logical blocks up to the short
write. I will add a fix for this in my series.


Kind regards,
Niklas