On 8/12/23 00:49, Bart Van Assche wrote: > On 8/10/23 17:45, Damien Le Moal wrote: >> With that, mq-deadline can even be simplified to not even look at the >> q zoned model and simply us q->limits.use_zone_write_lock to >> determine if locking a zone is needed or not. > > Hi Damien, > > I think implementing the above proposal requires splitting > 'use_zone_write_lock' into two variables: > (a) One member variable that indicates whether the zone write lock > should be used. > (b) Another member variable that indicates whether or not the LLD > preserves the order of SCSI commands. > > Member variable (b) should be set by the LLD and member variable (a) can > be set by disk_set_zoned(). > > Do you want me to implement this approach? > Looking at your v8, this seems better. I will have a more in-depth look Monday. -- Damien Le Moal Western Digital Research
