> > Create master device without partition when
> > CONFIG_MTD_PARTITIONED_MASTER flag is unset.
> >
> > This streamlines device tree and allows to anchor
> > runtime power management on master device in all cases.
>
> Please explain in more detail why this is needed.
> If this change makes the overall situation better and breaks
> no userspace, I'm happy. :-)
>
The rest of the series is a driver that need runtime power management.
Absence of the master device breaks power management logic,
as kernel automatically propagates state from children to parent.
I initially hooked runtime_pm on chip auxiliary device, but this is a hack,
not a solution.
> From skimming over the patch I think the mtd_master device completely
> useless for userspace, right?
>
As of today, yes.
In future we can add curated sysfs with common parameters for all partitions,
so user-space can query master device instead of one of the partitions.
> > int add_mtd_device(struct mtd_info *mtd)
> > {
> > struct device_node *np = mtd_get_of_node(mtd);
> > struct mtd_info *master = mtd_get_master(mtd);
> > struct mtd_notifier *not;
> > + bool partitioned = true;
> > int i, error, ofidx;
> >
> > /*
> > @@ -655,6 +678,11 @@ int add_mtd_device(struct mtd_info *mtd)
> > if (WARN_ONCE(mtd->dev.type, "MTD already registered\n"))
> > return -EEXIST;
> >
> > + if ((master == mtd) &&
> !IS_ENABLED(CONFIG_MTD_PARTITIONED_MASTER)) {
> > + partitioned = false;
> > + pr_debug("mtd: unpartitioned master %s\n", mtd->name);
> > + }
>
> So, when CONFIG_MTD_PARTITIONED_MASTER is not set and a driver like
> MTDRAM
> does mtd_device_register(mtd, NULL, 0) we end up here with partitioned =
> false,
> and allocate just a master device but no real mtd because with zero
> parts the mtd_device_parse_register() function will not call add_mtd_device().
> :-(
Yep, missed this.
I think that we can create master after partitions and condition it
on master not created in partition phase.
>
> Thanks,
> //Richard
- -
Thanks,
Sasha
