Re: [PATCH v6 06/10] mm/memory_hotplug: add offline_and_remove_memory_ranges()

On Thu, Jul 09, 2026 at 04:45:51PM +0800, Richard Cheng wrote:
> On Tue, Jun 30, 2026 at 05:18:38PM +0800, Gregory Price wrote:
> > +/**
> > + * offline_and_remove_memory_ranges - offline and remove multiple memory ranges
> > + * @ranges: array of physical address ranges to offline and remove
> > + * @nr_ranges: number of entries in @ranges
> > + *
> > + * Offline and remove several memory ranges as one operation, serialized
> > + * against other hotplug operations by a single lock_device_hotplug().
> > + *
> > + * This offlines all ranges before removing any of them.  If offlining any
> > + * range fails, the entire process is reverted and nothing is removed.
> > + * This provides a fully atomic semantic for unplugging an entire device.
> > + *
> > + * Each range must be memory-block aligned in start and size.
> > + *
> > + * Return: 0 on success, negative errno otherwise.  On failure no range has
> > + * been removed.
> > + */
> 
> I think this can return 1, and it shouldn't.
> device_offline() returns 1 when a block is already offline, and phase 1 passes that value through as-is.
> 
> Easy to hit with patch 0, offline one memory block via memoryN/state, then write
> "unplugged" to daxX.Y/state. The store returns 1, userspace treats it as a partial write of 1 byte,
> and retries the write with the rest of the string.
> 
> Maybe
> """
> if (rc > 0)
>     rc = -EBUSY;
> """

Oh weird, I thought I'd solved this problem at some point.  I must have
botched a rebase or something.  Good catch.

I'd originaly had some tests that race memoryN/state and dax/state, but
I dropped them because it caused long testing cycles (mostly because i
was testing force unplugging).  Maybe I'll re-introduce some of these
partial tests and just avoid the destructive ones.

~Gregory