Re: [PATCH RFC v4 08/13] mm/memory_hotplug: Introduce offline_and_remove_memory()

From: Michal Hocko <mhocko@kernel.org>
To: David Hildenbrand <david@redhat.com>
Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	virtio-dev@lists.oasis-open.org,
	virtualization@lists.linux-foundation.org, kvm@vger.kernel.org,
	Andrew Morton <akpm@linux-foundation.org>,
	"Michael S . Tsirkin" <mst@redhat.com>,
	Oscar Salvador <osalvador@suse.com>,
	Pavel Tatashin <pasha.tatashin@soleen.com>,
	Wei Yang <richard.weiyang@gmail.com>,
	Dan Williams <dan.j.williams@intel.com>, Qian Cai <cai@lca.pw>
Subject: Re: [PATCH RFC v4 08/13] mm/memory_hotplug: Introduce offline_and_remove_memory()
Date: Mon, 2 Mar 2020 13:48:52 +0100	[thread overview]
Message-ID: <20200302124852.GJ4380@dhcp22.suse.cz> (raw)
In-Reply-To: <d1dbb687-7959-f4f1-6a64-33ee039782ef@redhat.com>

On Tue 25-02-20 15:27:28, David Hildenbrand wrote:
> On 25.02.20 15:11, Michal Hocko wrote:
> > On Thu 12-12-19 18:11:32, David Hildenbrand wrote:
> >> virtio-mem wants to offline and remove a memory block once it unplugged
> >> all subblocks (e.g., using alloc_contig_range()). Let's provide
> >> an interface to do that from a driver. virtio-mem already supports to
> >> offline partially unplugged memory blocks. Offlining a fully unplugged
> >> memory block will not require to migrate any pages. All unplugged
> >> subblocks are PageOffline() and have a reference count of 0 - so
> >> offlining code will simply skip them.
> >>
> >> All we need an interface to trigger the "offlining" and the removing in a
> >> single operation - to make sure the memory block cannot get onlined by
> >> user space again before it gets removed.
> > 
> > Why does that matter? Is it really likely that the userspace would
> > interfere? What would be the scenario?
> 
> I guess it's not that relevant after all (I think this comment dates
> back to the times where we didn't have try_remove_memory() and could
> actually BUG_ON() in remove_memory() if there would have been a race).
> Can drop that part.
> 
> > 
> > Or is still mostly about not requiring callers to open code this general
> > patter?
> 
> From kernel module context, I cannot get access to the actual memory
> block device (find_memory_block()) and call the device_unregister().
> 
> Especially, also the device hotplug lock is not exported. So this is a
> clean helper function to be used from kernel module context. (e.g., also
> hyper-v showed interest for using that)

Fair enough.
-- 
Michal Hocko
SUSE Labs