On Tue 10-03-20 12:46:05, David Hildenbrand wrote:

On 10.03.20 12:43, Michal Hocko wrote:

quoted

On Mon 02-03-20 14:49:37, David Hildenbrand wrote:
[...]

quoted

+static void virtio_mem_notify_going_offline(struct virtio_mem *vm,
+					    unsigned long mb_id)
+{
+	const unsigned long nr_pages = PFN_DOWN(vm->subblock_size);
+	unsigned long pfn;
+	int sb_id, i;
+
+	for (sb_id = 0; sb_id < vm->nb_sb_per_mb; sb_id++) {
+		if (virtio_mem_mb_test_sb_plugged(vm, mb_id, sb_id, 1))
+			continue;
+		/*
+		 * Drop our reference to the pages so the memory can get
+		 * offlined and add the unplugged pages to the managed
+		 * page counters (so offlining code can correctly subtract
+		 * them again).
+		 */
+		pfn = PFN_DOWN(virtio_mem_mb_id_to_phys(mb_id) +
+			       sb_id * vm->subblock_size);
+		adjust_managed_page_count(pfn_to_page(pfn), nr_pages);
+		for (i = 0; i < nr_pages; i++)
+			page_ref_dec(pfn_to_page(pfn + i));

Is there ever situation this might be a different than 1->0 transition?

Only if some other code would be taking a reference. At least not from
virtio-mem perspective.

OK, so that is essentially an error condition. I think it shouldn't go
silent and you want something like
	if (WARN_ON(!page_ref_sub_and_test(page)))
		dump_page(pfn_to_page(pfn + i), "YOUR REASON");

-- 
Michal Hocko
SUSE Labs