On Mon, Jul 27, 2020 at 08:59:50PM +0800, Hillf Danton wrote: > Can you elaborate on the difference between the two dumps? You didn't trim anything, so I have no idea which two dumps you mean. I'll annotate below ... > > > On Sun, Jul 26, 2020 at 05:49:04PM +0100, Matthew Wilcox wrote: > > > > 1457 086 (20181): drop_caches: 3 > > > > 1457 page:00000000a216ae9a refcount:2 mapcount:0 mapping:000000009ba7bfed index:0x2227 pfn:0x229e7 > > > > 1457 aops:def_blk_aops ino:0 > > > > 1457 flags: 0x4000000000002030(lru|active|private) > > > > 1457 raw: 4000000000002030 fffff5b4416b5a48 fffff5b4408a7988 ffff9e9c34848578 > > > > 1457 raw: 0000000000002227 ffff9e9bd18f0d00 00000002ffffffff 0000000000000000 > > > > 1457 page dumped because: not locked > > > > 1457 swap entry 30.229e7 This is a dump of the page that was found when looking up the migration entry. > On Mon, 27 Jul 2020 13:03:10 +0100 Matthew Wilcox wrote: > > It's not mapped with a PMD. I tweaked my debugging slightly: > > > > static inline swp_entry_t make_migration_entry(struct page *page, int write) > > { > > - BUG_ON(!PageLocked(compound_head(page))); > > + VM_BUG_ON_PAGE(!PageLocked(page), page); > > > > +if (PageHead(page)) dump_page(page, "make entry"); > > +if (PageTail(page)) printk("pfn %lx order %d\n", page_to_pfn(page), thp_order(thp_head(page))); > > > > 1523 page:0000000006f62206 refcount:490 mapcount:1 mapping:0000000000000000 index:0x562b12a00 pfn:0x1dc00 > > 1523 head:0000000006f62206 order:9 compound_mapcount:0 compound_pincount:0 > > 1523 anon flags: 0x400000000009003d(locked|uptodate|dirty|lru|active|head|swapbacked) > > 1523 raw: 400000000009003d ffffecfd41301308 ffffecfd41b08008 ffff9e9971c00059 > > 1523 raw: 0000000562b12a00 0000000000000000 000001ea00000000 0000000000000000 > > 1523 page dumped because: make entry This is dumping the page when we create the entry. For completeness, here's the page that we find from the same run. 1523 page:00000000a18100e6 refcount:0 mapcount:0 mapping:0000000000000000 index:0x1 pfn:0x1ddde 1523 flags: 0x4000000000000000() 1523 raw: 4000000000000000 dead000000000100 dead000000000122 0000000000000000 1523 raw: 0000000000000001 0000000000000000 00000000ffffffff 0000000000000000 1523 page dumped because: not locked (an order-9 page will occupy PFNs 0x1dc00-0x1ddff) It's clearly been freed and is still sitting on the per-CPU free list. I've also seen them as PageBuddy and, as in the first example above, reallocated to a different user.