Re: [PATCH v2 2/2] mm: fix initialization of struct page for holes in memory layout

[Mike Rapoport <rppt@kernel.org>]

On Mon, Jan 11, 2021 at 10:06:43AM -0500, Qian Cai wrote:
> On Sun, 2021-01-10 at 17:39 +0200, Mike Rapoport wrote:
> > On Wed, Jan 06, 2021 at 04:04:21PM -0500, Qian Cai wrote:
> > > On Wed, 2021-01-06 at 10:05 +0200, Mike Rapoport wrote:
> > > > I think we trigger PF_POISONED_CHECK() in PageSlab(), then
> > > > fffffffffffffffe
> > > > is "accessed" from VM_BUG_ON_PAGE().
> > > > 
> > > > It seems to me that we are not initializing struct pages for holes at the
> > > > node
> > > > boundaries because zones are already clamped to exclude those holes.
> > > > 
> > > > Can you please try to see if the patch below will produce any useful info:
> > > 
> > > [    0.000000] init_unavailable_range: spfn: 8c, epfn: 9b, zone: DMA, node:
> > > 0
> > > [    0.000000] init_unavailable_range: spfn: 1f7be, epfn: 1f9fe, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 28784, epfn: 288e4, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 298b9, epfn: 298bd, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 29923, epfn: 29931, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 29933, epfn: 29941, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 29945, epfn: 29946, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 29ff9, epfn: 2a823, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 33a23, epfn: 33a53, zone:
> > > DMA32, node: 0
> > > [    0.000000] init_unavailable_range: spfn: 78000, epfn: 100000, zone:
> > > DMA32, node: 0
> > > ...
> > > [  572.222563][ T2302] kpagecount_read: pfn 47f380 is poisoned
> > ...
> > > [  590.570032][ T2302] kpagecount_read: pfn 47ffff is poisoned
> > > [  604.268653][ T2302] kpagecount_read: pfn 87ff80 is poisoned
> > ...
> > > [  604.611698][ T2302] kpagecount_read: pfn 87ffbc is poisoned
> > > [  617.484205][ T2302] kpagecount_read: pfn c7ff80 is poisoned
> > ...
> > > [  618.212344][ T2302] kpagecount_read: pfn c7ffff is poisoned
> > > [  633.134228][ T2302] kpagecount_read: pfn 107ff80 is poisoned
> > ...
> > > [  633.874087][ T2302] kpagecount_read: pfn 107ffff is poisoned
> > > [  647.686412][ T2302] kpagecount_read: pfn 147ff80 is poisoned
> > ...
> > > [  648.425548][ T2302] kpagecount_read: pfn 147ffff is poisoned
> > > [  663.692630][ T2302] kpagecount_read: pfn 187ff80 is poisoned
> > ...
> > > [  664.432671][ T2302] kpagecount_read: pfn 187ffff is poisoned
> > > [  675.462757][ T2302] kpagecount_read: pfn 1c7ff80 is poisoned
> > ...
> > > [  676.202548][ T2302] kpagecount_read: pfn 1c7ffff is poisoned
> > > [  687.121605][ T2302] kpagecount_read: pfn 207ff80 is poisoned
> > ...
> > > [  687.860981][ T2302] kpagecount_read: pfn 207ffff is poisoned
> > 
> > The e820 map has a hole near the end of each node and these holes are not
> > initialized with init_unavailable_range() after it was interleaved with
> > memmap initialization because such holes are not accounted by
> > zone->spanned_pages.
> > 
> > Yet, I'm still cannot really understand how this never triggered 
> > 
> > 	VM_BUG_ON_PAGE(!zone_spans_pfn(page_zone(page), pfn), page);
> > 
> > before v5.7 as all the struct pages for these holes would have zone=0 and
> > node=0 ... 
> > 
> > @Qian, can you please boot your system with memblock=debug and share the
> > logs?
> > 
> 
> http://people.redhat.com/qcai/memblock.txt

Thanks!

So, we have these large allocations for the memory maps:

memblock_alloc_exact_nid_raw: 266338304 bytes align=0x200000 nid=0 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x000000046f400000-0x000000047f1fffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=1 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x000000086fe00000-0x000000087fdfffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=2 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x0000000c6fe00000-0x0000000c7fdfffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=3 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x000000106fe00000-0x000000107fdfffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=4 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x000000146fe00000-0x000000147fdfffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=5 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x000000186fe00000-0x000000187fdfffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=6 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x0000001c6fe00000-0x0000001c7fdfffff] memblock_alloc_range_nid+0x108/0x1b6
memblock_alloc_exact_nid_raw: 268435456 bytes align=0x200000 nid=7 from=0x0000000001000000 max_addr=0x0000000000000000 sparse_init_nid+0x13b/0x519
memblock_reserve: [0x000000206fc00000-0x000000207fbfffff] memblock_alloc_range_nid+0x108/0x1b6

that will be always next to the end of each node and so the last several
pageblocks in a node will never be used by the page allocator.
That masks wrong zone=0 links in the pages corresponding to the holes near
each node and the VM_BUG_ON_PAGE(!zone_spans_pfn) never triggers.

I'm going to send v3 soon that should take better care of the zone links.

--
Sincerely yours,
Mike.