Re: [PATCH v3] arm64/mm: fix incorrect file_map_count for invalid pmd

[Will Deacon]

On Mon, Nov 28, 2022 at 05:26:14PM +0100, Denys Vlasenko wrote:
> On 11/21/22 19:18, Will Deacon wrote:
> > On Mon, Nov 21, 2022 at 03:36:08PM +0800, Liu Shixin wrote:
> > > The page table check trigger BUG_ON() unexpectedly when split hugepage:
> > > 
> > >   ------------[ cut here ]------------
> > >   kernel BUG at mm/page_table_check.c:119!
> > >   Internal error: Oops - BUG: 00000000f2000800 [#1] SMP
> > >   Dumping ftrace buffer:
> > >      (ftrace buffer empty)
> > >   Modules linked in:
> > >   CPU: 7 PID: 210 Comm: transhuge-stres Not tainted 6.1.0-rc3+ #748
> > >   Hardware name: linux,dummy-virt (DT)
> > >   pstate: 20000005 (nzCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> > >   pc : page_table_check_set.isra.0+0x398/0x468
> > >   lr : page_table_check_set.isra.0+0x1c0/0x468
> > > [...]
> > >   Call trace:
> > >    page_table_check_set.isra.0+0x398/0x468
> > >    __page_table_check_pte_set+0x160/0x1c0
> > >    __split_huge_pmd_locked+0x900/0x1648
> > >    __split_huge_pmd+0x28c/0x3b8
> > >    unmap_page_range+0x428/0x858
> > >    unmap_single_vma+0xf4/0x1c8
> > >    zap_page_range+0x2b0/0x410
> > >    madvise_vma_behavior+0xc44/0xe78
> > >    do_madvise+0x280/0x698
> > >    __arm64_sys_madvise+0x90/0xe8
> > >    invoke_syscall.constprop.0+0xdc/0x1d8
> > >    do_el0_svc+0xf4/0x3f8
> > >    el0_svc+0x58/0x120
> > >    el0t_64_sync_handler+0xb8/0xc0
> > >    el0t_64_sync+0x19c/0x1a0
> > > [...]
> > > 
> > > On arm64, pmd_leaf() will return true even if the pmd is invalid due to
> > > pmd_present_invalid() check. So in pmdp_invalidate() the file_map_count
> > > will not only decrease once but also increase once. Then in set_pte_at(),
> > > the file_map_count increase again, and so trigger BUG_ON() unexpectedly.
> > > 
> > > Add !pmd_present_invalid() check in pmd_user_accessible_page() to fix the
> > > problem.
> > > 
> > > Fixes: 42b2547137f5 ("arm64/mm: enable ARCH_SUPPORTS_PAGE_TABLE_CHECK")
> > > Reported-by: Denys Vlasenko <dvlasenk@xxxxxxxxxx>
> > > Signed-off-by: Liu Shixin <liushixin2@xxxxxxxxxx>
> > > Acked-by: Pasha Tatashin <pasha.tatashin@xxxxxxxxxx>
> > > Acked-by: David Hildenbrand <david@xxxxxxxxxx>
> > > Reviewed-by: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
> > > ---
> > > v1->v2: Update comment and optimize the code by moving p?d_valid() at
> > > 	first place suggested by Mark.
> > > v2->v3: Replace pmd_valid() with pmd_present_invalid() suggested by Will.
> > > 
> > >   arch/arm64/include/asm/pgtable.h | 2 +-
> > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> > > index edf6625ce965..17afb09f386f 100644
> > > --- a/arch/arm64/include/asm/pgtable.h
> > > +++ b/arch/arm64/include/asm/pgtable.h
> > > @@ -863,7 +863,7 @@ static inline bool pte_user_accessible_page(pte_t pte)
> > >   static inline bool pmd_user_accessible_page(pmd_t pmd)
> > >   {
> > > -	return pmd_leaf(pmd) && (pmd_user(pmd) || pmd_user_exec(pmd));
> > > +	return pmd_leaf(pmd) && !pmd_present_invalid(pmd) && (pmd_user(pmd) || pmd_user_exec(pmd));
> > >   }
> > 
> > Acked-by: Will Deacon <will@xxxxxxxxxx>
> > 
> > But please see my comment on v2 about pud_user_exec() for the PUD case.
> 
> Can you be more specific? Do you ask for pud_user_exec() to be defined
> and used here? Or something else?

So we now have three patches, all from Liu, that are tripping over each
other:

1. 5b47348fc0b1 ("arm64/mm: fix incorrect file_map_count for non-leaf pmd/pud")

	Merged upstream in -rc6

2. This patch ("arm64/mm: fix incorrect file_map_count for invalid pmd")

	This could land for -rc8 (I acked it), but I'd be more comfortable
	queuing it at -rc1 seeing it as it isn't a recent regression,
	it explodes in the page-table check code and it will conflict with
	(1).

3. https://lore.kernel.org/r/20221122123137.429686-1-liushixin2@xxxxxxxxxx
   ("arm64/mm: add pud_user_exec() check in pud_user_accessible_page()")

	This was just found by inspection, so it can definitely wait for
	next time (i.e. 6.3).

> Until this patch lands, arm64 PAGE_TABLE_CHECK + THP remains broken...

It's unfortunate, but I don't think it's new breakage and it's failing a
synthetic check so it's hard to justify squeezing it in this late.

Will