2023-02-24 21:23:30

by Liam R. Howlett

Subject: [PATCH] mm/mprotect: Fix successful vma_merge() of next in do_mprotect_pkey()

If mprotect_fixup() successfully calls vma_merge() and replaces vma and
the next vma, then the tmp variable in the do_mprotect_pkey() is not
updated to point to the new vma end. This results in the loop detecting
a gap between VMAs that does not exist. Fix the faulty value of tmp by
setting it to the end location of the vma iterator at the end of the
loop.

Reported-by: Bert Karwatzki <[email protected]>
Bug: https://bugzilla.kernel.org/show_bug.cgi?id=217061
Fixes: 2286a6914c77 ("mm: change mprotect_fixup to vma iterator")
Link: https://lore.kernel.org/linux-mm/[email protected]/
Signed-off-by: Liam R. Howlett <[email protected]>
---
mm/mprotect.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/mm/mprotect.c b/mm/mprotect.c
index 1d4843c97c2a..231929f119d9 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -832,6 +832,7 @@ static int do_mprotect_pkey(unsigned long start, size_t len,
if (error)
break;

+ tmp = vma_iter_end(&vmi);
nstart = tmp;
prot = reqprot;
}
--
2.39.0

2023-02-24 22:54:18

by Bert Karwatzki

Subject: Re: [PATCH] mm/mprotect: Fix successful vma_merge() of next in do_mprotect_pkey()

Am Freitag, dem 24.02.2023 um 16:20 -0500 schrieb Liam R. Howlett:
> If mprotect_fixup() successfully calls vma_merge() and replaces vma
> and
> the next vma, then the tmp variable in the do_mprotect_pkey() is not
> updated to point to the new vma end.  This results in the loop
> detecting
> a gap between VMAs that does not exist.  Fix the faulty value of tmp
> by
> setting it to the end location of the vma iterator at the end of the
> loop.
>
> Reported-by: Bert Karwatzki <[email protected]>
> Bug: https://bugzilla.kernel.org/show_bug.cgi?id=217061
> Fixes: 2286a6914c77 ("mm: change mprotect_fixup to vma iterator")
> Link:
> https://lore.kernel.org/linux-mm/[email protected]/
> Signed-off-by: Liam R. Howlett <[email protected]>
> ---
>  mm/mprotect.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/mm/mprotect.c b/mm/mprotect.c
> index 1d4843c97c2a..231929f119d9 100644
> --- a/mm/mprotect.c
> +++ b/mm/mprotect.c
> @@ -832,6 +832,7 @@ static int do_mprotect_pkey(unsigned long start,
> size_t len,
>                 if (error)
>                         break;
>  
> +               tmp = vma_iter_end(&vmi);
>                 nstart = tmp;
>                 prot = reqprot;
>         }
I tested this in linux-next-20230224 by starting stellaris from steam
(which is the only way I managed to trigger the bug) and it fixes the
issue for me.
Bert Karwatzki