2017-06-20 09:23:36

by Hugh Dickins

Subject: [PATCH] mm: fix new crash in unmapped_area_topdown()

Trinity gets kernel BUG at mm/mmap.c:1963! in about 3 minutes of
mmap testing. That's the VM_BUG_ON(gap_end < gap_start) at the
end of unmapped_area_topdown(). Linus points out how MAP_FIXED
(which does not have to respect our stack guard gap intentions)
could result in gap_end below gap_start there. Fix that, and
the similar case in its alternative, unmapped_area().

Cc: [email protected]
Fixes: 1be7107fbe18 ("mm: larger stack guard gap, between vmas")
Reported-by: Dave Jones <[email protected]>
Debugged-by: Linus Torvalds <[email protected]>
Signed-off-by: Hugh Dickins <[email protected]>
---

mm/mmap.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

--- 4.12-rc6/mm/mmap.c 2017-06-19 09:06:10.035407505 -0700
+++ linux/mm/mmap.c 2017-06-19 21:09:28.616707311 -0700
@@ -1817,7 +1817,8 @@ unsigned long unmapped_area(struct vm_un
/* Check if current node has a suitable gap */
if (gap_start > high_limit)
return -ENOMEM;
- if (gap_end >= low_limit && gap_end - gap_start >= length)
+ if (gap_end >= low_limit &&
+ gap_end > gap_start && gap_end - gap_start >= length)
goto found;

/* Visit right subtree if it looks promising */
@@ -1920,7 +1921,8 @@ unsigned long unmapped_area_topdown(stru
gap_end = vm_start_gap(vma);
if (gap_end < low_limit)
return -ENOMEM;
- if (gap_start <= high_limit && gap_end - gap_start >= length)
+ if (gap_start <= high_limit &&
+ gap_end > gap_start && gap_end - gap_start >= length)
goto found;

/* Visit left subtree if it looks promising */

2017-06-21 09:35:53

by Michal Hocko

Subject: Re: [PATCH] mm: fix new crash in unmapped_area_topdown()

On Tue 20-06-17 02:10:44, Hugh Dickins wrote:
> Trinity gets kernel BUG at mm/mmap.c:1963! in about 3 minutes of
> mmap testing. That's the VM_BUG_ON(gap_end < gap_start) at the
> end of unmapped_area_topdown(). Linus points out how MAP_FIXED
> (which does not have to respect our stack guard gap intentions)
> could result in gap_end below gap_start there. Fix that, and
> the similar case in its alternative, unmapped_area().

I finally found some more time to look at this and the fix looks good to
me. I have checked and it seems to be complete. I was even wondering
wheter we should warn when MAP_FIXED is too close to a stack area. Maybe
somebody does that intentionally, though (I can certainly imagine
PROT_NONE mapping under the stack to protect from {over,under}flows).

> Cc: [email protected]
> Fixes: 1be7107fbe18 ("mm: larger stack guard gap, between vmas")
> Reported-by: Dave Jones <[email protected]>
> Debugged-by: Linus Torvalds <[email protected]>
> Signed-off-by: Hugh Dickins <[email protected]>

Anyway feel free to add
Acked-by: Michal Hocko <[email protected]>

> ---
>
> mm/mmap.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> --- 4.12-rc6/mm/mmap.c 2017-06-19 09:06:10.035407505 -0700
> +++ linux/mm/mmap.c 2017-06-19 21:09:28.616707311 -0700
> @@ -1817,7 +1817,8 @@ unsigned long unmapped_area(struct vm_un
> /* Check if current node has a suitable gap */
> if (gap_start > high_limit)
> return -ENOMEM;
> - if (gap_end >= low_limit && gap_end - gap_start >= length)
> + if (gap_end >= low_limit &&
> + gap_end > gap_start && gap_end - gap_start >= length)
> goto found;
>
> /* Visit right subtree if it looks promising */
> @@ -1920,7 +1921,8 @@ unsigned long unmapped_area_topdown(stru
> gap_end = vm_start_gap(vma);
> if (gap_end < low_limit)
> return -ENOMEM;
> - if (gap_start <= high_limit && gap_end - gap_start >= length)
> + if (gap_start <= high_limit &&
> + gap_end > gap_start && gap_end - gap_start >= length)
> goto found;
>
> /* Visit left subtree if it looks promising */

--
Michal Hocko
SUSE Labs