On Fri 15-12-17 13:51:04, Arnd Bergmann wrote:
> When the down_read_trylock() fails, 'vma' has not been initialized
> yet, which gcc now warns about:
>
> mm/khugepaged.c: In function 'khugepaged':
> mm/khugepaged.c:1659:25: error: 'vma' may be used uninitialized in this
> function [-Werror=maybe-uninitialized]
ups missed that
>
> Presumable we are not supposed to call find_vma() without the mmap_sem
> either, so setting it to NULL for this case seems appropriate.
yes
> Fixes: 0951b59acf3a ("mm: thp: use down_read_trylock() in khugepaged to avoid
> long block")
This sha is not stable because this is a mmotm tree. I assume Andrew will fold
it to mm-thp-use-down_read_trylock-in-khugepaged-to-avoid-long-block.patch.
The patch looks good to me. I would initialize the vma in the
declaration, but that is a minor thing.
> Signed-off-by: Arnd Bergmann <[email protected]>
Acked-by: Michal Hocko <[email protected]>
> ---
> I'm not completely sure this patch is sufficient, it gets rid of
> the warning, but it would be good to have the code reviewed better
> to see if other problems remain that result from down_read_trylock()
> patch.
> ---
> mm/khugepaged.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 521b908f9600..b7e2268dfc9a 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -1677,11 +1677,10 @@ static unsigned int khugepaged_scan_mm_slot(unsigned
> int pages,
> * Don't wait for semaphore (to avoid long wait times). Just move to
> * the next mm on the list.
> */
> + vma = NULL;
> if (unlikely(!down_read_trylock(&mm->mmap_sem)))
> goto breakouterloop_mmap_sem;
> - if (unlikely(khugepaged_test_exit(mm)))
> - vma = NULL;
> - else
> + if (likely(!khugepaged_test_exit(mm)))
> vma = find_vma(mm, khugepaged_scan.address);
>
> progress++;
> --
> 2.9.0
>
--
Michal Hocko
SUSE Labs
