Re: [PATCH] ksm: fix race between ksmd and exiting task

[Andrea Righi]

On Thu, Jun 02, 2011 at 10:35:49AM -0700, Chris Wright wrote:
> Andrea Righi reported a case where an exiting task can race against
> ksmd.
> 
> ksm_scan.mm_slot == the only registered mm
> CPU 1 (bug program)		CPU 2 (ksmd)
>  				list_empty() is false
> lock
> ksm_scan.mm_slot
> list_del
> unlock
>  				slot == &ksm_mm_head (but list is now empty_)
> 
> Close this race by revalidating that the new slot is not simply the list
> head again.

I confirm this fixes the problem on my side.

Tested-by: Andrea Righi <andrea@xxxxxxxxxxxxxxx>

> 
> Reported-by: Andrea Righi <andrea@xxxxxxxxxxxxxxx>
> Cc: Hugh Dickins <hughd@xxxxxxxxxx>
> Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> Signed-off-by: Chris Wright <chrisw@xxxxxxxxxxxx>
> ---
>  mm/ksm.c |    3 +++
>  1 files changed, 3 insertions(+), 0 deletions(-)
> 
> diff --git a/mm/ksm.c b/mm/ksm.c
> index 942dfc7..0373ce4 100644
> --- a/mm/ksm.c
> +++ b/mm/ksm.c
> @@ -1301,6 +1301,9 @@ static struct rmap_item *scan_get_next_rmap_item(struct page **page)
>  		slot = list_entry(slot->mm_list.next, struct mm_slot, mm_list);
>  		ksm_scan.mm_slot = slot;
>  		spin_unlock(&ksm_mmlist_lock);
> +		/* We raced against exit of last slot on the list */
> +		if (slot == &ksm_mm_head)
> +			return NULL;
>  next_mm:
>  		ksm_scan.address = 0;
>  		ksm_scan.rmap_list = &slot->rmap_list;
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/