[Drbd-dev] [PATCH] block/drbd: Fix a sleep-in-atomic bug in drbd_bcast_event

Roland Kammerer roland.kammerer at linbit.com
Mon Oct 9 16:08:27 CEST 2017


On Wed, Oct 04, 2017 at 09:33:18AM +0800, Jia-Ju Bai wrote:
> The driver may sleep under a RCU lock, and the function call path is:
> drbd_sync_handshake (acquire the RCU lock)
>   drbd_asb_recover_1p
>     drbd_khelper
>       drbd_bcast_event
>         genlmsg_new(GFP_NOIO) --> may sleep
> 
> To fix it, GFP_NOIO is replaced with GFP_ATOMIC.
> This bug is found by my static analysis tool and my code review.
> 
> 
> Signed-off-by: Jia-Ju Bai <baijiaju1990 at 163.com>
> ---
>  drivers/block/drbd/drbd_nl.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
> index a12f77e..713c965 100644
> --- a/drivers/block/drbd/drbd_nl.c
> +++ b/drivers/block/drbd/drbd_nl.c
> @@ -4537,7 +4537,7 @@ void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib)
>  	int err = -ENOMEM;
>  
>  	seq = atomic_inc_return(&drbd_genl_seq);
> -	msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO);
> +	msg = genlmsg_new(NLMSG_GOODSIZE, GFP_ATOMIC);
>  	if (!msg)
>  		goto failed;

Hi Jia-Ju,

thanks for reporting that. I fixed that by narrowing the scope of the
rcu_lock in drbd_sync_handshake() while keeping the code down the rabbit
hole untouched. This has the benefit that it fixes all your 3 cases.

As usual, we will send that upstream soon.

Thanks, rck


More information about the drbd-dev mailing list