I wonder why don't you use cancel_delayed_work_sync() here (and in a
few other places), like in bond_work_cancel_all() from patch 2/4?
In bond_close(), you can't use cancel_delayed_work_sync() because you can't
unlock RTNL in it. (So, on current implementation, it becomes work cancel is
not ensur
On 15-01-2008 07:36, Makito SHIOKAWA wrote:
...
> --- a/drivers/net/bonding/bond_sysfs.c
> +++ b/drivers/net/bonding/bond_sysfs.c
> @@ -643,10 +643,8 @@ static ssize_t bonding_store_arp_interva
> "%s Disabling MII monitoring.\n",
> bond->dev->name, bond->de
delayed_work_pending() is used to check whether work is being queued or not,
when queueing or cancelling corresponding work. However,
delayed_work_pending() returns 0 when work is just running, and in that case,
work_struct will be destroyed or work won't be cancelled. So remove all
delayed_work_pe
