bonding: fix rtnl locking merge error
authorJay Vosburgh <fubar@us.ibm.com>
Tue, 6 Nov 2007 21:33:28 +0000 (13:33 -0800)
committerJeff Garzik <jeff@garzik.org>
Sat, 10 Nov 2007 09:25:14 +0000 (04:25 -0500)
Looks like I incorrectly merged one of the rtnl lock changes,
so that one function, bonding_show_active_slave, held rtnl but didn't
release it, and another, bonding_store_active_slave, never held rtnl but
did release it.

Fixed so the first function doesn't mess with rtnl, and the
second correctly acquires and releases rtnl.

Bug reported by Moni Shoua <monis@voltaire.com>

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
drivers/net/bonding/bond_sysfs.c

index 7a06ade..b29330d 100644 (file)
@@ -1193,8 +1193,6 @@ static ssize_t bonding_show_active_slave(struct device *d,
        struct bonding *bond = to_bond(d);
        int count;
 
-       rtnl_lock();
-
        read_lock(&bond->curr_slave_lock);
        curr = bond->curr_active_slave;
        read_unlock(&bond->curr_slave_lock);
@@ -1216,7 +1214,9 @@ static ssize_t bonding_store_active_slave(struct device *d,
         struct slave *new_active = NULL;
        struct bonding *bond = to_bond(d);
 
+       rtnl_lock();
        write_lock_bh(&bond->lock);
+
        if (!USES_PRIMARY(bond->params.mode)) {
                printk(KERN_INFO DRV_NAME
                       ": %s: Unable to change active slave; %s is in mode %d\n",