bonding: fix set mode race conditions

Changing the mode without any locking can result in multiple races (e.g.
upping a bond, enslaving/releasing). Depending on which race is hit the
impact can vary from incosistent bond state to kernel crash.
Use RTNL to synchronize the mode setting with the dangerous races.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index ea7a388f484306710a33375f3a553fd1ecc5b621..77ea237de90044f5cb43ed9ff37d7d0f4c9f2970 100644 (file)
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -316,6 +316,9 @@ static ssize_t bonding_store_mode(struct device *d,
        int new_value, ret = count;
        struct bonding *bond = to_bond(d);
 
+       if (!rtnl_trylock())
+               return restart_syscall();
+
        if (bond->dev->flags & IFF_UP) {
                pr_err("unable to update mode of %s because interface is up.\n",
                       bond->dev->name);
@@ -352,6 +355,7 @@ static ssize_t bonding_store_mode(struct device *d,
                bond->dev->name, bond_mode_tbl[new_value].modename,
                new_value);
 out:
+       rtnl_unlock();
        return ret;
 }
 static DEVICE_ATTR(mode, S_IRUGO | S_IWUSR,