phylib: fix PAL state machine restart on resume
authorSimon Guinot <sguinot@lacie.com>
Mon, 13 Sep 2010 22:12:01 +0000 (22:12 +0000)
committerDavid S. Miller <davem@davemloft.net>
Tue, 14 Sep 2010 21:31:03 +0000 (14:31 -0700)
On resume, before starting the PAL state machine, check if the
adjust_link() method is well supplied. If not, this would lead to a
NULL pointer dereference in the phy_state_machine() function.

This scenario can happen if the Ethernet driver call manually the PHY
functions instead of using the PAL state machine. The mv643xx_eth driver
is a such example.

Signed-off-by: Simon Guinot <sguinot@lacie.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/phy/mdio_bus.c

index 6a6b8199a0d6862467260384e32979584f24d95a..6c58da2b882c845e453ddd7b6bb0b544f4705b83 100644 (file)
@@ -308,7 +308,7 @@ static int mdio_bus_suspend(struct device *dev)
         * may call phy routines that try to grab the same lock, and that may
         * lead to a deadlock.
         */
-       if (phydev->attached_dev)
+       if (phydev->attached_dev && phydev->adjust_link)
                phy_stop_machine(phydev);
 
        if (!mdio_bus_phy_may_suspend(phydev))
@@ -331,7 +331,7 @@ static int mdio_bus_resume(struct device *dev)
                return ret;
 
 no_resume:
-       if (phydev->attached_dev)
+       if (phydev->attached_dev && phydev->adjust_link)
                phy_start_machine(phydev, NULL);
 
        return 0;