drivers-net-at91-make-mdio-protection-rt-safe.patch

From: Thomas Gleixner <tglx@linutronix.de>
Date: Tue, 17 Nov 2009 12:02:43 +0100
Subject: drivers: net: at91_ether: Make mdio protection -rt safe

Neither the phy interrupt nor the timer callback which updates the
link status in absense of a phy interrupt are taking lp->lock which
serializes the MDIO access. This works on mainline as at91 is an UP
machine. On preempt-rt the timer callback can run even in the
spin_lock_irq(&lp->lock) protected code pathes because spin_lock_irq
is neither disabling interrupts nor disabling preemption.

Fix this by adding proper locking to at91ether_phy_interrupt() and
at91_check_ether() which serializes the access on -rt.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

---
 drivers/net/ethernet/cadence/at91_ether.c |    5 +++++
 1 file changed, 5 insertions(+)

Index: linux-stable/drivers/net/ethernet/cadence/at91_ether.c
===================================================================
--- linux-stable.orig/drivers/net/ethernet/cadence/at91_ether.c
+++ linux-stable/drivers/net/ethernet/cadence/at91_ether.c
@@ -199,7 +199,9 @@ static irqreturn_t at91ether_phy_interru
 	struct net_device *dev = (struct net_device *) dev_id;
 	struct at91_private *lp = netdev_priv(dev);
 	unsigned int phy;
+	unsigned long flags;
 
+	spin_lock_irqsave(&lp->lock, flags);
 	/*
 	 * This hander is triggered on both edges, but the PHY chips expect
 	 * level-triggering.  We therefore have to check if the PHY actually has
@@ -241,6 +243,7 @@ static irqreturn_t at91ether_phy_interru
 
 done:
 	disable_mdi(lp);
+	spin_unlock_irqrestore(&lp->lock, flags);
 
 	return IRQ_HANDLED;
 }
@@ -397,9 +400,11 @@ static void at91ether_check_link(unsigne
 	struct net_device *dev = (struct net_device *) dev_id;
 	struct at91_private *lp = netdev_priv(dev);
 
+	spin_lock_irq(&lp->lock);
 	enable_mdi(lp);
 	update_linkspeed(dev, 1);
 	disable_mdi(lp);
+	spin_unlock_irq(&lp->lock);
 
 	mod_timer(&lp->check_timer, jiffies + LINK_POLL_INTERVAL);
 }