[PATCH] 8139cp: Prevent dev_close/cp_interrupt race on MTU change

commit:  cb64edb6b89491edfdbae52ba7db9a8b8391d339 upstream

Above commit may introduce a race between cp_interrupt and dev_close
/ change MTU / dev_open up state. Changes cp_interrupt to tolerate
this.  Change spin_locking in cp_interrupt to avoid possible
but unobserved race.

Reported-by: "Francois Romieu" <romieu@fr.zoreil.com>

Tested on virtual hardware, Tx MTU size up to 4096, max tx payload
    was ping -s 4068 for MTU of 4096. No real hardware, need test
    assist.

Signed-off-by: "John Greene" <redacted>
CC: "David S. Miller" <davem@davemloft.net>
CC: "David Woodhouse" <redacted>
---
 drivers/net/ethernet/realtek/8139cp.c | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/realtek/8139cp.c b/drivers/net/ethernet/realtek/8139cp.c
index 0da3f5e..585c35c 100644
--- a/drivers/net/ethernet/realtek/8139cp.c
+++ b/drivers/net/ethernet/realtek/8139cp.c

@@ -577,28 +577,30 @@ static irqreturn_t cp_interrupt (int irq, void *dev_instance)
 {
 	struct net_device *dev = dev_instance;
 	struct cp_private *cp;
+	int handled = 0;
 	u16 status;
 
 	if (unlikely(dev == NULL))
 		return IRQ_NONE;
 	cp = netdev_priv(dev);
 
+	spin_lock(&cp->lock);
+
 	status = cpr16(IntrStatus);
 	if (!status || (status == 0xFFFF))
-		return IRQ_NONE;
+		goto out_unlock;
+
+	handled = 1;
 
 	netif_dbg(cp, intr, dev, "intr, status %04x cmd %02x cpcmd %04x\n",
 		  status, cpr8(Cmd), cpr16(CpCmd));
 
 	cpw16(IntrStatus, status & ~cp_rx_intr_mask);
 
-	spin_lock(&cp->lock);
-
 	/* close possible race's with dev_close */
 	if (unlikely(!netif_running(dev))) {
 		cpw16(IntrMask, 0);
-		spin_unlock(&cp->lock);
-		return IRQ_HANDLED;
+		goto out_unlock;
 	}
 
 	if (status & (RxOK | RxErr | RxEmpty | RxFIFOOvr))

@@ -612,7 +614,6 @@ static irqreturn_t cp_interrupt (int irq, void *dev_instance)
 	if (status & LinkChg)
 		mii_check_media(&cp->mii_if, netif_msg_link(cp), false);
 
-	spin_unlock(&cp->lock);
 
 	if (status & PciErr) {
 		u16 pci_status;

@@ -625,7 +626,10 @@ static irqreturn_t cp_interrupt (int irq, void *dev_instance)
 		/* TODO: reset hardware */
 	}
 
-	return IRQ_HANDLED;
+out_unlock:
+	spin_unlock(&cp->lock);
+
+	return IRQ_RETVAL(handled);
 }
 
 #ifdef CONFIG_NET_POLL_CONTROLLER

-- 
1.7.11.7

From: John Greene <redacted>
Date: Wed, 19 Dec 2012 14:47:48 -0500

commit:  cb64edb6b89491edfdbae52ba7db9a8b8391d339 upstream

Above commit may introduce a race between cp_interrupt and dev_close
/ change MTU / dev_open up state. Changes cp_interrupt to tolerate
this.  Change spin_locking in cp_interrupt to avoid possible
but unobserved race.

Reported-by: "Francois Romieu" <romieu@fr.zoreil.com>

Tested on virtual hardware, Tx MTU size up to 4096, max tx payload
    was ping -s 4068 for MTU of 4096. No real hardware, need test
    assist.

Signed-off-by: "John Greene" <redacted>

You sent this as a "request for testing" last week, but I saw
no testing on real hardware whatsoever.

On Wed, 2012-12-19 at 12:40 -0800, David Miller wrote:

You sent this as a "request for testing" last week, but I saw
no testing on real hardware whatsoever.

Thanks for the reminder :)

Seems to work fine here. I haven't confirmed whether I actually see the
race or not but changing MTU on a live device works fine, even when it's
being ping-flooded.

Tested-by: David Woodhouse <redacted>

-- 
dwmw2

From: David Woodhouse <dwmw2@infradead.org>
Date: Wed, 19 Dec 2012 20:55:47 +0000

On Wed, 2012-12-19 at 12:40 -0800, David Miller wrote:

quoted

You sent this as a "request for testing" last week, but I saw
no testing on real hardware whatsoever.

Thanks for the reminder :)

Seems to work fine here. I haven't confirmed whether I actually see the
race or not but changing MTU on a live device works fine, even when it's
being ping-flooded.

Tested-by: David Woodhouse <redacted>

That's more like it, applied, thanks everyone. :-)

On 12/19/2012 03:55 PM, David Woodhouse wrote:

On Wed, 2012-12-19 at 12:40 -0800, David Miller wrote:

quoted

You sent this as a "request for testing" last week, but I saw
no testing on real hardware whatsoever.

Thanks for the reminder :)

Seems to work fine here. I haven't confirmed whether I actually see the
race or not but changing MTU on a live device works fine, even when it's
being ping-flooded.

Tested-by: David Woodhouse <redacted>

Thanks all. Happy holidays!

-- 
John Greene

On 12/19/2012 02:47 PM, John Greene wrote:

Can I get a quick update on this? Seems to have fallen thru the cracks. 
  Thanks.

-- 
John Greene