]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commitdiff
e1000: fix tx waking queue after queue stopped during shutdown
authorJesse Brandeburg <jesse.brandeburg@intel.com>
Fri, 25 Sep 2009 12:18:07 +0000 (12:18 +0000)
committerDavid S. Miller <davem@davemloft.net>
Sun, 27 Sep 2009 03:15:42 +0000 (20:15 -0700)
This fix closes a race where the adapter can be shutting down while
hard_start_xmit is being called and interrupts are being handled.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/e1000/e1000_main.c

index 11508afdfdb9c25fc4c64328b7c77512557ebf31..8c64363faaf3ab15baa15e1efbceb097fb518248 100644 (file)
@@ -2733,8 +2733,9 @@ static int e1000_tx_map(struct e1000_adapter *adapter,
                        size -= 4;
 
                buffer_info->length = size;
-               buffer_info->dma = skb_shinfo(skb)->dma_head + offset;
+               /* set time_stamp *before* dma to help avoid a possible race */
                buffer_info->time_stamp = jiffies;
+               buffer_info->dma = skb_shinfo(skb)->dma_head + offset;
                buffer_info->next_to_watch = i;
 
                len -= size;
@@ -2774,8 +2775,8 @@ static int e1000_tx_map(struct e1000_adapter *adapter,
                                size -= 4;
 
                        buffer_info->length = size;
-                       buffer_info->dma = map[f] + offset;
                        buffer_info->time_stamp = jiffies;
+                       buffer_info->dma = map[f] + offset;
                        buffer_info->next_to_watch = i;
 
                        len -= size;
@@ -3459,7 +3460,9 @@ static bool e1000_clean_tx_irq(struct e1000_adapter *adapter,
                 * sees the new next_to_clean.
                 */
                smp_mb();
-               if (netif_queue_stopped(netdev)) {
+
+               if (netif_queue_stopped(netdev) &&
+                   !(test_bit(__E1000_DOWN, &adapter->flags))) {
                        netif_wake_queue(netdev);
                        ++adapter->restart_queue;
                }
@@ -3469,8 +3472,8 @@ static bool e1000_clean_tx_irq(struct e1000_adapter *adapter,
                /* Detect a transmit hang in hardware, this serializes the
                 * check with the clearing of time_stamp and movement of i */
                adapter->detect_tx_hung = false;
-               if (tx_ring->buffer_info[i].time_stamp &&
-                   time_after(jiffies, tx_ring->buffer_info[i].time_stamp +
+               if (tx_ring->buffer_info[eop].time_stamp &&
+                   time_after(jiffies, tx_ring->buffer_info[eop].time_stamp +
                               (adapter->tx_timeout_factor * HZ))
                    && !(er32(STATUS) & E1000_STATUS_TXOFF)) {
 
@@ -3492,7 +3495,7 @@ static bool e1000_clean_tx_irq(struct e1000_adapter *adapter,
                                readl(hw->hw_addr + tx_ring->tdt),
                                tx_ring->next_to_use,
                                tx_ring->next_to_clean,
-                               tx_ring->buffer_info[i].time_stamp,
+                               tx_ring->buffer_info[eop].time_stamp,
                                eop,
                                jiffies,
                                eop_desc->upper.fields.status);