igb: Remove artificial restriction on RQDPC stat reading
Alexander Duyck [Wed, 8 Aug 2012 05:23:22 +0000 (05:23 +0000)]
commit ae1c07a6b7ced6c0c94c99e3b53f4e7856fa8bff upstream.

For some reason the reading of the RQDPC register was being artificially
limited to 4K.  Instead of limiting the value we should read the value and
add the full amount.  Otherwise this can lead to a misleading number of
dropped packets when the actual value is in fact much higher.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Jeff Pieper   <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Vinson Lee <vlee@twitter.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/net/ethernet/intel/igb/igb_main.c

index 6b4f014..1fb180d 100644 (file)
@@ -4649,11 +4649,13 @@ void igb_update_stats(struct igb_adapter *adapter,
        bytes = 0;
        packets = 0;
        for (i = 0; i < adapter->num_rx_queues; i++) {
-               u32 rqdpc_tmp = rd32(E1000_RQDPC(i)) & 0x0FFF;
+               u32 rqdpc = rd32(E1000_RQDPC(i));
                struct igb_ring *ring = adapter->rx_ring[i];
 
-               ring->rx_stats.drops += rqdpc_tmp;
-               net_stats->rx_fifo_errors += rqdpc_tmp;
+               if (rqdpc) {
+                       ring->rx_stats.drops += rqdpc;
+                       net_stats->rx_fifo_errors += rqdpc;
+               }
 
                do {
                        start = u64_stats_fetch_begin_bh(&ring->rx_syncp);