From: Stanislaw Gruszka <sgruszka@redhat.com>
Date: Fri, 30 Apr 2010 08:26:46 -0400
Subject: [net] bnx2x: Added GRO support
Message-id: <1272616006-3012-2-git-send-email-sgruszka@redhat.com>
Patchwork-id: 24704
O-Subject: [RHEL5.6 PATCH 2/2] bnx2x: Added GRO support
Bugzilla: 573114
RH-Acked-by: David S. Miller <davem@redhat.com>

BZ#573114

From:
Dmitry Kravkov <dmitry@broadcom.com>

Description (from upstream commit):
Adding GRO support on top of the HW LRO (TPA) support –
there is no measurable performance drawback of adding GRO
on top of it, and it allows better performance when LRO (TPA)
is turned off for virtualization or bridging.

Upstream:
Patch is in net-next tree
commit 4fd89b7af28292e190650b9b9bc4308658d81dd1
Author: Dmitry Kravkov <dmitry@broadcom.com>
Date:   Thu Apr 1 19:45:34 2010 -0700

    bnx2x: Added GRO support

Testing:
Tested by me by doing ssh and ftp with different module and ethtool
options. Additionally I run iperf on xen and patch shows perfomance
improvements.

diff --git a/drivers/net/bnx2x.h b/drivers/net/bnx2x.h
index 9144aa2..e3f97ed 100644
--- a/drivers/net/bnx2x.h
+++ b/drivers/net/bnx2x.h
@@ -260,6 +260,7 @@ struct bnx2x_eth_q_stats {
 struct bnx2x_fastpath {
 
 	struct net_device	dummy_netdev;
+	struct napi_struct	napi;
 
 	struct host_status_block *status_blk;
 	dma_addr_t		status_blk_mapping;
diff --git a/drivers/net/bnx2x_main.c b/drivers/net/bnx2x_main.c
index 2202ad3..b8b5266 100644
--- a/drivers/net/bnx2x_main.c
+++ b/drivers/net/bnx2x_main.c
@@ -56,7 +56,7 @@
 #include "bnx2x_init_ops.h"
 #include "bnx2x_dump.h"
 
-#define DRV_MODULE_VERSION	"1.52.1-7"
+#define DRV_MODULE_VERSION	"1.52.1-8"
 #define DRV_MODULE_RELDATE	"2010/02/28"
 #define BNX2X_BC_VER		0x040200
 
@@ -1434,12 +1434,13 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 #ifdef BCM_VLAN
 			if ((bp->vlgrp != NULL) && is_vlan_cqe &&
 			    (!is_not_hwaccel_vlan_cqe))
-				vlan_hwaccel_receive_skb(skb, bp->vlgrp,
-						le16_to_cpu(cqe->fast_path_cqe.
-							    vlan_tag));
+				vlan_gro_receive(&fp->napi, bp->vlgrp,
+						 le16_to_cpu(cqe->fast_path_cqe.
+							     vlan_tag), skb);
+
 			else
 #endif
-				netif_receive_skb(skb);
+				napi_gro_receive(&fp->napi, skb);
 		} else {
 			DP(NETIF_MSG_RX_STATUS, "Failed to allocate new pages"
 			   " - dropping packet!\n");
@@ -1691,11 +1692,11 @@ reuse_rx:
 		if ((bp->vlgrp != NULL) && (bp->flags & HW_VLAN_RX_FLAG) &&
 		    (le16_to_cpu(cqe->fast_path_cqe.pars_flags.flags) &
 		     PARSING_FLAGS_VLAN))
-			vlan_hwaccel_receive_skb(skb, bp->vlgrp,
-				le16_to_cpu(cqe->fast_path_cqe.vlan_tag));
+			vlan_gro_receive(&fp->napi, bp->vlgrp,
+				le16_to_cpu(cqe->fast_path_cqe.vlan_tag), skb);
 		else
 #endif
-			netif_receive_skb(skb);
+			napi_gro_receive(&fp->napi, skb);
 
 		bp->dev->last_rx = jiffies;
 
@@ -7383,9 +7384,12 @@ static void bnx2x_alloc_napi(struct bnx2x *bp)
 	/* initialize net_device for each interrupt */
 	for_each_queue(bp, i) {
 		struct net_device *dummy_netdev = &bnx2x_fp(bp, i, dummy_netdev);
+		struct napi_struct *napi = &bnx2x_fp(bp, i, napi);
+
 		dummy_netdev->poll = bnx2x_poll;
 		dummy_netdev->priv = &bp->fp[i];
 		dummy_netdev->weight = 128;
+		napi->dev = bp->dev;
 		set_bit(__LINK_STATE_START, &dummy_netdev->state);
 	}
 }
@@ -8950,6 +8954,7 @@ static int __devinit bnx2x_init_bp(struct bnx2x *bp)
 	}
 	bp->multi_mode = multi_mode;
 
+	bp->dev->features |= NETIF_F_GRO;
 
 	/* Set TPA flags */
 	if (disable_tpa) {
@@ -11000,6 +11005,7 @@ static int bnx2x_poll(struct net_device *dev, int *budget)
 #ifdef BNX2X_STOP_ON_ERROR
 poll_panic:
 #endif
+		napi_gro_flush(&fp->napi);
 		netif_rx_complete(dev);
 
 		bnx2x_ack_sb(bp, fp->sb_id, USTORM_ID,