From 712b97a63019863268edb2fe9067dc26ee1faf4a Mon Sep 17 00:00:00 2001
From: Sean Bruno <sbruno@FreeBSD.org>
Date: Thu, 7 Jan 2016 16:20:55 +0000
Subject: [PATCH] Switch em(4) to the extended RX descriptor format.  This
 matches the e1000/e1000e split in linux.

MFC after:	2 weeks
Sponsored by:	Limelight Networks
Differential Revision:	https://reviews.freebsd.org/D3447
---
 sys/dev/e1000/if_em.c | 108 ++++++++++++++++++++++++------------------
 1 file changed, 63 insertions(+), 45 deletions(-)

diff --git a/sys/dev/e1000/if_em.c b/sys/dev/e1000/if_em.c
index d9b3ca559d6e..9fa5ee40898b 100644
--- a/sys/dev/e1000/if_em.c
+++ b/sys/dev/e1000/if_em.c
@@ -3733,29 +3733,38 @@ em_transmit_checksum_setup(struct tx_ring *txr, struct mbuf *mp, int ip_off,
  		offload |= CSUM_TCP;
  		tucss = hdr_len;
  		tucso = hdr_len + offsetof(struct tcphdr, th_sum);
- 		/*
- 		 * Setting up new checksum offload context for every frames
- 		 * takes a lot of processing time for hardware. This also
- 		 * reduces performance a lot for small sized frames so avoid
- 		 * it if driver can use previously configured checksum
- 		 * offload context.
- 		 */
- 		if (txr->last_hw_offload == offload) {
- 			if (offload & CSUM_IP) {
- 				if (txr->last_hw_ipcss == ipcss &&
- 				    txr->last_hw_ipcso == ipcso &&
- 				    txr->last_hw_tucss == tucss &&
- 				    txr->last_hw_tucso == tucso)
- 					return;
- 			} else {
- 				if (txr->last_hw_tucss == tucss &&
- 				    txr->last_hw_tucso == tucso)
- 					return;
- 			}
-  		}
- 		txr->last_hw_offload = offload;
- 		txr->last_hw_tucss = tucss;
- 		txr->last_hw_tucso = tucso;
+		/*
+		 * The 82574L can only remember the *last* context used
+		 * regardless of queue that it was use for.  We cannot reuse
+		 * contexts on this hardware platform and must generate a new
+		 * context every time.  82574L hardware spec, section 7.2.6,
+		 * second note.
+		 */
+		if (adapter->num_queues < 2) {
+ 			/*
+ 		 	* Setting up new checksum offload context for every
+			* frames takes a lot of processing time for hardware.
+			* This also reduces performance a lot for small sized
+			* frames so avoid it if driver can use previously
+			* configured checksum offload context.
+ 		 	*/
+ 			if (txr->last_hw_offload == offload) {
+ 				if (offload & CSUM_IP) {
+ 					if (txr->last_hw_ipcss == ipcss &&
+ 				    	txr->last_hw_ipcso == ipcso &&
+ 				    	txr->last_hw_tucss == tucss &&
+ 				    	txr->last_hw_tucso == tucso)
+ 						return;
+ 				} else {
+ 					if (txr->last_hw_tucss == tucss &&
+ 				    	txr->last_hw_tucso == tucso)
+ 						return;
+ 				}
+  			}
+ 			txr->last_hw_offload = offload;
+ 			txr->last_hw_tucss = tucss;
+ 			txr->last_hw_tucso = tucso;
+		}
  		/*
  		 * Start offset for payload checksum calculation.
  		 * End offset for payload checksum calculation.
@@ -3771,29 +3780,38 @@ em_transmit_checksum_setup(struct tx_ring *txr, struct mbuf *mp, int ip_off,
  		*txd_upper |= E1000_TXD_POPTS_TXSM << 8;
  		tucss = hdr_len;
  		tucso = hdr_len + offsetof(struct udphdr, uh_sum);
- 		/*
- 		 * Setting up new checksum offload context for every frames
- 		 * takes a lot of processing time for hardware. This also
- 		 * reduces performance a lot for small sized frames so avoid
- 		 * it if driver can use previously configured checksum
- 		 * offload context.
- 		 */
- 		if (txr->last_hw_offload == offload) {
- 			if (offload & CSUM_IP) {
- 				if (txr->last_hw_ipcss == ipcss &&
- 				    txr->last_hw_ipcso == ipcso &&
- 				    txr->last_hw_tucss == tucss &&
- 				    txr->last_hw_tucso == tucso)
- 					return;
- 			} else {
- 				if (txr->last_hw_tucss == tucss &&
- 				    txr->last_hw_tucso == tucso)
- 					return;
+		/*
+		 * The 82574L can only remember the *last* context used
+		 * regardless of queue that it was use for.  We cannot reuse
+		 * contexts on this hardware platform and must generate a new
+		 * context every time.  82574L hardware spec, section 7.2.6,
+		 * second note.
+		 */
+		if (adapter->num_queues < 2) {
+ 			/*
+ 		 	* Setting up new checksum offload context for every
+			* frames takes a lot of processing time for hardware.
+			* This also reduces performance a lot for small sized
+			* frames so avoid it if driver can use previously
+			* configured checksum offload context.
+ 		 	*/
+ 			if (txr->last_hw_offload == offload) {
+ 				if (offload & CSUM_IP) {
+ 					if (txr->last_hw_ipcss == ipcss &&
+ 				    	txr->last_hw_ipcso == ipcso &&
+ 				    	txr->last_hw_tucss == tucss &&
+ 				    	txr->last_hw_tucso == tucso)
+ 						return;
+ 				} else {
+ 					if (txr->last_hw_tucss == tucss &&
+ 				    	txr->last_hw_tucso == tucso)
+ 						return;
+ 				}
  			}
- 		}
- 		txr->last_hw_offload = offload;
- 		txr->last_hw_tucss = tucss;
- 		txr->last_hw_tucso = tucso;
+ 			txr->last_hw_offload = offload;
+ 			txr->last_hw_tucss = tucss;
+ 			txr->last_hw_tucso = tucso;
+		}
  		/*
  		 * Start offset for header checksum calculation.
  		 * End offset for header checksum calculation.