[S390] qeth: preallocated qeth header for hiper socket

For hiper socket devices this patch will economize the reallocation
of the tx skb data segment by allocating separate memory for the qdio
transport information (qeth header).

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

4 files changed, 126 insertions, 39 deletions

diff --git a/drivers/s390/net/qeth_core.h b/drivers/s390/net/qeth_core.h
index 1895dbb553c..80971c21ea1 100644
--- a/drivers/s390/net/qeth_core.h
+++ b/drivers/s390/net/qeth_core.h
@@ -419,6 +419,7 @@ struct qeth_qdio_out_buffer {
 	int next_element_to_fill;
 	struct sk_buff_head skb_list;
 	struct list_head ctx_list;
+	int is_header[16];
 };
 
 struct qeth_card;
@@ -785,7 +786,7 @@ void qeth_core_remove_osn_attributes(struct device *);
 
 /* exports for qeth discipline device drivers */
 extern struct qeth_card_list_struct qeth_core_card_list;
-
+extern struct kmem_cache *qeth_core_header_cache;
 extern struct qeth_dbf_info qeth_dbf[QETH_DBF_INFOS];
 
 void qeth_set_allowed_threads(struct qeth_card *, unsigned long , int);
@@ -843,7 +844,7 @@ int qeth_get_priority_queue(struct qeth_card *, struct sk_buff *, int, int);
 int qeth_get_elements_no(struct qeth_card *, void *, struct sk_buff *, int);
 int qeth_do_send_packet_fast(struct qeth_card *, struct qeth_qdio_out_q *,
 			struct sk_buff *, struct qeth_hdr *, int,
-			struct qeth_eddp_context *);
+			struct qeth_eddp_context *, int, int);
 int qeth_do_send_packet(struct qeth_card *, struct qeth_qdio_out_q *,
 		    struct sk_buff *, struct qeth_hdr *,
 		    int, struct qeth_eddp_context *);
diff --git a/drivers/s390/net/qeth_core_main.c b/drivers/s390/net/qeth_core_main.c
index cebb25e36e8..7a554991971 100644
--- a/drivers/s390/net/qeth_core_main.c
+++ b/drivers/s390/net/qeth_core_main.c
@@ -48,6 +48,8 @@ EXPORT_SYMBOL_GPL(qeth_dbf);
 
 struct qeth_card_list_struct qeth_core_card_list;
 EXPORT_SYMBOL_GPL(qeth_core_card_list);
+struct kmem_cache *qeth_core_header_cache;
+EXPORT_SYMBOL_GPL(qeth_core_header_cache);
 
 static struct device *qeth_core_root_dev;
 static unsigned int known_devices[][10] = QETH_MODELLIST_ARRAY;
@@ -933,6 +935,10 @@ static void qeth_clear_output_buffer(struct qeth_qdio_out_q *queue,
 	}
 	qeth_eddp_buf_release_contexts(buf);
 	for (i = 0; i < QETH_MAX_BUFFER_ELEMENTS(queue->card); ++i) {
+		if (buf->buffer->element[i].addr && buf->is_header[i])
+			kmem_cache_free(qeth_core_header_cache,
+				buf->buffer->element[i].addr);
+		buf->is_header[i] = 0;
 		buf->buffer->element[i].length = 0;
 		buf->buffer->element[i].addr = NULL;
 		buf->buffer->element[i].flags = 0;
@@ -3002,8 +3008,8 @@ int qeth_get_elements_no(struct qeth_card *card, void *hdr,
 	if (skb_shinfo(skb)->nr_frags > 0)
 		elements_needed = (skb_shinfo(skb)->nr_frags + 1);
 	if (elements_needed == 0)
-		elements_needed = 1 + (((((unsigned long) hdr) % PAGE_SIZE)
-			+ skb->len) >> PAGE_SHIFT);
+		elements_needed = 1 + (((((unsigned long) skb->data) %
+				PAGE_SIZE) + skb->len) >> PAGE_SHIFT);
 	if ((elements_needed + elems) > QETH_MAX_BUFFER_ELEMENTS(card)) {
 		QETH_DBF_MESSAGE(2, "Invalid size of IP packet "
 			"(Number=%d / Length=%d). Discarded.\n",
@@ -3015,7 +3021,8 @@ int qeth_get_elements_no(struct qeth_card *card, void *hdr,
 EXPORT_SYMBOL_GPL(qeth_get_elements_no);
 
 static inline void __qeth_fill_buffer(struct sk_buff *skb,
-	struct qdio_buffer *buffer, int is_tso, int *next_element_to_fill)
+	struct qdio_buffer *buffer, int is_tso, int *next_element_to_fill,
+	int offset)
 {
 	int length = skb->len;
 	int length_here;
@@ -3027,6 +3034,11 @@ static inline void __qeth_fill_buffer(struct sk_buff *skb,
 	data = skb->data;
 	first_lap = (is_tso == 0 ? 1 : 0);
 
+	if (offset >= 0) {
+		data = skb->data + offset;
+		first_lap = 0;
+	}
+
 	while (length > 0) {
 		/* length_here is the remaining amount of data in this page */
 		length_here = PAGE_SIZE - ((unsigned long) data % PAGE_SIZE);
@@ -3058,22 +3070,22 @@ static inline void __qeth_fill_buffer(struct sk_buff *skb,
 }
 
 static inline int qeth_fill_buffer(struct qeth_qdio_out_q *queue,
-		struct qeth_qdio_out_buffer *buf, struct sk_buff *skb)
+		struct qeth_qdio_out_buffer *buf, struct sk_buff *skb,
+		struct qeth_hdr *hdr, int offset, int hd_len)
 {
 	struct qdio_buffer *buffer;
-	struct qeth_hdr_tso *hdr;
 	int flush_cnt = 0, hdr_len, large_send = 0;
 
 	buffer = buf->buffer;
 	atomic_inc(&skb->users);
 	skb_queue_tail(&buf->skb_list, skb);
 
-	hdr  = (struct qeth_hdr_tso *) skb->data;
 	/*check first on TSO ....*/
-	if (hdr->hdr.hdr.l3.id == QETH_HEADER_TYPE_TSO) {
+	if (hdr->hdr.l3.id == QETH_HEADER_TYPE_TSO) {
 		int element = buf->next_element_to_fill;
 
-		hdr_len = sizeof(struct qeth_hdr_tso) + hdr->ext.dg_hdr_len;
+		hdr_len = sizeof(struct qeth_hdr_tso) +
+			((struct qeth_hdr_tso *)hdr)->ext.dg_hdr_len;
 		/*fill first buffer entry only with header information */
 		buffer->element[element].addr = skb->data;
 		buffer->element[element].length = hdr_len;
@@ -3083,9 +3095,20 @@ static inline int qeth_fill_buffer(struct qeth_qdio_out_q *queue,
 		skb->len  -= hdr_len;
 		large_send = 1;
 	}
+
+	if (offset >= 0) {
+		int element = buf->next_element_to_fill;
+		buffer->element[element].addr = hdr;
+		buffer->element[element].length = sizeof(struct qeth_hdr) +
+							hd_len;
+		buffer->element[element].flags = SBAL_FLAGS_FIRST_FRAG;
+		buf->is_header[element] = 1;
+		buf->next_element_to_fill++;
+	}
+
 	if (skb_shinfo(skb)->nr_frags == 0)
 		__qeth_fill_buffer(skb, buffer, large_send,
-				   (int *)&buf->next_element_to_fill);
+				(int *)&buf->next_element_to_fill, offset);
 	else
 		__qeth_fill_buffer_frag(skb, buffer, large_send,
 					(int *)&buf->next_element_to_fill);
@@ -3115,7 +3138,7 @@ static inline int qeth_fill_buffer(struct qeth_qdio_out_q *queue,
 int qeth_do_send_packet_fast(struct qeth_card *card,
 		struct qeth_qdio_out_q *queue, struct sk_buff *skb,
 		struct qeth_hdr *hdr, int elements_needed,
-		struct qeth_eddp_context *ctx)
+		struct qeth_eddp_context *ctx, int offset, int hd_len)
 {
 	struct qeth_qdio_out_buffer *buffer;
 	int buffers_needed = 0;
@@ -3148,7 +3171,7 @@ int qeth_do_send_packet_fast(struct qeth_card *card,
 	}
 	atomic_set(&queue->state, QETH_OUT_Q_UNLOCKED);
 	if (ctx == NULL) {
-		qeth_fill_buffer(queue, buffer, skb);
+		qeth_fill_buffer(queue, buffer, skb, hdr, offset, hd_len);
 		qeth_flush_buffers(queue, index, 1);
 	} else {
 		flush_cnt = qeth_eddp_fill_buffer(queue, ctx, index);
@@ -3224,7 +3247,7 @@ int qeth_do_send_packet(struct qeth_card *card, struct qeth_qdio_out_q *queue,
 		}
 	}
 	if (ctx == NULL)
-		tmp = qeth_fill_buffer(queue, buffer, skb);
+		tmp = qeth_fill_buffer(queue, buffer, skb, hdr, -1, 0);
 	else {
 		tmp = qeth_eddp_fill_buffer(queue, ctx,
 						queue->next_buf_to_fill);
@@ -4443,8 +4466,17 @@ static int __init qeth_core_init(void)
 	rc = IS_ERR(qeth_core_root_dev) ? PTR_ERR(qeth_core_root_dev) : 0;
 	if (rc)
 		goto register_err;
-	return 0;
 
+	qeth_core_header_cache = kmem_cache_create("qeth_hdr",
+			sizeof(struct qeth_hdr) + ETH_HLEN, 64, 0, NULL);
+	if (!qeth_core_header_cache) {
+		rc = -ENOMEM;
+		goto slab_err;
+	}
+
+	return 0;
+slab_err:
+	s390_root_dev_unregister(qeth_core_root_dev);
 register_err:
 	driver_remove_file(&qeth_core_ccwgroup_driver.driver,
 			   &driver_attr_group);
@@ -4466,6 +4498,7 @@ static void __exit qeth_core_exit(void)
 			   &driver_attr_group);
 	ccwgroup_driver_unregister(&qeth_core_ccwgroup_driver);
 	ccw_driver_unregister(&qeth_ccw_driver);
+	kmem_cache_destroy(qeth_core_header_cache);
 	qeth_unregister_dbf_views();
 	PRINT_INFO("core functions removed\n");
 }
diff --git a/drivers/s390/net/qeth_l2_main.c b/drivers/s390/net/qeth_l2_main.c
index a8b069cd9a4..b3cee032f57 100644
--- a/drivers/s390/net/qeth_l2_main.c
+++ b/drivers/s390/net/qeth_l2_main.c
@@ -243,8 +243,7 @@ static void qeth_l2_get_packet_type(struct qeth_card *card,
 static void qeth_l2_fill_header(struct qeth_card *card, struct qeth_hdr *hdr,
 			struct sk_buff *skb, int ipv, int cast_type)
 {
-	struct vlan_ethhdr *veth = (struct vlan_ethhdr *)((skb->data) +
-					QETH_HEADER_SIZE);
+	struct vlan_ethhdr *veth = (struct vlan_ethhdr *)skb_mac_header(skb);
 
 	memset(hdr, 0, sizeof(struct qeth_hdr));
 	hdr->hdr.l2.id = QETH_HEADER_TYPE_LAYER2;
@@ -621,6 +620,9 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	int tx_bytes = skb->len;
 	enum qeth_large_send_types large_send = QETH_LARGE_SEND_NO;
 	struct qeth_eddp_context *ctx = NULL;
+	int data_offset = -1;
+	int elements_needed = 0;
+	int hd_len = 0;
 
 	if ((card->state != CARD_STATE_UP) || !card->lan_online) {
 		card->stats.tx_carrier_errors++;
@@ -643,13 +645,32 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	if (card->info.type == QETH_CARD_TYPE_OSN)
 		hdr = (struct qeth_hdr *)skb->data;
 	else {
-		/* create a clone with writeable headroom */
-		new_skb = skb_realloc_headroom(skb, sizeof(struct qeth_hdr));
-		if (!new_skb)
-			goto tx_drop;
-		hdr = (struct qeth_hdr *)skb_push(new_skb,
+		if ((card->info.type == QETH_CARD_TYPE_IQD) && (!large_send) &&
+		    (skb_shinfo(skb)->nr_frags == 0)) {
+			new_skb = skb;
+			data_offset = ETH_HLEN;
+			hd_len = ETH_HLEN;
+			hdr = kmem_cache_alloc(qeth_core_header_cache,
+						GFP_ATOMIC);
+			if (!hdr)
+				goto tx_drop;
+			elements_needed++;
+			skb_reset_mac_header(new_skb);
+			qeth_l2_fill_header(card, hdr, new_skb, ipv, cast_type);
+			hdr->hdr.l2.pkt_length = new_skb->len;
+			memcpy(((char *)hdr) + sizeof(struct qeth_hdr),
+				skb_mac_header(new_skb), ETH_HLEN);
+		} else {
+			/* create a clone with writeable headroom */
+			new_skb = skb_realloc_headroom(skb,
+						sizeof(struct qeth_hdr));
+			if (!new_skb)
+				goto tx_drop;
+			hdr = (struct qeth_hdr *)skb_push(new_skb,
 						sizeof(struct qeth_hdr));
-		qeth_l2_fill_header(card, hdr, new_skb, ipv, cast_type);
+			skb_set_mac_header(new_skb, sizeof(struct qeth_hdr));
+			qeth_l2_fill_header(card, hdr, new_skb, ipv, cast_type);
+		}
 	}
 
 	if (large_send == QETH_LARGE_SEND_EDDP) {
@@ -660,9 +681,13 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 			goto tx_drop;
 		}
 	} else {
-		elements = qeth_get_elements_no(card, (void *)hdr, new_skb, 0);
-		if (!elements)
+		elements = qeth_get_elements_no(card, (void *)hdr, new_skb,
+						elements_needed);
+		if (!elements) {
+			if (data_offset >= 0)
+				kmem_cache_free(qeth_core_header_cache, hdr);
 			goto tx_drop;
+		}
 	}
 
 	if ((large_send == QETH_LARGE_SEND_NO) &&
@@ -674,7 +699,7 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 					 elements, ctx);
 	else
 		rc = qeth_do_send_packet_fast(card, queue, new_skb, hdr,
-					      elements, ctx);
+					elements, ctx, data_offset, hd_len);
 	if (!rc) {
 		card->stats.tx_packets++;
 		card->stats.tx_bytes += tx_bytes;
@@ -701,6 +726,9 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		if (ctx != NULL)
 			qeth_eddp_put_context(ctx);
 
+		if (data_offset >= 0)
+			kmem_cache_free(qeth_core_header_cache, hdr);
+
 		if (rc == -EBUSY) {
 			if (new_skb != skb)
 				dev_kfree_skb_any(new_skb);
diff --git a/drivers/s390/net/qeth_l3_main.c b/drivers/s390/net/qeth_l3_main.c
index 3e1d1385735..dd72c3c2016 100644
--- a/drivers/s390/net/qeth_l3_main.c
+++ b/drivers/s390/net/qeth_l3_main.c
@@ -2604,6 +2604,7 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	int tx_bytes = skb->len;
 	enum qeth_large_send_types large_send = QETH_LARGE_SEND_NO;
 	struct qeth_eddp_context *ctx = NULL;
+	int data_offset = -1;
 
 	if ((card->info.type == QETH_CARD_TYPE_IQD) &&
 	    (skb->protocol != htons(ETH_P_IPV6)) &&
@@ -2624,14 +2625,28 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		card->perf_stats.outbound_start_time = qeth_get_micros();
 	}
 
-	/* create a clone with writeable headroom */
-	new_skb = skb_realloc_headroom(skb, sizeof(struct qeth_hdr_tso) +
-					VLAN_HLEN);
-	if (!new_skb)
-		goto tx_drop;
+	if (skb_is_gso(skb))
+		large_send = card->options.large_send;
+
+	if ((card->info.type == QETH_CARD_TYPE_IQD) && (!large_send) &&
+	    (skb_shinfo(skb)->nr_frags == 0)) {
+		new_skb = skb;
+		data_offset = ETH_HLEN;
+		hdr = kmem_cache_alloc(qeth_core_header_cache, GFP_ATOMIC);
+		if (!hdr)
+			goto tx_drop;
+		elements_needed++;
+	} else {
+		/* create a clone with writeable headroom */
+		new_skb = skb_realloc_headroom(skb, sizeof(struct qeth_hdr_tso)
+					+ VLAN_HLEN);
+		if (!new_skb)
+			goto tx_drop;
+	}
 
 	if (card->info.type == QETH_CARD_TYPE_IQD) {
-		skb_pull(new_skb, ETH_HLEN);
+		if (data_offset < 0)
+			skb_pull(new_skb, ETH_HLEN);
 	} else {
 		if (new_skb->protocol == htons(ETH_P_IP)) {
 			if (card->dev->type == ARPHRD_IEEE802_TR)
@@ -2657,9 +2672,6 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	netif_stop_queue(dev);
 
-	if (skb_is_gso(new_skb))
-		large_send = card->options.large_send;
-
 	/* fix hardware limitation: as long as we do not have sbal
 	 * chaining we can not send long frag lists so we temporary
 	 * switch to EDDP
@@ -2677,9 +2689,16 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		qeth_tso_fill_header(card, hdr, new_skb);
 		elements_needed++;
 	} else {
-		hdr = (struct qeth_hdr *)skb_push(new_skb,
+		if (data_offset < 0) {
+			hdr = (struct qeth_hdr *)skb_push(new_skb,
 						sizeof(struct qeth_hdr));
-		qeth_l3_fill_header(card, hdr, new_skb, ipv, cast_type);
+			qeth_l3_fill_header(card, hdr, new_skb, ipv,
+						cast_type);
+		} else {
+			qeth_l3_fill_header(card, hdr, new_skb, ipv,
+						cast_type);
+			hdr->hdr.l3.length = new_skb->len - data_offset;
+		}
 	}
 
 	if (large_send == QETH_LARGE_SEND_EDDP) {
@@ -2695,8 +2714,11 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	} else {
 		int elems = qeth_get_elements_no(card, (void *)hdr, new_skb,
 						 elements_needed);
-		if (!elems)
+		if (!elems) {
+			if (data_offset >= 0)
+				kmem_cache_free(qeth_core_header_cache, hdr);
 			goto tx_drop;
+		}
 		elements_needed += elems;
 	}
 
@@ -2709,7 +2731,7 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 					 elements_needed, ctx);
 	else
 		rc = qeth_do_send_packet_fast(card, queue, new_skb, hdr,
-					      elements_needed, ctx);
+					elements_needed, ctx, data_offset, 0);
 
 	if (!rc) {
 		card->stats.tx_packets++;
@@ -2737,6 +2759,9 @@ static int qeth_l3_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		if (ctx != NULL)
 			qeth_eddp_put_context(ctx);
 
+		if (data_offset >= 0)
+			kmem_cache_free(qeth_core_header_cache, hdr);
+
 		if (rc == -EBUSY) {
 			if (new_skb != skb)
 				dev_kfree_skb_any(new_skb);