[SK_BUFF]: Convert skb->tail to sk_buff_data_t

So that it is also an offset from skb->head, reduces its size from 8 to 4 bytes
on 64bit architectures, allowing us to combine the 4 bytes hole left by the
layer headers conversion, reducing struct sk_buff size to 256 bytes, i.e. 4
64byte cachelines, and since the sk_buff slab cache is SLAB_HWCACHE_ALIGN...
:-)

Many calculations that previously required that skb->{transport,network,
mac}_header be first converted to a pointer now can be done directly, being
meaningful as offsets or pointers.

Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

arch/ia64/sn/kernel/xpnet.c

@@ -264,15 +264,15 @@ xpnet_receive(partid_t partid, int channel, struct xpnet_message *msg)
 
 	dev_dbg(xpnet, "<skb->head=0x%p skb->data=0x%p skb->tail=0x%p "
 		"skb->end=0x%p skb->len=%d\n", (void *) skb->head,
-		(void *) skb->data, (void *) skb->tail, (void *) skb->end,
+		(void *)skb->data, skb_tail_pointer(skb), (void *)skb->end,
 		skb->len);
 
 	skb->protocol = eth_type_trans(skb, xpnet_device);
 	skb->ip_summed = CHECKSUM_UNNECESSARY;
 
 	dev_dbg(xpnet, "passing skb to network layer; \n\tskb->head=0x%p "
 		"skb->data=0x%p skb->tail=0x%p skb->end=0x%p skb->len=%d\n",
-		(void *) skb->head, (void *) skb->data, (void *) skb->tail,
+		(void *)skb->head, (void *)skb->data, skb_tail_pointer(skb),
 		(void *) skb->end, skb->len);
 
 
@@ -475,7 +475,7 @@ xpnet_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	dev_dbg(xpnet, ">skb->head=0x%p skb->data=0x%p skb->tail=0x%p "
 		"skb->end=0x%p skb->len=%d\n", (void *) skb->head,
-		(void *) skb->data, (void *) skb->tail, (void *) skb->end,
+		(void *)skb->data, skb_tail_pointer(skb), (void *)skb->end,
 		skb->len);
 
 
@@ -497,7 +497,7 @@ xpnet_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	/* get the beginning of the first cacheline and end of last */
 	start_addr = ((u64) skb->data & ~(L1_CACHE_BYTES - 1));
-	end_addr = L1_CACHE_ALIGN((u64) skb->tail);
+	end_addr = L1_CACHE_ALIGN((u64)skb_tail_pointer(skb));
 
 	/* calculate how many bytes to embed in the XPC message */
 	embedded_bytes = 0;
@@ -573,7 +573,7 @@ xpnet_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		msg->magic = XPNET_MAGIC;
 		msg->size = end_addr - start_addr;
 		msg->leadin_ignore = (u64) skb->data - start_addr;
-		msg->tailout_ignore = end_addr - (u64) skb->tail;
+		msg->tailout_ignore = end_addr - (u64)skb_tail_pointer(skb);
 		msg->buf_pa = __pa(start_addr);
 
 		dev_dbg(xpnet, "sending XPC message to %d:%d\nmsg->buf_pa="

drivers/atm/he.c

@@ -1901,13 +1901,13 @@ he_service_rbrq(struct he_dev *he_dev, int group)
 			case ATM_AAL0:
 				/* 2.10.1.5 raw cell receive */
 				skb->len = ATM_AAL0_SDU;
-				skb->tail = skb->data + skb->len;
+				skb_set_tail_pointer(skb, skb->len);
 				break;
 			case ATM_AAL5:
 				/* 2.10.1.2 aal5 receive */
 
 				skb->len = AAL5_LEN(skb->data, he_vcc->pdu_len);
-				skb->tail = skb->data + skb->len;
+				skb_set_tail_pointer(skb, skb->len);
 #ifdef USE_CHECKSUM_HW
 				if (vcc->vpi == 0 && vcc->vci >= ATM_NOT_RSV_VCI) {
 					skb->ip_summed = CHECKSUM_COMPLETE;

drivers/atm/idt77252.c

@@ -1816,7 +1816,8 @@ push_rx_skb(struct idt77252_dev *card, struct sk_buff *skb, int queue)
 	u32 handle;
 	u32 addr;
 
-	skb->data = skb->tail = skb->head;
+	skb->data = skb->head;
+	skb_reset_tail_pointer(skb);
 	skb->len = 0;
 
 	skb_reserve(skb, 16);

drivers/atm/nicstar.c

@@ -2208,7 +2208,7 @@ static void dequeue_rx(ns_dev *card, ns_rsqe *rsqe)
          if (i == 1 && ns_rsqe_eopdu(rsqe))
             *((u32 *) sb->data) |= 0x00000002;
          skb_put(sb, NS_AAL0_HEADER);
-         memcpy(sb->tail, cell, ATM_CELL_PAYLOAD);
+         memcpy(skb_tail_pointer(sb), cell, ATM_CELL_PAYLOAD);
          skb_put(sb, ATM_CELL_PAYLOAD);
          ATM_SKB(sb)->vcc = vcc;
 	 __net_timestamp(sb);
@@ -2252,7 +2252,8 @@ static void dequeue_rx(ns_dev *card, ns_rsqe *rsqe)
       vc->rx_iov = iovb;
       NS_SKB(iovb)->iovcnt = 0;
       iovb->len = 0;
-      iovb->tail = iovb->data = iovb->head;
+      iovb->data = iovb->head;
+      skb_reset_tail_pointer(iovb);
       NS_SKB(iovb)->vcc = vcc;
       /* IMPORTANT: a pointer to the sk_buff containing the small or large
                     buffer is stored as iovec base, NOT a pointer to the 
@@ -2265,7 +2266,8 @@ static void dequeue_rx(ns_dev *card, ns_rsqe *rsqe)
       recycle_iovec_rx_bufs(card, (struct iovec *) iovb->data, NS_MAX_IOVECS);
       NS_SKB(iovb)->iovcnt = 0;
       iovb->len = 0;
-      iovb->tail = iovb->data = iovb->head;
+      iovb->data = iovb->head;
+      skb_reset_tail_pointer(iovb);
       NS_SKB(iovb)->vcc = vcc;
    }
    iov = &((struct iovec *) iovb->data)[NS_SKB(iovb)->iovcnt++];
@@ -2489,7 +2491,7 @@ static void dequeue_rx(ns_dev *card, ns_rsqe *rsqe)
             {
                lb = (struct sk_buff *) iov->iov_base;
                tocopy = min_t(int, remaining, iov->iov_len);
-               memcpy(hb->tail, lb->data, tocopy);
+               memcpy(skb_tail_pointer(hb), lb->data, tocopy);
                skb_put(hb, tocopy);
                iov++;
                remaining -= tocopy;

drivers/infiniband/hw/amso1100/c2.c

@@ -439,7 +439,8 @@ static void c2_rx_error(struct c2_port *c2_port, struct c2_element *elem)
 	}
 
 	/* Setup the skb for reuse since we're dropping this pkt */
-	elem->skb->tail = elem->skb->data = elem->skb->head;
+	elem->skb->data = elem->skb->head;
+	skb_reset_tail_pointer(elem->skb);
 
 	/* Zero out the rxp hdr in the sk_buff */
 	memset(elem->skb->data, 0, sizeof(*rxp_hdr));
@@ -521,7 +522,7 @@ static void c2_rx_interrupt(struct net_device *netdev)
 		 * "sizeof(struct c2_rxp_hdr)".
 		 */
 		skb->data += sizeof(*rxp_hdr);
-		skb->tail = skb->data + buflen;
+		skb_set_tail_pointer(skb, buflen);
 		skb->len = buflen;
 		skb->protocol = eth_type_trans(skb, netdev);
 

drivers/isdn/i4l/isdn_net.c

@@ -881,7 +881,7 @@ isdn_net_log_skb(struct sk_buff * skb, isdn_net_local * lp)
 
 	addinfo[0] = '\0';
 	/* This check stolen from 2.1.72 dev_queue_xmit_nit() */
-	if (p < skb->data || p >= skb->tail) {
+	if (p < skb->data || skb->network_header >= skb->tail) {
 		/* fall back to old isdn_net_log_packet method() */
 		char * buf = skb->data;
 

drivers/media/dvb/dvb-core/dvb_net.c

@@ -600,6 +600,7 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 			/* Check CRC32, we've got it in our skb already. */
 			unsigned short ulen = htons(priv->ule_sndu_len);
 			unsigned short utype = htons(priv->ule_sndu_type);
+			const u8 *tail;
 			struct kvec iov[3] = {
 				{ &ulen, sizeof ulen },
 				{ &utype, sizeof utype },
@@ -613,10 +614,11 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 			}
 
 			ule_crc = iov_crc32(ule_crc, iov, 3);
-			expected_crc = *((u8 *)priv->ule_skb->tail - 4) << 24 |
-				       *((u8 *)priv->ule_skb->tail - 3) << 16 |
-				       *((u8 *)priv->ule_skb->tail - 2) << 8 |
-				       *((u8 *)priv->ule_skb->tail - 1);
+			tail = skb_tail_pointer(priv->ule_skb);
+			expected_crc = *(tail - 4) << 24 |
+				       *(tail - 3) << 16 |
+				       *(tail - 2) << 8 |
+				       *(tail - 1);
 			if (ule_crc != expected_crc) {
 				printk(KERN_WARNING "%lu: CRC32 check FAILED: %08x / %08x, SNDU len %d type %#x, ts_remain %d, next 2: %x.\n",
 				       priv->ts_count, ule_crc, expected_crc, priv->ule_sndu_len, priv->ule_sndu_type, ts_remain, ts_remain > 2 ? *(unsigned short *)from_where : 0);

drivers/net/cris/eth_v10.c

@@ -1348,7 +1348,7 @@ e100_rx(struct net_device *dev)
 
 #ifdef ETHDEBUG
 		printk("head = 0x%x, data = 0x%x, tail = 0x%x, end = 0x%x\n",
-		  skb->head, skb->data, skb->tail, skb->end);
+		  skb->head, skb->data, skb_tail_pointer(skb), skb->end);
 		printk("copying packet to 0x%x.\n", skb_data_ptr);
 #endif
 

drivers/net/cxgb3/sge.c

@@ -1325,13 +1325,13 @@ static void write_ofld_wr(struct adapter *adap, struct sk_buff *skb,
 	flits = skb_transport_offset(skb) / 8;
 	sgp = ndesc == 1 ? (struct sg_ent *)&d->flit[flits] : sgl;
 	sgl_flits = make_sgl(skb, sgp, skb_transport_header(skb),
-			     skb->tail - skb_transport_header(skb),
+			     skb->tail - skb->transport_header,
 			     adap->pdev);
 	if (need_skb_unmap()) {
 		setup_deferred_unmapping(skb, adap->pdev, sgp, sgl_flits);
 		skb->destructor = deferred_unmap_destructor;
 		((struct unmap_info *)skb->cb)->len = (skb->tail -
-						       skb_transport_header(skb));
+						       skb->transport_header);
 	}
 
 	write_wr_hdr_sgl(ndesc, skb, d, pidx, q, sgl, flits, sgl_flits,
@@ -1353,7 +1353,7 @@ static inline unsigned int calc_tx_descs_ofld(const struct sk_buff *skb)
 		return 1;	/* packet fits as immediate data */
 
 	flits = skb_transport_offset(skb) / 8;	/* headers */
-	if (skb->tail != skb_transport_header(skb))
+	if (skb->tail != skb->transport_header)
 		cnt++;
 	return flits_to_desc(flits + sgl_len(cnt));
 }

drivers/net/e1000/e1000_main.c

@@ -3304,7 +3304,7 @@ e1000_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
 				 * NOTE: this is a TSO only workaround
 				 * if end byte alignment not correct move us
 				 * into the next dword */
-				if ((unsigned long)(skb->tail - 1) & 4)
+				if ((unsigned long)(skb_tail_pointer(skb) - 1) & 4)
 					break;
 				/* fall through */
 			case e1000_82571:
@@ -4388,7 +4388,7 @@ e1000_clean_rx_irq_ps(struct e1000_adapter *adapter,
 				PCI_DMA_FROMDEVICE);
 			vaddr = kmap_atomic(ps_page->ps_page[0],
 			                    KM_SKB_DATA_SOFTIRQ);
-			memcpy(skb->tail, vaddr, l1);
+			memcpy(skb_tail_pointer(skb), vaddr, l1);
 			kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
 			pci_dma_sync_single_for_device(pdev,
 				ps_page_dma->ps_page_dma[0],

drivers/net/ibm_emac/ibm_emac_core.c

@@ -1338,7 +1338,7 @@ static inline int emac_rx_sg_append(struct ocp_enet_private *dev, int slot)
 			dev_kfree_skb(dev->rx_sg_skb);
 			dev->rx_sg_skb = NULL;
 		} else {
-			cacheable_memcpy(dev->rx_sg_skb->tail,
+			cacheable_memcpy(skb_tail_pointer(dev->rx_sg_skb),
 					 dev->rx_skb[slot]->data, len);
 			skb_put(dev->rx_sg_skb, len);
 			emac_recycle_rx_skb(dev, slot, len);

drivers/net/macb.c

@@ -575,7 +575,8 @@ static int macb_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	int i;
 	dev_dbg(&bp->pdev->dev,
 		"start_xmit: len %u head %p data %p tail %p end %p\n",
-		skb->len, skb->head, skb->data, skb->tail, skb->end);
+		skb->len, skb->head, skb->data,
+		skb_tail_pointer(skb), skb->end);
 	dev_dbg(&bp->pdev->dev,
 		"data:");
 	for (i = 0; i < 16; i++)

drivers/net/pcmcia/nmclan_cs.c

@@ -1185,7 +1185,7 @@ static int mace_rx(struct net_device *dev, unsigned char RxCnt)
 	skb_reserve(skb, 2);
 	insw(ioaddr + AM2150_RCV, skb_put(skb, pkt_len), pkt_len>>1);
 	if (pkt_len & 1)
-	    *(skb->tail-1) = inb(ioaddr + AM2150_RCV);
+	    *(skb_tail_pointer(skb) - 1) = inb(ioaddr + AM2150_RCV);
 	skb->protocol = eth_type_trans(skb, dev);
 
 	netif_rx(skb); /* Send the packet to the upper (protocol) layers. */

drivers/net/s2io.c

@@ -2195,7 +2195,7 @@ static int fill_rxd_3buf(struct s2io_nic *nic, struct RxD_t *rxdp, struct \
 	frag_list->next = NULL;
 	tmp = (void *)ALIGN((long)frag_list->data, ALIGN_SIZE + 1);
 	frag_list->data = tmp;
-	frag_list->tail = tmp;
+	skb_reset_tail_pointer(frag_list);
 
 	/* Buffer-2 receives L4 data payload */
 	((struct RxD3*)rxdp)->Buffer2_ptr = pci_map_single(nic->pdev,
@@ -2349,7 +2349,7 @@ static int fill_rx_buffers(struct s2io_nic *nic, int ring_no)
 			tmp += ALIGN_SIZE;
 			tmp &= ~ALIGN_SIZE;
 			skb->data = (void *) (unsigned long)tmp;
-			skb->tail = (void *) (unsigned long)tmp;
+			skb_reset_tail_pointer(skb);
 
 			if (!(((struct RxD3*)rxdp)->Buffer0_ptr))
 				((struct RxD3*)rxdp)->Buffer0_ptr =

drivers/net/tulip/uli526x.c

@@ -829,7 +829,9 @@ static void uli526x_rx_packet(struct net_device *dev, struct uli526x_board_info
 					!= NULL) ) {
 					/* size less than COPY_SIZE, allocate a rxlen SKB */
 					skb_reserve(skb, 2); /* 16byte align */
-					memcpy(skb_put(skb, rxlen), rxptr->rx_skb_ptr->tail, rxlen);
+					memcpy(skb_put(skb, rxlen),
+					       skb_tail_pointer(rxptr->rx_skb_ptr),
+					       rxlen);
 					uli526x_reuse_skb(db, rxptr->rx_skb_ptr);
 				} else
 					skb_put(skb, rxlen);
@@ -1175,7 +1177,10 @@ static void uli526x_reuse_skb(struct uli526x_board_info *db, struct sk_buff * sk
 
 	if (!(rxptr->rdes0 & cpu_to_le32(0x80000000))) {
 		rxptr->rx_skb_ptr = skb;
-		rxptr->rdes2 = cpu_to_le32( pci_map_single(db->pdev, skb->tail, RX_ALLOC_SIZE, PCI_DMA_FROMDEVICE) );
+		rxptr->rdes2 = cpu_to_le32(pci_map_single(db->pdev,
+							  skb_tail_pointer(skb),
+							  RX_ALLOC_SIZE,
+							  PCI_DMA_FROMDEVICE));
 		wmb();
 		rxptr->rdes0 = cpu_to_le32(0x80000000);
 		db->rx_avail_cnt++;
@@ -1339,7 +1344,10 @@ static void allocate_rx_buffer(struct uli526x_board_info *db)
 		if ( ( skb = dev_alloc_skb(RX_ALLOC_SIZE) ) == NULL )
 			break;
 		rxptr->rx_skb_ptr = skb; /* FIXME (?) */
-		rxptr->rdes2 = cpu_to_le32( pci_map_single(db->pdev, skb->tail, RX_ALLOC_SIZE, PCI_DMA_FROMDEVICE) );
+		rxptr->rdes2 = cpu_to_le32(pci_map_single(db->pdev,
+							  skb_tail_pointer(skb),
+							  RX_ALLOC_SIZE,
+							  PCI_DMA_FROMDEVICE));
 		wmb();
 		rxptr->rdes0 = cpu_to_le32(0x80000000);
 		rxptr = rxptr->next_rx_desc;

drivers/net/wan/hdlc_fr.c

@@ -533,7 +533,7 @@ static void fr_lmi_send(struct net_device *dev, int fullrep)
 		skb->protocol = __constant_htons(NLPID_CCITT_ANSI_LMI);
 		fr_hard_header(&skb, LMI_CCITT_ANSI_DLCI);
 	}
-	data = skb->tail;
+	data = skb_tail_pointer(skb);
 	data[i++] = LMI_CALLREF;
 	data[i++] = dce ? LMI_STATUS : LMI_STATUS_ENQUIRY;
 	if (lmi == LMI_ANSI)

drivers/net/wan/lmc/lmc_main.c

@@ -1636,7 +1636,7 @@ static int lmc_rx (struct net_device *dev) /*fold00*/
             if (nsb) {
                 sc->lmc_rxq[i] = nsb;
                 nsb->dev = dev;
-                sc->lmc_rxring[i].buffer1 = virt_to_bus (nsb->tail);
+                sc->lmc_rxring[i].buffer1 = virt_to_bus(skb_tail_pointer(nsb));
             }
             sc->failed_recv_alloc = 1;
             goto skip_packet;
@@ -1679,7 +1679,7 @@ static int lmc_rx (struct net_device *dev) /*fold00*/
             if (nsb) {
                 sc->lmc_rxq[i] = nsb;
                 nsb->dev = dev;
-                sc->lmc_rxring[i].buffer1 = virt_to_bus (nsb->tail);
+                sc->lmc_rxring[i].buffer1 = virt_to_bus(skb_tail_pointer(nsb));
                 /* Transferred to 21140 below */
             }
             else {

drivers/net/wireless/hostap/hostap_80211_rx.c

@@ -922,7 +922,7 @@ void hostap_80211_rx(struct net_device *dev, struct sk_buff *skb,
 		if (frag != 0)
 			flen -= hdrlen;
 
-		if (frag_skb->tail + flen > frag_skb->end) {
+		if (skb_tail_pointer(frag_skb) + flen > frag_skb->end) {
 			printk(KERN_WARNING "%s: host decrypted and "
 			       "reassembled frame did not fit skb\n",
 			       dev->name);

drivers/s390/net/ctcmain.c

@@ -706,7 +706,8 @@ ch_action_txdone(fsm_instance * fi, int event, void *arg)
 			spin_unlock(&ch->collect_lock);
 			return;
 		}
-		ch->trans_skb->tail = ch->trans_skb->data = ch->trans_skb_data;
+		ch->trans_skb->data = ch->trans_skb_data;
+		skb_reset_tail_pointer(ch->trans_skb);
 		ch->trans_skb->len = 0;
 		if (ch->prof.maxmulti < (ch->collect_len + 2))
 			ch->prof.maxmulti = ch->collect_len + 2;
@@ -831,7 +832,8 @@ ch_action_rx(fsm_instance * fi, int event, void *arg)
 		ctc_unpack_skb(ch, skb);
 	}
  again:
-	skb->data = skb->tail = ch->trans_skb_data;
+	skb->data = ch->trans_skb_data;
+	skb_reset_tail_pointer(skb);
 	skb->len = 0;
 	if (ctc_checkalloc_buffer(ch, 1))
 		return;
@@ -2226,7 +2228,8 @@ transmit_skb(struct channel *ch, struct sk_buff *skb)
 		 * IDAL support in CTC is broken, so we have to
 		 * care about skb's above 2G ourselves.
 		 */
-		hi = ((unsigned long) skb->tail + LL_HEADER_LENGTH) >> 31;
+		hi = ((unsigned long)skb_tail_pointer(skb) +
+		      LL_HEADER_LENGTH) >> 31;
 		if (hi) {
 			nskb = alloc_skb(skb->len, GFP_ATOMIC | GFP_DMA);
 			if (!nskb) {
@@ -2262,7 +2265,7 @@ transmit_skb(struct channel *ch, struct sk_buff *skb)
 				return -EBUSY;
 			}
 
-			ch->trans_skb->tail = ch->trans_skb->data;
+			skb_reset_tail_pointer(ch->trans_skb);
 			ch->trans_skb->len = 0;
 			ch->ccw[1].count = skb->len;
 			memcpy(skb_put(ch->trans_skb, skb->len), skb->data,

drivers/s390/net/netiucv.c

@@ -689,7 +689,8 @@ static void conn_action_rx(fsm_instance *fi, int event, void *arg)
 			       msg->length, conn->max_buffsize);
 		return;
 	}
-	conn->rx_buff->data = conn->rx_buff->tail = conn->rx_buff->head;
+	conn->rx_buff->data = conn->rx_buff->head;
+	skb_reset_tail_pointer(conn->rx_buff);
 	conn->rx_buff->len = 0;
 	rc = iucv_message_receive(conn->path, msg, 0, conn->rx_buff->data,
 				  msg->length, NULL);
@@ -735,7 +736,8 @@ static void conn_action_txdone(fsm_instance *fi, int event, void *arg)
 			}
 		}
 	}
-	conn->tx_buff->data = conn->tx_buff->tail = conn->tx_buff->head;
+	conn->tx_buff->data = conn->tx_buff->head;
+	skb_reset_tail_pointer(conn->tx_buff);
 	conn->tx_buff->len = 0;
 	spin_lock_irqsave(&conn->collect_lock, saveflags);
 	while ((skb = skb_dequeue(&conn->collect_queue))) {
@@ -1164,8 +1166,8 @@ static int netiucv_transmit_skb(struct iucv_connection *conn,
 		 * Copy the skb to a new allocated skb in lowmem only if the
 		 * data is located above 2G in memory or tailroom is < 2.
 		 */
-		unsigned long hi =
-			((unsigned long)(skb->tail + NETIUCV_HDRLEN)) >> 31;
+		unsigned long hi = ((unsigned long)(skb_tail_pointer(skb) +
+				    NETIUCV_HDRLEN)) >> 31;
 		int copied = 0;
 		if (hi || (skb_tailroom(skb) < 2)) {
 			nskb = alloc_skb(skb->len + NETIUCV_HDRLEN +