net: TX_RING and packet mmap

New packet socket feature that makes packet socket more efficient for
transmission.

- It reduces number of system call through a PACKET_TX_RING mechanism,
  based on PACKET_RX_RING (Circular buffer allocated in kernel space
  which is mmapped from user space).

- It minimizes CPU copy using fragmented SKB (almost zero copy).
Signed-off-by: Johann Baudy <johann.baudy@gnu-log.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/packet_mmap.txt

@@ -4,16 +4,18 @@
 
 This file documents the CONFIG_PACKET_MMAP option available with the PACKET
 socket interface on 2.4 and 2.6 kernels. This type of sockets is used for 
-capture network traffic with utilities like tcpdump or any other that uses 
-the libpcap library. 
-
-You can find the latest version of this document at
+capture network traffic with utilities like tcpdump or any other that needs
+raw access to network interface.
 
+You can find the latest version of this document at:
     http://pusa.uv.es/~ulisses/packet_mmap/
 
-Please send me your comments to
+Howto can be found at:
+    http://wiki.gnu-log.net (packet_mmap)
 
+Please send your comments to
     Ulisses Alonso Camaró <uaca@i.hate.spam.alumni.uv.es>
+    Johann Baudy <johann.baudy@gnu-log.net>
 
 -------------------------------------------------------------------------------
 + Why use PACKET_MMAP
@@ -25,19 +27,24 @@ to capture each packet, it requires two if you want to get packet's
 timestamp (like libpcap always does).
 
 In the other hand PACKET_MMAP is very efficient. PACKET_MMAP provides a size 
-configurable circular buffer mapped in user space. This way reading packets just 
-needs to wait for them, most of the time there is no need to issue a single 
-system call. By using a shared buffer between the kernel and the user 
-also has the benefit of minimizing packet copies.
-
-It's fine to use PACKET_MMAP to improve the performance of the capture process, 
-but it isn't everything. At least, if you are capturing at high speeds (this 
-is relative to the cpu speed), you should check if the device driver of your 
-network interface card supports some sort of interrupt load mitigation or 
-(even better) if it supports NAPI, also make sure it is enabled.
+configurable circular buffer mapped in user space that can be used to either
+send or receive packets. This way reading packets just needs to wait for them,
+most of the time there is no need to issue a single system call. Concerning
+transmission, multiple packets can be sent through one system call to get the
+highest bandwidth.
+By using a shared buffer between the kernel and the user also has the benefit
+of minimizing packet copies.
+
+It's fine to use PACKET_MMAP to improve the performance of the capture and
+transmission process, but it isn't everything. At least, if you are capturing
+at high speeds (this is relative to the cpu speed), you should check if the
+device driver of your network interface card supports some sort of interrupt
+load mitigation or (even better) if it supports NAPI, also make sure it is
+enabled. For transmission, check the MTU (Maximum Transmission Unit) used and
+supported by devices of your network.
 
 --------------------------------------------------------------------------------
-+ How to use CONFIG_PACKET_MMAP
++ How to use CONFIG_PACKET_MMAP to improve capture process
 --------------------------------------------------------------------------------
 
 From the user standpoint, you should use the higher level libpcap library, which
@@ -57,7 +64,7 @@ the low level details or want to improve libpcap by including PACKET_MMAP
 support.
 
 --------------------------------------------------------------------------------
-+ How to use CONFIG_PACKET_MMAP directly
++ How to use CONFIG_PACKET_MMAP directly to improve capture process
 --------------------------------------------------------------------------------
 
 From the system calls stand point, the use of PACKET_MMAP involves
@@ -66,6 +73,7 @@ the following process:
 
 [setup]     socket() -------> creation of the capture socket
             setsockopt() ---> allocation of the circular buffer (ring)
+                              option: PACKET_RX_RING
             mmap() ---------> mapping of the allocated buffer to the
                               user process
 
@@ -96,6 +104,65 @@ Next I will describe PACKET_MMAP settings and it's constraints,
 also the mapping of the circular buffer in the user process and 
 the use of this buffer.
 
+--------------------------------------------------------------------------------
++ How to use CONFIG_PACKET_MMAP directly to improve transmission process
+--------------------------------------------------------------------------------
+Transmission process is similar to capture as shown below.
+
+[setup]          socket() -------> creation of the transmission socket
+                 setsockopt() ---> allocation of the circular buffer (ring)
+                                   option: PACKET_TX_RING
+                 bind() ---------> bind transmission socket with a network interface
+                 mmap() ---------> mapping of the allocated buffer to the
+                                   user process
+
+[transmission]   poll() ---------> wait for free packets (optional)
+                 send() ---------> send all packets that are set as ready in
+                                   the ring
+                                   The flag MSG_DONTWAIT can be used to return
+                                   before end of transfer.
+
+[shutdown]  close() --------> destruction of the transmission socket and
+                              deallocation of all associated resources.
+
+Binding the socket to your network interface is mandatory (with zero copy) to
+know the header size of frames used in the circular buffer.
+
+As capture, each frame contains two parts:
+
+ --------------------
+| struct tpacket_hdr | Header. It contains the status of
+|                    | of this frame
+|--------------------|
+| data buffer        |
+.                    .  Data that will be sent over the network interface.
+.                    .
+ --------------------
+
+ bind() associates the socket to your network interface thanks to
+ sll_ifindex parameter of struct sockaddr_ll.
+
+ Initialization example:
+
+ struct sockaddr_ll my_addr;
+ struct ifreq s_ifr;
+ ...
+
+ strncpy (s_ifr.ifr_name, "eth0", sizeof(s_ifr.ifr_name));
+
+ /* get interface index of eth0 */
+ ioctl(this->socket, SIOCGIFINDEX, &s_ifr);
+
+ /* fill sockaddr_ll struct to prepare binding */
+ my_addr.sll_family = AF_PACKET;
+ my_addr.sll_protocol = ETH_P_ALL;
+ my_addr.sll_ifindex =  s_ifr.ifr_ifindex;
+
+ /* bind socket to eth0 */
+ bind(this->socket, (struct sockaddr *)&my_addr, sizeof(struct sockaddr_ll));
+
+ A complete tutorial is available at: http://wiki.gnu-log.net/
+
 --------------------------------------------------------------------------------
 + PACKET_MMAP settings
 --------------------------------------------------------------------------------
@@ -103,7 +170,10 @@ the use of this buffer.
 
 To setup PACKET_MMAP from user level code is done with a call like
 
+ - Capture process
      setsockopt(fd, SOL_PACKET, PACKET_RX_RING, (void *) &req, sizeof(req))
+ - Transmission process
+     setsockopt(fd, SOL_PACKET, PACKET_TX_RING, (void *) &req, sizeof(req))
 
 The most significant argument in the previous call is the req parameter, 
 this parameter must to have the following structure:
@@ -117,11 +187,11 @@ this parameter must to have the following structure:
     };
 
 This structure is defined in /usr/include/linux/if_packet.h and establishes a 
-circular buffer (ring) of unswappable memory mapped in the capture process. 
+circular buffer (ring) of unswappable memory.
 Being mapped in the capture process allows reading the captured frames and 
 related meta-information like timestamps without requiring a system call.
 
-Captured frames are grouped in blocks. Each block is a physically contiguous 
+Frames are grouped in blocks. Each block is a physically contiguous
 region of memory and holds tp_block_size/tp_frame_size frames. The total number 
 of blocks is tp_block_nr. Note that tp_frame_nr is a redundant parameter because
 
@@ -336,6 +406,7 @@ struct tpacket_hdr). If this field is 0 means that the frame is ready
 to be used for the kernel, If not, there is a frame the user can read 
 and the following flags apply:
 
++++ Capture process:
      from include/linux/if_packet.h
 
      #define TP_STATUS_COPY          2 
@@ -391,6 +462,37 @@ packets are in the ring:
 It doesn't incur in a race condition to first check the status value and 
 then poll for frames.
 
+
+++ Transmission process
+Those defines are also used for transmission:
+
+     #define TP_STATUS_AVAILABLE        0 // Frame is available
+     #define TP_STATUS_SEND_REQUEST     1 // Frame will be sent on next send()
+     #define TP_STATUS_SENDING          2 // Frame is currently in transmission
+     #define TP_STATUS_WRONG_FORMAT     4 // Frame format is not correct
+
+First, the kernel initializes all frames to TP_STATUS_AVAILABLE. To send a
+packet, the user fills a data buffer of an available frame, sets tp_len to
+current data buffer size and sets its status field to TP_STATUS_SEND_REQUEST.
+This can be done on multiple frames. Once the user is ready to transmit, it
+calls send(). Then all buffers with status equal to TP_STATUS_SEND_REQUEST are
+forwarded to the network device. The kernel updates each status of sent
+frames with TP_STATUS_SENDING until the end of transfer.
+At the end of each transfer, buffer status returns to TP_STATUS_AVAILABLE.
+
+    header->tp_len = in_i_size;
+    header->tp_status = TP_STATUS_SEND_REQUEST;
+    retval = send(this->socket, NULL, 0, 0);
+
+The user can also use poll() to check if a buffer is available:
+(status == TP_STATUS_SENDING)
+
+    struct pollfd pfd;
+    pfd.fd = fd;
+    pfd.revents = 0;
+    pfd.events = POLLOUT;
+    retval = poll(&pfd, 1, timeout);
+
 --------------------------------------------------------------------------------
 + THANKS
 --------------------------------------------------------------------------------

include/linux/if_packet.h

@@ -46,6 +46,8 @@ struct sockaddr_ll
 #define PACKET_VERSION			10
 #define PACKET_HDRLEN			11
 #define PACKET_RESERVE			12
+#define PACKET_TX_RING			13
+#define PACKET_LOSS			14
 
 struct tpacket_stats
 {
@@ -63,14 +65,22 @@ struct tpacket_auxdata
 	__u16		tp_vlan_tci;
 };
 
+/* Rx ring - header status */
+#define TP_STATUS_KERNEL	0x0
+#define TP_STATUS_USER		0x1
+#define TP_STATUS_COPY		0x2
+#define TP_STATUS_LOSING	0x4
+#define TP_STATUS_CSUMNOTREADY	0x8
+
+/* Tx ring - header status */
+#define TP_STATUS_AVAILABLE	0x0
+#define TP_STATUS_SEND_REQUEST	0x1
+#define TP_STATUS_SENDING	0x2
+#define TP_STATUS_WRONG_FORMAT	0x4
+
 struct tpacket_hdr
 {
 	unsigned long	tp_status;
-#define TP_STATUS_KERNEL	0
-#define TP_STATUS_USER		1
-#define TP_STATUS_COPY		2
-#define TP_STATUS_LOSING	4
-#define TP_STATUS_CSUMNOTREADY	8
 	unsigned int	tp_len;
 	unsigned int	tp_snaplen;
 	unsigned short	tp_mac;

include/linux/skbuff.h

@@ -203,6 +203,9 @@ struct skb_shared_info {
 #ifdef CONFIG_HAS_DMA
 	dma_addr_t	dma_maps[MAX_SKB_FRAGS + 1];
 #endif
+	/* Intermediate layers must ensure that destructor_arg
+	 * remains valid until skb destructor */
+	void *		destructor_arg;
 };
 
 /* We divide dataref into two halves.  The higher 16 bits hold references

net/packet/af_packet.c

@@ -39,6 +39,7 @@
  *					will simply extend the hardware address
  *					byte arrays at the end of sockaddr_ll
  *					and packet_mreq.
+ *		Johann Baudy	:	Added TX RING.
  *
  *		This program is free software; you can redistribute it and/or
  *		modify it under the terms of the GNU General Public License
@@ -157,7 +158,25 @@ struct packet_mreq_max
 };
 
 #ifdef CONFIG_PACKET_MMAP
-static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing);
+static int packet_set_ring(struct sock *sk, struct tpacket_req *req,
+		int closing, int tx_ring);
+
+struct packet_ring_buffer {
+	char *			*pg_vec;
+	unsigned int		head;
+	unsigned int		frames_per_block;
+	unsigned int		frame_size;
+	unsigned int		frame_max;
+
+	unsigned int		pg_vec_order;
+	unsigned int		pg_vec_pages;
+	unsigned int		pg_vec_len;
+
+	atomic_t		pending;
+};
+
+struct packet_sock;
+static int tpacket_snd(struct packet_sock *po, struct msghdr *msg);
 #endif
 
 static void packet_flush_mclist(struct sock *sk);
@@ -167,11 +186,8 @@ struct packet_sock {
 	struct sock		sk;
 	struct tpacket_stats	stats;
 #ifdef CONFIG_PACKET_MMAP
-	char *			*pg_vec;
-	unsigned int		head;
-	unsigned int            frames_per_block;
-	unsigned int		frame_size;
-	unsigned int		frame_max;
+	struct packet_ring_buffer	rx_ring;
+	struct packet_ring_buffer	tx_ring;
 	int			copy_thresh;
 #endif
 	struct packet_type	prot_hook;
@@ -185,12 +201,10 @@ struct packet_sock {
 	struct packet_mclist	*mclist;
 #ifdef CONFIG_PACKET_MMAP
 	atomic_t		mapped;
-	unsigned int            pg_vec_order;
-	unsigned int		pg_vec_pages;
-	unsigned int		pg_vec_len;
 	enum tpacket_versions	tp_version;
 	unsigned int		tp_hdrlen;
 	unsigned int		tp_reserve;
+	unsigned int		tp_loss:1;
 #endif
 };
 
@@ -206,36 +220,33 @@ struct packet_skb_cb {
 
 #ifdef CONFIG_PACKET_MMAP
 
-static void *packet_lookup_frame(struct packet_sock *po, unsigned int position,
-				 int status)
+static void __packet_set_status(struct packet_sock *po, void *frame, int status)
 {
-	unsigned int pg_vec_pos, frame_offset;
 	union {
 		struct tpacket_hdr *h1;
 		struct tpacket2_hdr *h2;
 		void *raw;
 	} h;
 
-	pg_vec_pos = position / po->frames_per_block;
-	frame_offset = position % po->frames_per_block;
-
-	h.raw = po->pg_vec[pg_vec_pos] + (frame_offset * po->frame_size);
+	h.raw = frame;
 	switch (po->tp_version) {
 	case TPACKET_V1:
-		if (status != (h.h1->tp_status ? TP_STATUS_USER :
-						TP_STATUS_KERNEL))
-			return NULL;
+		h.h1->tp_status = status;
+		flush_dcache_page(virt_to_page(&h.h1->tp_status));
 		break;
 	case TPACKET_V2:
-		if (status != (h.h2->tp_status ? TP_STATUS_USER :
-						TP_STATUS_KERNEL))
-			return NULL;
+		h.h2->tp_status = status;
+		flush_dcache_page(virt_to_page(&h.h2->tp_status));
 		break;
+	default:
+		printk(KERN_ERR "TPACKET version not supported\n");
+		BUG();
 	}
-	return h.raw;
+
+	smp_wmb();
 }
 
-static void __packet_set_status(struct packet_sock *po, void *frame, int status)
+static int __packet_get_status(struct packet_sock *po, void *frame)
 {
 	union {
 		struct tpacket_hdr *h1;
@@ -243,16 +254,66 @@ static void __packet_set_status(struct packet_sock *po, void *frame, int status)
 		void *raw;
 	} h;
 
+	smp_rmb();
+
 	h.raw = frame;
 	switch (po->tp_version) {
 	case TPACKET_V1:
-		h.h1->tp_status = status;
-		break;
+		flush_dcache_page(virt_to_page(&h.h1->tp_status));
+		return h.h1->tp_status;
 	case TPACKET_V2:
-		h.h2->tp_status = status;
-		break;
+		flush_dcache_page(virt_to_page(&h.h2->tp_status));
+		return h.h2->tp_status;
+	default:
+		printk(KERN_ERR "TPACKET version not supported\n");
+		BUG();
+		return 0;
 	}
 }
+
+static void *packet_lookup_frame(struct packet_sock *po,
+		struct packet_ring_buffer *rb,
+		unsigned int position,
+		int status)
+{
+	unsigned int pg_vec_pos, frame_offset;
+	union {
+		struct tpacket_hdr *h1;
+		struct tpacket2_hdr *h2;
+		void *raw;
+	} h;
+
+	pg_vec_pos = position / rb->frames_per_block;
+	frame_offset = position % rb->frames_per_block;
+
+	h.raw = rb->pg_vec[pg_vec_pos] + (frame_offset * rb->frame_size);
+
+	if (status != __packet_get_status(po, h.raw))
+		return NULL;
+
+	return h.raw;
+}
+
+static inline void *packet_current_frame(struct packet_sock *po,
+		struct packet_ring_buffer *rb,
+		int status)
+{
+	return packet_lookup_frame(po, rb, rb->head, status);
+}
+
+static inline void *packet_previous_frame(struct packet_sock *po,
+		struct packet_ring_buffer *rb,
+		int status)
+{
+	unsigned int previous = rb->head ? rb->head - 1 : rb->frame_max;
+	return packet_lookup_frame(po, rb, previous, status);
+}
+
+static inline void packet_increment_head(struct packet_ring_buffer *buff)
+{
+	buff->head = buff->head != buff->frame_max ? buff->head+1 : 0;
+}
+
 #endif
 
 static inline struct packet_sock *pkt_sk(struct sock *sk)
@@ -648,7 +709,7 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, struct packe
 		macoff = netoff - maclen;
 	}
 
-	if (macoff + snaplen > po->frame_size) {
+	if (macoff + snaplen > po->rx_ring.frame_size) {
 		if (po->copy_thresh &&
 		    atomic_read(&sk->sk_rmem_alloc) + skb->truesize <
 		    (unsigned)sk->sk_rcvbuf) {
@@ -661,16 +722,16 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, struct packe
 			if (copy_skb)
 				skb_set_owner_r(copy_skb, sk);
 		}
-		snaplen = po->frame_size - macoff;
+		snaplen = po->rx_ring.frame_size - macoff;
 		if ((int)snaplen < 0)
 			snaplen = 0;
 	}
 
 	spin_lock(&sk->sk_receive_queue.lock);
-	h.raw = packet_lookup_frame(po, po->head, TP_STATUS_KERNEL);
+	h.raw = packet_current_frame(po, &po->rx_ring, TP_STATUS_KERNEL);
 	if (!h.raw)
 		goto ring_is_full;
-	po->head = po->head != po->frame_max ? po->head+1 : 0;
+	packet_increment_head(&po->rx_ring);
 	po->stats.tp_packets++;
 	if (copy_skb) {
 		status |= TP_STATUS_COPY;
@@ -727,7 +788,6 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, struct packe
 
 	__packet_set_status(po, h.raw, status);
 	smp_mb();
-
 	{
 		struct page *p_start, *p_end;
 		u8 *h_end = h.raw + macoff + snaplen - 1;
@@ -760,10 +820,249 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, struct packe
 	goto drop_n_restore;
 }
 
-#endif
+static void tpacket_destruct_skb(struct sk_buff *skb)
+{
+	struct packet_sock *po = pkt_sk(skb->sk);
+	void * ph;
 
+	BUG_ON(skb == NULL);
 
-static int packet_sendmsg(struct kiocb *iocb, struct socket *sock,
+	if (likely(po->tx_ring.pg_vec)) {
+		ph = skb_shinfo(skb)->destructor_arg;
+		BUG_ON(__packet_get_status(po, ph) != TP_STATUS_SENDING);
+		BUG_ON(atomic_read(&po->tx_ring.pending) == 0);
+		atomic_dec(&po->tx_ring.pending);
+		__packet_set_status(po, ph, TP_STATUS_AVAILABLE);
+	}
+
+	sock_wfree(skb);
+}
+
+static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff * skb,
+		void * frame, struct net_device *dev, int size_max,
+		__be16 proto, unsigned char * addr)
+{
+	union {
+		struct tpacket_hdr *h1;
+		struct tpacket2_hdr *h2;
+		void *raw;
+	} ph;
+	int to_write, offset, len, tp_len, nr_frags, len_max;
+	struct socket *sock = po->sk.sk_socket;
+	struct page *page;
+	void *data;
+	int err;
+
+	ph.raw = frame;
+
+	skb->protocol = proto;
+	skb->dev = dev;
+	skb->priority = po->sk.sk_priority;
+	skb_shinfo(skb)->destructor_arg = ph.raw;
+
+	switch (po->tp_version) {
+	case TPACKET_V2:
+		tp_len = ph.h2->tp_len;
+		break;
+	default:
+		tp_len = ph.h1->tp_len;
+		break;
+	}
+	if (unlikely(tp_len > size_max)) {
+		printk(KERN_ERR "packet size is too long (%d > %d)\n",
+				tp_len, size_max);
+		return -EMSGSIZE;
+	}
+
+	skb_reserve(skb, LL_RESERVED_SPACE(dev));
+	skb_reset_network_header(skb);
+
+	data = ph.raw + po->tp_hdrlen - sizeof(struct sockaddr_ll);
+	to_write = tp_len;
+
+	if (sock->type == SOCK_DGRAM) {
+		err = dev_hard_header(skb, dev, ntohs(proto), addr,
+				NULL, tp_len);
+		if (unlikely(err < 0))
+			return -EINVAL;
+	} else if (dev->hard_header_len ) {
+		/* net device doesn't like empty head */
+		if (unlikely(tp_len <= dev->hard_header_len)) {
+			printk(KERN_ERR "packet size is too short "
+					"(%d < %d)\n", tp_len,
+					dev->hard_header_len);
+			return -EINVAL;
+		}
+
+		skb_push(skb, dev->hard_header_len);
+		err = skb_store_bits(skb, 0, data,
+				dev->hard_header_len);
+		if (unlikely(err))
+			return err;
+
+		data += dev->hard_header_len;
+		to_write -= dev->hard_header_len;
+	}
+
+	err = -EFAULT;
+	page = virt_to_page(data);
+	offset = offset_in_page(data);
+	len_max = PAGE_SIZE - offset;
+	len = ((to_write > len_max) ? len_max : to_write);
+
+	skb->data_len = to_write;
+	skb->len += to_write;
+	skb->truesize += to_write;
+	atomic_add(to_write, &po->sk.sk_wmem_alloc);
+
+	while (likely(to_write)) {
+		nr_frags = skb_shinfo(skb)->nr_frags;
+
+		if (unlikely(nr_frags >= MAX_SKB_FRAGS)) {
+			printk(KERN_ERR "Packet exceed the number "
+					"of skb frags(%lu)\n",
+					MAX_SKB_FRAGS);
+			return -EFAULT;
+		}
+
+		flush_dcache_page(page);
+		get_page(page);
+		skb_fill_page_desc(skb,
+				nr_frags,
+				page++, offset, len);
+		to_write -= len;
+		offset = 0;
+		len_max = PAGE_SIZE;
+		len = ((to_write > len_max) ? len_max : to_write);
+	}
+
+	return tp_len;
+}
+
+static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
+{
+	struct socket *sock;
+	struct sk_buff *skb;
+	struct net_device *dev;
+	__be16 proto;
+	int ifindex, err, reserve = 0;
+	void * ph;
+	struct sockaddr_ll *saddr=(struct sockaddr_ll *)msg->msg_name;
+	int tp_len, size_max;
+	unsigned char *addr;
+	int len_sum = 0;
+	int status = 0;
+
+	sock = po->sk.sk_socket;
+
+	mutex_lock(&po->pg_vec_lock);
+
+	err = -EBUSY;
+	if (saddr == NULL) {
+		ifindex	= po->ifindex;
+		proto	= po->num;
+		addr	= NULL;
+	} else {
+		err = -EINVAL;
+		if (msg->msg_namelen < sizeof(struct sockaddr_ll))
+			goto out;
+		if (msg->msg_namelen < (saddr->sll_halen
+					+ offsetof(struct sockaddr_ll,
+						sll_addr)))
+			goto out;
+		ifindex	= saddr->sll_ifindex;
+		proto	= saddr->sll_protocol;
+		addr	= saddr->sll_addr;
+	}
+
+	dev = dev_get_by_index(sock_net(&po->sk), ifindex);
+	err = -ENXIO;
+	if (unlikely(dev == NULL))
+		goto out;
+
+	reserve = dev->hard_header_len;
+
+	err = -ENETDOWN;
+	if (unlikely(!(dev->flags & IFF_UP)))
+		goto out_put;
+
+	size_max = po->tx_ring.frame_size
+		- sizeof(struct skb_shared_info)
+		- po->tp_hdrlen
+		- LL_ALLOCATED_SPACE(dev)
+		- sizeof(struct sockaddr_ll);
+
+	if (size_max > dev->mtu + reserve)
+		size_max = dev->mtu + reserve;
+
+	do {
+		ph = packet_current_frame(po, &po->tx_ring,
+				TP_STATUS_SEND_REQUEST);
+
+		if (unlikely(ph == NULL)) {
+			schedule();
+			continue;
+		}
+
+		status = TP_STATUS_SEND_REQUEST;
+		skb = sock_alloc_send_skb(&po->sk,
+				LL_ALLOCATED_SPACE(dev)
+				+ sizeof(struct sockaddr_ll),
+				0, &err);
+
+		if (unlikely(skb == NULL))
+			goto out_status;
+
+		tp_len = tpacket_fill_skb(po, skb, ph, dev, size_max, proto,
+				addr);
+
+		if (unlikely(tp_len < 0)) {
+			if (po->tp_loss) {
+				__packet_set_status(po, ph,
+						TP_STATUS_AVAILABLE);
+				packet_increment_head(&po->tx_ring);
+				kfree_skb(skb);
+				continue;
+			} else {
+				status = TP_STATUS_WRONG_FORMAT;
+				err = tp_len;
+				goto out_status;
+			}
+		}
+
+		skb->destructor = tpacket_destruct_skb;
+		__packet_set_status(po, ph, TP_STATUS_SENDING);
+		atomic_inc(&po->tx_ring.pending);
+
+		status = TP_STATUS_SEND_REQUEST;
+		err = dev_queue_xmit(skb);
+		if (unlikely(err > 0 && (err = net_xmit_errno(err)) != 0))
+			goto out_xmit;
+		packet_increment_head(&po->tx_ring);
+		len_sum += tp_len;
+	}
+	while (likely((ph != NULL) || ((!(msg->msg_flags & MSG_DONTWAIT))
+					&& (atomic_read(&po->tx_ring.pending))))
+	      );
+
+	err = len_sum;
+	goto out_put;
+
+out_xmit:
+	skb->destructor = sock_wfree;
+	atomic_dec(&po->tx_ring.pending);
+out_status:
+	__packet_set_status(po, ph, status);
+	kfree_skb(skb);
+out_put:
+	dev_put(dev);
+out:
+	mutex_unlock(&po->pg_vec_lock);
+	return err;
+}
+#endif
+
+static int packet_snd(struct socket *sock,
 			  struct msghdr *msg, size_t len)
 {
 	struct sock *sk = sock->sk;
@@ -854,6 +1153,19 @@ static int packet_sendmsg(struct kiocb *iocb, struct socket *sock,
 	return err;
 }
 
+static int packet_sendmsg(struct kiocb *iocb, struct socket *sock,
+		struct msghdr *msg, size_t len)
+{
+#ifdef CONFIG_PACKET_MMAP
+	struct sock *sk = sock->sk;
+	struct packet_sock *po = pkt_sk(sk);
+	if (po->tx_ring.pg_vec)
+		return tpacket_snd(po, msg);
+	else
+#endif
+		return packet_snd(sock, msg, len);
+}
+
 /*
  *	Close a PACKET socket. This is fairly simple. We immediately go
  *	to 'closed' state and remove our protocol entry in the device list.
@@ -864,6 +1176,9 @@ static int packet_release(struct socket *sock)
 	struct sock *sk = sock->sk;
 	struct packet_sock *po;
 	struct net *net;
+#ifdef CONFIG_PACKET_MMAP
+	struct tpacket_req req;
+#endif
 
 	if (!sk)
 		return 0;
@@ -893,11 +1208,13 @@ static int packet_release(struct socket *sock)
 	packet_flush_mclist(sk);
 
 #ifdef CONFIG_PACKET_MMAP
-	if (po->pg_vec) {
-		struct tpacket_req req;
-		memset(&req, 0, sizeof(req));
-		packet_set_ring(sk, &req, 1);
-	}
+	memset(&req, 0, sizeof(req));
+
+	if (po->rx_ring.pg_vec)
+		packet_set_ring(sk, &req, 1, 0);
+
+	if (po->tx_ring.pg_vec)
+		packet_set_ring(sk, &req, 1, 1);
 #endif
 
 	/*
@@ -1391,7 +1708,7 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 	if (level != SOL_PACKET)
 		return -ENOPROTOOPT;
 
-	switch(optname)	{
+	switch (optname) {
 	case PACKET_ADD_MEMBERSHIP:
 	case PACKET_DROP_MEMBERSHIP:
 	{
@@ -1415,6 +1732,7 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 #ifdef CONFIG_PACKET_MMAP
 	case PACKET_RX_RING:
+	case PACKET_TX_RING:
 	{
 		struct tpacket_req req;
 
@@ -1422,7 +1740,7 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 			return -EINVAL;
 		if (copy_from_user(&req,optval,sizeof(req)))
 			return -EFAULT;
-		return packet_set_ring(sk, &req, 0);
+		return packet_set_ring(sk, &req, 0, optname == PACKET_TX_RING);
 	}
 	case PACKET_COPY_THRESH:
 	{
@@ -1442,7 +1760,7 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 		if (optlen != sizeof(val))
 			return -EINVAL;
-		if (po->pg_vec)
+		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
 			return -EBUSY;
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
@@ -1461,13 +1779,26 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 		if (optlen != sizeof(val))
 			return -EINVAL;
-		if (po->pg_vec)
+		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
 			return -EBUSY;
 		if (copy_from_user(&val, optval, sizeof(val)))
 			return -EFAULT;
 		po->tp_reserve = val;
 		return 0;
 	}
+	case PACKET_LOSS:
+	{
+		unsigned int val;
+
+		if (optlen != sizeof(val))
+			return -EINVAL;
+		if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
+			return -EBUSY;
+		if (copy_from_user(&val, optval, sizeof(val)))
+			return -EFAULT;
+		po->tp_loss = !!val;
+		return 0;
+	}
 #endif
 	case PACKET_AUXDATA:
 	{
@@ -1517,7 +1848,7 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
 	if (len < 0)
 		return -EINVAL;
 
-	switch(optname)	{
+	switch (optname) {
 	case PACKET_STATISTICS:
 		if (len > sizeof(struct tpacket_stats))
 			len = sizeof(struct tpacket_stats);
@@ -1573,6 +1904,12 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
 		val = po->tp_reserve;
 		data = &val;
 		break;
+	case PACKET_LOSS:
+		if (len > sizeof(unsigned int))
+			len = sizeof(unsigned int);
+		val = po->tp_loss;
+		data = &val;
+		break;
 #endif
 	default:
 		return -ENOPROTOOPT;
@@ -1643,7 +1980,7 @@ static int packet_ioctl(struct socket *sock, unsigned int cmd,
 {
 	struct sock *sk = sock->sk;
 
-	switch(cmd) {
+	switch (cmd) {
 		case SIOCOUTQ:
 		{
 			int amount = atomic_read(&sk->sk_wmem_alloc);
@@ -1705,13 +2042,17 @@ static unsigned int packet_poll(struct file * file, struct socket *sock,
 	unsigned int mask = datagram_poll(file, sock, wait);
 
 	spin_lock_bh(&sk->sk_receive_queue.lock);
-	if (po->pg_vec) {
-		unsigned last = po->head ? po->head-1 : po->frame_max;
-
-		if (packet_lookup_frame(po, last, TP_STATUS_USER))
+	if (po->rx_ring.pg_vec) {
+		if (!packet_previous_frame(po, &po->rx_ring, TP_STATUS_KERNEL))
 			mask |= POLLIN | POLLRDNORM;
 	}
 	spin_unlock_bh(&sk->sk_receive_queue.lock);
+	spin_lock_bh(&sk->sk_write_queue.lock);
+	if (po->tx_ring.pg_vec) {
+		if (packet_current_frame(po, &po->tx_ring, TP_STATUS_AVAILABLE))
+			mask |= POLLOUT | POLLWRNORM;
+	}
+	spin_unlock_bh(&sk->sk_write_queue.lock);
 	return mask;
 }
 
@@ -1788,21 +2129,33 @@ static char **alloc_pg_vec(struct tpacket_req *req, int order)
 	goto out;
 }
 
-static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing)
+static int packet_set_ring(struct sock *sk, struct tpacket_req *req,
+		int closing, int tx_ring)
 {
 	char **pg_vec = NULL;
 	struct packet_sock *po = pkt_sk(sk);
 	int was_running, order = 0;
+	struct packet_ring_buffer *rb;
+	struct sk_buff_head *rb_queue;
 	__be16 num;
-	int err = 0;
+	int err;
 
-	if (req->tp_block_nr) {
-		int i;
+	rb = tx_ring ? &po->tx_ring : &po->rx_ring;
+	rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue;
 
-		/* Sanity tests and some calculations */
+	err = -EBUSY;
+	if (!closing) {
+		if (atomic_read(&po->mapped))
+			goto out;
+		if (atomic_read(&rb->pending))
+			goto out;
+	}
 
-		if (unlikely(po->pg_vec))
-			return -EBUSY;
+	if (req->tp_block_nr) {
+		/* Sanity tests and some calculations */
+		err = -EBUSY;
+		if (unlikely(rb->pg_vec))
+			goto out;
 
 		switch (po->tp_version) {
 		case TPACKET_V1:
@@ -1813,42 +2166,35 @@ static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing
 			break;
 		}
 
+		err = -EINVAL;
 		if (unlikely((int)req->tp_block_size <= 0))
-			return -EINVAL;
+			goto out;
 		if (unlikely(req->tp_block_size & (PAGE_SIZE - 1)))
-			return -EINVAL;
+			goto out;
 		if (unlikely(req->tp_frame_size < po->tp_hdrlen +
-						  po->tp_reserve))
-			return -EINVAL;
+					po->tp_reserve))
+			goto out;
 		if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1)))
-			return -EINVAL;
+			goto out;
 
-		po->frames_per_block = req->tp_block_size/req->tp_frame_size;
-		if (unlikely(po->frames_per_block <= 0))
-			return -EINVAL;
-		if (unlikely((po->frames_per_block * req->tp_block_nr) !=
-			     req->tp_frame_nr))
-			return -EINVAL;
+		rb->frames_per_block = req->tp_block_size/req->tp_frame_size;
+		if (unlikely(rb->frames_per_block <= 0))
+			goto out;
+		if (unlikely((rb->frames_per_block * req->tp_block_nr) !=
+					req->tp_frame_nr))
+			goto out;
 
 		err = -ENOMEM;
 		order = get_order(req->tp_block_size);
 		pg_vec = alloc_pg_vec(req, order);
 		if (unlikely(!pg_vec))
 			goto out;
-
-		for (i = 0; i < req->tp_block_nr; i++) {
-			void *ptr = pg_vec[i];
-			int k;
-
-			for (k = 0; k < po->frames_per_block; k++) {
-				__packet_set_status(po, ptr, TP_STATUS_KERNEL);
-				ptr += req->tp_frame_size;
-			}
-		}
-		/* Done */
-	} else {
+	}
+	/* Done */
+	else {
+		err = -EINVAL;
 		if (unlikely(req->tp_frame_nr))
-			return -EINVAL;
+			goto out;
 	}
 
 	lock_sock(sk);
@@ -1872,23 +2218,24 @@ static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing
 	if (closing || atomic_read(&po->mapped) == 0) {
 		err = 0;
 #define XC(a, b) ({ __typeof__ ((a)) __t; __t = (a); (a) = (b); __t; })
-
-		spin_lock_bh(&sk->sk_receive_queue.lock);
-		pg_vec = XC(po->pg_vec, pg_vec);
-		po->frame_max = (req->tp_frame_nr - 1);
-		po->head = 0;
-		po->frame_size = req->tp_frame_size;
-		spin_unlock_bh(&sk->sk_receive_queue.lock);
-
-		order = XC(po->pg_vec_order, order);
-		req->tp_block_nr = XC(po->pg_vec_len, req->tp_block_nr);
-
-		po->pg_vec_pages = req->tp_block_size/PAGE_SIZE;
-		po->prot_hook.func = po->pg_vec ? tpacket_rcv : packet_rcv;
-		skb_queue_purge(&sk->sk_receive_queue);
+		spin_lock_bh(&rb_queue->lock);
+		pg_vec = XC(rb->pg_vec, pg_vec);
+		rb->frame_max = (req->tp_frame_nr - 1);
+		rb->head = 0;
+		rb->frame_size = req->tp_frame_size;
+		spin_unlock_bh(&rb_queue->lock);
+
+		order = XC(rb->pg_vec_order, order);
+		req->tp_block_nr = XC(rb->pg_vec_len, req->tp_block_nr);
+
+		rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE;
+		po->prot_hook.func = (po->rx_ring.pg_vec) ?
+						tpacket_rcv : packet_rcv;
+		skb_queue_purge(rb_queue);
 #undef XC
 		if (atomic_read(&po->mapped))
-			printk(KERN_DEBUG "packet_mmap: vma is busy: %d\n", atomic_read(&po->mapped));
+			printk(KERN_DEBUG "packet_mmap: vma is busy: %d\n",
+						atomic_read(&po->mapped));
 	}
 	mutex_unlock(&po->pg_vec_lock);
 
@@ -1909,11 +2256,13 @@ static int packet_set_ring(struct sock *sk, struct tpacket_req *req, int closing
 	return err;
 }
 
-static int packet_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
+static int packet_mmap(struct file *file, struct socket *sock,
+		struct vm_area_struct *vma)
 {
 	struct sock *sk = sock->sk;
 	struct packet_sock *po = pkt_sk(sk);
-	unsigned long size;
+	unsigned long size, expected_size;
+	struct packet_ring_buffer *rb;
 	unsigned long start;
 	int err = -EINVAL;
 	int i;
@@ -1921,26 +2270,43 @@ static int packet_mmap(struct file *file, struct socket *sock, struct vm_area_st
 	if (vma->vm_pgoff)
 		return -EINVAL;
 
-	size = vma->vm_end - vma->vm_start;
-
 	mutex_lock(&po->pg_vec_lock);
-	if (po->pg_vec == NULL)
+
+	expected_size = 0;
+	for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) {
+		if (rb->pg_vec) {
+			expected_size += rb->pg_vec_len
+						* rb->pg_vec_pages
+						* PAGE_SIZE;
+		}
+	}
+
+	if (expected_size == 0)
 		goto out;
-	if (size != po->pg_vec_len*po->pg_vec_pages*PAGE_SIZE)
+
+	size = vma->vm_end - vma->vm_start;
+	if (size != expected_size)
 		goto out;
 
 	start = vma->vm_start;
-	for (i = 0; i < po->pg_vec_len; i++) {
-		struct page *page = virt_to_page(po->pg_vec[i]);
-		int pg_num;
-
-		for (pg_num = 0; pg_num < po->pg_vec_pages; pg_num++, page++) {
-			err = vm_insert_page(vma, start, page);
-			if (unlikely(err))
-				goto out;
-			start += PAGE_SIZE;
+	for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) {
+		if (rb->pg_vec == NULL)
+			continue;
+
+		for (i = 0; i < rb->pg_vec_len; i++) {
+			struct page *page = virt_to_page(rb->pg_vec[i]);
+			int pg_num;
+
+			for (pg_num = 0; pg_num < rb->pg_vec_pages;
+					pg_num++,page++) {
+				err = vm_insert_page(vma, start, page);
+				if (unlikely(err))
+					goto out;
+				start += PAGE_SIZE;
+			}
 		}
 	}
+
 	atomic_inc(&po->mapped);
 	vma->vm_ops = &packet_mmap_ops;
 	err = 0;