Merge branch 'TPACKET_V3-TX_RING-support'

Sowmini Varadhan says:

====================
TPACKET_V3 TX_RING support

This patch series allows an application to use a single PF_PACKET
descriptor and leverage the best implementations of TX_RING
and RX_RING that exist today.

Patch 1 adds the kernel/Documentation changes for TX_RING
support and patch2 adds the associated test case in selftests.

Changes since v2: additional sanity checks for setsockopt
input for TX_RING/TPACKET_V3. Refactored psock_tpacket.c
test code to avoid code duplication from V2.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/packet_mmap.txt

@@ -565,7 +565,7 @@ TPACKET_V1 --> TPACKET_V2:
 		   (void *)hdr + TPACKET_ALIGN(sizeof(struct tpacket_hdr))
 
 TPACKET_V2 --> TPACKET_V3:
-	- Flexible buffer implementation:
+	- Flexible buffer implementation for RX_RING:
 		1. Blocks can be configured with non-static frame-size
 		2. Read/poll is at a block-level (as opposed to packet-level)
 		3. Added poll timeout to avoid indefinite user-space wait
@@ -574,7 +574,12 @@ TPACKET_V2 --> TPACKET_V3:
 			4.1 block::timeout
 			4.2 tpkt_hdr::sk_rxhash
 	- RX Hash data available in user space
-	- Currently only RX_RING available
+	- TX_RING semantics are conceptually similar to TPACKET_V2;
+	  use tpacket3_hdr instead of tpacket2_hdr, and TPACKET3_HDRLEN
+	  instead of TPACKET2_HDRLEN. In the current implementation,
+	  the tp_next_offset field in the tpacket3_hdr MUST be set to
+	  zero, indicating that the ring does not hold variable sized frames.
+	  Packets with non-zero values of tp_next_offset will be dropped.
 
 -------------------------------------------------------------------------------
 + AF_PACKET fanout mode

net/packet/af_packet.c

@@ -409,6 +409,9 @@ static void __packet_set_status(struct packet_sock *po, void *frame, int status)
 		flush_dcache_page(pgv_to_page(&h.h2->tp_status));
 		break;
 	case TPACKET_V3:
+		h.h3->tp_status = status;
+		flush_dcache_page(pgv_to_page(&h.h3->tp_status));
+		break;
 	default:
 		WARN(1, "TPACKET version not supported.\n");
 		BUG();
@@ -432,6 +435,8 @@ static int __packet_get_status(struct packet_sock *po, void *frame)
 		flush_dcache_page(pgv_to_page(&h.h2->tp_status));
 		return h.h2->tp_status;
 	case TPACKET_V3:
+		flush_dcache_page(pgv_to_page(&h.h3->tp_status));
+		return h.h3->tp_status;
 	default:
 		WARN(1, "TPACKET version not supported.\n");
 		BUG();
@@ -2497,6 +2502,13 @@ static int tpacket_parse_header(struct packet_sock *po, void *frame,
 	ph.raw = frame;
 
 	switch (po->tp_version) {
+	case TPACKET_V3:
+		if (ph.h3->tp_next_offset != 0) {
+			pr_warn_once("variable sized slot not supported");
+			return -EINVAL;
+		}
+		tp_len = ph.h3->tp_len;
+		break;
 	case TPACKET_V2:
 		tp_len = ph.h2->tp_len;
 		break;
@@ -2516,6 +2528,9 @@ static int tpacket_parse_header(struct packet_sock *po, void *frame,
 		off_max = po->tx_ring.frame_size - tp_len;
 		if (po->sk.sk_type == SOCK_DGRAM) {
 			switch (po->tp_version) {
+			case TPACKET_V3:
+				off = ph.h3->tp_net;
+				break;
 			case TPACKET_V2:
 				off = ph.h2->tp_net;
 				break;
@@ -2525,6 +2540,9 @@ static int tpacket_parse_header(struct packet_sock *po, void *frame,
 			}
 		} else {
 			switch (po->tp_version) {
+			case TPACKET_V3:
+				off = ph.h3->tp_mac;
+				break;
 			case TPACKET_V2:
 				off = ph.h2->tp_mac;
 				break;
@@ -4113,11 +4131,6 @@ static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u,
 	struct tpacket_req *req = &req_u->req;
 
 	lock_sock(sk);
-	/* Opening a Tx-ring is NOT supported in TPACKET_V3 */
-	if (!closing && tx_ring && (po->tp_version > TPACKET_V2)) {
-		net_warn_ratelimited("Tx-ring is not supported.\n");
-		goto out;
-	}
 
 	rb = tx_ring ? &po->tx_ring : &po->rx_ring;
 	rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue;
@@ -4177,11 +4190,19 @@ static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u,
 			goto out;
 		switch (po->tp_version) {
 		case TPACKET_V3:
-		/* Transmit path is not supported. We checked
-		 * it above but just being paranoid
-		 */
-			if (!tx_ring)
+			/* Block transmit is not supported yet */
+			if (!tx_ring) {
 				init_prb_bdqc(po, rb, pg_vec, req_u);
+			} else {
+				struct tpacket_req3 *req3 = &req_u->req3;
+
+				if (req3->tp_retire_blk_tov ||
+				    req3->tp_sizeof_priv ||
+				    req3->tp_feature_req_word) {
+					err = -EINVAL;
+					goto out;
+				}
+			}
 			break;
 		default:
 			break;

tools/testing/selftests/net/psock_tpacket.c

@@ -311,20 +311,33 @@ static inline void __v2_tx_user_ready(struct tpacket2_hdr *hdr)
 	__sync_synchronize();
 }
 
-static inline int __v1_v2_tx_kernel_ready(void *base, int version)
+static inline int __v3_tx_kernel_ready(struct tpacket3_hdr *hdr)
+{
+	return !(hdr->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING));
+}
+
+static inline void __v3_tx_user_ready(struct tpacket3_hdr *hdr)
+{
+	hdr->tp_status = TP_STATUS_SEND_REQUEST;
+	__sync_synchronize();
+}
+
+static inline int __tx_kernel_ready(void *base, int version)
 {
 	switch (version) {
 	case TPACKET_V1:
 		return __v1_tx_kernel_ready(base);
 	case TPACKET_V2:
 		return __v2_tx_kernel_ready(base);
+	case TPACKET_V3:
+		return __v3_tx_kernel_ready(base);
 	default:
 		bug_on(1);
 		return 0;
 	}
 }
 
-static inline void __v1_v2_tx_user_ready(void *base, int version)
+static inline void __tx_user_ready(void *base, int version)
 {
 	switch (version) {
 	case TPACKET_V1:
@@ -333,6 +346,9 @@ static inline void __v1_v2_tx_user_ready(void *base, int version)
 	case TPACKET_V2:
 		__v2_tx_user_ready(base);
 		break;
+	case TPACKET_V3:
+		__v3_tx_user_ready(base);
+		break;
 	}
 }
 
@@ -348,7 +364,22 @@ static void __v1_v2_set_packet_loss_discard(int sock)
 	}
 }
 
-static void walk_v1_v2_tx(int sock, struct ring *ring)
+static inline void *get_next_frame(struct ring *ring, int n)
+{
+	uint8_t *f0 = ring->rd[0].iov_base;
+
+	switch (ring->version) {
+	case TPACKET_V1:
+	case TPACKET_V2:
+		return ring->rd[n].iov_base;
+	case TPACKET_V3:
+		return f0 + (n * ring->req3.tp_frame_size);
+	default:
+		bug_on(1);
+	}
+}
+
+static void walk_tx(int sock, struct ring *ring)
 {
 	struct pollfd pfd;
 	int rcv_sock, ret;
@@ -360,9 +391,19 @@ static void walk_v1_v2_tx(int sock, struct ring *ring)
 		.sll_family = PF_PACKET,
 		.sll_halen = ETH_ALEN,
 	};
+	int nframes;
+
+	/* TPACKET_V{1,2} sets up the ring->rd* related variables based
+	 * on frames (e.g., rd_num is tp_frame_nr) whereas V3 sets these
+	 * up based on blocks (e.g, rd_num is  tp_block_nr)
+	 */
+	if (ring->version <= TPACKET_V2)
+		nframes = ring->rd_num;
+	else
+		nframes = ring->req3.tp_frame_nr;
 
 	bug_on(ring->type != PACKET_TX_RING);
-	bug_on(ring->rd_num < NUM_PACKETS);
+	bug_on(nframes < NUM_PACKETS);
 
 	rcv_sock = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
 	if (rcv_sock == -1) {
@@ -388,10 +429,11 @@ static void walk_v1_v2_tx(int sock, struct ring *ring)
 	create_payload(packet, &packet_len);
 
 	while (total_packets > 0) {
-		while (__v1_v2_tx_kernel_ready(ring->rd[frame_num].iov_base,
-					       ring->version) &&
+		void *next = get_next_frame(ring, frame_num);
+
+		while (__tx_kernel_ready(next, ring->version) &&
 		       total_packets > 0) {
-			ppd.raw = ring->rd[frame_num].iov_base;
+			ppd.raw = next;
 
 			switch (ring->version) {
 			case TPACKET_V1:
@@ -413,14 +455,27 @@ static void walk_v1_v2_tx(int sock, struct ring *ring)
 				       packet_len);
 				total_bytes += ppd.v2->tp_h.tp_snaplen;
 				break;
+			case TPACKET_V3: {
+				struct tpacket3_hdr *tx = next;
+
+				tx->tp_snaplen = packet_len;
+				tx->tp_len = packet_len;
+				tx->tp_next_offset = 0;
+
+				memcpy((uint8_t *)tx + TPACKET3_HDRLEN -
+				       sizeof(struct sockaddr_ll), packet,
+				       packet_len);
+				total_bytes += tx->tp_snaplen;
+				break;
+			}
 			}
 
 			status_bar_update();
 			total_packets--;
 
-			__v1_v2_tx_user_ready(ppd.raw, ring->version);
+			__tx_user_ready(next, ring->version);
 
-			frame_num = (frame_num + 1) % ring->rd_num;
+			frame_num = (frame_num + 1) % nframes;
 		}
 
 		poll(&pfd, 1, 1);
@@ -460,7 +515,7 @@ static void walk_v1_v2(int sock, struct ring *ring)
 	if (ring->type == PACKET_RX_RING)
 		walk_v1_v2_rx(sock, ring);
 	else
-		walk_v1_v2_tx(sock, ring);
+		walk_tx(sock, ring);
 }
 
 static uint64_t __v3_prev_block_seq_num = 0;
@@ -583,7 +638,7 @@ static void walk_v3(int sock, struct ring *ring)
 	if (ring->type == PACKET_RX_RING)
 		walk_v3_rx(sock, ring);
 	else
-		bug_on(1);
+		walk_tx(sock, ring);
 }
 
 static void __v1_v2_fill(struct ring *ring, unsigned int blocks)
@@ -602,12 +657,13 @@ static void __v1_v2_fill(struct ring *ring, unsigned int blocks)
 	ring->flen = ring->req.tp_frame_size;
 }
 
-static void __v3_fill(struct ring *ring, unsigned int blocks)
+static void __v3_fill(struct ring *ring, unsigned int blocks, int type)
 {
-	ring->req3.tp_retire_blk_tov = 64;
-	ring->req3.tp_sizeof_priv = 0;
-	ring->req3.tp_feature_req_word = TP_FT_REQ_FILL_RXHASH;
-
+	if (type == PACKET_RX_RING) {
+		ring->req3.tp_retire_blk_tov = 64;
+		ring->req3.tp_sizeof_priv = 0;
+		ring->req3.tp_feature_req_word = TP_FT_REQ_FILL_RXHASH;
+	}
 	ring->req3.tp_block_size = getpagesize() << 2;
 	ring->req3.tp_frame_size = TPACKET_ALIGNMENT << 7;
 	ring->req3.tp_block_nr = blocks;
@@ -641,7 +697,7 @@ static void setup_ring(int sock, struct ring *ring, int version, int type)
 		break;
 
 	case TPACKET_V3:
-		__v3_fill(ring, blocks);
+		__v3_fill(ring, blocks, type);
 		ret = setsockopt(sock, SOL_PACKET, type, &ring->req3,
 				 sizeof(ring->req3));
 		break;
@@ -796,6 +852,7 @@ int main(void)
 	ret |= test_tpacket(TPACKET_V2, PACKET_TX_RING);
 
 	ret |= test_tpacket(TPACKET_V3, PACKET_RX_RING);
+	ret |= test_tpacket(TPACKET_V3, PACKET_TX_RING);
 
 	if (ret)
 		return 1;