Commit cd9e72b6 authored by Jay Jayatheerthan's avatar Jay Jayatheerthan Committed by Alexei Starovoitov
Browse files

samples/bpf: xdpsock: Add option to specify batch size 



New option to specify batch size for tx, rx and l2fwd has been added. This
allows fine tuning to maximize performance. It is specified using '-b' or
'--batch_size' options. When not specified default is 64.

Signed-off-by: default avatarJay Jayatheerthan <jay.jayatheerthan@intel.com>
Signed-off-by: default avatarAlexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20191220085530.4980-4-jay.jayatheerthan@intel.com
parent 69525588

samples/bpf/xdpsock_user.c

+29 −23
Original line number Diff line number Diff line
@@ -45,7 +45,6 @@ 
#endif



#define NUM_FRAMES (4 * 1024)

#define BATCH_SIZE 64



#define DEBUG_HEXDUMP 0
@@ -68,6 +67,7 @@ static int opt_queue; 
static unsigned long opt_duration;

static unsigned long start_time;

static bool benchmark_done;

static u32 opt_batch_size = 64;

static int opt_poll;

static int opt_interval = 1;

static u32 opt_xdp_bind_flags = XDP_USE_NEED_WAKEUP;
@@ -237,7 +237,6 @@ static void __exit_with_error(int error, const char *file, const char *func, 


#define exit_with_error(error) __exit_with_error(error, __FILE__, __func__, \

						 __LINE__)



static const char pkt_data[] =

	"\x3c\xfd\xfe\x9e\x7f\x71\xec\xb1\xd7\x98\x3a\xc0\x08\x00\x45\x00"

	"\x00\x2e\x00\x00\x00\x00\x40\x11\x88\x97\x05\x08\x07\x08\xc8\x14"
@@ -291,11 +290,10 @@ static void hex_dump(void *pkt, size_t length, u64 addr) 
	printf("\n");

}



static size_t gen_eth_frame(struct xsk_umem_info *umem, u64 addr)

static void gen_eth_frame(struct xsk_umem_info *umem, u64 addr)

{

	memcpy(xsk_umem__get_data(umem->buffer, addr), pkt_data,

	       sizeof(pkt_data) - 1);

	return sizeof(pkt_data) - 1;

}



static struct xsk_umem_info *xsk_configure_umem(void *buffer, u64 size)
@@ -393,6 +391,7 @@ static struct option long_options[] = { 
	{"shared-umem", no_argument, 0, 'M'},

	{"force", no_argument, 0, 'F'},

	{"duration", required_argument, 0, 'd'},

	{"batch-size", required_argument, 0, 'b'},

	{0, 0, 0, 0}

};
@@ -419,8 +418,11 @@ static void usage(const char *prog) 
		"  -F, --force		Force loading the XDP prog\n"

		"  -d, --duration=n	Duration in secs to run command.\n"

		"			Default: forever.\n"

		"  -b, --batch-size=n	Batch size for sending or receiving\n"

		"			packets. Default: %d\n"

		"\n";

	fprintf(stderr, str, prog, XSK_UMEM__DEFAULT_FRAME_SIZE);

	fprintf(stderr, str, prog, XSK_UMEM__DEFAULT_FRAME_SIZE,

		opt_batch_size);

	exit(EXIT_FAILURE);

}
@@ -431,7 +433,7 @@ static void parse_command_line(int argc, char **argv) 
	opterr = 0;



	for (;;) {

		c = getopt_long(argc, argv, "Frtli:q:psSNn:czf:muMd:",

		c = getopt_long(argc, argv, "Frtli:q:pSNn:czf:muMd:b:",

				long_options, &option_index);

		if (c == -1)

			break;
@@ -493,6 +495,9 @@ static void parse_command_line(int argc, char **argv) 
			opt_duration = atoi(optarg);

			opt_duration *= 1000000000;

			break;

		case 'b':

			opt_batch_size = atoi(optarg);

			break;

		default:

			usage(basename(argv[0]));

		}
@@ -540,7 +545,7 @@ static inline void complete_tx_l2fwd(struct xsk_socket_info *xsk, 
	if (!opt_need_wakeup || xsk_ring_prod__needs_wakeup(&xsk->tx))

		kick_tx(xsk);



	ndescs = (xsk->outstanding_tx > BATCH_SIZE) ? BATCH_SIZE :

	ndescs = (xsk->outstanding_tx > opt_batch_size) ? opt_batch_size :

		xsk->outstanding_tx;



	/* re-add completed Tx buffers */
@@ -580,7 +585,7 @@ static inline void complete_tx_only(struct xsk_socket_info *xsk) 
	if (!opt_need_wakeup || xsk_ring_prod__needs_wakeup(&xsk->tx))

		kick_tx(xsk);



	rcvd = xsk_ring_cons__peek(&xsk->umem->cq, BATCH_SIZE, &idx);

	rcvd = xsk_ring_cons__peek(&xsk->umem->cq, opt_batch_size, &idx);

	if (rcvd > 0) {

		xsk_ring_cons__release(&xsk->umem->cq, rcvd);

		xsk->outstanding_tx -= rcvd;
@@ -594,7 +599,7 @@ static void rx_drop(struct xsk_socket_info *xsk, struct pollfd *fds) 
	u32 idx_rx = 0, idx_fq = 0;

	int ret;



	rcvd = xsk_ring_cons__peek(&xsk->rx, BATCH_SIZE, &idx_rx);

	rcvd = xsk_ring_cons__peek(&xsk->rx, opt_batch_size, &idx_rx);

	if (!rcvd) {

		if (xsk_ring_prod__needs_wakeup(&xsk->umem->fq))

			ret = poll(fds, num_socks, opt_timeout);
@@ -655,23 +660,24 @@ static void rx_drop_all(void) 
static void tx_only(struct xsk_socket_info *xsk, u32 frame_nb)

{

	u32 idx;



	if (xsk_ring_prod__reserve(&xsk->tx, BATCH_SIZE, &idx) == BATCH_SIZE) {

	unsigned int i;



		for (i = 0; i < BATCH_SIZE; i++) {

			xsk_ring_prod__tx_desc(&xsk->tx, idx + i)->addr	=

				(frame_nb + i) << XSK_UMEM__DEFAULT_FRAME_SHIFT;

			xsk_ring_prod__tx_desc(&xsk->tx, idx + i)->len =

				sizeof(pkt_data) - 1;

	while (xsk_ring_prod__reserve(&xsk->tx, opt_batch_size, &idx) <

				      opt_batch_size) {

		complete_tx_only(xsk);

	}



		xsk_ring_prod__submit(&xsk->tx, BATCH_SIZE);

		xsk->outstanding_tx += BATCH_SIZE;

		frame_nb += BATCH_SIZE;

		frame_nb %= NUM_FRAMES;

	for (i = 0; i < opt_batch_size; i++) {

		struct xdp_desc *tx_desc = xsk_ring_prod__tx_desc(&xsk->tx,

								  idx + i);

		tx_desc->addr = (frame_nb + i) << XSK_UMEM__DEFAULT_FRAME_SHIFT;

		tx_desc->len = sizeof(pkt_data) - 1;

	}



	xsk_ring_prod__submit(&xsk->tx, opt_batch_size);

	xsk->outstanding_tx += opt_batch_size;

	frame_nb += opt_batch_size;

	frame_nb %= NUM_FRAMES;

	complete_tx_only(xsk);

}
@@ -712,7 +718,7 @@ static void l2fwd(struct xsk_socket_info *xsk, struct pollfd *fds) 


	complete_tx_l2fwd(xsk, fds);



	rcvd = xsk_ring_cons__peek(&xsk->rx, BATCH_SIZE, &idx_rx);

	rcvd = xsk_ring_cons__peek(&xsk->rx, opt_batch_size, &idx_rx);

	if (!rcvd) {

		if (xsk_ring_prod__needs_wakeup(&xsk->umem->fq))

			ret = poll(fds, num_socks, opt_timeout);

CRA Git | Maintained and supported by SUSTech CRA and CCSE