Message-ID: <1272548980.2222.87.camel@edumazet-laptop>
Date:	Thu, 29 Apr 2010 15:49:40 +0200
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	hadi@...erus.ca
Cc:	Changli Gao <xiaosuo@...il.com>,
	David Miller <davem@...emloft.net>, therbert@...gle.com,
	shemminger@...tta.com, netdev@...r.kernel.org,
	Eilon Greenstein <eilong@...adcom.com>,
	Brian Bloniarz <bmb@...enacr.com>
Subject: Re: [PATCH net-next-2.6] net: speedup udp receive path

Le jeudi 29 avril 2010 à 09:37 -0400, jamal a écrit :
> On Thu, 2010-04-29 at 15:21 +0200, Eric Dumazet wrote:
> 
> 
> > 
> > You could try following program :
> > 
> 
> Will do later today (test machine is not on the network and is about 20
> minutes from here; so worst case i will get you results by end of day)
> I guess this program is good enough since it tells me the system wide
> ipi count - what my patch did was also to break it down by which cpu got
> how many IPIs (served to check if there was uneven distribution)
> 
> > 
> > Is your application mono threaded and receiving data to 8 sockets ?
> > 
> 
> I fork one instance per detected cpu and bind to different ports each
> time. Example bind to port 8200 on cpu0, 8201 on cpu1, etc.
> 

I guess this is the problem ;)

With RPS, you should not bind your threads to cpu.
This is the rps hash who will decide for you.


I am using following program :

/*
 *  Usage: udpsink [ -p baseport] nbports
 *
 */
#include <sys/socket.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>
#include <unistd.h>
#include <stdlib.h>
#include <fcntl.h>

struct worker_data {
	int fd;
	unsigned long pack_count;
	unsigned long bytes_count;
	unsigned long _padd[16 - 3]; /* alignment */ 
};

void usage(int code)
{
	fprintf(stderr, "Usage: udpsink [-p baseport] nbports\n");
	exit(code);
}

void *worker_func(void *arg)
{
	struct worker_data *wdata = (struct worker_data *)arg;
	char buffer[4096];
	struct sockaddr_in addr;
	int lu;

	while (1) {
		socklen_t len = sizeof(addr);
		lu = recvfrom(wdata->fd, buffer, sizeof(buffer), 0, (struct sockaddr *)&addr, &len);
		if (lu > 0) {
			wdata->pack_count++;
			wdata->bytes_count += lu;
		}
	}
}

int main(int argc, char *argv[])
{
int c;
int baseport = 4000;
int nbthreads;
struct worker_data *wdata;
unsigned long ototal = 0;
int concurrent = 0;
int verbose = 0;
int i;
	while ((c = getopt(argc, argv, "cvp:")) != -1) {
		if (c == 'p')
			baseport = atoi(optarg);
		else if (c == 'c')
			concurrent = 1;
		else if (c == 'v')
			verbose++;
		else usage(1);
	}
	if (optind == argc)
		usage(1);
	nbthreads = atoi(argv[optind]);
	wdata = calloc(sizeof(struct worker_data), nbthreads);
	if (!wdata) {
		perror("calloc");
		return 1;
	}
	for (i = 0; i < nbthreads; i++) {
		struct sockaddr_in addr;
		pthread_t tid;

		if (i && concurrent) {
			wdata[i].fd = wdata[0].fd ;
		} else {
			wdata[i].fd = socket(PF_INET, SOCK_DGRAM, 0);
			if (wdata[i].fd == -1) {
				perror("socket");
				return 1;
			}
			memset(&addr, 0, sizeof(addr));
			addr.sin_family = AF_INET;
//			addr.sin_addr.s_addr = inet_addr(argv[optind]);
			addr.sin_port = htons(baseport + i);
			if (bind(wdata[i].fd, (struct sockaddr *) &addr, sizeof(addr)) < 0) {
				perror("bind");
				return 1;
				}
//			fcntl(wdata[i].fd, F_SETFL, O_NDELAY);
			}
		pthread_create(&tid, NULL, worker_func, wdata + i);
	}
	for (;;) {
		unsigned long total;
		long delta;

		sleep(1);
		total = 0;
		for (i = 0; i < nbthreads;i++) {
			total += wdata[i].pack_count;
		}
		delta = total - ototal;
		if (delta) {
			printf("%lu pps (%lu", delta, total);
			if (verbose) {
				for (i = 0; i < nbthreads;i++) { 
					if (wdata[i].pack_count)
						printf(" %d:%lu", i, wdata[i].pack_count);
				}
			}
			printf(")\n");
		}
		ototal = total;
	}
}



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists