@@ -20,11 +20,14 @@
#include <string.h>
#include <errno.h>
#include <unistd.h>
+#include <sys/ioctl.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/wait.h>
#include <sys/time.h>
#include <sys/poll.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
#include <limits.h>
#include <getopt.h>
#include <signal.h>
@@ -46,6 +49,7 @@ static unsigned int fifo = 0;
static unsigned int process_mode = PROCESS_MODE;
static int use_pipes = 0;
+static int use_inet = 0;
struct sender_context {
unsigned int num_fds;
@@ -99,6 +103,7 @@ static void print_usage_exit(int error)
"-h --help print this message\n"
"-l --loops=LOOPS how many message should be send\n"
"-p --pipe send data via a pipe\n"
+ "-i --inet send data via a inet tcp connection\n"
"-s --datasize=SIZE message size\n"
"-T --threads use POSIX threads\n"
"-P --process use fork (default)\n"
@@ -106,11 +111,52 @@ static void print_usage_exit(int error)
exit(error);
}
+static int inet_socketpair(int fds[2])
+{
+ int s1, s2;
+ struct sockaddr_in sin;
+ unsigned long ul = 1;
+
+ if ((s1 = socket(AF_INET, SOCK_STREAM, 0)) < 0)
+ barf("socket");
+ if ((s2 = socket(AF_INET, SOCK_STREAM, 0)) < 0)
+ barf("socket");
+
+ socklen_t len = sizeof(sin);
+ bzero(&sin, len);
+ sin.sin_family = AF_INET;
+ sin.sin_port = 0;
+ sin.sin_addr.s_addr = inet_addr("127.0.0.1");
+
+ if (bind(s1, &sin, len) < 0)
+ barf("bind");
+ getsockname(s1, &sin, &len);
+ if (listen(s1, 10) < 0)
+ barf("listen");
+ if (ioctl(s2, FIONBIO, &ul) < 0)
+ barf("ioctl");
+ if (ioctl(s1, FIONBIO, &ul) < 0)
+ barf("ioctl");
+ connect(s2, &sin, len);
+ if ((fds[0] = accept(s1, &sin, &len)) < 0)
+ barf("accept");
+ ul = 0;
+ if (ioctl(s2, FIONBIO, &ul) < 0)
+ barf("ioctl");
+ fds[1] = s2;
+ close(s1);
+
+ return 0;
+}
+
static void fdpair(int fds[2])
{
if (use_pipes) {
if (pipe(fds) == 0)
return;
+ } else if (use_inet) {
+ if (inet_socketpair(fds) == 0)
+ return;
} else {
if (socketpair(AF_UNIX, SOCK_STREAM, 0, fds) == 0)
return;
@@ -364,13 +410,14 @@ static void process_options(int argc, char *argv[])
{"help", no_argument, NULL, 'h'},
{"loops", required_argument, NULL, 'l'},
{"pipe", no_argument, NULL, 'p'},
+ {"inet", no_argument, NULL, 'i'},
{"datasize", required_argument, NULL, 's'},
{"threads", no_argument, NULL, 'T'},
{"processes", no_argument, NULL, 'P'},
{NULL, 0, NULL, 0}
};
- int c = getopt_long(argc, argv, "f:Fg:hl:ps:TP",
+ int c = getopt_long(argc, argv, "f:Fg:hl:pis:TP",
longopts, &optind);
if (c == -1) {
break;
@@ -402,6 +449,9 @@ static void process_options(int argc, char *argv[])
case 'p':
use_pipes = 1;
break;
+ case 'i':
+ use_inet = 1;
+ break;
case 's':
if (!(argv[optind] && (datasize = atoi(optarg)) > 0)) {
fprintf(stderr, "%s: --datasize|-s requires an integer > 0\n",
We observerved strange performance regression with hackbench, but only with its default mode instead of pipe mode. It turned out it was unix domain socket to blame rather than the scheduler. Because on older kernels, there was a big lock. And modifications to the scheduler made the contension worse. However pipe mode is much different from the default af_unix mode. So it's hard to prove it's really the case. So we think it might be a good idea to have af_inet mode for hackbench. After implementing it, we compare these three modes. And the result shows as what we expect, that only af_unix mode has regression, the others don't. This proves adding af_inet mode is valuable. Signed-off-by: Yihao Wu <wuyihao@linux.alibaba.com> --- src/hackbench/hackbench.c | 57 ++++++++++++++++++++++++++++++++++++++- 1 file changed, 56 insertions(+), 1 deletion(-) argv[0]); @@ -418,6 +468,11 @@ static void process_options(int argc, char *argv[]) print_usage_exit(1); } } + + if (use_pipes && use_inet) { + fprintf(stderr, "%s: --pipe|-p and --inet|-i cannot be used together\n", argv[0]); + print_usage_exit(1); + } } void sigcatcher(int sig) { -- 2.32.0.604.gb1f3e1269