passt: Align pkt_buf to PAGE_SIZE (start and size), try to fit in huge pages

If transparent huge pages are available, madvise() will do the trick.

While at it, decrease EPOLL_EVENTS for the main loop from 10 to 8,
for slightly better socket fairness.

Signed-off-by: Stefano Brivio <sbrivio@redhat.com>

4 files changed, 10 insertions, 3 deletions

diff --git a/Makefile b/Makefile
index c7cbbb7..ac809df 100644
--- a/Makefile
+++ b/Makefile
@@ -1,5 +1,6 @@
 CFLAGS += -Wall -Wextra -pedantic
 CFLAGS += -DRLIMIT_STACK_VAL=$(shell ulimit -s)
+CFLAGS += -DPAGE_SIZE=$(shell getconf PAGE_SIZE)
 
 prefix ?= /usr/local
 
diff --git a/passt.c b/passt.c
index 3604fa1..6645e45 100644
--- a/passt.c
+++ b/passt.c
@@ -28,6 +28,7 @@
 #include <sys/stat.h>
 #include <dirent.h>
 #include <fcntl.h>
+#include <sys/mman.h>
 #include <sys/resource.h>
 #include <sys/uio.h>
 #include <sys/wait.h>
@@ -61,12 +62,12 @@
 #include "tap.h"
 #include "conf.h"
 
-#define EPOLL_EVENTS		10
+#define EPOLL_EVENTS		8
 
 #define __TIMER_INTERVAL	MIN(TCP_TIMER_INTERVAL, UDP_TIMER_INTERVAL)
 #define TIMER_INTERVAL		MIN(__TIMER_INTERVAL, ICMP_TIMER_INTERVAL)
 
-char pkt_buf			[PKT_BUF_BYTES];
+char pkt_buf[PKT_BUF_BYTES]	__attribute__ ((aligned(PAGE_SIZE)));
 
 char *ip_proto_str[IPPROTO_SCTP + 1] = {
 	[IPPROTO_ICMP]		= "ICMP",
@@ -323,6 +324,9 @@ int main(int argc, char **argv)
 		log_name = "passt";
 	}
 
+	if (madvise(pkt_buf, TAP_BUF_BYTES, MADV_HUGEPAGE))
+		perror("madvise");
+
 	openlog(log_name, 0, LOG_DAEMON);
 
 	setlogmask(LOG_MASK(LOG_EMERG));
diff --git a/passt.h b/passt.h
index 649d4fc..bb41e90 100644
--- a/passt.h
+++ b/passt.h
@@ -45,7 +45,8 @@ union epoll_ref {
 	uint64_t u64;
 };
 
-#define TAP_BUF_BYTES		((ETH_MAX_MTU + sizeof(uint32_t)) * 256)
+#define TAP_BUF_BYTES							\
+	ROUND_DOWN(((ETH_MAX_MTU + sizeof(uint32_t)) * 256), PAGE_SIZE)
 #define TAP_BUF_FILL		(TAP_BUF_BYTES - ETH_MAX_MTU - sizeof(uint32_t))
 #define TAP_MSGS							\
 	DIV_ROUND_UP(TAP_BUF_BYTES, ETH_ZLEN - 2 * ETH_ALEN + sizeof(uint32_t))
diff --git a/util.h b/util.h
index 2e7699c..6402f41 100644
--- a/util.h
+++ b/util.h
@@ -25,6 +25,7 @@ void debug(const char *format, ...);
 #endif
 
 #define DIV_ROUND_UP(n, d)	(((n) + (d) - 1) / (d))
+#define ROUND_DOWN(x, y)	((x) & ~((y) - 1))
 
 #define SWAP(a, b)							\
 	do {								\