git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v2] git daemon: avoid waking up too often

From
Johannes Schindelin <johannes.schindelin@gmx.de>
Date
Jul 22, 2008, 22:03 UTC
Message-ID
<alpine.DEB.1.00.0807222302440.8986@racer>
In-Reply-To
<alpine.DEB.1.00.0807222251570.8986@racer>

To avoid waking up unnecessarily, a pipe is set up that is only ever written to by child_handler(), when a child disconnects, as suggested per Junio.

This avoids waking up the main process every second to see if a child was disconnected.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
	Darn.  Forgot to commit after fixing the compiler warning:
	The now-unused variable timeout is gone.
	That is the only difference to the version I have running in
	production ;-)
 daemon.c |   25 +++++++++++--------------
 1 files changed, 11 insertions(+), 14 deletions(-)
diff --git a/daemon.c b/daemon.c
index 7df41a6..4540e8d 100644
--- a/daemon.c
+++ b/daemon.c
@@ -16,6 +16,7 @@
 static int log_syslog;
 static int verbose;
 static int reuseaddr;
+static int child_handler_pipe[2];
 
 static const char daemon_usage[] =
 "git daemon [--verbose] [--syslog] [--export-all]\n"
@@ -788,6 +789,7 @@ static void child_handler(int signo)
 				pid = -pid;
 			dead_child[reaped % MAX_CHILDREN] = pid;
 			children_reaped = reaped + 1;
+			write(child_handler_pipe[1], &status, 1);
 			continue;
 		}
 		break;
@@ -933,29 +935,24 @@ static int service_loop(int socknum, int *socklist)
 	struct pollfd *pfd;
 	int i;
 
-	pfd = xcalloc(socknum, sizeof(struct pollfd));
+	if (pipe(child_handler_pipe) < 0)
+		die ("Could not set up pipe for child handler");
+
+	pfd = xcalloc(socknum + 1, sizeof(struct pollfd));
 
 	for (i = 0; i < socknum; i++) {
 		pfd[i].fd = socklist[i];
 		pfd[i].events = POLLIN;
 	}
+	pfd[socknum].fd = child_handler_pipe[0];
+	pfd[socknum].events = POLLIN;
 
 	signal(SIGCHLD, child_handler);
 
 	for (;;) {
 		int i;
-		int timeout;
 
-		/*
-		 * This 1-sec timeout could lead to idly looping but it is
-		 * here so that children culled in child_handler() are reported
-		 * without too much delay.  We could probably set up a pipe
-		 * to ourselves that we poll, and write to the fd from child_handler()
-		 * to wake us up (and consume it when the poll() returns...
-		 */
-		timeout = (children_spawned != children_deleted) ? 1000 : -1;
-		i = poll(pfd, socknum, timeout);
-		if (i < 0) {
+		if (poll(pfd, socknum + 1, -1) < 0) {
 			if (errno != EINTR) {
 				error("poll failed, resuming: %s",
 				      strerror(errno));
@@ -963,9 +960,9 @@ static int service_loop(int socknum, int *socklist)
 			}
 			continue;
 		}
-		if (i == 0) {
+		if (pfd[socknum].revents & POLLIN) {
+			read(child_handler_pipe[0], &i, 1);
 			check_dead_children();
-			continue;
 		}
 
 		for (i = 0; i < socknum; i++) {
-- 
1.6.0.rc0.22.gf2096d.dirty
Previous: Johannes SchindelinNext: Johannes Sixt
Message 2 of 6 in “git daemon: avoid waking up too often”
  1. git daemon: avoid waking up too oftenJohannes Schindelin, Jul 22, 2008
  2. git daemon: avoid waking up too oftenJohannes Schindelin, Jul 22, 2008
  3. Johannes SixtJul 23, 2008
  4. Johannes SchindelinJul 23, 2008
  5. Avery PennarunJul 23, 2008
  6. Johannes SchindelinJul 23, 2008

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.