[mod_fastcgi] detect child exit, restart proactively

(instead of detecting upon a subsequent HTTP request)

(for backends spawned by mod_fastcgi)

src/mod_fastcgi.c

@@ -3550,6 +3550,49 @@ TRIGGER_FUNC(mod_fastcgi_handle_trigger) {
 
 				host = ex->hosts[n];
 
+				for (proc = host->first; proc; proc = proc->next) {
+					int status;
+
+					if (proc->pid == 0) continue;
+
+					switch (waitpid(proc->pid, &status, WNOHANG)) {
+					case 0:
+						/* child still running after timeout, good */
+						break;
+					case -1:
+						if (errno != EINTR) {
+							/* no PID found ? should never happen */
+							log_error_write(srv, __FILE__, __LINE__, "sddss",
+									"pid ", proc->pid, proc->state,
+									"not found:", strerror(errno));
+						}
+						break;
+					default:
+						/* the child should not terminate at all */
+						if (WIFEXITED(status)) {
+							if (proc->state != PROC_STATE_KILLED) {
+								log_error_write(srv, __FILE__, __LINE__, "sdb",
+										"child exited:",
+										WEXITSTATUS(status), proc->connection_name);
+							}
+						} else if (WIFSIGNALED(status)) {
+							if (WTERMSIG(status) != SIGTERM) {
+								log_error_write(srv, __FILE__, __LINE__, "sd",
+										"child signaled:",
+										WTERMSIG(status));
+							}
+						} else {
+							log_error_write(srv, __FILE__, __LINE__, "sd",
+									"child died somehow:",
+									status);
+						}
+						proc->pid = 0;
+						if (proc->state == PROC_STATE_RUNNING) host->active_procs--;
+						proc->state = PROC_STATE_DIED;
+						host->max_id--;
+					}
+				}
+
 				fcgi_restart_dead_procs(srv, p, host);
 
 				for (proc = host->unused_procs; proc; proc = proc->next) {

tests/fcgi-responder.c

@@ -11,7 +11,7 @@
 #include <string.h>
 
 int main (void) {
-	int num_requests = 2;
+	int num_requests = 1;
 
 	while (num_requests > 0 &&
 	       FCGI_Accept() >= 0) {

tests/mod-fastcgi.t

@@ -7,7 +7,7 @@ BEGIN {
 }
 
 use strict;
-use Test::More tests => 60;
+use Test::More tests => 59;
 use LightyTest;
 
 my $tf = LightyTest->new();
@@ -409,17 +409,7 @@ EOF
 	$t->{RESPONSE} = [ { 'HTTP-Protocol' => 'HTTP/1.0', 'HTTP-Status' => 200, 'HTTP-Content' => 'test123' } ];
 	ok($tf->handle_http($t) == 0, 'killing fastcgi and wait for restart');
 
-	select(undef, undef, undef, .2);
-	$t->{REQUEST}  = ( <<EOF
-GET /index.fcgi?die-at-end HTTP/1.0
-Host: www.example.org
-EOF
- );
-	$t->{RESPONSE} = [ { 'HTTP-Protocol' => 'HTTP/1.0', 'HTTP-Status' => 200, 'HTTP-Content' => 'test123' } ];
-	ok($tf->handle_http($t) == 0, 'killing fastcgi and wait for restart');
-
-
-	select(undef, undef, undef, .2);
+	select(undef, undef, undef, .1) while (!$tf->listening_on(10000));
 	$t->{REQUEST}  = ( <<EOF
 GET /index.fcgi?crlf HTTP/1.0
 Host: www.example.org