[1/3] service: allow to execute page-server as a child process

Submitted by Andrei Vagin on May 17, 2017, 5:07 a.m.

Details

Message ID 20170517050713.759-2-avagin@openvz.org
State New
Series "service: allow to run page-server as a child process"
Headers show

Commit Message

Andrei Vagin May 17, 2017, 5:07 a.m.
From: Andrei Vagin <avagin@virtuozzo.com>

In this case we can wait it and get an exit code.

For example, it will be useful for p.haul where one connection
is used several times, so we need a way how to understand  that
page-server exited unexpectedly.

Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
---
 criu/cr-service.c        | 79 ++++++++++++++++++++++++++++++++++--------------
 criu/include/page-xfer.h |  5 +++
 criu/page-xfer.c         | 13 ++++++++
 images/rpc.proto         |  8 +++++
 4 files changed, 82 insertions(+), 23 deletions(-)

Patch hide | download patch | download mbox

diff --git a/criu/cr-service.c b/criu/cr-service.c
index 8cf76dd..18ba2fd 100644
--- a/criu/cr-service.c
+++ b/criu/cr-service.c
@@ -705,12 +705,7 @@  static int pre_dump_loop(int sk, CriuReq *msg)
 	return dump_using_req(sk, msg->opts);
 }
 
-struct ps_info {
-	int pid;
-	unsigned short port;
-};
-
-static int start_page_server_req(int sk, CriuOpts *req)
+static int start_page_server_req(int sk, CriuOpts *req, bool daemon_mode)
 {
 	int ret = -1, pid, start_pipe[2];
 	ssize_t count;
@@ -735,35 +730,40 @@  static int start_page_server_req(int sk, CriuOpts *req)
 
 		pr_debug("Starting page server\n");
 
-		pid = cr_page_server(true, false, start_pipe[1]);
+		pid = cr_page_server(daemon_mode, false, start_pipe[1]);
 		if (pid < 0)
 			goto out_ch;
 
-		info.pid = pid;
-		info.port = opts.port;
+		if (daemon_mode) {
+			info.pid = pid;
+			info.port = opts.port;
 
-		count = write(start_pipe[1], &info, sizeof(info));
-		if (count != sizeof(info))
-			goto out_ch;
+			count = write(start_pipe[1], &info, sizeof(info));
+			if (count != sizeof(info))
+				goto out_ch;
+		}
 
 		ret = 0;
 out_ch:
-		if (ret < 0 && pid > 0)
+		if (daemon_mode && ret < 0 && pid > 0)
 			kill(pid, SIGKILL);
 		close(start_pipe[1]);
 		exit(ret);
 	}
 
 	close(start_pipe[1]);
-	wait(&ret);
-	if (WIFEXITED(ret)) {
-		if (WEXITSTATUS(ret)) {
-			pr_err("Child exited with an error\n");
+
+	if (daemon_mode) {
+		wait(&ret);
+		if (WIFEXITED(ret)) {
+			if (WEXITSTATUS(ret)) {
+				pr_err("Child exited with an error\n");
+				goto out;
+			}
+		} else {
+			pr_err("Child wasn't terminated normally\n");
 			goto out;
 		}
-	} else {
-		pr_err("Child wasn't terminated normally\n");
-		goto out;
 	}
 
 	count = read(start_pipe[0], &info, sizeof(info));
@@ -771,11 +771,12 @@  out_ch:
 	if (count != sizeof(info))
 		goto out;
 
-	success = true;
-	ps.has_pid = true;
 	ps.pid = info.pid;
 	ps.has_port = true;
 	ps.port = info.port;
+
+	success = true;
+	ps.has_pid = true;
 	resp.ps = &ps;
 
 	pr_debug("Page server started\n");
@@ -802,6 +803,9 @@  static int chk_keepopen_req(CriuReq *msg)
 	if (msg->type == CRIU_REQ_TYPE__PAGE_SERVER)
 		/* This just fork()-s so no leaks */
 		return 0;
+	if (msg->type == CRIU_REQ_TYPE__PAGE_SERVER_CHLD)
+		/* This just fork()-s so no leaks */
+		return 0;
 	else if (msg->type == CRIU_REQ_TYPE__CPUINFO_DUMP ||
 		 msg->type == CRIU_REQ_TYPE__CPUINFO_CHECK)
 		return 0;
@@ -946,6 +950,29 @@  out:
 	return send_criu_msg(sk, &resp);
 }
 
+static int handle_wait_pid(int sk, int pid)
+{
+	CriuResp resp = CRIU_RESP__INIT;
+	bool success = false;
+	int status;
+
+	if (waitpid(pid, &status, 0) == -1) {
+		resp.cr_errno = errno;
+		pr_perror("Unable to wait %d", pid);
+		goto out;
+	}
+
+	resp.status = status;
+	resp.has_status = true;
+
+	success = true;
+out:
+	resp.type = CRIU_REQ_TYPE__WAIT_PID;
+	resp.success = success;
+
+	return send_criu_msg(sk, &resp);
+}
+
 static int handle_cpuinfo(int sk, CriuReq *msg)
 {
 	CriuResp resp = CRIU_RESP__INIT;
@@ -1031,7 +1058,13 @@  more:
 		ret = pre_dump_loop(sk, msg);
 		break;
 	case CRIU_REQ_TYPE__PAGE_SERVER:
-		ret =  start_page_server_req(sk, msg->opts);
+		ret = start_page_server_req(sk, msg->opts, true);
+		break;
+	case CRIU_REQ_TYPE__PAGE_SERVER_CHLD:
+		ret = start_page_server_req(sk, msg->opts, false);
+		break;
+	case CRIU_REQ_TYPE__WAIT_PID:
+		ret =  handle_wait_pid(sk, msg->pid);
 		break;
 	case CRIU_REQ_TYPE__CPUINFO_DUMP:
 	case CRIU_REQ_TYPE__CPUINFO_CHECK:
diff --git a/criu/include/page-xfer.h b/criu/include/page-xfer.h
index 1f59bdb..0b92ec9 100644
--- a/criu/include/page-xfer.h
+++ b/criu/include/page-xfer.h
@@ -2,6 +2,11 @@ 
 #define __CR_PAGE_XFER__H__
 #include "pagemap.h"
 
+struct ps_info {
+	int pid;
+	unsigned short port;
+};
+
 extern int cr_page_server(bool daemon_mode, bool lazy_dump, int cfd);
 
 /*
diff --git a/criu/page-xfer.c b/criu/page-xfer.c
index fa3aacc..418bdc4 100644
--- a/criu/page-xfer.c
+++ b/criu/page-xfer.c
@@ -961,6 +961,19 @@  int cr_page_server(bool daemon_mode, bool lazy_dump, int cfd)
 	if (sk == -1)
 		return -1;
 no_server:
+
+	if (!daemon_mode) {
+		struct ps_info info = {.pid = getpid(), .port = opts.port};
+		int count;
+
+		count = write(cfd, &info, sizeof(info));
+		close_safe(&cfd);
+		if (count != sizeof(info)) {
+			pr_perror("Unable to write ps_info");
+			exit(1);
+		}
+	}
+
 	ret = run_tcp_server(daemon_mode, &ask, cfd, sk);
 	if (ret != 0)
 		return ret > 0 ? 0 : -1;
diff --git a/images/rpc.proto b/images/rpc.proto
index 48e42e2..71f47d5 100644
--- a/images/rpc.proto
+++ b/images/rpc.proto
@@ -142,6 +142,9 @@  enum criu_req_type {
 	FEATURE_CHECK	= 9;
 
 	VERSION		= 10;
+
+	WAIT_PID	= 11;
+	PAGE_SERVER_CHLD = 12;
 }
 
 /*
@@ -176,6 +179,9 @@  message criu_req {
 	 * via RPC.
 	 */
 	optional criu_features		features	= 5;
+
+	/* 'pid' is used for WAIT_PID */
+	optional uint32			pid		= 6;
 }
 
 /*
@@ -196,6 +202,8 @@  message criu_resp {
 	optional criu_features		features	= 8;
 	optional string			cr_errmsg	= 9;
 	optional criu_version		version		= 10;
+
+	optional int32			status		= 11;
 }
 
 /* Answer for criu_req_type.VERSION requests */