[Sheepdog] [PATCH 3/3] check protocol version

[MORITA Kazutaka]

We should avoid wrong version members coming in the cluster.

Signed-off-by: MORITA Kazutaka <morita.kazutaka at lab.ntt.co.jp>
---
 collie/group.c           |   20 +++++++++++++++-----
 include/collie.h         |    8 ++++++++
 include/sheepdog_proto.h |    3 +++
 shepherd/shepherd.c      |    1 +
 4 files changed, 27 insertions(+), 5 deletions(-)

diff --git a/collie/group.c b/collie/group.c
index c41f918..91b0f4e 100644
--- a/collie/group.c
+++ b/collie/group.c
@@ -43,9 +43,10 @@ enum deliver_msg_state {
 };
 
 struct message_header {
+	uint8_t proto_ver;
+	uint8_t pad;
 	uint8_t op;
 	uint8_t state;
-	uint8_t pad[2];
 	uint32_t msg_length;
 	uint32_t nodeid;
 	uint32_t pid;
@@ -475,6 +476,12 @@ static void join(struct join_message *msg)
 	struct sheepdog_node_list_entry entry[SD_MAX_NODES];
 	int i;
 
+	if (msg->header.proto_ver != SD_COLLIE_PROTO_VER) {
+		eprintf("joining node send a wrong version message\n");
+		msg->result = SD_RES_VER_MISMATCH;
+		return;
+	}
+
 	if (msg->nr_sobjs)
 		sys->nr_sobjs = msg->nr_sobjs;
 
@@ -581,9 +588,8 @@ static void update_cluster_info(struct join_message *msg)
 		if (is_myself(&msg->header.from)) {
 			eprintf("failed to join sheepdog, %d\n", msg->result);
 			sys->status = SD_STATUS_JOIN_FAILED;
-			return;
-		} else
-			return;
+		}
+		return;
 	}
 
 	if (!sys->nr_sobjs)
@@ -656,7 +662,7 @@ out:
 			sys->epoch = get_latest_epoch();
 	}
 
-	if (sys->status != SD_STATUS_INCONSISTENT_EPOCHS) {
+	if (sys->status != SD_STATUS_INCONSISTENT_EPOCHS || sys->status != SD_STATUS_JOIN_FAILED) {
 		if (msg->cluster_status == SD_STATUS_OK) {
 			get_vdi_bitmap_from_all();
 			set_global_nr_copies(sys->nr_sobjs);
@@ -690,6 +696,10 @@ static void vdi_op(struct vdi_op_message *msg)
 		break;
 	case SD_OP_LOCK_VDI:
 	case SD_OP_GET_VDI_INFO:
+		if (hdr->proto_ver != SD_PROTO_VER) {
+			ret = SD_RES_VER_MISMATCH;
+			break;
+		}
 		ret = lookup_vdi(hdr->epoch, data, hdr->data_length, &vid, hdr->snapid);
 		if (ret != SD_RES_SUCCESS)
 			break;
diff --git a/include/collie.h b/include/collie.h
index b3c8264..e4019d2 100644
--- a/include/collie.h
+++ b/include/collie.h
@@ -15,9 +15,12 @@
 #include "util.h"
 #include "list.h"
 
+#define SD_COLLIE_PROTO_VER 0x01
+
 #define SD_MAX_NODES 1024
 #define SD_MAX_VMS   4096 /* FIXME: should be removed */
 
+#define SD_OP_COLLIE         0x80
 #define SD_OP_DEL_VDI        0x81
 #define SD_OP_GET_NODE_LIST  0x82
 #define SD_OP_GET_VM_LIST    0x83
@@ -172,4 +175,9 @@ static inline void print_node_list_entry(struct sheepdog_node_list_entry *e,
 		 e->addr[14], e->addr[15], e->port);
 }
 
+static inline int is_collie_op(uint8_t op)
+{
+	return op & SD_OP_COLLIE;
+}
+
 #endif
diff --git a/include/sheepdog_proto.h b/include/sheepdog_proto.h
index 5b77544..19ace97 100644
--- a/include/sheepdog_proto.h
+++ b/include/sheepdog_proto.h
@@ -15,6 +15,8 @@
 #include <stdint.h>
 #include "util.h"
 
+#define SD_PROTO_VER 0x01
+
 #define SD_LISTEN_PORT 7000
 
 #define SD_OP_CREATE_AND_WRITE_OBJ  0x01
@@ -50,6 +52,7 @@
 #define SD_RES_SHUTDOWN      0x11 /* Sheepdog is shutting down */
 #define SD_RES_NO_MEM        0x12 /* Cannot allocate memory */
 #define SD_RES_FULL_VDI      0x13 /* we already have the maximum vdis */
+#define SD_RES_VER_MISMATCH  0x14 /* Protocol version mismatch */
 
 /*
  * Object ID rules
diff --git a/shepherd/shepherd.c b/shepherd/shepherd.c
index cb5d5e3..3b59a61 100644
--- a/shepherd/shepherd.c
+++ b/shepherd/shepherd.c
@@ -843,6 +843,7 @@ static int vdi_lock(int argc, char **argv)
 
 	memset(&hdr, 0, sizeof(hdr));
 
+	hdr.proto_ver = SD_PROTO_VER; /* version is checked when locking */
 	hdr.opcode = opcode;
 	hdr.data_length = wlen;
 	hdr.flags = flags;
-- 
1.5.6.5