[sheepdog] [PATCH] dog: let sheeps prevent inode update requests during snapshot creation

Hitoshi Mitake mitake.hitoshi at lab.ntt.co.jp
Fri Aug 22 06:32:42 CEST 2014


The recent new GC algorithm requires mutual exclusion between inode
update and and snapshot creation. This patch let sheeps prevent inode
update requests during snapshot creation with VDI locking information.

Signed-off-by: Hitoshi Mitake <mitake.hitoshi at lab.ntt.co.jp>
---
 dog/vdi.c | 131 +++++++++++++++++++++++++++++++++++++++++++-------------------
 1 file changed, 91 insertions(+), 40 deletions(-)

diff --git a/dog/vdi.c b/dog/vdi.c
index cd85b65..19da73e 100644
--- a/dog/vdi.c
+++ b/dog/vdi.c
@@ -517,6 +517,47 @@ out:
 	return ret;
 }
 
+static struct vdi_state *get_vdi_state(int *count)
+{
+	int ret;
+	struct sd_req hdr;
+	struct sd_rsp *rsp = (struct sd_rsp *)&hdr;
+	struct vdi_state *vs = NULL;
+	unsigned int rlen;
+
+#define DEFAULT_VDI_STATE_COUNT 512
+	rlen = DEFAULT_VDI_STATE_COUNT * sizeof(struct vdi_state);
+	vs = xzalloc(rlen);
+retry:
+	sd_init_req(&hdr, SD_OP_GET_VDI_COPIES);
+	hdr.data_length = rlen;
+
+	ret = dog_exec_req(&sd_nid, &hdr, (char *)vs);
+	if (ret < 0)
+		goto fail;
+
+	switch (ret) {
+	case SD_RES_SUCCESS:
+		break;
+	case SD_RES_BUFFER_SMALL:
+		rlen *= 2;
+		vs = xrealloc(vs, rlen);
+		goto retry;
+	default:
+		sd_err("failed to execute SD_OP_GET_VDI_COPIES: %s",
+		       sd_strerror(ret));
+		goto fail;
+	}
+
+	*count = rsp->data_length / sizeof(*vs);
+	return vs;
+
+fail:
+	free(vs);
+	vs = NULL;
+	return NULL;
+}
+
 static int vdi_snapshot(int argc, char **argv)
 {
 	const char *vdiname = argv[optind++];
@@ -525,6 +566,10 @@ static int vdi_snapshot(int argc, char **argv)
 	char buf[SD_INODE_HEADER_SIZE];
 	struct sd_inode *inode = (struct sd_inode *)buf;
 	struct sd_req hdr;
+	struct vdi_state *vs = NULL;
+	int vs_count = 0;
+	struct node_id owners[SD_MAX_COPIES];
+	int nr_owners = 0, nr_issued_prevent_inode_update = 0;
 
 	if (vdi_cmd_data.snapshot_id != 0) {
 		sd_err("Please specify a non-integer value for "
@@ -541,11 +586,43 @@ static int vdi_snapshot(int argc, char **argv)
 		return EXIT_FAILURE;
 	}
 
-	sd_init_req(&hdr, SD_OP_PREVENT_INODE_UPDATE);
-	ret = dog_exec_req(&sd_nid, &hdr, NULL);
-	if (ret < 0) {
-		sd_err("preventing inode update failed");
+	vs = get_vdi_state(&vs_count);
+	if (!vs)
 		return EXIT_FAILURE;
+
+	for (int i = 0; i < vs_count; i++) {
+		struct vdi_state *s = &vs[i];
+
+		if (s->vid != vid)
+			continue;
+
+		if (s->lock_state == LOCK_STATE_LOCKED) {
+			/* QEMU is using it */
+			memset(&owners[0], 0, sizeof(owners[0]));
+			memcpy(&owners[0], &s->lock_owner, sizeof(owners[0]));
+			nr_owners = 1;
+		} else {
+			/* tgt is using it */
+			for (int j = 0; j < s->nr_participants; i++) {
+				memset(&owners[nr_owners], 0,
+				       sizeof(owners[nr_owners]));
+				memcpy(&owners[nr_owners],
+				       &s->participants[nr_owners],
+				       sizeof(owners[nr_owners]));
+				nr_owners++;
+			}
+		}
+	}
+
+	for (int i = 0; i < nr_owners; i++) {
+		sd_init_req(&hdr, SD_OP_PREVENT_INODE_UPDATE);
+		ret = dog_exec_req(&owners[i], &hdr, NULL);
+		if (ret < 0) {
+			sd_err("preventing inode update failed");
+			goto out;
+		}
+
+		nr_issued_prevent_inode_update++;
 	}
 
 	ret = dog_write_object(vid_to_vdi_oid(vid), 0,
@@ -555,7 +632,7 @@ static int vdi_snapshot(int argc, char **argv)
 			       0, inode->nr_copies, inode->copy_policy,
 			       false, false);
 	if (ret != SD_RES_SUCCESS)
-		return EXIT_FAILURE;
+		goto out;
 
 	ret = do_vdi_create(vdiname, inode->vdi_size, vid, &new_vid, true,
 			    inode->nr_copies, inode->copy_policy,
@@ -569,11 +646,12 @@ static int vdi_snapshot(int argc, char **argv)
 			       " VDI ID of newly created snapshot: %x\n", new_vid, vid);
 	}
 
-	sd_init_req(&hdr, SD_OP_ALLOW_INODE_UPDATE);
-	ret = dog_exec_req(&sd_nid, &hdr, NULL);
-	if (ret < 0) {
-		sd_err("allowing inode update failed");
-		return EXIT_FAILURE;
+out:
+	for (int i = 0; i < nr_issued_prevent_inode_update; i++) {
+		sd_init_req(&hdr, SD_OP_ALLOW_INODE_UPDATE);
+		ret = dog_exec_req(&owners[i], &hdr, NULL);
+		if (ret < 0)
+			sd_err("allowing inode update failed");
 	}
 
 	return ret;
@@ -2716,41 +2794,14 @@ static int vdi_alter_copy(int argc, char **argv)
 
 static int lock_list(int argc, char **argv)
 {
-	struct sd_req hdr;
-	struct sd_rsp *rsp = (struct sd_rsp *)&hdr;
 	struct vdi_state *vs = NULL;
-	unsigned int rlen;
-	int ret, count;
+	int ret = 0, count = 0;
 
-#define DEFAULT_VDI_STATE_COUNT 512
-	rlen = DEFAULT_VDI_STATE_COUNT * sizeof(struct vdi_state);
-	vs = xzalloc(rlen);
-retry:
-	sd_init_req(&hdr, SD_OP_GET_VDI_COPIES);
-	hdr.data_length = rlen;
-
-	ret = dog_exec_req(&sd_nid, &hdr, (char *)vs);
-	if (ret < 0)
-		return EXIT_SYSFAIL;
-
-	switch (ret) {
-	case SD_RES_SUCCESS:
-		break;
-	case SD_RES_BUFFER_SMALL:
-		rlen *= 2;
-		vs = xrealloc(vs, rlen);
-		goto retry;
-	default:
-		sd_err("failed to execute SD_OP_GET_VDI_COPIES: %s",
-		       sd_strerror(ret));
-		goto out;
-	}
+	vs = get_vdi_state(&count);
 
 	init_tree();
 	if (parse_vdi(construct_vdi_tree, SD_INODE_HEADER_SIZE, NULL) < 0)
-		return EXIT_SYSFAIL;
-
-	count = rsp->data_length / sizeof(*vs);
+		goto out;
 
 	printf("VDI | owner node\n");
 	for (int i = 0; i < count; i++) {
-- 
1.8.3.2




More information about the sheepdog mailing list