[dm-devel] [PATCH 2 of 2] dm log userspace group clear and mark requests

[Jonathan Brassow]

Patch name: dm-log-userspace-group-clear-and-mark-requests.patch

This patch allows the device-mapper log 'mark' and 'clear'
requests to be grouped and processed in a batch.  This can
significantly reduce the amount of traffic going between
the kernel and userspace (where the processing daemon resides).
Although the ability to group these requests is new, it does
not deviate from the API described for these types of requests
as defined in include/linux/dm-log-userspace.h.

Signed-off-by: Jonathan Brassow <jbrassow at redhat.com>

Index: linux-2.6/drivers/md/dm-log-userspace-base.c
===================================================================
--- linux-2.6.orig/drivers/md/dm-log-userspace-base.c
+++ linux-2.6/drivers/md/dm-log-userspace-base.c
@@ -18,6 +18,14 @@ struct flush_entry {
 	struct list_head list;
 };
 
+/*
+ * This limit on the number of mark and clear request is, to a degree,
+ * arbitrary.  However, there is some basis for the choice in the limits
+ * imposed on the size of data payload by dm-log-userspace-transfer.c:
+ * dm_consult_userspace().
+ */
+#define MAX_FLUSH_GROUP_COUNT 32
+
 struct log_c {
 	struct dm_target *ti;
 	uint32_t region_size;
@@ -46,6 +54,7 @@ struct log_c {
 	spinlock_t flush_lock;
 	struct list_head mark_list;
 	struct list_head clear_list;
+	uint64_t group[MAX_FLUSH_GROUP_COUNT];
 };
 
 static mempool_t *flush_entry_pool;
@@ -349,6 +358,66 @@ static int userspace_in_sync(struct dm_d
 	return (r) ? 0 : (int)in_sync;
 }
 
+static int flush_one_by_one(struct log_c *lc, struct list_head *flush_list)
+{
+	int r = 0;
+	struct flush_entry *fe;
+
+	list_for_each_entry(fe, flush_list, list) {
+		r = userspace_do_request(lc, lc->uuid, fe->type,
+					 (char *)&fe->region,
+					 sizeof(fe->region),
+					 NULL, NULL);
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+static int flush_by_group(struct log_c *lc, struct list_head *flush_list)
+{
+	int r = 0;
+	int count;
+	uint32_t type;
+	struct flush_entry *fe, *tmp_fe;
+	LIST_HEAD(tmp_list);
+
+	/*
+	 * Group process the requests
+	 */
+	while (!list_empty(flush_list)) {
+		count = 0;
+
+		list_for_each_entry_safe(fe, tmp_fe, flush_list, list) {
+			lc->group[count] = fe->region;
+			count++;
+
+			list_del(&fe->list);
+			list_add(&fe->list, &tmp_list);
+
+			type = fe->type;
+			if (count >= MAX_FLUSH_GROUP_COUNT)
+				break;
+		}
+
+		r = userspace_do_request(lc, lc->uuid, type,
+					 (char *)(lc->group),
+					 count * sizeof(uint64_t),
+					 NULL, NULL);
+		if (r) {
+			/* Group send failed.  Attempt one-by-one. */
+			list_splice_init(&tmp_list, flush_list);
+			r = flush_one_by_one(lc, flush_list);
+			break;
+		}
+	}
+
+	list_splice_init(&tmp_list, flush_list);
+
+	return r;
+}
+
 /*
  * userspace_flush
  *
@@ -383,30 +452,13 @@ static int userspace_flush(struct dm_dir
 	if (list_empty(&mark_list) && list_empty(&clear_list))
 		return 0;
 
-	/*
-	 * FIXME: Count up requests, group request types,
-	 * allocate memory to stick all requests in and
-	 * send to server in one go.  Failing the allocation,
-	 * do it one by one.
-	 */
-
-	list_for_each_entry(fe, &mark_list, list) {
-		r = userspace_do_request(lc, lc->uuid, fe->type,
-					 (char *)&fe->region,
-					 sizeof(fe->region),
-					 NULL, NULL);
-		if (r)
-			goto fail;
-	}
+	r = flush_by_group(lc, &mark_list);
+	if (r)
+		goto fail;
 
-	list_for_each_entry(fe, &clear_list, list) {
-		r = userspace_do_request(lc, lc->uuid, fe->type,
-					 (char *)&fe->region,
-					 sizeof(fe->region),
-					 NULL, NULL);
-		if (r)
-			goto fail;
-	}
+	r = flush_by_group(lc, &clear_list);
+	if (r)
+		goto fail;
 
 	r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH,
 				 NULL, 0, NULL, NULL);