btrfs: add hybrid latency-rr read policy

This mode combines latency and round-robin modes by considering all
stripes within 120% of the minimum latency. It falls back to round-robin
if all stripes have no latency recorded yet.

Signed-off-by: Kai Krakow <[email protected]>

Author: kakra

fs/btrfs/sysfs.c

@@ -1310,6 +1310,7 @@ static const char *btrfs_read_policy_name[] = {
 #ifdef CONFIG_BTRFS_EXPERIMENTAL
 	"round-robin",
 	"latency",
+	"latency-rr",
 	"devid",
 #endif
 };
@@ -1325,7 +1326,7 @@ char *btrfs_get_mod_read_policy(void)
 /* Set perm 0, disable sys/module/btrfs/parameter/read_policy interface */
 module_param(read_policy, charp, 0);
 MODULE_PARM_DESC(read_policy,
-"Global read policy; pid (default), round-robin[:min_contig_read], latency, devid[:devid]");
+"Global read policy; pid (default), round-robin[:min_contig_read], latency, latency-rr[:min_contig_read], devid[:devid]");
 #endif
 
 int btrfs_read_policy_to_enum(const char *str, s64 *value)
@@ -1383,6 +1384,10 @@ static ssize_t btrfs_read_policy_show(struct kobject *kobj,
 		ret += sysfs_emit_at(buf, ret, "%s", btrfs_read_policy_name[i]);
 
 #ifdef CONFIG_BTRFS_EXPERIMENTAL
+		if (i == BTRFS_READ_POLICY_LATENCY_RR)
+			ret += sysfs_emit_at(buf, ret, ":%d",
+				 READ_ONCE(fs_devices->rr_min_contig_read));
+
 		if (i == BTRFS_READ_POLICY_RR)
 			ret += sysfs_emit_at(buf, ret, ":%d",
 				 READ_ONCE(fs_devices->rr_min_contig_read));
@@ -1418,7 +1423,11 @@ static ssize_t btrfs_read_policy_store(struct kobject *kobj,
 	    index != BTRFS_READ_POLICY_RR)
 		fs_devices->fs_stats = false;
 
-	if (index == BTRFS_READ_POLICY_RR) {
+	if (fs_devices->read_policy == BTRFS_READ_POLICY_LATENCY_RR &&
+	    index != BTRFS_READ_POLICY_LATENCY_RR)
+		fs_devices->fs_stats = false;
+
+	if ((index == BTRFS_READ_POLICY_RR) || (index == BTRFS_READ_POLICY_LATENCY_RR)) {
 		if (value != -1) {
 			u32 sectorsize = fs_devices->fs_info->sectorsize;
 

fs/btrfs/volumes.c

@@ -6143,6 +6143,40 @@ static int btrfs_read_rr(struct btrfs_chunk_map *map, int first, int num_stripes
 
 	return ret_stripe;
 }
+
+/*
+ * btrfs_read_fastest_rr.
+ *
+ * Select a stripe for reading using a hybrid algorithm:
+ *
+ *  1. Determine the fastest stripe using btrfs_best_stripe.
+ *  2. Add 20% headroom to the selected latency.
+ *  3. Select a stripe using btrfs_read_rr filtered by latency.
+ */
+static int btrfs_read_fastest_rr(struct btrfs_fs_info *fs_info,
+                                 struct btrfs_chunk_map *map, int first,
+                                 int num_stripes)
+{
+	u64 min_latency;
+	int ret_stripe = -1;
+
+	btrfs_best_stripe(fs_info, map, first, num_stripes, &min_latency,
+	                     &ret_stripe);
+
+	/* min_latency will be 0 if no latency has been recorded yet,
+	 * add 20% headroom otherwise.
+	 */
+	if (likely(min_latency)) {
+		min_latency = min_latency * 6;
+		min_latency = div_u64(min_latency, 5);
+		ret_stripe = btrfs_read_rr(map, first, num_stripes, min_latency);
+	}
+
+	if (ret_stripe < 0)
+		ret_stripe = btrfs_read_rr(map, first, num_stripes, 0);
+
+	return ret_stripe;
+}
 #endif
 
 static int find_live_mirror(struct btrfs_fs_info *fs_info,
@@ -6185,6 +6219,10 @@ static int find_live_mirror(struct btrfs_fs_info *fs_info,
 		preferred_mirror = btrfs_read_fastest(fs_info, map, first,
 								num_stripes);
 		break;
+	case BTRFS_READ_POLICY_LATENCY_RR:
+		preferred_mirror = btrfs_read_fastest_rr(fs_info, map, first,
+		                                         num_stripes);
+		break;
 #endif
 	}
 

fs/btrfs/volumes.h

@@ -310,6 +310,8 @@ enum btrfs_read_policy {
 	BTRFS_READ_POLICY_RR,
 	/* Use the lowest-latency device dynamically */
 	BTRFS_READ_POLICY_LATENCY,
+	/* Use hybrid approach of lowest-latency and round-robin */
+	BTRFS_READ_POLICY_LATENCY_RR,
 	/* Read from the specific device */
 	BTRFS_READ_POLICY_DEVID,
 #endif