[Ocfs2-tools-devel] [PATCH 5/7] mkfs.ocfs2: Allocate groups to the extent allocator

[Sunil Mushran]

Extent allocator comes into play when an inode can no longer store all the
data extent pointers in the inode block and needs an indirect block that is
allocated from the extent allocator. This typically happens when the free
space is very fragmented. However, when that is the case, the file system
is also unable to find 4MB contiguous free space that is typically required
to extend this allocator.

To address this issue, we are now preallocating the extent allocator for
all slots and also teaching the file system to steal extents from other
slots when it is unable to extend its own slot allocator.

Signed-off-by: Sunil Mushran <sunil.mushran at oracle.com>
---
 mkfs.ocfs2/mkfs.c |   78 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 mkfs.ocfs2/mkfs.h |    2 +
 2 files changed, 80 insertions(+), 0 deletions(-)

diff --git a/mkfs.ocfs2/mkfs.c b/mkfs.ocfs2/mkfs.c
index 3d4698b..ca88a68 100644
--- a/mkfs.ocfs2/mkfs.c
+++ b/mkfs.ocfs2/mkfs.c
@@ -363,6 +363,28 @@ error:
 	exit(1);
 }
 
+static void grow_extent_allocator(State *s, ocfs2_filesys *fs)
+{
+	errcode_t ret;
+	int i;
+
+	for (i = 0; i < OCFS2_RAW_SB(fs->fs_super)->s_max_slots; i++) {
+		ret = ocfs2_grow_chain_allocator(fs, EXTENT_ALLOC_SYSTEM_INODE,
+						 i, s->extent_alloc_size_in_clusters);
+		if (ret) {
+			com_err(s->progname, ret, "while growing the extent "
+				"allocator for slot %d by %d clusters",
+				i, s->extent_alloc_size_in_clusters);
+			goto error;
+		}
+	}
+	return;
+
+error:
+	clear_both_ends(s);
+	exit(1);
+}
+
 static void finish_normal_format(State *s)
 {
 	errcode_t ret;
@@ -402,6 +424,12 @@ static void finish_normal_format(State *s)
 		printf("done\n");
 
 	if (!s->quiet)
+		printf("Growing extent allocator: ");
+	grow_extent_allocator(s, fs);
+	if (!s->quiet)
+		printf("done\n");
+
+	if (!s->quiet)
 		printf("Formatting slot map: ");
 	format_slotmap(s, fs);
 	if (!s->quiet)
@@ -1344,6 +1372,42 @@ static uint32_t cluster_size_datafiles(State *s)
 	return cluster_size * 1024;
 }
 
+static uint32_t figure_extent_alloc_size(State *s)
+{
+	uint32_t cpg, numgroups;
+	uint64_t unitsize, totalsize;
+	double curr_percent, target_percent;
+
+	if (!s->initial_slots)
+		return 0;
+
+	switch (s->fs_type) {
+	case OCFS2_MKFSTYPE_DATAFILES:
+	case OCFS2_MKFSTYPE_VMSTORE:
+		target_percent = 0.3;
+		break;
+	case OCFS2_MKFSTYPE_MAIL:
+	default:
+		target_percent = 0.1;
+		break;
+	}
+
+	cpg = ocfs2_clusters_per_group(s->blocksize, s->cluster_size_bits);
+
+	/* size of the allocator across all slots with one group */
+	unitsize = cpg * s->cluster_size * s->initial_slots;
+
+	totalsize = unitsize;
+	for (numgroups = 1; ; ++numgroups) {
+		curr_percent = (double)totalsize * 100 / s->volume_size_in_bytes;
+		if (curr_percent >= target_percent)
+			break;
+		totalsize += unitsize;
+	}
+
+	return cpg * numgroups;
+}
+
 static void
 fill_defaults(State *s)
 {
@@ -1528,6 +1592,8 @@ fill_defaults(State *s)
 	}
 
 	s->journal_size_in_bytes = figure_journal_size(s->journal_size_in_bytes, s);
+
+	s->extent_alloc_size_in_clusters = figure_extent_alloc_size(s);
 }
 
 static int
@@ -2598,10 +2664,20 @@ print_state(State *s)
 {
 	int i;
 	char buf[PATH_MAX] = "\0";
+	uint64_t extsize = 0;
+	uint32_t numgrps = 0;
 
 	if (s->quiet)
 		return;
 
+	if (s->extent_alloc_size_in_clusters) {
+		numgrps = s->extent_alloc_size_in_clusters /
+			ocfs2_clusters_per_group(s->blocksize,
+						 s->cluster_size_bits);
+		extsize = (uint64_t)s->extent_alloc_size_in_clusters *
+			s->cluster_size;
+	}
+
 	ocfs2_snprint_feature_flags(buf, PATH_MAX, &s->feature_flags);
 
 	if (s->fs_type != OCFS2_MKFSTYPE_DEFAULT) {
@@ -2623,6 +2699,8 @@ print_state(State *s)
 	printf("Cluster groups: %u (tail covers %u clusters, rest cover %u "
 	       "clusters)\n", s->nr_cluster_groups, s->tail_group_bits,
 	       s->global_cpg);
+	printf("Extent allocator size: %"PRIu64" (%u groups)\n",
+	       extsize, numgrps);
 	if (s->hb_dev)
 		printf("Heartbeat device\n");
 	else
diff --git a/mkfs.ocfs2/mkfs.h b/mkfs.ocfs2/mkfs.h
index 5049e9f..9b50a9f 100644
--- a/mkfs.ocfs2/mkfs.h
+++ b/mkfs.ocfs2/mkfs.h
@@ -209,6 +209,8 @@ struct _State {
 	uint64_t journal_size_in_bytes;
 	int journal64;
 
+	uint32_t extent_alloc_size_in_clusters;
+
 	char *vol_label;
 	char *device_name;
 	unsigned char *uuid;
-- 
1.6.3.3