[Cluster-devel] GFS2: Use kmalloc when possible for ->readdir() (try #2)

[Steven Whitehouse <swhiteho@redhat.com>]

If we don't need a huge amount of memory in ->readdir() then
we can use kmalloc rather than vmalloc to allocate it. This
should cut down on the greater overheads associated with
vmalloc for smaller directories.

We may be able to eliminate vmalloc entirely at some stage,
but this is easy to do right away.

Also using GFP_NOFS to avoid any issues wrt to deleting inodes
while under a glock, and suggestion from Linus to factor out
the alloc/dealloc.

I've given this a test with a variety of different sized
directories and it seems to work ok.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 401deaf..b9dd88a 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1238,6 +1238,25 @@ static int do_filldir_main(struct gfs2_inode *dip, u64 *offset,
 	return 0;
 }
 
+static void *gfs2_alloc_sort_buffer(unsigned size)
+{
+	void *ptr = NULL;
+
+	if (size < KMALLOC_MAX_SIZE)
+		ptr = kmalloc(size, GFP_NOFS | __GFP_NOWARN);
+	if (!ptr)
+		ptr = __vmalloc(size, GFP_NOFS, PAGE_KERNEL);
+	return ptr;
+}
+
+static void gfs2_free_sort_buffer(void *ptr)
+{
+	if (is_vmalloc_addr(ptr))
+		vfree(ptr);
+	else
+		kfree(ptr);
+}
+
 static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
 			      filldir_t filldir, int *copied, unsigned *depth,
 			      u64 leaf_no)
@@ -1278,7 +1297,7 @@ static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
 	 * 99 is the maximum number of entries that can fit in a single
 	 * leaf block.
 	 */
-	larr = vmalloc((leaves + entries + 99) * sizeof(void *));
+	larr = gfs2_alloc_sort_buffer((leaves + entries + 99) * sizeof(void *));
 	if (!larr)
 		goto out;
 	darr = (const struct gfs2_dirent **)(larr + leaves);
@@ -1289,7 +1308,7 @@ static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
 	do {
 		error = get_leaf(ip, lfn, &bh);
 		if (error)
-			goto out_kfree;
+			goto out_free;
 		lf = (struct gfs2_leaf *)bh->b_data;
 		lfn = be64_to_cpu(lf->lf_next);
 		if (lf->lf_entries) {
@@ -1298,7 +1317,7 @@ static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
 						gfs2_dirent_gather, NULL, &g);
 			error = PTR_ERR(dent);
 			if (IS_ERR(dent))
-				goto out_kfree;
+				goto out_free;
 			if (entries2 != g.offset) {
 				fs_warn(sdp, "Number of entries corrupt in dir "
 						"leaf %llu, entries2 (%u) != "
@@ -1307,7 +1326,7 @@ static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
 					entries2, g.offset);
 					
 				error = -EIO;
-				goto out_kfree;
+				goto out_free;
 			}
 			error = 0;
 			larr[leaf++] = bh;
@@ -1319,10 +1338,10 @@ static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
 	BUG_ON(entries2 != entries);
 	error = do_filldir_main(ip, offset, opaque, filldir, darr,
 				entries, copied);
-out_kfree:
+out_free:
 	for(i = 0; i < leaf; i++)
 		brelse(larr[i]);
-	vfree(larr);
+	gfs2_free_sort_buffer(larr);
 out:
 	return error;
 }