ocfs2: reflink: fix slow unlink for refcounted file

When running ocfs2 test suite multiple nodes reflink stress test, for a
4 nodes cluster, every unlink() for refcounted file needs about 700s.

The slow unlink is caused by the contention of refcount tree lock since
all nodes are unlink files using the same refcount tree.  When the
unlinking file have many extents(over 1600 in our test), most of the
extents has refcounted flag set.  In ocfs2_commit_truncate(), it will
execute the following call trace for every extents.  This means it needs
get and released refcount tree lock about 1600 times.  And when several
nodes are do this at the same time, the performance will be very low.

  ocfs2_remove_btree_range()
  --  ocfs2_lock_refcount_tree()
  ----  ocfs2_refcount_lock()
  ------  __ocfs2_cluster_lock()

ocfs2_refcount_lock() is costly, move it to ocfs2_commit_truncate() to
do lock/unlock once can improve a lot performance.

Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Wengang <wen.gang.wang@oracle.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index a93bf98922565ab120d85f995f0cb565deb1e144..fcae9ef1a328750ca08cda37691bb51dad5293ef 100644 (file)
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -5662,7 +5662,7 @@ int ocfs2_remove_btree_range(struct inode *inode,
                             struct ocfs2_extent_tree *et,
                             u32 cpos, u32 phys_cpos, u32 len, int flags,
                             struct ocfs2_cached_dealloc_ctxt *dealloc,
-                            u64 refcount_loc)
+                            u64 refcount_loc, bool refcount_tree_locked)
 {
        int ret, credits = 0, extra_blocks = 0;
        u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
@@ -5676,11 +5676,13 @@ int ocfs2_remove_btree_range(struct inode *inode,
                BUG_ON(!(OCFS2_I(inode)->ip_dyn_features &
                         OCFS2_HAS_REFCOUNT_FL));
 
-               ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1,
-                                              &ref_tree, NULL);
-               if (ret) {
-                       mlog_errno(ret);
-                       goto bail;
+               if (!refcount_tree_locked) {
+                       ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1,
+                                                      &ref_tree, NULL);
+                       if (ret) {
+                               mlog_errno(ret);
+                               goto bail;
+                       }
                }
 
                ret = ocfs2_prepare_refcount_change_for_del(inode,
@@ -7021,6 +7023,7 @@ int ocfs2_commit_truncate(struct ocfs2_super *osb,
        u64 refcount_loc = le64_to_cpu(di->i_refcount_loc);
        struct ocfs2_extent_tree et;
        struct ocfs2_cached_dealloc_ctxt dealloc;
+       struct ocfs2_refcount_tree *ref_tree = NULL;
 
        ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
        ocfs2_init_dealloc_ctxt(&dealloc);
@@ -7130,9 +7133,18 @@ start:
 
        phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
 
+       if ((flags & OCFS2_EXT_REFCOUNTED) && trunc_len && !ref_tree) {
+               status = ocfs2_lock_refcount_tree(osb, refcount_loc, 1,
+                               &ref_tree, NULL);
+               if (status) {
+                       mlog_errno(status);
+                       goto bail;
+               }
+       }
+
        status = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
                                          phys_cpos, trunc_len, flags, &dealloc,
-                                         refcount_loc);
+                                         refcount_loc, true);
        if (status < 0) {
                mlog_errno(status);
                goto bail;
@@ -7147,6 +7159,8 @@ start:
        goto start;
 
 bail:
+       if (ref_tree)
+               ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
 
        ocfs2_schedule_truncate_log_flush(osb, 1);
 

diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h
index ca381c5841273433d279cd6aca8137ec01d03bbd..fb09b97db162d802e6dbedc495d595a1eee23315 100644 (file)
--- a/fs/ocfs2/alloc.h
+++ b/fs/ocfs2/alloc.h
@@ -142,7 +142,7 @@ int ocfs2_remove_btree_range(struct inode *inode,
                             struct ocfs2_extent_tree *et,
                             u32 cpos, u32 phys_cpos, u32 len, int flags,
                             struct ocfs2_cached_dealloc_ctxt *dealloc,
-                            u64 refcount_loc);
+                            u64 refcount_loc, bool refcount_tree_locked);
 
 int ocfs2_num_free_extents(struct ocfs2_super *osb,
                           struct ocfs2_extent_tree *et);

diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index 79d56dc981bc4b9fd6b927a4bce04d3f79ede736..319e786175afed345660719cfeb4c245741a2461 100644 (file)
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -4479,7 +4479,7 @@ int ocfs2_dx_dir_truncate(struct inode *dir, struct buffer_head *di_bh)
                p_cpos = ocfs2_blocks_to_clusters(dir->i_sb, blkno);
 
                ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen, 0,
-                                              &dealloc, 0);
+                                              &dealloc, 0, false);
                if (ret) {
                        mlog_errno(ret);
                        goto out;

diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 69fb9f75b0825f14527a8e2ec3f8e99c72f5372d..3950693dd0f6d9026587ac71e85037c555790f77 100644 (file)
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -1803,7 +1803,7 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 
                ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
                                               phys_cpos, trunc_len, flags,
-                                              &dealloc, refcount_loc);
+                                              &dealloc, refcount_loc, false);
                if (ret < 0) {
                        mlog_errno(ret);
                        goto out;