btrfs_search_slot: reduce lock contention by cowing in two stages

A btree block cow has two parts, the first is to allocate a destination
block and the second is to copy the old bock over.

The first part needs locks in the extent allocation tree, and may need to
do IO.  This changeset splits that into a separate function that can be
called without any tree locks held.

btrfs_search_slot is changed to drop its path and start over if it has
to COW a contended block.  This often means that many writers will
pre-alloc a new destination for a the same contended block, but they
cache their prealloc for later use on lower levels in the tree.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

6 files changed, 173 insertions, 48 deletions

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index c4792062dd53..7114faafa9d4 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -181,7 +181,8 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
                              struct extent_buffer *buf,
                              struct extent_buffer *parent, int parent_slot,
                              struct extent_buffer **cow_ret,
-                             u64 search_start, u64 empty_size)
+                             u64 search_start, u64 empty_size,
+                             u64 prealloc_dest)
 {
         u64 root_gen;
         struct extent_buffer *cow;
@@ -216,10 +217,27 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
         } else {
                 first_key.objectid = 0;
         }
-        cow = btrfs_alloc_free_block(trans, root, buf->len,
-                                     root->root_key.objectid,
-                                     root_gen, first_key.objectid, level,
-                                     search_start, empty_size);
+        if (prealloc_dest) {
+                struct btrfs_key ins;
+
+                ins.objectid = prealloc_dest;
+                ins.offset = buf->len;
+                ins.type = BTRFS_EXTENT_ITEM_KEY;
+
+                ret = btrfs_alloc_reserved_extent(trans, root,
+                                                  root->root_key.objectid,
+                                                  root_gen, level,
+                                                  first_key.objectid,
+                                                  &ins);
+                BUG_ON(ret);
+                cow = btrfs_init_new_buffer(trans, root, prealloc_dest,
+                                            buf->len);
+        } else {
+                cow = btrfs_alloc_free_block(trans, root, buf->len,
+                                             root->root_key.objectid,
+                                             root_gen, first_key.objectid,
+                                             level, search_start, empty_size);
+        }
         if (IS_ERR(cow))
                 return PTR_ERR(cow);
 
@@ -279,7 +297,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
                     struct btrfs_root *root, struct extent_buffer *buf,
                     struct extent_buffer *parent, int parent_slot,
-                    struct extent_buffer **cow_ret)
+                    struct extent_buffer **cow_ret, u64 prealloc_dest)
 {
         u64 search_start;
         u64 header_trans;
@@ -302,12 +320,14 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
             !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
                 *cow_ret = buf;
                 spin_unlock(&root->fs_info->hash_lock);
+                WARN_ON(prealloc_dest);
                 return 0;
         }
         spin_unlock(&root->fs_info->hash_lock);
         search_start = buf->start & ~((u64)(1024 * 1024 * 1024) - 1);
         ret = __btrfs_cow_block(trans, root, buf, parent,
-                                 parent_slot, cow_ret, search_start, 0);
+                                 parent_slot, cow_ret, search_start, 0,
+                                 prealloc_dest);
         return ret;
 }
 
@@ -451,7 +471,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
                 err = __btrfs_cow_block(trans, root, cur, parent, i,
                                         &cur, search_start,
                                         min(16 * blocksize,
-                                            (end_slot - i) * blocksize));
+                                            (end_slot - i) * blocksize), 0);
                 if (err) {
                         btrfs_tree_unlock(cur);
                         free_extent_buffer(cur);
@@ -803,7 +823,7 @@ static int balance_level(struct btrfs_trans_handle *trans,
                 child = read_node_slot(root, mid, 0);
                 btrfs_tree_lock(child);
                 BUG_ON(!child);
-                ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
+                ret = btrfs_cow_block(trans, root, child, mid, 0, &child, 0);
                 BUG_ON(ret);
 
                 spin_lock(&root->node_lock);
@@ -836,7 +856,7 @@ static int balance_level(struct btrfs_trans_handle *trans,
         if (left) {
                 btrfs_tree_lock(left);
                 wret = btrfs_cow_block(trans, root, left,
-                                       parent, pslot - 1, &left);
+                                       parent, pslot - 1, &left, 0);
                 if (wret) {
                         ret = wret;
                         goto enospc;
@@ -846,7 +866,7 @@ static int balance_level(struct btrfs_trans_handle *trans,
         if (right) {
                 btrfs_tree_lock(right);
                 wret = btrfs_cow_block(trans, root, right,
-                                       parent, pslot + 1, &right);
+                                       parent, pslot + 1, &right, 0);
                 if (wret) {
                         ret = wret;
                         goto enospc;
@@ -1021,7 +1041,7 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans,
                         wret = 1;
                 } else {
                         ret = btrfs_cow_block(trans, root, left, parent,
-                                              pslot - 1, &left);
+                                              pslot - 1, &left, 0);
                         if (ret)
                                 wret = 1;
                         else {
@@ -1069,7 +1089,7 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans,
                 } else {
                         ret = btrfs_cow_block(trans, root, right,
                                               parent, pslot + 1,
-                                              &right);
+                                              &right, 0);
                         if (ret)
                                 wret = 1;
                         else {
@@ -1245,6 +1265,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
         u8 lowest_level = 0;
         u64 blocknr;
         u64 gen;
+        struct btrfs_key prealloc_block;
 
         lowest_level = p->lowest_level;
         WARN_ON(lowest_level && ins_len);
@@ -1253,6 +1274,9 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
                 !mutex_is_locked(&root->fs_info->alloc_mutex));
         if (ins_len < 0)
                 lowest_unlock = 2;
+
+        prealloc_block.objectid = 0;
+
 again:
         if (p->skip_locking)
                 b = btrfs_root_node(root);
@@ -1261,27 +1285,82 @@ again:
 
         while (b) {
                 level = btrfs_header_level(b);
+
+                /*
+                 * setup the path here so we can release it under lock
+                 * contention with the cow code
+                 */
+                p->nodes[level] = b;
+                if (!p->skip_locking)
+                        p->locks[level] = 1;
+
                 if (cow) {
                         int wret;
+
+                        /* is a cow on this block not required */
+                        spin_lock(&root->fs_info->hash_lock);
+                        if (btrfs_header_generation(b) == trans->transid &&
+                            !btrfs_header_flag(b, BTRFS_HEADER_FLAG_WRITTEN)) {
+                                spin_unlock(&root->fs_info->hash_lock);
+                                goto cow_done;
+                        }
+                        spin_unlock(&root->fs_info->hash_lock);
+
+                        /* ok, we have to cow, is our old prealloc the right
+                         * size?
+                         */
+                        if (prealloc_block.objectid &&
+                            prealloc_block.offset != b->len) {
+                                btrfs_free_reserved_extent(root,
+                                           prealloc_block.objectid,
+                                           prealloc_block.offset);
+                                prealloc_block.objectid = 0;
+                        }
+
+                        /*
+                         * for higher level blocks, try not to allocate blocks
+                         * with the block and the parent locks held.
+                         */
+                        if (level > 1 && !prealloc_block.objectid &&
+                            btrfs_path_lock_waiting(p, level)) {
+                                u32 size = b->len;
+                                u64 hint = b->start;
+
+                                btrfs_release_path(root, p);
+                                ret = btrfs_reserve_extent(trans, root,
+                                                           size, size, 0,
+                                                           hint, (u64)-1,
+                                                           &prealloc_block, 0);
+                                BUG_ON(ret);
+                                goto again;
+                        }
+
                         wret = btrfs_cow_block(trans, root, b,
                                                p->nodes[level + 1],
                                                p->slots[level + 1],
-                                               &b);
+                                               &b, prealloc_block.objectid);
+                        prealloc_block.objectid = 0;
                         if (wret) {
                                 free_extent_buffer(b);
-                                return wret;
+                                ret = wret;
+                                goto done;
                         }
                 }
+cow_done:
                 BUG_ON(!cow && ins_len);
                 if (level != btrfs_header_level(b))
                         WARN_ON(1);
                 level = btrfs_header_level(b);
+
                 p->nodes[level] = b;
                 if (!p->skip_locking)
                         p->locks[level] = 1;
+
                 ret = check_block(root, p, level);
-                if (ret)
-                        return -1;
+                if (ret) {
+                        ret = -1;
+                        goto done;
+                }
 
                 ret = bin_search(b, key, level, &slot);
                 if (level != 0) {
@@ -1292,15 +1371,19 @@ again:
                             BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
                                 int sret = split_node(trans, root, p, level);
                                 BUG_ON(sret > 0);
-                                if (sret)
-                                        return sret;
+                                if (sret) {
+                                        ret = sret;
+                                        goto done;
+                                }
                                 b = p->nodes[level];
                                 slot = p->slots[level];
                         } else if (ins_len < 0) {
                                 int sret = balance_level(trans, root, p,
                                                          level);
-                                if (sret)
-                                        return sret;
+                                if (sret) {
+                                        ret = sret;
+                                        goto done;
+                                }
                                 b = p->nodes[level];
                                 if (!b) {
                                         btrfs_release_path(NULL, p);
@@ -1362,14 +1445,24 @@ again:
                                 int sret = split_leaf(trans, root, key,
                                                       p, ins_len, ret == 0);
                                 BUG_ON(sret > 0);
-                                if (sret)
-                                        return sret;
+                                if (sret) {
+                                        ret = sret;
+                                        goto done;
+                                }
                         }
                         unlock_up(p, level, lowest_unlock);
-                        return ret;
+                        goto done;
                 }
         }
-        return 1;
+        ret = 1;
+done:
+        if (prealloc_block.objectid) {
+                btrfs_free_reserved_extent(root,
+                           prealloc_block.objectid,
+                           prealloc_block.offset);
+        }
+
+        return ret;
 }
 
 /*
@@ -1840,7 +1933,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 
         /* cow and double check */
         ret = btrfs_cow_block(trans, root, right, upper,
-                              slot + 1, &right);
+                              slot + 1, &right, 0);
         if (ret)
                 goto out_unlock;
 
@@ -2021,7 +2114,7 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
         /* cow and double check */
         ret = btrfs_cow_block(trans, root, left,
-                              path->nodes[1], slot - 1, &left);
+                              path->nodes[1], slot - 1, &left, 0);
         if (ret) {
                 /* we hit -ENOSPC, but it isn't fatal here */
                 ret = 1;
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index d788ab0dcd96..9b025960bbde 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1421,6 +1421,9 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
                                              int level,
                                              u64 hint,
                                              u64 empty_size);
+struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
+                                            struct btrfs_root *root,
+                                            u64 bytenr, u32 blocksize);
 int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size);
 int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans,
                                  struct btrfs_root *root,
@@ -1451,6 +1454,7 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
                       *root, u64 bytenr, u64 num_bytes,
                       u64 root_objectid, u64 ref_generation,
                       u64 owner_objectid, u64 owner_offset, int pin);
+int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
                                struct btrfs_root *root,
                                struct extent_io_tree *unpin);
@@ -1484,7 +1488,7 @@ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
                     struct btrfs_root *root, struct extent_buffer *buf,
                     struct extent_buffer *parent, int parent_slot,
-                    struct extent_buffer **cow_ret);
+                    struct extent_buffer **cow_ret, u64 prealloc_dest);
 int btrfs_copy_root(struct btrfs_trans_handle *trans,
                       struct btrfs_root *root,
                       struct extent_buffer *buf,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 74bcd48a9c43..98a1c0faedae 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2118,6 +2118,15 @@ again:
         return 0;
 }
 
+int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
+{
+        maybe_lock_mutex(root);
+        set_extent_dirty(&root->fs_info->free_space_cache,
+                         start, start + len - 1, GFP_NOFS);
+        maybe_unlock_mutex(root);
+        return 0;
+}
+
 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
                                   struct btrfs_root *root,
                                   u64 num_bytes, u64 min_alloc_size,
@@ -2267,6 +2276,26 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
         maybe_unlock_mutex(root);
         return ret;
 }
+
+struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
+                                            struct btrfs_root *root,
+                                            u64 bytenr, u32 blocksize)
+{
+        struct extent_buffer *buf;
+
+        buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
+        if (!buf)
+                return ERR_PTR(-ENOMEM);
+        btrfs_set_header_generation(buf, trans->transid);
+        btrfs_tree_lock(buf);
+        clean_tree_block(trans, root, buf);
+        btrfs_set_buffer_uptodate(buf);
+        set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
+                         buf->start + buf->len - 1, GFP_NOFS);
+        trans->blocks_used++;
+        return buf;
+}
+
 /*
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
@@ -2293,26 +2322,8 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
                 BUG_ON(ret > 0);
                 return ERR_PTR(ret);
         }
-        buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize);
-        if (!buf) {
-                btrfs_free_extent(trans, root, ins.objectid, blocksize,
-                                  root->root_key.objectid, ref_generation,
-                                  0, 0, 0);
-                return ERR_PTR(-ENOMEM);
-        }
-        btrfs_set_header_generation(buf, trans->transid);
-        btrfs_tree_lock(buf);
-        clean_tree_block(trans, root, buf);
-        btrfs_set_buffer_uptodate(buf);
-
-        if (PageDirty(buf->first_page)) {
-                printk("page %lu dirty\n", buf->first_page->index);
-                WARN_ON(1);
-        }
 
-        set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
-                         buf->start + buf->len - 1, GFP_NOFS);
-        trans->blocks_used++;
+        buf = btrfs_init_new_buffer(trans, root, ins.objectid, blocksize);
         return buf;
 }
 
diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c
index d43e14c7471a..0cc314c10d66 100644
--- a/fs/btrfs/locking.c
+++ b/fs/btrfs/locking.c
@@ -56,3 +56,19 @@ int btrfs_tree_locked(struct extent_buffer *eb)
 {
         return mutex_is_locked(&eb->mutex);
 }
+
+int btrfs_path_lock_waiting(struct btrfs_path *path, int level)
+{
+        int i;
+        struct extent_buffer *eb;
+        for (i = level; i <= level + 1 && i < BTRFS_MAX_LEVEL; i++) {
+                eb = path->nodes[i];
+                if (!eb)
+                        break;
+                smp_mb();
+                if (!list_empty(&eb->mutex.wait_list))
+                        return 1;
+        }
+        return 0;
+}
+
diff --git a/fs/btrfs/locking.h b/fs/btrfs/locking.h
index 2dab96d8280e..bc1faef12519 100644
--- a/fs/btrfs/locking.h
+++ b/fs/btrfs/locking.h
@@ -23,4 +23,5 @@ int btrfs_tree_lock(struct extent_buffer *eb);
 int btrfs_tree_unlock(struct extent_buffer *eb);
 int btrfs_tree_locked(struct extent_buffer *eb);
 int btrfs_try_tree_lock(struct extent_buffer *eb);
+int btrfs_path_lock_waiting(struct btrfs_path *path, int level);
 #endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index a68779499302..9d84daf10008 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -622,7 +622,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
         btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 
         old = btrfs_lock_root_node(root);
-        btrfs_cow_block(trans, root, old, NULL, 0, &old);
+        btrfs_cow_block(trans, root, old, NULL, 0, &old, 0);
 
         btrfs_copy_root(trans, root, old, &tmp, objectid);
         btrfs_tree_unlock(old);