1 files changed, 126 insertions, 66 deletions

diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 372b05ff1943..2b0a627cb5f9 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -274,18 +274,32 @@ struct io_ctl {
 };
 
 static int io_ctl_init(struct io_ctl *io_ctl, struct inode *inode,
-                       struct btrfs_root *root)
+                       struct btrfs_root *root, int write)
 {
+        int num_pages;
+        int check_crcs = 0;
+
+        num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
+                    PAGE_CACHE_SHIFT;
+
+        if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
+                check_crcs = 1;
+
+        /* Make sure we can fit our crcs into the first page */
+        if (write && check_crcs &&
+            (num_pages * sizeof(u32)) >= PAGE_CACHE_SIZE)
+                return -ENOSPC;
+
         memset(io_ctl, 0, sizeof(struct io_ctl));
-        io_ctl->num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
-                PAGE_CACHE_SHIFT;
-        io_ctl->pages = kzalloc(sizeof(struct page *) * io_ctl->num_pages,
-                                GFP_NOFS);
+
+        io_ctl->pages = kzalloc(sizeof(struct page *) * num_pages, GFP_NOFS);
         if (!io_ctl->pages)
                 return -ENOMEM;
+
+        io_ctl->num_pages = num_pages;
         io_ctl->root = root;
-        if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
-                io_ctl->check_crcs = 1;
+        io_ctl->check_crcs = check_crcs;
+
         return 0;
 }
 
@@ -666,6 +680,13 @@ static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
         generation = btrfs_free_space_generation(leaf, header);
         btrfs_release_path(path);
 
+        if (!BTRFS_I(inode)->generation) {
+                btrfs_info(root->fs_info,
+                           "The free space cache file (%llu) is invalid. skip it\n",
+                           offset);
+                return 0;
+        }
+
         if (BTRFS_I(inode)->generation != generation) {
                 btrfs_err(root->fs_info,
                         "free space inode generation (%llu) "
@@ -677,7 +698,7 @@ static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
         if (!num_entries)
                 return 0;
 
-        ret = io_ctl_init(&io_ctl, inode, root);
+        ret = io_ctl_init(&io_ctl, inode, root, 0);
         if (ret)
                 return ret;
 
@@ -957,19 +978,18 @@ fail:
 }
 
 static noinline_for_stack int
-add_ioctl_entries(struct btrfs_root *root,
-                  struct inode *inode,
-                  struct btrfs_block_group_cache *block_group,
-                  struct io_ctl *io_ctl,
-                  struct extent_state **cached_state,
-                  struct list_head *bitmap_list,
-                  int *entries)
+write_pinned_extent_entries(struct btrfs_root *root,
+                            struct btrfs_block_group_cache *block_group,
+                            struct io_ctl *io_ctl,
+                            int *entries)
 {
         u64 start, extent_start, extent_end, len;
-        struct list_head *pos, *n;
         struct extent_io_tree *unpin = NULL;
         int ret;
 
+        if (!block_group)
+                return 0;
+
         /*
          * We want to add any pinned extents to our free space cache
          * so we don't leak the space
@@ -979,23 +999,19 @@ add_ioctl_entries(struct btrfs_root *root,
          */
         unpin = root->fs_info->pinned_extents;
 
-        if (block_group)
-                start = block_group->key.objectid;
+        start = block_group->key.objectid;
 
-        while (block_group && (start < block_group->key.objectid +
-                               block_group->key.offset)) {
+        while (start < block_group->key.objectid + block_group->key.offset) {
                 ret = find_first_extent_bit(unpin, start,
                                             &extent_start, &extent_end,
                                             EXTENT_DIRTY, NULL);
-                if (ret) {
-                        ret = 0;
-                        break;
-                }
+                if (ret)
+                        return 0;
 
                 /* This pinned extent is out of our range */
                 if (extent_start >= block_group->key.objectid +
                     block_group->key.offset)
-                        break;
+                        return 0;
 
                 extent_start = max(extent_start, start);
                 extent_end = min(block_group->key.objectid +
@@ -1005,11 +1021,20 @@ add_ioctl_entries(struct btrfs_root *root,
                 *entries += 1;
                 ret = io_ctl_add_entry(io_ctl, extent_start, len, NULL);
                 if (ret)
-                        goto out_nospc;
+                        return -ENOSPC;
 
                 start = extent_end;
         }
 
+        return 0;
+}
+
+static noinline_for_stack int
+write_bitmap_entries(struct io_ctl *io_ctl, struct list_head *bitmap_list)
+{
+        struct list_head *pos, *n;
+        int ret;
+
         /* Write out the bitmaps */
         list_for_each_safe(pos, n, bitmap_list) {
                 struct btrfs_free_space *entry =
@@ -1017,36 +1042,24 @@ add_ioctl_entries(struct btrfs_root *root,
 
                 ret = io_ctl_add_bitmap(io_ctl, entry->bitmap);
                 if (ret)
-                        goto out_nospc;
+                        return -ENOSPC;
                 list_del_init(&entry->list);
         }
 
-        /* Zero out the rest of the pages just to make sure */
-        io_ctl_zero_remaining_pages(io_ctl);
-
-        ret = btrfs_dirty_pages(root, inode, io_ctl->pages, io_ctl->num_pages,
-                                0, i_size_read(inode), cached_state);
-        io_ctl_drop_pages(io_ctl);
-        unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
-                             i_size_read(inode) - 1, cached_state, GFP_NOFS);
+        return 0;
+}
 
-        if (ret)
-                goto fail;
+static int flush_dirty_cache(struct inode *inode)
+{
+        int ret;
 
         ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
-        if (ret) {
+        if (ret)
                 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
                                  EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
                                  GFP_NOFS);
-                goto fail;
-        }
-        return 0;
 
-fail:
-        return -1;
-
-out_nospc:
-        return -ENOSPC;
+        return ret;
 }
 
 static void noinline_for_stack
@@ -1056,6 +1069,7 @@ cleanup_write_cache_enospc(struct inode *inode,
                            struct list_head *bitmap_list)
 {
         struct list_head *pos, *n;
+
         list_for_each_safe(pos, n, bitmap_list) {
                 struct btrfs_free_space *entry =
                         list_entry(pos, struct btrfs_free_space, list);
@@ -1088,64 +1102,104 @@ static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
 {
         struct extent_state *cached_state = NULL;
         struct io_ctl io_ctl;
-        struct list_head bitmap_list;
+        LIST_HEAD(bitmap_list);
         int entries = 0;
         int bitmaps = 0;
         int ret;
-        int err = -1;
-
-        INIT_LIST_HEAD(&bitmap_list);
 
         if (!i_size_read(inode))
                 return -1;
 
-        ret = io_ctl_init(&io_ctl, inode, root);
+        ret = io_ctl_init(&io_ctl, inode, root, 1);
         if (ret)
                 return -1;
 
+        if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA)) {
+                down_write(&block_group->data_rwsem);
+                spin_lock(&block_group->lock);
+                if (block_group->delalloc_bytes) {
+                        block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+                        spin_unlock(&block_group->lock);
+                        up_write(&block_group->data_rwsem);
+                        BTRFS_I(inode)->generation = 0;
+                        ret = 0;
+                        goto out;
+                }
+                spin_unlock(&block_group->lock);
+        }
+
         /* Lock all pages first so we can lock the extent safely. */
         io_ctl_prepare_pages(&io_ctl, inode, 0);
 
         lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
                          0, &cached_state);
 
-
-        /* Make sure we can fit our crcs into the first page */
-        if (io_ctl.check_crcs &&
-            (io_ctl.num_pages * sizeof(u32)) >= PAGE_CACHE_SIZE)
-                goto out_nospc;
-
         io_ctl_set_generation(&io_ctl, trans->transid);
 
+        /* Write out the extent entries in the free space cache */
         ret = write_cache_extent_entries(&io_ctl, ctl,
                                          block_group, &entries, &bitmaps,
                                          &bitmap_list);
         if (ret)
                 goto out_nospc;
 
-        ret = add_ioctl_entries(root, inode, block_group, &io_ctl,
-                                &cached_state, &bitmap_list, &entries);
+        /*
+         * Some spaces that are freed in the current transaction are pinned,
+         * they will be added into free space cache after the transaction is
+         * committed, we shouldn't lose them.
+         */
+        ret = write_pinned_extent_entries(root, block_group, &io_ctl, &entries);
+        if (ret)
+                goto out_nospc;
 
-        if (ret == -ENOSPC)
+        /* At last, we write out all the bitmaps. */
+        ret = write_bitmap_entries(&io_ctl, &bitmap_list);
+        if (ret)
                 goto out_nospc;
-        else if (ret)
+
+        /* Zero out the rest of the pages just to make sure */
+        io_ctl_zero_remaining_pages(&io_ctl);
+
+        /* Everything is written out, now we dirty the pages in the file. */
+        ret = btrfs_dirty_pages(root, inode, io_ctl.pages, io_ctl.num_pages,
+                                0, i_size_read(inode), &cached_state);
+        if (ret)
+                goto out_nospc;
+
+        if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
+                up_write(&block_group->data_rwsem);
+        /*
+         * Release the pages and unlock the extent, we will flush
+         * them out later
+         */
+        io_ctl_drop_pages(&io_ctl);
+
+        unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
+                             i_size_read(inode) - 1, &cached_state, GFP_NOFS);
+
+        /* Flush the dirty pages in the cache file. */
+        ret = flush_dirty_cache(inode);
+        if (ret)
                 goto out;
 
-        err = update_cache_item(trans, root, inode, path, offset,
+        /* Update the cache item to tell everyone this cache file is valid. */
+        ret = update_cache_item(trans, root, inode, path, offset,
                                 entries, bitmaps);
-
 out:
         io_ctl_free(&io_ctl);
-        if (err) {
+        if (ret) {
                 invalidate_inode_pages2(inode->i_mapping);
                 BTRFS_I(inode)->generation = 0;
         }
         btrfs_update_inode(trans, root, inode);
-        return err;
+        return ret;
 
 out_nospc:
-
         cleanup_write_cache_enospc(inode, &io_ctl, &cached_state, &bitmap_list);
+
+        if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
+                up_write(&block_group->data_rwsem);
+
         goto out;
 }
 
@@ -1165,6 +1219,12 @@ int btrfs_write_out_cache(struct btrfs_root *root,
                 spin_unlock(&block_group->lock);
                 return 0;
         }
+
+        if (block_group->delalloc_bytes) {
+                block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+                spin_unlock(&block_group->lock);
+                return 0;
+        }
         spin_unlock(&block_group->lock);
 
         inode = lookup_free_space_inode(root, block_group, path);