Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs

Pull btrfs update from Chris Mason:
 "These are the usual mixture of bugs, cleanups and performance fixes.
  Miao has some really nice tuning of our crc code as well as our
  transaction commits.

  Josef is peeling off more and more problems related to early enospc,
  and has a number of important bug fixes in here too"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (81 commits)
  Btrfs: wait ordered range before doing direct io
  Btrfs: only do the tree_mod_log_free_eb if this is our last ref
  Btrfs: hold the tree mod lock in __tree_mod_log_rewind
  Btrfs: make backref walking code handle skinny metadata
  Btrfs: fix crash regarding to ulist_add_merge
  Btrfs: fix several potential problems in copy_nocow_pages_for_inode
  Btrfs: cleanup the code of copy_nocow_pages_for_inode()
  Btrfs: fix oops when recovering the file data by scrub function
  Btrfs: make the chunk allocator completely tree lockless
  Btrfs: cleanup orphaned root orphan item
  Btrfs: fix wrong mirror number tuning
  Btrfs: cleanup redundant code in btrfs_submit_direct()
  Btrfs: remove btrfs_sector_sum structure
  Btrfs: check if we can nocow if we don't have data space
  Btrfs: stop using try_to_writeback_inodes_sb_nr to flush delalloc
  Btrfs: use a percpu to keep track of possibly pinned bytes
  Btrfs: check for actual acls rather than just xattrs when caching no acl
  Btrfs: move btrfs_truncate_page to btrfs_cont_expand instead of btrfs_truncate
  Btrfs: optimize reada_for_balance
  Btrfs: optimize read_block_for_search
  ...

1 files changed, 52 insertions, 38 deletions

diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c
index 79bd479317cb..4ba2a69a60ad 100644
--- a/fs/btrfs/scrub.c
+++ b/fs/btrfs/scrub.c
@@ -2126,8 +2126,7 @@ static int scrub_find_csum(struct scrub_ctx *sctx, u64 logical, u64 len,
 			   u8 *csum)
 {
 	struct btrfs_ordered_sum *sum = NULL;
-	int ret = 0;
-	unsigned long i;
+	unsigned long index;
 	unsigned long num_sectors;
 
 	while (!list_empty(&sctx->csum_list)) {
@@ -2146,19 +2145,14 @@ static int scrub_find_csum(struct scrub_ctx *sctx, u64 logical, u64 len,
 	if (!sum)
 		return 0;
 
+	index = ((u32)(logical - sum->bytenr)) / sctx->sectorsize;
 	num_sectors = sum->len / sctx->sectorsize;
-	for (i = 0; i < num_sectors; ++i) {
-		if (sum->sums[i].bytenr == logical) {
-			memcpy(csum, &sum->sums[i].sum, sctx->csum_size);
-			ret = 1;
-			break;
-		}
-	}
-	if (ret && i == num_sectors - 1) {
+	memcpy(csum, sum->sums + index, sctx->csum_size);
+	if (index == num_sectors - 1) {
 		list_del(&sum->list);
 		kfree(sum);
 	}
-	return ret;
+	return 1;
 }
 
 /* scrub extent tries to collect up to 64 kB for each bio */
@@ -2505,6 +2499,7 @@ again:
 			if (ret)
 				goto out;
 
+			scrub_free_csums(sctx);
 			if (extent_logical + extent_len <
 			    key.objectid + bytes) {
 				logical += increment;
@@ -3204,16 +3199,18 @@ out:
 
 static int copy_nocow_pages_for_inode(u64 inum, u64 offset, u64 root, void *ctx)
 {
-	unsigned long index;
 	struct scrub_copy_nocow_ctx *nocow_ctx = ctx;
-	int ret = 0;
+	struct btrfs_fs_info *fs_info = nocow_ctx->sctx->dev_root->fs_info;
 	struct btrfs_key key;
-	struct inode *inode = NULL;
+	struct inode *inode;
+	struct page *page;
 	struct btrfs_root *local_root;
 	u64 physical_for_dev_replace;
 	u64 len;
-	struct btrfs_fs_info *fs_info = nocow_ctx->sctx->dev_root->fs_info;
+	unsigned long index;
 	int srcu_index;
+	int ret;
+	int err;
 
 	key.objectid = root;
 	key.type = BTRFS_ROOT_ITEM_KEY;
@@ -3227,6 +3224,11 @@ static int copy_nocow_pages_for_inode(u64 inum, u64 offset, u64 root, void *ctx)
 		return PTR_ERR(local_root);
 	}
 
+	if (btrfs_root_refs(&local_root->root_item) == 0) {
+		srcu_read_unlock(&fs_info->subvol_srcu, srcu_index);
+		return -ENOENT;
+	}
+
 	key.type = BTRFS_INODE_ITEM_KEY;
 	key.objectid = inum;
 	key.offset = 0;
@@ -3235,19 +3237,21 @@ static int copy_nocow_pages_for_inode(u64 inum, u64 offset, u64 root, void *ctx)
 	if (IS_ERR(inode))
 		return PTR_ERR(inode);
 
+	/* Avoid truncate/dio/punch hole.. */
+	mutex_lock(&inode->i_mutex);
+	inode_dio_wait(inode);
+
+	ret = 0;
 	physical_for_dev_replace = nocow_ctx->physical_for_dev_replace;
 	len = nocow_ctx->len;
 	while (len >= PAGE_CACHE_SIZE) {
-		struct page *page = NULL;
-		int ret_sub;
-
 		index = offset >> PAGE_CACHE_SHIFT;
-
+again:
 		page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
 		if (!page) {
 			pr_err("find_or_create_page() failed\n");
 			ret = -ENOMEM;
-			goto next_page;
+			goto out;
 		}
 
 		if (PageUptodate(page)) {
@@ -3255,39 +3259,49 @@ static int copy_nocow_pages_for_inode(u64 inum, u64 offset, u64 root, void *ctx)
 				goto next_page;
 		} else {
 			ClearPageError(page);
-			ret_sub = extent_read_full_page(&BTRFS_I(inode)->
+			err = extent_read_full_page(&BTRFS_I(inode)->
 							 io_tree,
 							page, btrfs_get_extent,
 							nocow_ctx->mirror_num);
-			if (ret_sub) {
-				ret = ret_sub;
+			if (err) {
+				ret = err;
 				goto next_page;
 			}
-			wait_on_page_locked(page);
+
+			lock_page(page);
+			/*
+			 * If the page has been remove from the page cache,
+			 * the data on it is meaningless, because it may be
+			 * old one, the new data may be written into the new
+			 * page in the page cache.
+			 */
+			if (page->mapping != inode->i_mapping) {
+				page_cache_release(page);
+				goto again;
+			}
 			if (!PageUptodate(page)) {
 				ret = -EIO;
 				goto next_page;
 			}
 		}
-		ret_sub = write_page_nocow(nocow_ctx->sctx,
-					   physical_for_dev_replace, page);
-		if (ret_sub) {
-			ret = ret_sub;
-			goto next_page;
-		}
-
+		err = write_page_nocow(nocow_ctx->sctx,
+				       physical_for_dev_replace, page);
+		if (err)
+			ret = err;
 next_page:
-		if (page) {
-			unlock_page(page);
-			put_page(page);
-		}
+		unlock_page(page);
+		page_cache_release(page);
+
+		if (ret)
+			break;
+
 		offset += PAGE_CACHE_SIZE;
 		physical_for_dev_replace += PAGE_CACHE_SIZE;
 		len -= PAGE_CACHE_SIZE;
 	}
-
-	if (inode)
-		iput(inode);
+out:
+	mutex_unlock(&inode->i_mutex);
+	iput(inode);
 	return ret;
 }