1 files changed, 49 insertions, 87 deletions

diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
index 384b6ebb655..2c6ccc49ba2 100644
--- a/fs/ext3/inode.c
+++ b/fs/ext3/inode.c
@@ -228,7 +228,7 @@ void ext3_evict_inode (struct inode *inode)
 		log_wait_commit(journal, commit_tid);
 		filemap_write_and_wait(&inode->i_data);
 	}
-	truncate_inode_pages(&inode->i_data, 0);
+	truncate_inode_pages_final(&inode->i_data);
 
 	ext3_discard_reservation(inode);
 	rsv = ei->i_block_alloc_info;
@@ -1559,56 +1559,17 @@ static int buffer_unmapped(handle_t *handle, struct buffer_head *bh)
 }
 
 /*
- * Note that we always start a transaction even if we're not journalling
- * data.  This is to preserve ordering: any hole instantiation within
- * __block_write_full_page -> ext3_get_block() should be journalled
- * along with the data so we don't crash and then get metadata which
+ * Note that whenever we need to map blocks we start a transaction even if
+ * we're not journalling data.  This is to preserve ordering: any hole
+ * instantiation within __block_write_full_page -> ext3_get_block() should be
+ * journalled along with the data so we don't crash and then get metadata which
  * refers to old data.
  *
  * In all journalling modes block_write_full_page() will start the I/O.
  *
- * Problem:
- *
- *	ext3_writepage() -> kmalloc() -> __alloc_pages() -> page_launder() ->
- *		ext3_writepage()
- *
- * Similar for:
- *
- *	ext3_file_write() -> generic_file_write() -> __alloc_pages() -> ...
- *
- * Same applies to ext3_get_block().  We will deadlock on various things like
- * lock_journal and i_truncate_mutex.
- *
- * Setting PF_MEMALLOC here doesn't work - too many internal memory
- * allocations fail.
- *
- * 16May01: If we're reentered then journal_current_handle() will be
- *	    non-zero. We simply *return*.
- *
- * 1 July 2001: @@@ FIXME:
- *   In journalled data mode, a data buffer may be metadata against the
- *   current transaction.  But the same file is part of a shared mapping
- *   and someone does a writepage() on it.
- *
- *   We will move the buffer onto the async_data list, but *after* it has
- *   been dirtied. So there's a small window where we have dirty data on
- *   BJ_Metadata.
- *
- *   Note that this only applies to the last partial page in the file.  The
- *   bit which block_write_full_page() uses prepare/commit for.  (That's
- *   broken code anyway: it's wrong for msync()).
- *
- *   It's a rare case: affects the final partial page, for journalled data
- *   where the file is subject to bith write() and writepage() in the same
- *   transction.  To fix it we'll need a custom block_write_full_page().
- *   We'll probably need that anyway for journalling writepage() output.
- *
  * We don't honour synchronous mounts for writepage().  That would be
  * disastrous.  Any write() or metadata operation will sync the fs for
  * us.
- *
- * AKPM2: if all the page's buffers are mapped to disk and !data=journal,
- * we don't need to open a transaction here.
  */
 static int ext3_ordered_writepage(struct page *page,
 				struct writeback_control *wbc)
@@ -1673,12 +1634,9 @@ static int ext3_ordered_writepage(struct page *page,
 	 * block_write_full_page() succeeded.  Otherwise they are unmapped,
 	 * and generally junk.
 	 */
-	if (ret == 0) {
-		err = walk_page_buffers(handle, page_bufs, 0, PAGE_CACHE_SIZE,
+	if (ret == 0)
+		ret = walk_page_buffers(handle, page_bufs, 0, PAGE_CACHE_SIZE,
 					NULL, journal_dirty_data_fn);
-		if (!ret)
-			ret = err;
-	}
 	walk_page_buffers(handle, page_bufs, 0,
 			PAGE_CACHE_SIZE, NULL, bput_one);
 	err = ext3_journal_stop(handle);
@@ -1758,17 +1716,17 @@ static int ext3_journalled_writepage(struct page *page,
 	WARN_ON_ONCE(IS_RDONLY(inode) &&
 		     !(EXT3_SB(inode->i_sb)->s_mount_state & EXT3_ERROR_FS));
 
-	if (ext3_journal_current_handle())
-		goto no_write;
-
 	trace_ext3_journalled_writepage(page);
-	handle = ext3_journal_start(inode, ext3_writepage_trans_blocks(inode));
-	if (IS_ERR(handle)) {
-		ret = PTR_ERR(handle);
-		goto no_write;
-	}
-
 	if (!page_has_buffers(page) || PageChecked(page)) {
+		if (ext3_journal_current_handle())
+			goto no_write;
+
+		handle = ext3_journal_start(inode,
+					    ext3_writepage_trans_blocks(inode));
+		if (IS_ERR(handle)) {
+			ret = PTR_ERR(handle);
+			goto no_write;
+		}
 		/*
 		 * It's mmapped pagecache.  Add buffers and journal it.  There
 		 * doesn't seem much point in redirtying the page here.
@@ -1791,17 +1749,18 @@ static int ext3_journalled_writepage(struct page *page,
 		atomic_set(&EXT3_I(inode)->i_datasync_tid,
 			   handle->h_transaction->t_tid);
 		unlock_page(page);
+		err = ext3_journal_stop(handle);
+		if (!ret)
+			ret = err;
 	} else {
 		/*
-		 * It may be a page full of checkpoint-mode buffers.  We don't
-		 * really know unless we go poke around in the buffer_heads.
-		 * But block_write_full_page will do the right thing.
+		 * It is a page full of checkpoint-mode buffers. Go and write
+		 * them. They should have been already mapped when they went
+		 * to the journal so provide NULL get_block function to catch
+		 * errors.
 		 */
-		ret = block_write_full_page(page, ext3_get_block, wbc);
+		ret = block_write_full_page(page, NULL, wbc);
 	}
-	err = ext3_journal_stop(handle);
-	if (!ret)
-		ret = err;
 out:
 	return ret;
 
@@ -1862,8 +1821,7 @@ static int ext3_releasepage(struct page *page, gfp_t wait)
  * VFS code falls back into buffered path in that case so we are safe.
  */
 static ssize_t ext3_direct_IO(int rw, struct kiocb *iocb,
-			const struct iovec *iov, loff_t offset,
-			unsigned long nr_segs)
+			struct iov_iter *iter, loff_t offset)
 {
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_mapping->host;
@@ -1871,10 +1829,10 @@ static ssize_t ext3_direct_IO(int rw, struct kiocb *iocb,
 	handle_t *handle;
 	ssize_t ret;
 	int orphan = 0;
-	size_t count = iov_length(iov, nr_segs);
+	size_t count = iov_iter_count(iter);
 	int retries = 0;
 
-	trace_ext3_direct_IO_enter(inode, offset, iov_length(iov, nr_segs), rw);
+	trace_ext3_direct_IO_enter(inode, offset, count, rw);
 
 	if (rw == WRITE) {
 		loff_t final_size = offset + count;
@@ -1898,15 +1856,14 @@ static ssize_t ext3_direct_IO(int rw, struct kiocb *iocb,
 	}
 
 retry:
-	ret = blockdev_direct_IO(rw, iocb, inode, iov, offset, nr_segs,
-				 ext3_get_block);
+	ret = blockdev_direct_IO(rw, iocb, inode, iter, offset, ext3_get_block);
 	/*
 	 * In case of error extending write may have instantiated a few
 	 * blocks outside i_size. Trim these off again.
 	 */
 	if (unlikely((rw & WRITE) && ret < 0)) {
 		loff_t isize = i_size_read(inode);
-		loff_t end = offset + iov_length(iov, nr_segs);
+		loff_t end = offset + count;
 
 		if (end > isize)
 			ext3_truncate_failed_direct_write(inode);
@@ -1925,6 +1882,8 @@ retry:
 			 * and pretend the write failed... */
 			ext3_truncate_failed_direct_write(inode);
 			ret = PTR_ERR(handle);
+			if (inode->i_nlink)
+				ext3_orphan_del(NULL, inode);
 			goto out;
 		}
 		if (inode->i_nlink)
@@ -1949,8 +1908,7 @@ retry:
 			ret = err;
 	}
 out:
-	trace_ext3_direct_IO_exit(inode, offset,
-				iov_length(iov, nr_segs), rw, ret);
+	trace_ext3_direct_IO_exit(inode, offset, count, rw, ret);
 	return ret;
 }
 
@@ -3212,21 +3170,20 @@ out_brelse:
  *
  * We are called from a few places:
  *
- * - Within generic_file_write() for O_SYNC files.
+ * - Within generic_file_aio_write() -> generic_write_sync() for O_SYNC files.
  *   Here, there will be no transaction running. We wait for any running
  *   transaction to commit.
  *
- * - Within sys_sync(), kupdate and such.
- *   We wait on commit, if tol to.
+ * - Within flush work (for sys_sync(), kupdate and such).
+ *   We wait on commit, if told to.
  *
- * - Within prune_icache() (PF_MEMALLOC == true)
- *   Here we simply return.  We can't afford to block kswapd on the
- *   journal commit.
+ * - Within iput_final() -> write_inode_now()
+ *   We wait on commit, if told to.
  *
  * In all cases it is actually safe for us to return without doing anything,
  * because the inode has been copied into a raw inode buffer in
- * ext3_mark_inode_dirty().  This is a correctness thing for O_SYNC and for
- * knfsd.
+ * ext3_mark_inode_dirty().  This is a correctness thing for WB_SYNC_ALL
+ * writeback.
  *
  * Note that we are absolutely dependent upon all inode dirtiers doing the
  * right thing: they *must* call mark_inode_dirty() after dirtying info in
@@ -3238,13 +3195,13 @@ out_brelse:
  *	stuff();
  *	inode->i_size = expr;
  *
- * is in error because a kswapd-driven write_inode() could occur while
- * `stuff()' is running, and the new i_size will be lost.  Plus the inode
- * will no longer be on the superblock's dirty inode list.
+ * is in error because write_inode() could occur while `stuff()' is running,
+ * and the new i_size will be lost.  Plus the inode will no longer be on the
+ * superblock's dirty inode list.
  */
 int ext3_write_inode(struct inode *inode, struct writeback_control *wbc)
 {
-	if (current->flags & PF_MEMALLOC)
+	if (WARN_ON_ONCE(current->flags & PF_MEMALLOC))
 		return 0;
 
 	if (ext3_journal_current_handle()) {
@@ -3253,7 +3210,12 @@ int ext3_write_inode(struct inode *inode, struct writeback_control *wbc)
 		return -EIO;
 	}
 
-	if (wbc->sync_mode != WB_SYNC_ALL)
+	/*
+	 * No need to force transaction in WB_SYNC_NONE mode. Also
+	 * ext3_sync_fs() will force the commit after everything is
+	 * written.
+	 */
+	if (wbc->sync_mode != WB_SYNC_ALL || wbc->for_sync)
 		return 0;
 
 	return ext3_force_commit(inode->i_sb);