Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel…

…/git/jack/linux-fs-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs-2.6:
  jbd: Fix comment to match the code in journal_start()
  jbd/jbd2: remove obsolete summarise_journal_usage.
  jbd: Fix forever sleeping process in do_get_write_access()
  ext2: fix error msg when mounting fs with too-large blocksize
  jbd: fix fsync() tid wraparound bug
  ext3: Fix fs corruption when make_indexed_dir() fails
  ext3: Fix lock inversion in ext3_symlink()

fs/ext2/super.c

@@ -898,7 +898,8 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 		brelse(bh);
 
 		if (!sb_set_blocksize(sb, blocksize)) {
-			ext2_msg(sb, KERN_ERR, "error: blocksize is too small");
+			ext2_msg(sb, KERN_ERR,
+				"error: bad blocksize %d", blocksize);
 			goto failed_sbi;
 		}
 

fs/ext3/namei.c

@@ -1416,10 +1416,19 @@ static int make_indexed_dir(handle_t *handle, struct dentry *dentry,
 	frame->at = entries;
 	frame->bh = bh;
 	bh = bh2;
+	/*
+	 * Mark buffers dirty here so that if do_split() fails we write a
+	 * consistent set of buffers to disk.
+	 */
+	ext3_journal_dirty_metadata(handle, frame->bh);
+	ext3_journal_dirty_metadata(handle, bh);
 	de = do_split(handle,dir, &bh, frame, &hinfo, &retval);
-	dx_release (frames);
-	if (!(de))
+	if (!de) {
+		ext3_mark_inode_dirty(handle, dir);
+		dx_release(frames);
 		return retval;
+	}
+	dx_release(frames);
 
 	return add_dirent_to_buf(handle, dentry, inode, de, bh);
 }
@@ -2189,17 +2198,34 @@ static int ext3_symlink (struct inode * dir,
 	handle_t *handle;
 	struct inode * inode;
 	int l, err, retries = 0;
+	int credits;
 
 	l = strlen(symname)+1;
 	if (l > dir->i_sb->s_blocksize)
 		return -ENAMETOOLONG;
 
 	dquot_initialize(dir);
 
+	if (l > EXT3_N_BLOCKS * 4) {
+		/*
+		 * For non-fast symlinks, we just allocate inode and put it on
+		 * orphan list in the first transaction => we need bitmap,
+		 * group descriptor, sb, inode block, quota blocks.
+		 */
+		credits = 4 + EXT3_MAXQUOTAS_INIT_BLOCKS(dir->i_sb);
+	} else {
+		/*
+		 * Fast symlink. We have to add entry to directory
+		 * (EXT3_DATA_TRANS_BLOCKS + EXT3_INDEX_EXTRA_TRANS_BLOCKS),
+		 * allocate new inode (bitmap, group descriptor, inode block,
+		 * quota blocks, sb is already counted in previous macros).
+		 */
+		credits = EXT3_DATA_TRANS_BLOCKS(dir->i_sb) +
+			  EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3 +
+			  EXT3_MAXQUOTAS_INIT_BLOCKS(dir->i_sb);
+	}
 retry:
-	handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS(dir->i_sb) +
-					EXT3_INDEX_EXTRA_TRANS_BLOCKS + 5 +
-					EXT3_MAXQUOTAS_INIT_BLOCKS(dir->i_sb));
+	handle = ext3_journal_start(dir, credits);
 	if (IS_ERR(handle))
 		return PTR_ERR(handle);
 
@@ -2211,21 +2237,45 @@ static int ext3_symlink (struct inode * dir,
 	if (IS_ERR(inode))
 		goto out_stop;
 
-	if (l > sizeof (EXT3_I(inode)->i_data)) {
+	if (l > EXT3_N_BLOCKS * 4) {
 		inode->i_op = &ext3_symlink_inode_operations;
 		ext3_set_aops(inode);
 		/*
-		 * page_symlink() calls into ext3_prepare/commit_write.
-		 * We have a transaction open.  All is sweetness.  It also sets
-		 * i_size in generic_commit_write().
+		 * We cannot call page_symlink() with transaction started
+		 * because it calls into ext3_write_begin() which acquires page
+		 * lock which ranks below transaction start (and it can also
+		 * wait for journal commit if we are running out of space). So
+		 * we have to stop transaction now and restart it when symlink
+		 * contents is written. 
+		 *
+		 * To keep fs consistent in case of crash, we have to put inode
+		 * to orphan list in the mean time.
 		 */
+		drop_nlink(inode);
+		err = ext3_orphan_add(handle, inode);
+		ext3_journal_stop(handle);
+		if (err)
+			goto err_drop_inode;
 		err = __page_symlink(inode, symname, l, 1);
+		if (err)
+			goto err_drop_inode;
+		/*
+		 * Now inode is being linked into dir (EXT3_DATA_TRANS_BLOCKS
+		 * + EXT3_INDEX_EXTRA_TRANS_BLOCKS), inode is also modified
+		 */
+		handle = ext3_journal_start(dir,
+				EXT3_DATA_TRANS_BLOCKS(dir->i_sb) +
+				EXT3_INDEX_EXTRA_TRANS_BLOCKS + 1);
+		if (IS_ERR(handle)) {
+			err = PTR_ERR(handle);
+			goto err_drop_inode;
+		}
+		inc_nlink(inode);
+		err = ext3_orphan_del(handle, inode);
 		if (err) {
+			ext3_journal_stop(handle);
 			drop_nlink(inode);
-			unlock_new_inode(inode);
-			ext3_mark_inode_dirty(handle, inode);
-			iput (inode);
-			goto out_stop;
+			goto err_drop_inode;
 		}
 	} else {
 		inode->i_op = &ext3_fast_symlink_inode_operations;
@@ -2239,6 +2289,10 @@ static int ext3_symlink (struct inode * dir,
 	if (err == -ENOSPC && ext3_should_retry_alloc(dir->i_sb, &retries))
 		goto retry;
 	return err;
+err_drop_inode:
+	unlock_new_inode(inode);
+	iput(inode);
+	return err;
 }
 
 static int ext3_link (struct dentry * old_dentry,

fs/jbd/commit.c

@@ -302,12 +302,6 @@ void journal_commit_transaction(journal_t *journal)
 	 * all outstanding updates to complete.
 	 */
 
-#ifdef COMMIT_STATS
-	spin_lock(&journal->j_list_lock);
-	summarise_journal_usage(journal);
-	spin_unlock(&journal->j_list_lock);
-#endif
-
 	/* Do we need to erase the effects of a prior journal_flush? */
 	if (journal->j_flags & JFS_FLUSHED) {
 		jbd_debug(3, "super block updated\n");
@@ -722,8 +716,13 @@ void journal_commit_transaction(journal_t *journal)
                    required. */
 		JBUFFER_TRACE(jh, "file as BJ_Forget");
 		journal_file_buffer(jh, commit_transaction, BJ_Forget);
-		/* Wake up any transactions which were waiting for this
-		   IO to complete */
+		/*
+		 * Wake up any transactions which were waiting for this
+		 * IO to complete. The barrier must be here so that changes
+		 * by journal_file_buffer() take effect before wake_up_bit()
+		 * does the waitqueue check.
+		 */
+		smp_mb();
 		wake_up_bit(&bh->b_state, BH_Unshadow);
 		JBUFFER_TRACE(jh, "brelse shadowed buffer");
 		__brelse(bh);

fs/jbd/journal.c

@@ -437,9 +437,12 @@ int __log_space_left(journal_t *journal)
 int __log_start_commit(journal_t *journal, tid_t target)
 {
 	/*
-	 * Are we already doing a recent enough commit?
+	 * The only transaction we can possibly wait upon is the
+	 * currently running transaction (if it exists).  Otherwise,
+	 * the target tid must be an old one.
 	 */
-	if (!tid_geq(journal->j_commit_request, target)) {
+	if (journal->j_running_transaction &&
+	    journal->j_running_transaction->t_tid == target) {
 		/*
 		 * We want a new commit: OK, mark the request and wakeup the
 		 * commit thread.  We do _not_ do the commit ourselves.
@@ -451,7 +454,14 @@ int __log_start_commit(journal_t *journal, tid_t target)
 			  journal->j_commit_sequence);
 		wake_up(&journal->j_wait_commit);
 		return 1;
-	}
+	} else if (!tid_geq(journal->j_commit_request, target))
+		/* This should never happen, but if it does, preserve
+		   the evidence before kjournald goes into a loop and
+		   increments j_commit_sequence beyond all recognition. */
+		WARN_ONCE(1, "jbd: bad log_start_commit: %u %u %u %u\n",
+		    journal->j_commit_request, journal->j_commit_sequence,
+		    target, journal->j_running_transaction ?
+		    journal->j_running_transaction->t_tid : 0);
 	return 0;
 }
 

fs/jbd/transaction.c

@@ -266,7 +266,8 @@ static handle_t *new_handle(int nblocks)
  * This function is visible to journal users (like ext3fs), so is not
  * called with the journal already locked.
  *
- * Return a pointer to a newly allocated handle, or NULL on failure
+ * Return a pointer to a newly allocated handle, or an ERR_PTR() value
+ * on failure.
  */
 handle_t *journal_start(journal_t *journal, int nblocks)
 {

fs/jbd2/commit.c

@@ -338,12 +338,6 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 	 * all outstanding updates to complete.
 	 */
 
-#ifdef COMMIT_STATS
-	spin_lock(&journal->j_list_lock);
-	summarise_journal_usage(journal);
-	spin_unlock(&journal->j_list_lock);
-#endif
-
 	/* Do we need to erase the effects of a prior jbd2_journal_flush? */
 	if (journal->j_flags & JBD2_FLUSHED) {
 		jbd_debug(3, "super block updated\n");