[RFC PATCH v3 3/3] ext4: data=journal: write-protect pages on j_submit_inode_data_buffers()

[Mauricio Faria de Oliveira <mfo@canonical.com>]

This implements journal callbacks j_submit|finish_inode_data_buffers()
with different behavior for data=journal: to write-protect pages under
commit, preventing changes to buffers writeably mapped to userspace.

If a buffer's content changes between commit's checksum calculation
and write-out to disk, it can cause journal recovery/mount failures
upon a kernel crash or power loss.

    [   27.334874] EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!
    [   27.339492] JBD2: Invalid checksum recovering data block 8705 in log
    [   27.342716] JBD2: recovery failed
    [   27.343316] EXT4-fs (loop0): error loading journal
    mount: /ext4: can't read superblock on /dev/loop0.

In j_submit_inode_data_buffers() we write-protect the inode's pages
with write_cache_pages() and redirty w/ writepage callback if needed.

In j_finish_inode_data_buffers() there is nothing do to.

And in order to use the callbacks, inodes are added to the inode list
in transaction in __ext4_journalled_writepage() and ext4_page_mkwrite().

In ext4_page_mkwrite() we must make sure that:

1) the inode is always added to the list;
   thus we skip the 'all buffers mapped' optimization on data=journal;

2) the buffers are attached to transaction as dirty;
   as already done in __ext4_journalled_writepage().

Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reported-by: Dann Frazier <dann.frazier@canonical.com>
---
 fs/ext4/inode.c | 29 ++++++++++++++------
 fs/ext4/super.c | 72 +++++++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 91 insertions(+), 10 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index bf596467c234..fa4109da056c 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1910,6 +1910,9 @@ static int __ext4_journalled_writepage(struct page *page,
 		err = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
 					     write_end_fn);
 	}
+	if (ret == 0)
+		ret = err;
+	err = ext4_jbd2_inode_add_write(handle, inode, 0, len);
 	if (ret == 0)
 		ret = err;
 	EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;
@@ -6004,9 +6007,12 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
 		len = PAGE_SIZE;
 	/*
 	 * Return if we have all the buffers mapped. This avoids the need to do
-	 * journal_start/journal_stop which can block and take a long time
+	 * journal_start/journal_stop which can block and take a long time.
+	 *
+	 * This cannot be done for data journalling, as we have to add the
+	 * inode to the transaction's list to writeprotect pages on commit.
 	 */
-	if (page_has_buffers(page)) {
+	if (page_has_buffers(page) && !ext4_should_journal_data(inode)) {
 		if (!ext4_walk_page_buffers(NULL, page_buffers(page),
 					    0, len, NULL,
 					    ext4_bh_unmapped)) {
@@ -6032,12 +6038,14 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
 	err = block_page_mkwrite(vma, vmf, get_block);
 	if (!err && ext4_should_journal_data(inode)) {
 		if (ext4_walk_page_buffers(handle, page_buffers(page), 0,
-			  PAGE_SIZE, NULL, do_journal_get_write_access)) {
-			unlock_page(page);
-			ret = VM_FAULT_SIGBUS;
-			ext4_journal_stop(handle);
-			goto out;
-		}
+			PAGE_SIZE, NULL, do_journal_get_write_access))
+			goto out_err;
+		/* Make sure buffers are attached to the transaction as dirty */
+		if (ext4_walk_page_buffers(handle, page_buffers(page), 0,
+			PAGE_SIZE, NULL, write_end_fn))
+			goto out_err;
+		if (ext4_jbd2_inode_add_write(handle, inode, 0, PAGE_SIZE))
+			goto out_err;
 		ext4_set_inode_state(inode, EXT4_STATE_JDATA);
 	}
 	ext4_journal_stop(handle);
@@ -6049,6 +6057,11 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
 	up_read(&EXT4_I(inode)->i_mmap_sem);
 	sb_end_pagefault(inode->i_sb);
 	return ret;
+out_err:
+	unlock_page(page);
+	ret = VM_FAULT_SIGBUS;
+	ext4_journal_stop(handle);
+	goto out;
 }
 
 vm_fault_t ext4_filemap_fault(struct vm_fault *vmf)
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 7303839d7ad9..528b5e20b71c 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -472,14 +472,82 @@ static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
 	spin_unlock(&sbi->s_md_lock);
 }
 
+/*
+ * This writepage callback for write_cache_pages()
+ * takes care of a few cases after page cleaning.
+ *
+ * write_cache_pages() already checks for dirty pages
+ * and calls clear_page_dirty_for_io(), which we want,
+ * to write protect the pages.
+ *
+ * However, we have to redirty a page in these cases:
+ * 1) some buffer is dirty (needs checkpointing)
+ * 2) some buffer is not part of the committing transaction
+ * 3) some buffer already has b_next_transaction set
+ */
+
+static int ext4_journalled_writepage_callback(struct page *page,
+					      struct writeback_control *wbc,
+					      void *data)
+{
+	transaction_t *transaction = (transaction_t *) data;
+	struct buffer_head *bh, *head;
+	struct journal_head *jh;
+
+	bh = head = page_buffers(page);
+	do {
+		jh = bh2jh(bh);
+		if (buffer_dirty(bh) ||
+			(jh && (jh->b_transaction != transaction ||
+				jh->b_next_transaction))) {
+			redirty_page_for_writepage(wbc, page);
+			goto out;
+		}
+	} while ((bh = bh->b_this_page) != head);
+
+out:
+	return AOP_WRITEPAGE_ACTIVATE;
+}
+
+static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
+{
+	struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
+	transaction_t *transaction = jinode->i_transaction;
+	loff_t dirty_start = jinode->i_dirty_start;
+	loff_t dirty_end = jinode->i_dirty_end;
+
+	struct writeback_control wbc = {
+		.sync_mode =  WB_SYNC_ALL,
+		.nr_to_write = ~0ULL,
+		.range_start = dirty_start,
+		.range_end = dirty_end,
+        };
+
+	return write_cache_pages(mapping, &wbc,
+				 ext4_journalled_writepage_callback,
+				 transaction);
+}
+
 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
 {
-	return jbd2_journal_submit_inode_data_buffers(jinode);
+	int ret;
+
+	if (ext4_should_journal_data(jinode->i_vfs_inode))
+		ret = ext4_journalled_submit_inode_data_buffers(jinode);
+	else
+		ret = jbd2_journal_submit_inode_data_buffers(jinode);
+
+	return ret;
 }
 
 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
 {
-	return jbd2_journal_finish_inode_data_buffers(jinode);
+	int ret = 0;
+
+	if (!ext4_should_journal_data(jinode->i_vfs_inode))
+		ret = jbd2_journal_finish_inode_data_buffers(jinode);
+
+	return ret;
 }
 
 static bool system_going_down(void)
-- 
2.17.1