[PATCH] ext4: fix directory index node split corruption

I send patch whose author is Denis Lukianov [off-list ref]
His messages can't reach this list somehow.
I only rebased it ontop of master's HEAD and tested that it fixes
the problem and reviewed.

----

Following commit b5776e7, a trivial sequential write of empty files to
an empty ext4 file system (with large_dir enabled) fails after just
over 26 million files. Depending on luck, file creation will give error
EEXIST or EUCLEAN.

Commit b5776e7 fixed the no-restart condition so that
ext4_handle_dirty_dx_node is always called, but it also broke the
restart condition. This is because when restart=1, the original
implementation correctly skipped do_split() but b5776e7 clobbered the
"if(restart)goto journal_error;" logic.

This complementary change protects do_split() from restart condition,
making it safe from both current and future ordering of goto statements
in earlier sections of the code.

Tested on 5.11.20 with handy testing script:

i = 0
while i <= 32000000:
    print (i)
    with open('tmpmnt/%d' % i, 'wb') as fout:
        i += 1

Google-Bug-Id: 176345532
Fixes: b5776e7 ("ext4: fix potential htree index checksum corruption")
Signed-off-by: Denis Lukianov <redacted>
Signed-off-by: Artem Blagodarenko <redacted>
---
 fs/ext4/namei.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 5fd56f616cf0..0bbff03d4167 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c

@@ -2542,13 +2542,16 @@ static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname,
 			goto journal_error;
 		}
 	}
-	de = do_split(handle, dir, &bh, frame, &fname->hinfo);
-	if (IS_ERR(de)) {
-		err = PTR_ERR(de);
+	if (!restart) {
+		de = do_split(handle, dir, &bh, frame, &fname->hinfo);
+		if (IS_ERR(de)) {
+			err = PTR_ERR(de);
+			goto cleanup;
+		}
+		err = add_dirent_to_buf(handle, fname, dir, inode, de,
+bh);
 		goto cleanup;
 	}
-	err = add_dirent_to_buf(handle, fname, dir, inode, de, bh);
-	goto cleanup;
 
 journal_error:
 	ext4_std_error(dir->i_sb, err); /* this is a no-op if err == 0 */

-- 
2.18.4

On Fri 30-07-21 14:24:03, Artem Blagodarenko wrote:

I send patch whose author is Denis Lukianov [off-list ref]
His messages can't reach this list somehow.
I only rebased it ontop of master's HEAD and tested that it fixes
the problem and reviewed.

----

Following commit b5776e7, a trivial sequential write of empty files to
an empty ext4 file system (with large_dir enabled) fails after just
over 26 million files. Depending on luck, file creation will give error
EEXIST or EUCLEAN.

Commit b5776e7 fixed the no-restart condition so that
ext4_handle_dirty_dx_node is always called, but it also broke the
restart condition. This is because when restart=1, the original
implementation correctly skipped do_split() but b5776e7 clobbered the
"if(restart)goto journal_error;" logic.

This complementary change protects do_split() from restart condition,
making it safe from both current and future ordering of goto statements
in earlier sections of the code.

Tested on 5.11.20 with handy testing script:

i = 0
while i <= 32000000:
    print (i)
    with open('tmpmnt/%d' % i, 'wb') as fout:
        i += 1

Google-Bug-Id: 176345532
Fixes: b5776e7 ("ext4: fix potential htree index checksum corruption")

Please use 12 characters from git commit when identifying it (7 characters 
is likely to become non-unique). Otherwise the fix looks good so feel free
to add:

Reviewed-by: Jan Kara <jack@suse.cz>

Also I think you should include Ted in 'To' or 'CC' to make it more likely
he sees the patch.

								Honza

-- 
Jan Kara [off-list ref]
SUSE Labs, CR