linux-kernel - [PATCH] ext4: fix inline data overflow when xattr value is empty

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <20251018161606.412713-1-kartikey406@gmail.com>
Date: Sat, 18 Oct 2025 21:46:06 +0530
From: Deepanshu Kartikey <kartikey406@...il.com>
To: tytso@....edu,
	adilger.kernel@...ger.ca
Cc: linux-ext4@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	Deepanshu Kartikey <kartikey406@...il.com>,
	syzbot+f3185be57d7e8dda32b8@...kaller.appspotmail.com
Subject: [PATCH] ext4: fix inline data overflow when xattr value is empty

When a file has inline data with an xattr entry but e_value_size is 0,
ext4_prepare_inline_data() incorrectly uses the theoretical maximum
inline size (128 bytes) instead of the actual current capacity (60 bytes
from i_block only). This causes it to accept writes that exceed the
actual capacity, leading to a kernel crash in ext4_write_inline_data_end()
when the BUG_ON(pos + len > EXT4_I(inode)->i_inline_size) is triggered.

This scenario occurs when:
1. A file is created with inline data
2. The file is truncated, leaving an xattr entry with e_value_size=0
3. A write is attempted that exceeds i_block capacity (>60 bytes)

The bug occurs because ext4_prepare_inline_data() calls
ext4_get_max_inline_size() which returns the theoretical maximum (128)
even when the xattr value space is not allocated. This leads to:
- ext4_prepare_inline_data() thinks the write will fit (120 < 128)
- Does not return -ENOSPC
- Inline write path is taken
- ext4_write_inline_data_end() detects overflow and crashes

The fix checks e_value_size in ext4_prepare_inline_data():
- If e_value_size is 0: xattr exists but has no data, cannot expand,
  use actual current capacity (i_inline_size)
- If e_value_size > 0: xattr has data, expansion possible,
  use theoretical maximum (ext4_get_max_inline_size)
- If no xattr entry: use theoretical maximum

This ensures the capacity check accurately reflects available space,
triggering proper conversion to extents when needed and preventing
the overflow crash.

Reported-by: syzbot+f3185be57d7e8dda32b8@...kaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=f3185be57d7e8dda32b8
Signed-off-by: Deepanshu Kartikey <kartikey406@...il.com>
---
 fs/ext4/inline.c | 25 ++++++++++++++++++++++++-
 1 file changed, 24 insertions(+), 1 deletion(-)

diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c
index 1b094a4f3866..3a3aa2d803db 100644
--- a/fs/ext4/inline.c
+++ b/fs/ext4/inline.c
@@ -413,7 +413,30 @@ static int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
 	if (!ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA))
 		return -ENOSPC;
 
-	size = ext4_get_max_inline_size(inode);
+	if (ei->i_inline_off) {
+		struct ext4_iloc iloc;
+		struct ext4_inode *raw_inode;
+		struct ext4_xattr_entry *entry;
+
+		ret = ext4_get_inode_loc(inode, &iloc);
+		if (ret)
+			return ret;
+
+		raw_inode = ext4_raw_inode(&iloc);
+		entry = (struct ext4_xattr_entry *)
+			((void *)raw_inode + ei->i_inline_off);
+
+		if (le32_to_cpu(entry->e_value_size) == 0) {
+			ext4_find_inline_data_nolock(inode);
+			size = ei->i_inline_size;
+		} else {
+			size = ext4_get_max_inline_size(inode);
+		}
+
+		brelse(iloc.bh);
+	} else {
+		size = ext4_get_max_inline_size(inode);
+	}
 	if (size < len)
 		return -ENOSPC;
 
-- 
2.43.0