[<prev] [next>] [day] [month] [year] [list]
Message-Id: <20241018162236.1553-1-andrew.kanner@gmail.com>
Date: Fri, 18 Oct 2024 18:22:36 +0200
From: Andrew Kanner <andrew.kanner@...il.com>
To: aivazian.tigran@...il.com
Cc: linux-kernel@...r.kernel.org,
Andrew Kanner <andrew.kanner@...il.com>,
syzbot+94891a5155abdf6821b7@...kaller.appspotmail.com,
stable@...r.kernel.org
Subject: [PATCH RESEND] fs/bfs: fix possible NULL pointer dereference caused by empty i_op/i_fop
Syzkaller reported and reproduced the following issue:
loop0: detected capacity change from 0 to 64
overlayfs: fs on './file0' does not support file handles, \
falling back to index=off,nfs_export=off.
BUG: kernel NULL pointer dereference, address: 0000000000000000
[...]
Comm: syz-executor169 Not tainted 6.11.0-rc5-syzkaller-00176-g20371ba12063 #0
[...]
Call Trace:
<TASK>
__lookup_slow+0x28c/0x3f0 fs/namei.c:1718
lookup_slow fs/namei.c:1735 [inline]
lookup_one_unlocked+0x1a4/0x290 fs/namei.c:2898
ovl_lookup_positive_unlocked fs/overlayfs/namei.c:210 [inline]
ovl_lookup_single+0x200/0xbd0 fs/overlayfs/namei.c:240
ovl_lookup_layer+0x417/0x510 fs/overlayfs/namei.c:333
ovl_lookup+0xcf7/0x2a60 fs/overlayfs/namei.c:1124
lookup_one_qstr_excl+0x11f/0x260 fs/namei.c:1633
filename_create+0x297/0x540 fs/namei.c:3980
do_mknodat+0x18b/0x5b0 fs/namei.c:4125
__do_sys_mknod fs/namei.c:4171 [inline]
__se_sys_mknod fs/namei.c:4169 [inline]
__x64_sys_mknod+0x8c/0xa0 fs/namei.c:4169
do_syscall_x64 arch/x86/entry/common.c:52 [inline]
do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7fc4b42b2839
However, the actual root cause is not related to overlayfs:
(gdb) p lower.dentry->d_inode->i_op
$6 = (const struct inode_operations *) 0xffffffff8242fcc0 <empty_iops>
(gdb) p lower.dentry->d_inode->i_op->lookup
$7 = (struct dentry *(*) \
(struct inode *, struct dentry *, unsigned int)) 0x0
The inode, which is passed to ovl_lookup(), has an empty i_op,
so the following __lookup_slow() hit NULL doing it's job:
old = inode->i_op->lookup(inode, dentry, flags);
bfs_fill_super()->bfs_iget() are skipping i_op/i_fop initialization
if vnode type is not BFS_VDIR or BFS_VREG (e.g. corrupted fs).
Adding extra error handling fixes the issue and syzkaller repro
doesn't trigger anything bad anymore.
Issue affects kernel builds with CONFIG_BFS_FS=y or =m only.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot+94891a5155abdf6821b7@...kaller.appspotmail.com
Closes: https://lore.kernel.org/all/0000000000003d5bc30617238b6d@google.com/T/
Signed-off-by: Andrew Kanner <andrew.kanner@...il.com>
Cc: stable@...r.kernel.org
---
fs/bfs/inode.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/fs/bfs/inode.c b/fs/bfs/inode.c
index db81570c9637..e590b231ad20 100644
--- a/fs/bfs/inode.c
+++ b/fs/bfs/inode.c
@@ -70,6 +70,10 @@ struct inode *bfs_iget(struct super_block *sb, unsigned long ino)
inode->i_op = &bfs_file_inops;
inode->i_fop = &bfs_file_operations;
inode->i_mapping->a_ops = &bfs_aops;
+ } else {
+ pr_err("Bad i_vtype for inode %s:%08lx\n", inode->i_sb->s_id, ino);
+ brelse(bh);
+ goto error;
}
BFS_I(inode)->i_sblock = le32_to_cpu(di->i_sblock);
--
2.39.3
Powered by blists - more mailing lists