On 2024/8/3 0:31, Kent Overstreet wrote:
On Tue, Jul 09, 2024 at 04:10:20PM GMT, Hongbo Li wrote:
The longest name length for file is BCH_NAME_MAX in bcachefs.
When looking up a longer path, it does not check this limit. So
we add the validation for this to avoid undetermined risk.
What's the actual bug this fixes?
The error code is inconsistent for this case. Actually, in internal the
BCH_NAME_MAX is defined and used. If this don't fix, it will give the
ENOENT error for long path.
Thanks,
Hongbo
Signed-off-by: Hongbo Li <[email protected]>
---
fs/bcachefs/fs.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index fd01a5402c28..f8da598ca9f8 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -474,6 +474,9 @@ static struct dentry *bch2_lookup(struct inode *vdir,
struct dentry *dentry,
struct bch_inode_info *dir = to_bch_ei(vdir);
struct bch_hash_info hash = bch2_hash_info_init(c, &dir->ei_inode);
+ if (dentry->d_name.len > BCH_NAME_MAX)
+ return ERR_PTR(-ENAMETOOLONG);
+
struct bch_inode_info *inode;
bch2_trans_do(c, NULL, NULL, 0,
PTR_ERR_OR_ZERO(inode = bch2_lookup_trans(trans,
inode_inum(dir),
--
2.34.1
