2014-04-11 07:16:54

by Zhang Zhen

[permalink] [raw]
Subject: [PATCH] ext4: avoid unneeded lookup when xattr name is invalid

In ext4_xattr_set_handle() we have checked the xattr name's length. So we should
also check it in ext4_xattr_get() to avoid unneeded lookup caused by invalid name.

In addition, we deleted the check of NULL in ext4_xattr_set_handle(), because in all
the callers of the ext4_xattr_set_handle(), the name can't be NULL.

Signed-off-by: Zhang Zhen <[email protected]>
---
fs/ext4/xattr.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index 1f5cf58..340bdfa 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -369,6 +369,9 @@ ext4_xattr_get(struct inode *inode, int name_index, const char *name,
{
int error;

+ if (strlen(name) > 255)
+ return -ERANGE;
+
down_read(&EXT4_I(inode)->xattr_sem);
error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
buffer_size);
@@ -1073,8 +1076,6 @@ ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
unsigned long no_expand;
int error;

- if (!name)
- return -EINVAL;
if (strlen(name) > 255)
return -ERANGE;
down_write(&EXT4_I(inode)->xattr_sem);
--
1.8.5.5


.






2014-05-12 14:01:35

by Theodore Ts'o

[permalink] [raw]
Subject: Re: [PATCH] ext4: avoid unneeded lookup when xattr name is invalid

On Fri, Apr 11, 2014 at 03:15:07PM +0800, Zhang Zhen wrote:
> In ext4_xattr_set_handle() we have checked the xattr name's length. So we should
> also check it in ext4_xattr_get() to avoid unneeded lookup caused by invalid name.
>
> Signed-off-by: Zhang Zhen <[email protected]>

Applied, although I dropped this bit:

> In addition, we deleted the check of NULL in ext4_xattr_set_handle(), because in all
> the callers of the ext4_xattr_set_handle(), the name can't be NULL.

Verifying this is non-trivial, since there are many non-local
users of this function, and in order to make sure this is true you
need to verify (at least) the callers of the callers. I could be
convinced to turn this into a BUG_ON(!name), or "if
(unlikely(!name))", but leaving the check in is a good idea, since
even if it is true today, it's not necessarily guarantted to be true N
years in the future.

- Ted