__gfs2_set_acl will not increase the reference count of acl if it fails.
In this case, posix_acl_release are called twice, one after
__gfs2_set_acl and one in the error handling block. This patch does not
release acl when __gfs2_set_acl fails.
Fixes: e01580bf9e4("gfs2: use generic posix ACL infrastructure")
Signed-off-by: Pan Bian <[email protected]>
---
fs/gfs2/inode.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 648f0ca..3a9041c 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -744,12 +744,13 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
the gfs2 structures. */
if (default_acl) {
error = __gfs2_set_acl(inode, default_acl, ACL_TYPE_DEFAULT);
- posix_acl_release(default_acl);
+ if (!error)
+ posix_acl_release(default_acl);
}
- if (acl) {
+ if (acl && !error) {
+ error = __gfs2_set_acl(inode, acl, ACL_TYPE_ACCESS);
if (!error)
- error = __gfs2_set_acl(inode, acl, ACL_TYPE_ACCESS);
- posix_acl_release(acl);
+ posix_acl_release(acl);
}
if (error)
--
2.7.4
On Tue, Nov 27, 2018 at 12:19:48AM +0100, Andreas Gruenbacher wrote:
> Pan,
>
> On Mon, 26 Nov 2018 at 09:44, Pan Bian <[email protected]> wrote:
>
> > __gfs2_set_acl will not increase the reference count of acl if it fails.
> > In this case, posix_acl_release are called twice, one after
> > __gfs2_set_acl and one in the error handling block. This patch does not
> > release acl when __gfs2_set_acl fails.
> >
>
> thanks for the bug report and patch; I think this should be fixed more
> clearly as in the patch I've posted earlier today.
I have seen the patch. It is really much clearer. Thank you!
Pan
>
> Andreas