Re: [PATCH] nfs: Fix potential posix_acl refcnt leak in nfs3_set_acl

From: Trond Myklebust
Date: Mon Apr 20 2020 - 09:00:02 EST


On Mon, 2020-04-20 at 14:51 +0200, Andreas Gruenbacher wrote:
> Am Mo., 20. Apr. 2020 um 14:15 Uhr schrieb Trond Myklebust <
> trondmy@xxxxxxxxxxxxxxx>:
> > I don't really see any alternative to adding a 'dfalloc' to track
> > the
> > allocated dfacl separately.
>
> Something like the attached patch should work as well.
>
> Thanks,
> Andreas
>
> ---
> fs/nfs/nfs3acl.c | 32 ++++++++++++++++++--------------
> 1 file changed, 18 insertions(+), 14 deletions(-)
>
> diff --git a/fs/nfs/nfs3acl.c b/fs/nfs/nfs3acl.c
> index c5c3fc6e6c60..f1581f11c220 100644
> --- a/fs/nfs/nfs3acl.c
> +++ b/fs/nfs/nfs3acl.c
> @@ -253,37 +253,41 @@ int nfs3_proc_setacls(struct inode *inode,
> struct posix_acl *acl,
>
> int nfs3_set_acl(struct inode *inode, struct posix_acl *acl, int
> type)
> {
> - struct posix_acl *alloc = NULL, *dfacl = NULL;
> + struct posix_acl *orig = acl, *dfacl = NULL;
> int status;
>
> if (S_ISDIR(inode->i_mode)) {
> switch(type) {
> case ACL_TYPE_ACCESS:
> - alloc = dfacl = get_acl(inode,
> ACL_TYPE_DEFAULT);
> - if (IS_ERR(alloc))
> - goto fail;
> + dfacl = get_acl(inode, ACL_TYPE_DEFAULT);
> + status = PTR_ERR(dfacl);
> + if (IS_ERR(dfacl))
> + goto out;
> break;
>
> case ACL_TYPE_DEFAULT:
> dfacl = acl;
> - alloc = acl = get_acl(inode, ACL_TYPE_ACCESS);
> - if (IS_ERR(alloc))
> - goto fail;
> + acl = get_acl(inode, ACL_TYPE_ACCESS);
> + status = PTR_ERR(acl);
> + if (IS_ERR(acl))
> + goto out;
> break;
> }
> }
>
> if (acl == NULL) {
> - alloc = acl = posix_acl_from_mode(inode->i_mode,
> GFP_KERNEL);
> - if (IS_ERR(alloc))
> - goto fail;
> + acl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
> + status = PTR_ERR(acl);
> + if (IS_ERR(acl))
> + goto out;
> }
> status = __nfs3_proc_setacls(inode, acl, dfacl);
> - posix_acl_release(alloc);
> +out:
> + if (acl != orig)
> + posix_acl_release(acl);
> + if (dfacl != orig)
> + posix_acl_release(dfacl);
> return status;
> -
> -fail:
> - return PTR_ERR(alloc);
> }
>
> const struct xattr_handler *nfs3_xattr_handlers[] = {
>
> base-commit: ae83d0b416db002fe95601e7f97f64b59514d936

Well, that should Oops when either IS_ERR(acl) or IS_ERR(dfacl)
triggers, shouldn't it? ð

--
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@xxxxxxxxxxxxxxx