[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20231009144340.418904-1-max.kellermann@ionos.com>
Date: Mon, 9 Oct 2023 16:43:39 +0200
From: Max Kellermann <max.kellermann@...os.com>
To: Xiubo Li <xiubli@...hat.com>, Ilya Dryomov <idryomov@...il.com>,
Jeff Layton <jlayton@...nel.org>, Jan Kara <jack@...e.com>,
Dave Kleikamp <shaggy@...nel.org>
Cc: Max Kellermann <max.kellermann@...os.com>,
ceph-devel@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-ext4@...r.kernel.org, jfs-discussion@...ts.sourceforge.net
Subject: [PATCH v2] fs/{posix_acl,ext2,jfs,ceph}: apply umask if ACL support is disabled
One important implementation detail of the posix_acl_create() function
is that it applies the umask to the "mode" parameter. If
CONFIG_FS_POSIX_ACL is disabled, this detail is missing and the umask
may not get applied.
This patch adds the missing code to posix_acl_create() and to three
filesystems that omit the posix_acl_create() call if their individual
ACL support is disabled (CONFIG_EXT2_FS_POSIX_ACL,
CONFIG_JFS_POSIX_ACL, CONFIG_CEPH_FS_POSIX_ACL). If
posix_acl_create() never gets called, the umask needs to be applied
anyway.
This bug used to be exploitable easily with O_TMPFILE (see
https://bugzilla.kernel.org/show_bug.cgi?id=203625) but that part was
fixed by commit ac6800e279a2 ("fs: Add missing umask strip in
vfs_tmpfile") last year. The bug may not be reachable by userspace
anymore, but since it is apparently still necessary to apply the umask
again in posix_acl_create(), there is no reason to assume it's not
necessary with ACL support is disabled.
Signed-off-by: Max Kellermann <max.kellermann@...os.com>
---
fs/ceph/super.h | 6 ++++++
fs/ext2/acl.h | 6 ++++++
fs/jfs/jfs_acl.h | 6 ++++++
include/linux/posix_acl.h | 1 +
4 files changed, 19 insertions(+)
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 51c7f2b14f6f..58349639bd57 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -1194,6 +1194,12 @@ static inline void ceph_forget_all_cached_acls(struct inode *inode)
static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
struct ceph_acl_sec_ctx *as_ctx)
{
+ /* usually, the umask is applied by posix_acl_create(), but if
+ * ACL support is disabled at compile time, we need to do it
+ * here, because posix_acl_create() will never be called
+ */
+ *mode &= ~current_umask();
+
return 0;
}
static inline void ceph_init_inode_acls(struct inode *inode,
diff --git a/fs/ext2/acl.h b/fs/ext2/acl.h
index 4a8443a2b8ec..0ecaa9c20c0c 100644
--- a/fs/ext2/acl.h
+++ b/fs/ext2/acl.h
@@ -67,6 +67,12 @@ extern int ext2_init_acl (struct inode *, struct inode *);
static inline int ext2_init_acl (struct inode *inode, struct inode *dir)
{
+ /* usually, the umask is applied by posix_acl_create(), but if
+ * ACL support is disabled at compile time, we need to do it
+ * here, because posix_acl_create() will never be called
+ */
+ inode->i_mode &= ~current_umask();
+
return 0;
}
#endif
diff --git a/fs/jfs/jfs_acl.h b/fs/jfs/jfs_acl.h
index f892e54d0fcd..64a05e663a45 100644
--- a/fs/jfs/jfs_acl.h
+++ b/fs/jfs/jfs_acl.h
@@ -17,6 +17,12 @@ int jfs_init_acl(tid_t, struct inode *, struct inode *);
static inline int jfs_init_acl(tid_t tid, struct inode *inode,
struct inode *dir)
{
+ /* usually, the umask is applied by posix_acl_create(), but if
+ * ACL support is disabled at compile time, we need to do it
+ * here, because posix_acl_create() will never be called
+ */
+ inode->i_mode &= ~current_umask();
+
return 0;
}
diff --git a/include/linux/posix_acl.h b/include/linux/posix_acl.h
index 0e65b3d634d9..54bc9b1061ca 100644
--- a/include/linux/posix_acl.h
+++ b/include/linux/posix_acl.h
@@ -128,6 +128,7 @@ static inline void cache_no_acl(struct inode *inode)
static inline int posix_acl_create(struct inode *inode, umode_t *mode,
struct posix_acl **default_acl, struct posix_acl **acl)
{
+ *mode &= ~current_umask();
*default_acl = *acl = NULL;
return 0;
}
--
2.39.2
Powered by blists - more mailing lists