fs/ceph/ioctl: add owner/capability checks for CEPH_IOC_SET_LAYOUT*
PATCH 1/1 fs/ceph/ioctl: add owner/capability checks for CEPH_IOC_SET_LAYOUT*
These permission checks were already missing in the initial impementation of these ioctls. This Ceph allows any user who owns a file descriptor to manipulate the layout of any file, even if they don't have write permissions.
It might be a good idea to guard other ioctls with permission checks as well or even disallow regular users (even if they own the file) to manipulate layout settings completely, as this may be abused to DoS the Ceph servers, but right now, I find it most urgent to have setter checks at all.
---
Note: this is a resend. I had already sent this to [email protected] and [email protected] on 2024-11-25, but the Ceph maintainers thought it was "not a big problem". It was never merged.
--- fs/ceph/ioctl.c | 6 ++++++ 1 file changed, 6 insertions(+)
fs/ceph/ioctl.c +6 -0
@@ -72,6 +72,9 @@ static long ceph_ioctl_set_layout(struct file *file, void __user *arg)struct ceph_ioctl_layout nl;int err; +if (!inode_owner_or_capable (&nop_mnt_idmap, inode)) +return -EACCES ; +if (copy_from_user (&l, arg,sizeof (l)))return -EFAULT ;
@@ -142,6 +145,9 @@ static long ceph_ioctl_set_layout_policy (struct file *file, void __user *arg)int err;struct ceph_mds_client *mdsc =ceph_sb_to_fs_client (inode->i_sb)->mdsc; +if (!inode_owner_or_capable (&nop_mnt_idmap, inode)) +return -EACCES ; +/* copy and validate */ if (copy_from_user (&l, arg,sizeof (l)))return -EFAULT ;
Signature · 1 line
2.47.3
Hi Max,
Quoted 6 lines · Max Kellermann
These permission checks were already missing in the initial impementation of these ioctls. This Ceph allows any user who owns a file descriptor to manipulate the layout of any file, even if they don't have write permissions.
Quoted 6 lines
It might be a good idea to guard other ioctls with permission checks as well or even disallow regular users (even if they own the file) to manipulate layout settings completely, as this may be abused to DoS the Ceph servers, but right now, I find it most urgent to have setter checks at all.
Thanks for bringing this up again. LGTM :-)
Quoted 6 lines
Fixes 8f4e91dee2a2 ("ceph: ioctls")Signed-off-by Max Kellermann <[email protected]>---
Note: this is a resend. I had already sent this to [email protected] and [email protected] on 2024-11-25, but the Ceph maintainers thought it was "not a big problem". It was never merged.
It should have just been missed.
The MDSAuthCaps::is_capable() in the MDS does validate the client's cephx caps. However, with the default cephx cap configuration ("allow rw" / "allow rwp"), grant.match.uid defaults to MDS_AUTH_UID_ANY, and is_capable() returns true immediately — all Unix user-level permission checks (caller_uid vs inode_uid, mode bits) are skipped entirely. The MDS only enforces client-level caps; it delegates user-level enforcement to the kernel client.
On a multi-user host with cephx 'rwp' credentials, any local user with a file descriptor can manipulate another user's file layout, redirect data to arbitrary RADOS pools, set directory quotas, or pin directories to specific MDS ranks. The kernel is the only place where per-user isolation can happen for these operations — defense in depth cannot be outsourced to the MDS here.
Thanks
Xiubo
Quoted 31 lines
--- fs/ceph/ioctl.c | 6 ++++++ 1 file changed, 6 insertions(+)fs/ceph/ioctl.c +6
-1@@ -72,6 +72,9 @@ static long ceph_ioctl_set_layout(struct file *file, void __user *arg)struct ceph_ioctl_layout nl;int err; +if (!inode_owner_or_capable (&nop_mnt_idmap, inode)) +return -EACCES ; +if (copy_from_user (&l, arg,sizeof (l)))return -EFAULT ;@@ -142,6 +145,9 @@ static long ceph_ioctl_set_layout_policy (struct file *file, void __user *arg)int err;struct ceph_mds_client *mdsc =ceph_sb_to_fs_client (inode->i_sb)->mdsc; +if (!inode_owner_or_capable (&nop_mnt_idmap, inode)) +return -EACCES ; +/* copy and validate */ if (copy_from_user (&l, arg,sizeof (l)))return -EFAULT ; --2.47.3