Re: [PATCH v2 3/3] fuse: Add support for mounts from user namespaces

[Serge Hallyn <serge.hallyn@ubuntu.com>]

Quoting Seth Forshee (seth.forshee@canonical.com):
> On Fri, Sep 05, 2014 at 04:48:11PM +0000, Serge Hallyn wrote:
> > Quoting Seth Forshee (seth.forshee@canonical.com):
> > > Update fuse to support mounts from within user namespaces. This
> > > is mostly a matter of translating uids and gids into the
> > > namespace of the process reading requests before handing the
> > > requests off to userspace.
> > > 
> > > Due to security concerns the namespace used should be fixed,
> > > otherwise a user might be able to pass the fuse fd to
> > > init_user_ns and inject suid files owned by a user outside the
> > > namespace in order to gain elevated privileges. For fuse we
> > > stash current_user_ns() when a filesystem is first mounted and
> > > abort the mount if this namespace is different than the one used
> > > to open the fd passed in the mount options.
> > > 
> > > The allow_others options could also be a problem, as a userns
> > > mount could bypass system policy for this option and thus open
> > > the possiblity of DoS attacks. This is prevented by restricting
> > > the scope of allow_other to apply only to that superblock's
> > > userns and its children, giving the expected behavior within the
> > > userns while preventing DoS attacks on more privileged contexts.
> > > 
> > > Signed-off-by: Seth Forshee <seth.forshee@canonical.com>

Acked-by: Serge E. Hallyn <serge.hallyn@ubuntu.com>

> > 
> > Thanks, Seth, just two little questions below.
> > 
> > > ---
> > >  fs/fuse/dev.c    |  4 ++--
> > >  fs/fuse/dir.c    | 46 +++++++++++++++++++++++++++++++++-------------
> > >  fs/fuse/fuse_i.h |  4 ++++
> > >  fs/fuse/inode.c  | 28 ++++++++++++++++++++--------
> > >  4 files changed, 59 insertions(+), 23 deletions(-)
> > > 
> > > diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
> > > index 839caebd34f1..03c8785ed731 100644
> > > --- a/fs/fuse/dev.c
> > > +++ b/fs/fuse/dev.c
> > > @@ -127,8 +127,8 @@ static void __fuse_put_request(struct fuse_req *req)
> > >  
> > >  static void fuse_req_init_context(struct fuse_conn *fc, struct fuse_req *req)
> > >  {
> > > -	req->in.h.uid = from_kuid_munged(&init_user_ns, current_fsuid());
> > > -	req->in.h.gid = from_kgid_munged(&init_user_ns, current_fsgid());
> > > +	req->in.h.uid = from_kuid_munged(fc->user_ns, current_fsuid());
> > > +	req->in.h.gid = from_kgid_munged(fc->user_ns, current_fsgid());
> > >  	req->in.h.pid = pid_nr_ns(task_pid(current), fc->pid_ns);
> > >  }
> > >  
> > > diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
> > > index de1d84af9f7c..c0b9968db6a1 100644
> > > --- a/fs/fuse/dir.c
> > > +++ b/fs/fuse/dir.c
> > > @@ -905,8 +905,8 @@ static void fuse_fillattr(struct inode *inode, struct fuse_attr *attr,
> > >  	stat->ino = attr->ino;
> > >  	stat->mode = (inode->i_mode & S_IFMT) | (attr->mode & 07777);
> > >  	stat->nlink = attr->nlink;
> > > -	stat->uid = make_kuid(&init_user_ns, attr->uid);
> > > -	stat->gid = make_kgid(&init_user_ns, attr->gid);
> > > +	stat->uid = make_kuid(fc->user_ns, attr->uid);
> > > +	stat->gid = make_kgid(fc->user_ns, attr->gid);
> > >  	stat->rdev = inode->i_rdev;
> > >  	stat->atime.tv_sec = attr->atime;
> > >  	stat->atime.tv_nsec = attr->atimensec;
> > > @@ -1085,12 +1085,20 @@ int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
> > >   */
> > >  int fuse_allow_current_process(struct fuse_conn *fc)
> > >  {
> > > -	const struct cred *cred;
> > > +	const struct cred *cred = current_cred();
> > >  
> > > -	if (fc->flags & FUSE_ALLOW_OTHER)
> > > -		return 1;
> > > +	if (fc->flags & FUSE_ALLOW_OTHER) {
> > > +		if (kuid_has_mapping(fc->user_ns, cred->euid) &&
> > > +		    kuid_has_mapping(fc->user_ns, cred->suid) &&
> > > +		    kuid_has_mapping(fc->user_ns, cred->uid) &&
> > > +		    kgid_has_mapping(fc->user_ns, cred->egid) &&
> > > +		    kgid_has_mapping(fc->user_ns, cred->sgid) &&
> > > +		    kgid_has_mapping(fc->user_ns, cred->gid))
> > 
> > Should fsuid be checked here?
> 
> The point of restricting access here is to prevent a DoS type of attack
> on a more privileged context by making a filesystem operation block
> indefinitely. Coming from that perspective I was thinking that these
> checks ought to be sufficient, but I could be wrong.

I supose it would, as euid 0 would already not be mapped.

> > > +			return 1;
> > > +
> > > +		return 0;
> > > +	}
> > >  
> > > -	cred = current_cred();
> > >  	if (uid_eq(cred->euid, fc->user_id) &&
> > >  	    uid_eq(cred->suid, fc->user_id) &&
> > >  	    uid_eq(cred->uid,  fc->user_id) &&
> > > @@ -1556,17 +1564,25 @@ static bool update_mtime(unsigned ivalid, bool trust_local_mtime)
> > >  	return true;
> > >  }
> > >  
> > > -static void iattr_to_fattr(struct iattr *iattr, struct fuse_setattr_in *arg,
> > > -			   bool trust_local_cmtime)
> > > +static int iattr_to_fattr(struct fuse_conn *fc, struct iattr *iattr,
> > > +			   struct fuse_setattr_in *arg, bool trust_local_cmtime)
> > >  {
> > >  	unsigned ivalid = iattr->ia_valid;
> > >  
> > >  	if (ivalid & ATTR_MODE)
> > >  		arg->valid |= FATTR_MODE,   arg->mode = iattr->ia_mode;
> > > -	if (ivalid & ATTR_UID)
> > > -		arg->valid |= FATTR_UID,    arg->uid = from_kuid(&init_user_ns, iattr->ia_uid);
> > > -	if (ivalid & ATTR_GID)
> > > -		arg->valid |= FATTR_GID,    arg->gid = from_kgid(&init_user_ns, iattr->ia_gid);
> > > +	if (ivalid & ATTR_UID) {
> > > +		arg->uid = from_kuid(fc->user_ns, iattr->ia_uid);
> > > +		if (arg->uid == (uid_t)-1)
> > 
> > Any reason not to use uid_valid() here (and gid_valid() below)?
> 
> Yes. arg->uid is a uid_t and not a kuid_t, so it wouldn't be valid to
> pass that to uid_valid(). And from_kuid() can return -1 for values other
> than INVALID_UID.

D'oh.  Right.