[PATCH v2 16/30] acl: add vfs_get_acl()

[Paul Moore]

On Wed, Sep 28, 2022 at 11:12 AM Christian Brauner <brauner at kernel.org> wrote:
> On Wed, Sep 28, 2022 at 10:58:33AM -0400, Paul Moore wrote:
> > On Wed, Sep 28, 2022 at 3:40 AM Christian Brauner <brauner at kernel.org> wrote:
> > >
> > > On Tue, Sep 27, 2022 at 06:55:25PM -0400, Paul Moore wrote:
> > > > On Mon, Sep 26, 2022 at 11:24 AM Christian Brauner <brauner at kernel.org> wrote:
> > > > >
> > > > > In previous patches we implemented get and set inode operations for all
> > > > > non-stacking filesystems that support posix acls but didn't yet
> > > > > implement get and/or set acl inode operations. This specifically
> > > > > affected cifs and 9p.
> > > > >
> > > > > Now we can build a posix acl api based solely on get and set inode
> > > > > operations. We add a new vfs_get_acl() api that can be used to get posix
> > > > > acls. This finally removes all type unsafety and type conversion issues
> > > > > explained in detail in [1] that we aim to get rid of.
> > > > >
> > > > > After we finished building the vfs api we can switch stacking
> > > > > filesystems to rely on the new posix api and then finally switch the
> > > > > xattr system calls themselves to rely on the posix acl api.
> > > > >
> > > > > Link: https://lore.kernel.org/all/20220801145520.1532837-1-brauner@kernel.org [1]
> > > > > Signed-off-by: Christian Brauner (Microsoft) <brauner at kernel.org>
> > > > > ---
> > > > >
> > > > > Notes:
> > > > >     /* v2 */
> > > > >     unchanged
> > > > >
> > > > >  fs/posix_acl.c                  | 131 ++++++++++++++++++++++++++++++--
> > > > >  include/linux/posix_acl.h       |   9 +++
> > > > >  include/linux/posix_acl_xattr.h |  10 +++
> > > > >  3 files changed, 142 insertions(+), 8 deletions(-)
> > > >
> > > > ...
> > > >
> > > > > diff --git a/fs/posix_acl.c b/fs/posix_acl.c
> > > > > index ef0908a4bc46..18873be583a9 100644
> > > > > --- a/fs/posix_acl.c
> > > > > +++ b/fs/posix_acl.c
> > > > > @@ -1369,3 +1439,48 @@ int vfs_set_acl(struct user_namespace *mnt_userns, struct dentry *dentry,
> > > > >         return error;
> > > > >  }
> > > > >  EXPORT_SYMBOL(vfs_set_acl);
> > > > > +
> > > > > +/**
> > > > > + * vfs_get_acl - get posix acls
> > > > > + * @mnt_userns: user namespace of the mount
> > > > > + * @dentry: the dentry based on which to retrieve the posix acls
> > > > > + * @acl_name: the name of the posix acl
> > > > > + *
> > > > > + * This function retrieves @kacl from the filesystem. The caller must all
> > > > > + * posix_acl_release() on @kacl.
> > > > > + *
> > > > > + * Return: On success POSIX ACLs in VFS format, on error negative errno.
> > > > > + */
> > > > > +struct posix_acl *vfs_get_acl(struct user_namespace *mnt_userns,
> > > > > +                             struct dentry *dentry, const char *acl_name)
> > > > > +{
> > > > > +       struct inode *inode = d_inode(dentry);
> > > > > +       struct posix_acl *acl;
> > > > > +       int acl_type, error;
> > > > > +
> > > > > +       acl_type = posix_acl_type(acl_name);
> > > > > +       if (acl_type < 0)
> > > > > +               return ERR_PTR(-EINVAL);
> > > > > +
> > > > > +       /*
> > > > > +        * The VFS has no restrictions on reading POSIX ACLs so calling
> > > > > +        * something like xattr_permission() isn't needed. Only LSMs get a say.
> > > > > +        */
> > > > > +       error = security_inode_getxattr(dentry, acl_name);
> > > > > +       if (error)
> > > > > +               return ERR_PTR(error);
> > > >
> > > > I understand the desire to reuse the security_inode_getxattr() hook
> > > > here, it makes perfect sense, but given that this patchset introduces
> > > > an ACL specific setter hook I think it makes sense to have a matching
> > > > getter hook.  It's arguably a little silly given the current crop of
> > > > LSMs and their approach to ACLs, but if we are going to differentiate
> > > > on the write side I think we might as well be consistent and
> > > > differentiate on the read side as well.
> > >
> > > Sure, I don't mind doing that. I'll add the infrastructure and then the
> > > individual LSMs can add their own hooks.
> >
> > Adding the ACL hook infrastructure, including the call in
> > vfs_get_acl(), without the LSM implementations would result in an
> > access control regression for both SELinux and Smack.  Similar issues
> > with the removexattr hook, although that looks to have IMA/EVM calls
> > too (which may be noops in the case of an ACL, I haven't checked).
> >
> > The good news is that the individual LSM implementations should be
> > trivial, and if you wanted to just have the new ACL hook
> > implementations call into the existing xattr implementations inside
> > each LSM I think that would be okay to start.
>
> Yeah, I realized right after I sent the mail that I'd need to implement
> them. I think I came up with something fairly minimal for all lsms and
> the integrity modules. I folded the trivial patches for adding get, set,
> and remove hooks for the individual modules together to not needlessly
> inflate the security portion of the patchset.

I'll keep an eye out for v3.

-- 
paul-moore.com