Re: [PATCH 21/28] security: Introduce inode_post_remove_acl hook

[Roberto Sassu]

On Mon, 2023-03-06 at 11:16 -0500, Stefan Berger wrote:
> 
> On 3/6/23 10:34, Roberto Sassu wrote:
> > On Mon, 2023-03-06 at 10:22 -0500, Stefan Berger wrote:
> > > On 3/3/23 13:18, Roberto Sassu wrote:
> > > > From: Roberto Sassu <roberto.sassu@xxxxxxxxxx>
> > > > 
> > > > In preparation for moving IMA and EVM to the LSM infrastructure, introduce
> > > > the inode_post_remove_acl hook.
> > > > 
> > > > Signed-off-by: Roberto Sassu <roberto.sassu@xxxxxxxxxx>
> > > > ---
> > > >    
> > > > +/**
> > > > + * security_inode_post_remove_acl() - Update inode sec after remove_acl op
> > > > + * @idmap: idmap of the mount
> > > > + * @dentry: file
> > > > + * @acl_name: acl name
> > > > + *
> > > > + * Update inode security field after successful remove_acl operation on @dentry
> > > > + * in @idmap. The posix acls are identified by @acl_name.
> > > > + */
> > > > +void security_inode_post_remove_acl(struct mnt_idmap *idmap,
> > > > +				    struct dentry *dentry, const char *acl_name)
> > > > +{
> > > > +	if (unlikely(IS_PRIVATE(d_backing_inode(dentry))))
> > > > +		return;
> > > 
> > > Was that a mistake before that EVM and IMA functions did not filtered out private inodes?
> > 
> > Looks like that. At least for hooks that are not called from
> > security.c.
> 
> It seems like that all security_* functions are filtering on private inodes. Anonymous inodes have them and some filesystem set the S_PRIVATE flag. So it may not make a difference fro IMA and EVM then.

Currently, what it would happen is that the HMAC would be updated
without check, since the check function is usually in security.c
(skipped) and the post elsewhere.

With this patch set, also the post function would not be executed for
private inodes. Maybe, it is worth mentioning it in the next version.

Thanks

Roberto