Hello Apple Developer Forums,
I am looking for a documented, supported user-space method on macOS to verify that an already-open file or directory has no extended ACL, while distinguishing that state from an unsupported filesystem or an ACL retrieval failure.
Environment and observations:
- macOS 15.7.7 on an Intel Mac; Command Line Tools SDK 15.5.
- In an isolated synthetic test, acl_get_fd_np(fd, ACL_TYPE_EXTENDED) returned NULL with errno == ENOENT for a newly created regular file and directory with no ACL added by the test.
- The same API returned an ACL containing one ACE after an ACL was added to a separate synthetic file.
- An invalid file descriptor returned NULL with errno == EBADF.
- These are observations on this OS and test filesystem only. I have not interpreted ENOENT as proof of ACL absence, and I have not tested fgetattrlist.
I could not establish from the available documentation whether NULL/ENOENT is a guaranteed indication of ACL absence. I also could not establish whether ATTR_CMN_RETURNED_ATTRS and KAUTH_FILESEC_NOACL provide a supported, stable user-space contract when using fgetattrlist.
Questions:
-
When acl_get_fd_np(fd, ACL_TYPE_EXTENDED) returns NULL with errno == ENOENT for an existing object, does that reliably mean the object has no extended ACL? Could it instead indicate unsupported or unavailable ACL data?
-
When requesting ATTR_CMN_EXTENDED_SECURITY through fgetattrlist, how can a caller confirm that the attribute was actually returned? Is ATTR_CMN_RETURNED_ATTRS supported for fgetattrlist? How should omitted or unsupported attributes be distinguished from ACL absence?
-
Is KAUTH_FILESEC_NOACL a supported, stable public user-space contract for parsing kauth_filesec returned by fgetattrlist? Under which documented conditions, if any, does it establish ACL absence?
-
Does VOL_CAP_INT_EXTENDED_SECURITY establish only volume capability, or can it help establish per-object ACL absence? What documented result distinguishes an ACL-free object from a filesystem that does not support extended security?
-
What guarantees apply when retrieving ACL information through an already-open descriptor if the pathname is renamed or replaced? Are ACL retrieval and fstat an atomic snapshot, particularly if another process changes the ACL?
-
What supported API and decision procedure would Apple recommend for positively confirming no extended ACL on an already-open file or directory, while treating unsupported, omitted, malformed, and error results as distinct fail-closed outcomes?
If this behavior is not covered by a supported public API contract, a clarification of that limitation and a reference to the closest documented API would be very helpful.
Thank you.