Re: [PATCH] Capabilities, this time in elf section

Albert D. Cahalan (acahalan@cs.uml.edu)
Fri, 9 Apr 1999 23:39:17 -0400 (EDT)


Jonathan Walther writes:
> On Fri, 9 Apr 1999, Daniel Taylor wrote:

>> Exactly.
>> Non-SUID binary, & capabilities with user,
>> SUID binary, own capabilities.
>
> I disagree. if you suid to some other user, then they become the
> effective user, so the effective capabilities should become
> and & of THAT users capabilities and the binaries capabilities,
> and not those of the invoker.

Any UID change caused by the exec must be independent of any
capability change caused by the exec.

The old way: capabilities are associated with a user (UID 0)
The new way: capabilities are orthogonal to identity

I may want a setuid-lpr program to operate with capabilities that
are different from the printing daemon. I don't want to be forced
to make the daemon have a superset or subset of the setuid program.
The daemon might get CAP_A and CAP_B, while the other program
gets CAP_B and CAP_C.

Anyway, the kernel does not look up user database entries to determine
what capabilities a user might have.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/