Re: Possible security hole? [was: verify_area(...) possible problem]

Jamie Lokier (lkd@tantalophile.demon.co.uk)
Mon, 19 Apr 1999 20:09:12 +0200

Ralf Baechle wrote:
> This is a known problem, it is pretty much similar to what happened
> under 2.0 for all all architectures, not just the i386 and one of
> reasons for the new user space access stuff introduced from 2.1.4 on.
> Now that only the 386 is still affected just nobody bothered because
> on the 386 there are other funnies left which may make running a
> system used by possibly hostile users a bad idea.

I haven't seen any security warnings against using a 386 for multiple
users. Should there be one?

Perhaps it would be possible to construct a patch which does this on a 386:

- Implement read-write semaphores.
- access_ok() grabs the mm semaphore, read-only, and sets a flag.
- On exit from system call, if the flag is set it is cleared,
and the mm read-only semaphore is released.
- The page fault handler uses the read-only semaphore, otherwise
deadlock.

-- Jamie

-
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/