DragonFlyBSD

The problem is with the AML Mutex handling function AcpiExReleaseMutex where the
function can fault under the very rare condition when the interpreter has
blocked, the interpreter lock is released, the interpreter is then reentered via
the same thread, and attempts to acquire an AML mutex that was previously acquired.

This is indeed fixed in the acpica-unix-20100806 snapshot.

Attached is the required fix from the acpica-unix-20100806 snapshot;
Please review and commit.

Fix delivered in commit 920c7d82e6a0b17caf88ae20355adbc230efd09b.

Cheers,
Edward.