futex: Don't include process MM in futex key on no-MMU

On no-MMU, all futexes are treated as private because there is no need
to map a virtual address to physical to match the futex across
processes. This doesn't quite work though, because private futexes
include the current process's mm_struct as part of their key. This makes
it impossible for one process to wake up a shared futex being waited on
in another process.

Fix this bug by excluding the mm_struct from the key. With
a single address space, the futex address is already a unique key.

Fixes: 784bdf3bb6 ("futex: Assume all mappings are private on !MMU systems")
Signed-off-by: Ben Wolsieffer <ben.wolsieffer@hefring.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: André Almeida <andrealmeid@igalia.com>
Link: https://lore.kernel.org/r/20231019204548.1236437-2-ben.wolsieffer@hefring.com

kernel/futex/core.c

@@ -252,7 +252,17 @@ int get_futex_key(u32 __user *uaddr, unsigned int flags, union futex_key *key,
 	 *        but access_ok() should be faster than find_vma()
 	 */
 	if (!fshared) {
-		key->private.mm = mm;
+		/*
+		 * On no-MMU, shared futexes are treated as private, therefore
+		 * we must not include the current process in the key. Since
+		 * there is only one address space, the address is a unique key
+		 * on its own.
+		 */
+		if (IS_ENABLED(CONFIG_MMU))
+			key->private.mm = mm;
+		else
+			key->private.mm = NULL;
+
 		key->private.address = address;
 		return 0;
 	}