Zephyr RTOS: timer race lets a local thread corrupt kernel memory

The kernel helper z_waitq_head() (kernel/include/wait_q.h) reads the head of a thread wait queue, but on SMP systems it was called from several places without holding the scheduler spinlock _sched_spinlock that serializes wait-queue mutation. The consequential call site is z_timer_expiration_handler() in kernel/timer.c, which held only the timer subsystem lock timer_lock while reading timer->wait_q and then unpending, setting the return value of, and readying the thread it obtained. halt_thread() in kernel/sched.c removes an aborted thread from its wait queue while holding only _sched_spinlock, so the two paths are not mutually exclusive.

When the last waiter is removed on another CPU inside that window, sys_dlist_peek_head() can return the list sentinel — a pointer into the k_timer object itself — typed as a struct k_thread * (with CONFIG_WAITQ_SCALABLE, rb_get_min() can return an arbitrary node of a tree being restructured), or it can return a pointer to a thread that has already been unpended and, for dynamically allocated threads, freed. The handler then calls unpend_thread_no_timeout(), which performs a sys_dlist_remove() through pointers read from that object, and arch_thread_return_value_set(), which writes unconditionally through the pointer.

The result is out-of-bounds and use-after-free writes in supervisor context, plus scheduler state corruption; a milder outcome is a thread that has re-pended elsewhere being woken with a success return value from an unrelated blocking call. All of the primitives needed to drive the interleaving — k_timer_start(), k_timer_status_sync(), k_thread_create() and k_thread_abort() — are system calls, so with CONFIG_USERSPACE enabled an unprivileged user-mode thread can attempt the race against kernel-owned data.

Exploitation requires a multi-CPU build (CONFIG_MP_MAX_NUM_CPUS greater than one; the affected region is a no-op on uniprocessor targets), local code execution, and winning a narrow timing window, so the practical outcome is most often a kernel crash rather than controlled corruption. The fix splits the helper into z_waitq_head_locked() for callers that already hold the scheduler spinlock and a z_waitq_head() that acquires it, and brings the timer expiration handler's head read, unpend and ready under a single LOCK_SCHED_SPINLOCK region.