CVE-2024-26737: bpf: Fix racing between bpf_timer_cancel_and_free and bpf_timer_cancel

In the Linux kernel, the following vulnerability has been resolved:

bpf: Fix racing between bpftimercancelandfree and bpftimercancel

The following race is possible between bpftimercancelandfree and bpftimercancel. It will lead a UAF on the timer->timer.

bpftimercancel(); spinlock(); t = timer->time; spinunlock();

bpftimercancelandfree(); spinlock(); t = timer->timer; timer->timer = NULL; spinunlock(); hrtimercancel(&t->timer); kfree(t);

/ UAF on t / hrtimercancel(&t->timer);

In bpftimercancelandfree, this patch frees the timer->timer after a rcu grace period. This requires a rcuhead addition to the "struct bpfhrtimer". Another kfree(t) happens in bpftimerinit, this does not need a kfreercu because it is still under the spinlock and timer->timer has not been visible by others yet.

In bpftimercancel, rcureadlock() is added because this helper can be used in a non rcu critical section context (e.g. from a sleepable bpf prog). Other timer->timer usages in helpers.c have been audited, bpftimercancel() is the only place where timer->timer is used outside of the spinlock.

Another solution considered is to mark a t->flag in bpftimercancel and clear it after hrtimercancel() is done. In bpftimercancelandfree, it busy waits for the flag to be cleared before kfree(t). This patch goes with a straight forward solution and frees timer->timer after a rcu grace period.