On Fri, Feb 05, 2021 at 10:00:21AM -0800, Yu, Yu-cheng wrote:

On 2/5/2021 5:59 AM, Borislav Petkov wrote:

quoted

On Wed, Feb 03, 2021 at 02:55:28PM -0800, Yu-cheng Yu wrote:

quoted

+DEFINE_IDTENTRY_ERRORCODE(exc_control_protection)
+{
+	static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL,
+				      DEFAULT_RATELIMIT_BURST);
+	struct task_struct *tsk;
+
+	if (!user_mode(regs)) {
+		pr_emerg("PANIC: unexpected kernel control protection fault\n");
+		die("kernel control protection fault", regs, error_code);
+		panic("Machine halted.");
+	}
+
+	cond_local_irq_enable(regs);
+
+	if (!boot_cpu_has(X86_FEATURE_CET))
+		WARN_ONCE(1, "Control protection fault with CET support disabled\n");
+
+	tsk = current;
+	tsk->thread.error_code = error_code;
+	tsk->thread.trap_nr = X86_TRAP_CP;
+
+	if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
+	    __ratelimit(&rs)) {

I can't find it written down anywhere why the ratelimiting is needed at
all?

The ratelimit here is only for #CP, and its rate is not counted together
with other types of faults.  If a task gets here, it will exit.  The only
condition the ratelimit will trigger is when multiple tasks hit #CP at once,
which is unlikely.  Are you suggesting that we do not need the ratelimit
here?

Since this is a potentially unprivileged-userspace-triggerable
condition, I tend to prefer having a ratelimit. I don't feel _strongly_
about it, but I find it better to be defensive against log spamming
(whether malicious or accidental).

-- 
Kees Cook