CHROMIUM: Reland "pm: re-program timer HW only when the exit latency …

…is not zero""

This reverts commit e939377, relanding
3a58885. (aka
6a962fb).

The original issue has been fixed in:

2bda7b8 driver: timer: npcx: bypass timer counter reading issue
bdf0500 driver: timer: npcx: fix possible vulnerabilities

(http://crrev.com/c/5041238 and http://crrev.com/c/5855298)

BUG=b:361773754
TEST=Verified on a Rex system, letting it idle with ap off for 4 days, no
crashes observed.

Change-Id: I1aa8cf1744c707d913c85f489d960ca883f36381
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/third_party/zephyr/+/6110606
Commit-Queue: Fabio Baltieri <fabiobaltieri@google.com>
Reviewed-by: Peter Marheine <pmarheine@chromium.org>
Tested-by: Fabio Baltieri <fabiobaltieri@google.com>
Reviewed-by: Jeremy Bettis <jbettis@chromium.org>

subsys/pm/pm.c

@@ -194,7 +194,8 @@ bool pm_system_suspend(int32_t kernel_ticks)
 	}
 #endif
 
-	if (ticks != K_TICKS_FOREVER) {
+	if ((z_cpus_pm_state[id].exit_latency_us != 0) &&
+	    (ticks != K_TICKS_FOREVER)) {
 		/*
 		 * We need to set the timer to interrupt a little bit early to
 		 * accommodate the time required by the CPU to fully wake up.