diff --git a/linuxkm/linuxkm_wc_port.h b/linuxkm/linuxkm_wc_port.h
index 3e6035980..78dbdb84d 100644
--- a/linuxkm/linuxkm_wc_port.h
+++ b/linuxkm/linuxkm_wc_port.h
@@ -360,7 +360,7 @@
     #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)
         /* for signal_pending() */
         #include <linux/sched/signal.h>
-        /* for sched_clock_cpu() */
+        /* for local_clock() */
         #include <linux/sched/clock.h>
     #endif
     #include <linux/random.h>
@@ -545,7 +545,6 @@
 
     #elif defined(WOLFSSL_LINUXKM_USE_SAVE_VECTOR_REGISTERS)
         #error WOLFSSL_LINUXKM_USE_SAVE_VECTOR_REGISTERS is set for an unsupported architecture.
-        #define RESTORE_VECTOR_REGISTERS() WC_RELAX_LONG_LOOP();
     #endif /* WOLFSSL_LINUXKM_USE_SAVE_VECTOR_REGISTERS */
 
     _Pragma("GCC diagnostic pop");
diff --git a/linuxkm/module_hooks.c b/linuxkm/module_hooks.c
index 18a8887a7..e0674868a 100644
--- a/linuxkm/module_hooks.c
+++ b/linuxkm/module_hooks.c
@@ -219,7 +219,7 @@ int wc_linuxkm_check_for_intr_signals(void) {
             if (sigismember(&current->pending.signal, intr_signals[i])) {
 #ifdef WOLFSSL_LINUXKM_VERBOSE_DEBUG
                 pr_err("INFO: wc_linuxkm_check_for_intr_signals returning "
-                       "-EINTR on signal %d\n", intr_signals[i]);
+                       "INTERRUPTED_E on signal %d\n", intr_signals[i]);
 #endif
                 return INTERRUPTED_E;
             }
@@ -229,25 +229,33 @@ int wc_linuxkm_check_for_intr_signals(void) {
 }
 
 void wc_linuxkm_relax_long_loop(void) {
-#if WC_LINUXKM_MAX_NS_WITHOUT_YIELD >= 0
+    #if WC_LINUXKM_MAX_NS_WITHOUT_YIELD >= 0
     if (preempt_count() == 0) {
-#if (WC_LINUXKM_MAX_NS_WITHOUT_YIELD == 0) || !defined(CONFIG_SCHED_INFO)
+        #if (WC_LINUXKM_MAX_NS_WITHOUT_YIELD == 0) || !defined(CONFIG_SCHED_INFO)
         cond_resched();
-#else
+        #else
+        /* note that local_clock() wraps a local_clock_noinstr() in a
+         * preempt_disable_notrace(), which sounds expensive but isn't --
+         * preempt_disable_notrace() is actually just a nonlocking integer
+         * increment of current_thread_info()->preempt.count, protected only by
+         * various compiler optimizer barriers.
+         */
         u64 now = local_clock();
         u64 current_last_arrival = current->sched_info.last_arrival;
         s64 delta = (s64)(now - current_last_arrival);
         if (delta > WC_LINUXKM_MAX_NS_WITHOUT_YIELD) {
             cond_resched();
-            /* note, if nothing else is runnable, cond_resched() is a no-op and
+            /* if nothing else is runnable, cond_resched() is a no-op and
              * doesn't even update .last_arrival.  we could force update by
              * sleeping, but there's no need.  we've been nice enough by just
-             * cond_resched()ing.
+             * cond_resched()ing, and it's actually preferable to call
+             * cond_resched() frequently once computation has looped
+             * continuously for longer than WC_LINUXKM_MAX_NS_WITHOUT_YIELD.
              */
         }
-#endif
+        #endif
     }
-#endif
+    #endif
 }
 
 #if defined(WOLFSSL_LINUXKM_USE_SAVE_VECTOR_REGISTERS) && defined(CONFIG_X86)