drm/i915/selftest: Fix engine timestamp and ktime disparity

While reading the engine timestamps there can be uncontrollable
concurrent mmio access via other i915 child drivers and by GuC,
which is not truly atomic context as expected by this selftest,
which may cause mmio latency to read the engine timestamps,
Account such latency to calculate time to read engine timestamp
such that selftest can validate the timestamp and ktime pair.

Cc: Chris Wilson <chris.p.wilson@intel.com>
Signed-off-by: Anshuman Gupta <anshuman.gupta@intel.com>
Reviewed-by: Badal Nilawar <badal.nilawar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230223100503.3323627-2-anshuman.gupta@intel.com

drivers/gpu/drm/i915/gt/selftest_gt_pm.c

@@ -63,8 +63,8 @@ static void measure_clocks(struct intel_engine_cs *engine,
 
 		udelay(1000);
 
-		dt[i] = ktime_sub(ktime_get(), dt[i]);
 		cycles[i] += read_timestamp(engine);
+		dt[i] = ktime_sub(ktime_get(), dt[i]);
 		local_irq_enable();
 	}
 

drivers/gpu/drm/i915/gt/selftest_rps.c

@@ -299,13 +299,13 @@ int live_rps_clock_interval(void *arg)
 			for (i = 0; i < 5; i++) {
 				preempt_disable();
 
-				dt_[i] = ktime_get();
 				cycles_[i] = -intel_uncore_read_fw(gt->uncore, GEN6_RP_CUR_UP_EI);
+				dt_[i] = ktime_get();
 
 				udelay(1000);
 
-				dt_[i] = ktime_sub(ktime_get(), dt_[i]);
 				cycles_[i] += intel_uncore_read_fw(gt->uncore, GEN6_RP_CUR_UP_EI);
+				dt_[i] = ktime_sub(ktime_get(), dt_[i]);
 
 				preempt_enable();
 			}