diff --git a/drivers/gpu/host1x/fence.c b/drivers/gpu/host1x/fence.c
index ac6e07c1..4374e249 100644
--- a/drivers/gpu/host1x/fence.c
+++ b/drivers/gpu/host1x/fence.c
@@ -72,7 +72,7 @@ const struct dma_fence_ops host1x_syncpt_fence_ops = {
 	.enable_signaling = host1x_syncpt_fence_enable_signaling,
 };
 
-void host1x_fence_signal(struct host1x_syncpt_fence *f)
+void host1x_fence_signal(struct host1x_syncpt_fence *f, ktime_t ts)
 {
 	if (atomic_xchg(&f->signaling, 1)) {
 		/*
@@ -91,7 +91,7 @@ void host1x_fence_signal(struct host1x_syncpt_fence *f)
 		dma_fence_put(&f->base);
 	}
 
-	dma_fence_signal_locked(&f->base);
+	dma_fence_signal_timestamp_locked(&f->base, ts);
 	dma_fence_put(&f->base);
 }
 
diff --git a/drivers/gpu/host1x/fence.h b/drivers/gpu/host1x/fence.h
index f3c644c7..0a2e6b3f 100644
--- a/drivers/gpu/host1x/fence.h
+++ b/drivers/gpu/host1x/fence.h
@@ -25,6 +25,6 @@ struct host1x_fence_list {
 	struct list_head list;
 };
 
-void host1x_fence_signal(struct host1x_syncpt_fence *fence);
+void host1x_fence_signal(struct host1x_syncpt_fence *fence, ktime_t ts);
 
 #endif
diff --git a/drivers/gpu/host1x/hw/intr_hw.c b/drivers/gpu/host1x/hw/intr_hw.c
index b915ef7d..d6cdc489 100644
--- a/drivers/gpu/host1x/hw/intr_hw.c
+++ b/drivers/gpu/host1x/hw/intr_hw.c
@@ -9,6 +9,7 @@
 #include <linux/interrupt.h>
 #include <linux/irq.h>
 #include <linux/io.h>
+#include <linux/timekeeping.h>
 
 #include "../intr.h"
 #include "../dev.h"
@@ -18,6 +19,9 @@ static irqreturn_t syncpt_thresh_isr(int irq, void *dev_id)
 	struct host1x *host = dev_id;
 	unsigned long reg;
 	unsigned int i, id;
+	ktime_t ts;
+
+	ts = ktime_get();
 
 	for (i = 0; i < DIV_ROUND_UP(host->info->nb_pts, 32); i++) {
 		reg = host1x_sync_readl(host,
@@ -29,7 +33,7 @@ static irqreturn_t syncpt_thresh_isr(int irq, void *dev_id)
 			HOST1X_SYNC_SYNCPT_THRESH_CPU0_INT_STATUS(i));
 
 		for_each_set_bit(id, &reg, 32)
-			host1x_intr_handle_interrupt(host, i * 32 + id);
+			host1x_intr_handle_interrupt(host, i * 32 + id, ts);
 	}
 
 	return IRQ_HANDLED;
diff --git a/drivers/gpu/host1x/include/linux/host1x-next.h b/drivers/gpu/host1x/include/linux/host1x-next.h
index 6fc0bb28..b9ab6475 100644
--- a/drivers/gpu/host1x/include/linux/host1x-next.h
+++ b/drivers/gpu/host1x/include/linux/host1x-next.h
@@ -10,6 +10,7 @@
 #include <linux/dma-direction.h>
 #include <linux/dma-fence.h>
 #include <linux/spinlock.h>
+#include <linux/timekeeping.h>
 #include <linux/types.h>
 
 enum host1x_class {
@@ -211,6 +212,8 @@ u32 host1x_syncpt_read_max(struct host1x_syncpt *sp);
 u32 host1x_syncpt_read(struct host1x_syncpt *sp);
 int host1x_syncpt_incr(struct host1x_syncpt *sp);
 u32 host1x_syncpt_incr_max(struct host1x_syncpt *sp, u32 incrs);
+int host1x_syncpt_wait_ts(struct host1x_syncpt *sp, u32 thresh, long timeout,
+			  u32 *value, ktime_t *ts);
 int host1x_syncpt_wait(struct host1x_syncpt *sp, u32 thresh, long timeout,
 		       u32 *value);
 struct host1x_syncpt *host1x_syncpt_request(struct host1x_client *client,
diff --git a/drivers/gpu/host1x/intr.c b/drivers/gpu/host1x/intr.c
index 995bfa98..6f4b1820 100644
--- a/drivers/gpu/host1x/intr.c
+++ b/drivers/gpu/host1x/intr.c
@@ -72,7 +72,7 @@ bool host1x_intr_remove_fence(struct host1x *host, struct host1x_syncpt_fence *f
 	return true;
 }
 
-void host1x_intr_handle_interrupt(struct host1x *host, unsigned int id)
+void host1x_intr_handle_interrupt(struct host1x *host, unsigned int id, ktime_t ts)
 {
 	struct host1x_syncpt *sp = &host->syncpt[id];
 	struct host1x_syncpt_fence *fence, *tmp;
@@ -89,7 +89,7 @@ void host1x_intr_handle_interrupt(struct host1x *host, unsigned int id)
 		}
 
 		list_del_init(&fence->list);
-		host1x_fence_signal(fence);
+		host1x_fence_signal(fence, ts);
 	}
 
 	/* Re-enable interrupt if necessary */
diff --git a/drivers/gpu/host1x/intr.h b/drivers/gpu/host1x/intr.h
index 3b5610b5..e4f307ab 100644
--- a/drivers/gpu/host1x/intr.h
+++ b/drivers/gpu/host1x/intr.h
@@ -8,6 +8,8 @@
 #ifndef __HOST1X_INTR_H
 #define __HOST1X_INTR_H
 
+#include <linux/timekeeping.h>
+
 struct host1x;
 struct host1x_syncpt_fence;
 
@@ -23,7 +25,7 @@ void host1x_intr_start(struct host1x *host);
 /* Disable host1x sync point interrupt */
 void host1x_intr_stop(struct host1x *host);
 
-void host1x_intr_handle_interrupt(struct host1x *host, unsigned int id);
+void host1x_intr_handle_interrupt(struct host1x *host, unsigned int id, ktime_t ts);
 
 void host1x_intr_add_fence_locked(struct host1x *host, struct host1x_syncpt_fence *fence);
 
diff --git a/drivers/gpu/host1x/syncpt.c b/drivers/gpu/host1x/syncpt.c
index 2d200776..098ff002 100644
--- a/drivers/gpu/host1x/syncpt.c
+++ b/drivers/gpu/host1x/syncpt.c
@@ -9,6 +9,7 @@
 #include <linux/device.h>
 #include <linux/dma-fence.h>
 #include <linux/slab.h>
+#include <linux/timekeeping.h>
 
 #include <trace/events/host1x.h>
 
@@ -211,14 +212,15 @@ int host1x_syncpt_incr(struct host1x_syncpt *sp)
 EXPORT_SYMBOL(host1x_syncpt_incr);
 
 /**
- * host1x_syncpt_wait() - wait for a syncpoint to reach a given value
+ * host1x_syncpt_wait_ts() - wait for a syncpoint to reach a given value
  * @sp: host1x syncpoint
  * @thresh: threshold
  * @timeout: maximum time to wait for the syncpoint to reach the given value
  * @value: return location for the syncpoint value
+ * @ts: return location for completion timestamp
  */
-int host1x_syncpt_wait(struct host1x_syncpt *sp, u32 thresh, long timeout,
-		       u32 *value)
+int host1x_syncpt_wait_ts(struct host1x_syncpt *sp, u32 thresh, long timeout, u32 *value,
+			  ktime_t *ts)
 {
 	struct dma_fence *fence;
 	long wait_err;
@@ -227,6 +229,8 @@ int host1x_syncpt_wait(struct host1x_syncpt *sp, u32 thresh, long timeout,
 
 	if (value)
 		*value = host1x_syncpt_load(sp);
+	if (ts)
+		*ts = ktime_get();
 
 	if (host1x_syncpt_is_expired(sp, thresh))
 		return 0;
@@ -243,10 +247,13 @@ int host1x_syncpt_wait(struct host1x_syncpt *sp, u32 thresh, long timeout,
 	wait_err = dma_fence_wait_timeout(fence, true, timeout);
 	if (wait_err == 0)
 		host1x_fence_cancel(fence);
-	dma_fence_put(fence);
 
 	if (value)
 		*value = host1x_syncpt_load(sp);
+	if (ts)
+		*ts = fence->timestamp;
+
+	dma_fence_put(fence);
 
 	if (wait_err == 0)
 		return -EAGAIN;
@@ -255,6 +262,19 @@ int host1x_syncpt_wait(struct host1x_syncpt *sp, u32 thresh, long timeout,
 	else
 		return 0;
 }
+EXPORT_SYMBOL(host1x_syncpt_wait_ts);
+
+/**
+ * host1x_syncpt_wait() - wait for a syncpoint to reach a given value
+ * @sp: host1x syncpoint
+ * @thresh: threshold
+ * @timeout: maximum time to wait for the syncpoint to reach the given value
+ * @value: return location for the syncpoint value
+ */
+int host1x_syncpt_wait(struct host1x_syncpt *sp, u32 thresh, long timeout, u32 *value)
+{
+	return host1x_syncpt_wait_ts(sp, thresh, timeout, value, NULL);
+}
 EXPORT_SYMBOL(host1x_syncpt_wait);
 
 /*