diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c
index 7f33767..4357f8f 100644
--- a/drivers/gpu/drm/radeon/evergreen.c
+++ b/drivers/gpu/drm/radeon/evergreen.c
@@ -4377,6 +4377,10 @@ int evergreen_rlc_resume(struct radeon_device *rdev)
 
 u32 evergreen_get_vblank_counter(struct radeon_device *rdev, int crtc)
 {
+	u32 adl = RREG32(0x6ecc + crtc_offsets[crtc]);
+	u32 fvr = RREG32(CRTC_STATUS_FRAME_COUNT + 4 + crtc_offsets[crtc]);
+	DRM_DEBUG_VBL("crtc %d: adl0: %d adl8: %d adls: %d Fd %d Vd %d\n", crtc, adl & (1 << 0), adl & (1 << 8), (adl >> 16) & 0xf, fvr & 0xffff, fvr >> 16);
+
 	if (crtc >= rdev->num_crtc)
 		return 0;
 	else
diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index a8d9927..76b3449 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/drivers/gpu/drm/radeon/radeon_display.c
@@ -1910,11 +1910,22 @@ int radeon_get_crtc_scanoutpos(struct drm_device *dev, unsigned int pipe,
 		vbl_start = mode->crtc_vdisplay;
 		vbl_end = 0;
 	}
-
+	DRM_DEBUG_VBL("crtc %d: vpos %d : vs %d , ve %d, vs2 %d, vt %d\n", pipe, *vpos, vbl_start, vbl_end, mode->crtc_vdisplay, mode->crtc_vtotal);
 	/* Test scanout position against vblank region. */
 	if ((*vpos < vbl_start) && (*vpos >= vbl_end))
 		in_vbl = false;
 
+	/* In vblank? */
+	if (in_vbl)
+	    ret |= DRM_SCANOUTPOS_IN_VBLANK;
+
+	/* Called from driver internal vblank counter query code? */
+	if (flags & 0x80000000) {
+		/* Caller wants distance from vbl_start */
+		*vpos -= vbl_start;
+		return ret;
+	}
+
 	/* Check if inside vblank area and apply corrective offsets:
 	 * vpos will then be >=0 in video scanout area, but negative
 	 * within vblank area, counting down the number of lines until
@@ -1930,10 +1941,6 @@ int radeon_get_crtc_scanoutpos(struct drm_device *dev, unsigned int pipe,
 	/* Correct for shifted end of vbl at vbl_end. */
 	*vpos = *vpos - vbl_end;
 
-	/* In vblank? */
-	if (in_vbl)
-		ret |= DRM_SCANOUTPOS_IN_VBLANK;
-
 	/* Is vpos outside nominal vblank area, but less than
 	 * 1/100 of a frame height away from start of vblank?
 	 * If so, assume this isn't a massively delayed vblank
diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
index 0ec6fcc..885f934 100644
--- a/drivers/gpu/drm/radeon/radeon_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_kms.c
@@ -755,6 +755,8 @@ void radeon_driver_preclose_kms(struct drm_device *dev,
  */
 u32 radeon_get_vblank_counter_kms(struct drm_device *dev, int crtc)
 {
+	int vpos, hpos, stat, bump_line;
+	u32 count;
 	struct radeon_device *rdev = dev->dev_private;
 
 	if (crtc < 0 || crtc >= rdev->num_crtc) {
@@ -762,7 +764,57 @@ u32 radeon_get_vblank_counter_kms(struct drm_device *dev, int crtc)
 		return -EINVAL;
 	}
 
-	return radeon_get_vblank_counter(rdev, crtc);
+	/* If called from hw vblank irq then we need hw quirk treatment.
+	 * The hw fires its vblank irq typically 2-3 scanlines before
+	 * the vblank, triggering sampling of our hw counter by DRM
+	 * core before it increments at start of vblank, which is bad
+	 * for vblank counting and timestamping, causing frequent
+	 * off-by-one errors. Try to cook the hw count here to make it
+	 * appear to the caller as if it was sampled after it incremented.
+	 */
+	if (rdev->mode_info.crtcs[crtc]) {
+		do {
+			count = radeon_get_vblank_counter(rdev, crtc);
+			/* Ask function to return vpos as distance to start of
+			 * vblank, instead of regular vertical scanout pos.
+			 */
+			stat = radeon_get_crtc_scanoutpos(
+				dev, crtc,
+				0x80000000, &vpos, &hpos, NULL, NULL,
+				&rdev->mode_info.crtcs[crtc]->base.hwmode);
+		} while (count != radeon_get_vblank_counter(rdev, crtc));
+
+		if (((stat & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE)) !=
+		    (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE))) {
+			DRM_DEBUG_VBL("Query failed! stat %d\n", stat);
+		}
+		else {
+			/* Bump counter if we are at >= leading edge of vblank,
+			 * but before wraparound to zero where the hw counter
+			 * increments. Bump counter already 10 scanlines before
+			 * vblank if we are called from vblank irq handler, as
+			 * that one fires a few scanlines before vblank but
+			 * should get a count as if it was called inside vblank.
+			 */
+			bump_line = in_irq() ? -10 : 0;
+			if (vpos >= bump_line) {
+				count++;
+				DRM_DEBUG_VBL("crtc %d: vpos %d >= bump_line %d -> Bumped.\n",
+					      crtc, vpos, bump_line);
+			}
+			else {
+				DRM_DEBUG_VBL("crtc %d: vpos %d\n",
+					      crtc, vpos);
+			}
+		}
+	}
+	else {
+	    /* Fallback to use value as is. */
+	    count = radeon_get_vblank_counter(rdev, crtc);
+	    DRM_DEBUG_VBL("NULL mode info!\n");
+	}
+
+	return count;
 }
 
 /**
diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c
index 97a9048..460ca2f 100644
--- a/drivers/gpu/drm/radeon/rs600.c
+++ b/drivers/gpu/drm/radeon/rs600.c
@@ -835,6 +835,9 @@ int rs600_irq_process(struct radeon_device *rdev)
 
 u32 rs600_get_vblank_counter(struct radeon_device *rdev, int crtc)
 {
+	u32 fvr = RREG32((crtc == 0) ? (R_0060A4_D1CRTC_STATUS_FRAME_COUNT + 4) : (R_0068A4_D2CRTC_STATUS_FRAME_COUNT + 4));
+	DRM_DEBUG_VBL("crtc %d: Fd%d Vd %d\n", crtc, fvr & 0xffff, fvr >> 16);
+
 	if (crtc == 0)
 		return RREG32(R_0060A4_D1CRTC_STATUS_FRAME_COUNT);
 	else