Attachment 119628 Details for Bug 92826 – [PATCH] drm/i915: Always keep cursor WM at its highest on VLV/CHV

[patch] [PATCH] drm/i915: Always keep cursor WM at its highest on VLV/CHV

0001-drm-i915-Always-keep-cursor-WM-at-its-highest-on-VLV.patch (text/plain), 1.83 KB, created by Ville Syrjala on 2015-11-13 10:54:50 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Ville Syrjala

Created: 2015-11-13 10:54:50 UTC

Size: 1.83 KB

patch

obsolete

>From c2c7feec67ea544599fc147f11aa2ed510e1339d Mon Sep 17 00:00:00 2001
>From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= <ville.syrjala@linux.intel.com>
>Date: Fri, 13 Nov 2015 12:44:43 +0200
>Subject: [PATCH] drm/i915: Always keep cursor WM at its highest on VLV/CHV
>MIME-Version: 1.0
>Content-Type: text/plain; charset=UTF-8
>Content-Transfer-Encoding: 8bit
>
>Cursor updates don't do any vblank waits, and so we risk underruns
>when turning the cursor off due to the watermark update happening
>before the cursor has actually stopped scanning out.
>
>As an interim hack (until we get proper two stage watermark
>progarmming), just keep the cursor WM at its highest level
>regardless of the state of the cursor plane. This is quite OK
>actually since the cursor has a dedicated FIFO, and so we
>don't have to worry about having the watermark not fit in the
>FIFO.
>
>Signed-off-by: Ville SyrjÃ¤lÃ¤ <ville.syrjala@linux.intel.com>
>---
> drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++--
> 1 file changed, 10 insertions(+), 2 deletions(-)
>
>diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
>index dca7db2..2ea0b91 100644
>--- a/drivers/gpu/drm/i915/intel_pm.c
>+++ b/drivers/gpu/drm/i915/intel_pm.c
>@@ -936,8 +936,16 @@ static uint16_t vlv_compute_wm_level(struct intel_plane *plane,
> 	if (dev_priv->wm.pri_latency[level] == 0)
> 		return USHRT_MAX;
> 
>-	if (!state->visible)
>-		return 0;
>+	if (!state->visible) {
>+		/*
>+		 * FIXME need proper two stage watermark update,
>+		 * otherwise risk underruns when turning off the
>+		 * cursor. For other planes we supposedly have
>+		 * vblank waits around the plane disable to avoid
>+		 * this, but that too is sub-optimal.
>+		 */
>+		return plane->base.type == DRM_PLANE_TYPE_CURSOR ? 63 : 0;
>+	}
> 
> 	pixel_size = drm_format_plane_cpp(state->base.fb->pixel_format, 0);
> 	clock = crtc->config->base.adjusted_mode.crtc_clock;
>-- 
>2.4.10
>

Actions: View | Splinter Review

Attachments on bug 92826: 119412 | 119413 | 119414 | 119598 | 119599 | 119600 | 119601 | 119602 | 119628 | 120441 | 120531 | 120549

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.