Attachment 99940 Details for Bug 77104 – [PATCH] drm/i915: Don't use RCS flips on gen7+ when multiple pipes are enabled

[patch] [PATCH] drm/i915: Don't use RCS flips on gen7+ when multiple pipes are enabled

0001-drm-i915-Don-t-use-RCS-flips-on-gen7-when-multiple-p.patch (text/plain), 2.60 KB, created by Ville Syrjala on 2014-05-27 10:29:45 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Ville Syrjala

Created: 2014-05-27 10:29:45 UTC

Size: 2.60 KB

patch

obsolete

>From f53fe967f2b47eef16eda03f342db44a62a40693 Mon Sep 17 00:00:00 2001
>From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= <ville.syrjala@linux.intel.com>
>Date: Wed, 26 Feb 2014 12:43:39 +0200
>Subject: [PATCH] drm/i915: Don't use RCS flips on gen7+ when multiple pipes
> are enabled
>MIME-Version: 1.0
>Content-Type: text/plain; charset=UTF-8
>Content-Transfer-Encoding: 8bit
>
>We have some reports which suggest that using RCS flips with multiple
>planes simultaneosly can lead to hangs. This might have something to
>do with the warnings in the spec that unmasking multiple events
>simultaneosly in DERRMR can lead to problems.
>
>So let's just fall back to BCS flips whenever multiple pipes are
>active. BCS flips were in use for a long time before RCS flips were
>introduced, so the hope is that BCS is more resilient in this regard.
>
>We will be looking at crtc->active for all the crtcs without locking
>all the crtc mutexes or the mode_config.mutex. We do hold the crtc
>mutex for the current crtc however, and since a full modeset will grab
>all the crtc mutexes, we can't have a full modeset happning in parallel.
>That means we can safely examine crtc->active for all crtcs as long
>as we hold at least one crtc mutex. We can ignore load detection since
>it's not used for this class of hardware, and in any case load detection
>wouldn't be issuing page flips.
>
>v2: Rebase and use for_each_intel_crtc()
>
>Cc: Enrico Tagliavini <enrico.tagliavini@gmail.com>
>Signed-off-by: Ville SyrjÃ¤lÃ¤ <ville.syrjala@linux.intel.com>
>---
> drivers/gpu/drm/i915/intel_display.c | 14 +++++++++++++-
> 1 file changed, 13 insertions(+), 1 deletion(-)
>
>diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
>index 8cbce75..3829fd7 100644
>--- a/drivers/gpu/drm/i915/intel_display.c
>+++ b/drivers/gpu/drm/i915/intel_display.c
>@@ -9184,6 +9184,17 @@ static int intel_default_queue_flip(struct drm_device *dev,
> 	return -ENODEV;
> }
> 
>+static int intel_num_active_pipes(struct drm_device *dev)
>+{
>+	struct intel_crtc *crtc;
>+	int num_active_pipes = 0;
>+
>+	for_each_intel_crtc(dev, crtc)
>+		num_active_pipes += crtc->active;
>+
>+	return num_active_pipes;
>+}
>+
> static int intel_crtc_page_flip(struct drm_crtc *crtc,
> 				struct drm_framebuffer *fb,
> 				struct drm_pending_vblank_event *event,
>@@ -9268,7 +9279,8 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
> 		ring = &dev_priv->ring[BCS];
> 	} else if (INTEL_INFO(dev)->gen >= 7) {
> 		ring = obj->ring;
>-		if (ring == NULL || ring->id != RCS)
>+		if (ring == NULL || ring->id != RCS ||
>+		    intel_num_active_pipes(dev) > 1)
> 			ring = &dev_priv->ring[BCS];
> 	} else {
> 		ring = &dev_priv->ring[RCS];
>-- 
>1.8.5.5
>

Actions: View | Splinter Review

Attachments on bug 77104: 96978 | 98467 | 98837 | 99397 | 99940 | 100213

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.