Skip to contentSkip to conversation

Pipe-A underrun workaround for i830 chipset.

PATCH In review Thomas Richter 2 participants 3 messages Nov 14, 2013 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Pipe-A underrun workaround for i830 chipset.

Dear Daniel, dear intel-experts,

please find a patch for the dreadful pipe-A underruns on the i830 attached that works at least in the linear framebuffer mode. As far as the tiling mode is concerned, I'm still scratching my head about it. It as at this time unclear how precisely this works, but I'll keep looking. At least this is something.

So long,
Thomas

Quoted 1 line

From ba72c1e506bb162f8ac595af94cc6ed850439883 Mon Sep 17 00:00:00 2001

From: Thomas Richter <[email protected]>
Date: Thu, 14 Nov 2013 18:16:14 +0100
Subject: [PATCH 11/11] Added a workaround for pipe-A underruns on i830 with
panning.

---
 drivers/gpu/drm/i915/intel_display.c |   61 ++++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)
drivers/gpu/drm/i915/intel_display.c +61 -0
index 971f6b9..21895b8 100644
@@ -2002,6 +2002,64 @@ unsigned long intel_gen4_compute_page_offset(int *x, int *y,
 	}
 }
 
+/*
+** Update the linear frame pointer for i830 based devices. Due to some
+** unknown hardware feature, updating the frame pointer may cause
+** the unified FIFO of these chips run dry, then causing a flicker
+** and a potential shutdown of the video overlay, or worse.
+** thor, 14.11.2013
+*/
+static void i830_update_plane_offset(struct drm_crtc *crtc,
+				     struct drm_framebuffer *fb,
+				     unsigned long linear_offset)
+{
+	struct drm_i915_gem_object *obj;
+	struct intel_framebuffer *intel_fb;
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+	unsigned long planeadr, oldadr;
+	int plane = intel_crtc->plane;
+	u32 reg = DSPCNTR(plane);
+
+	intel_fb = to_intel_framebuffer(fb);
+	obj = intel_fb->obj;
+
+	planeadr = i915_gem_obj_ggtt_offset(obj) + linear_offset;
+	DRM_DEBUG_KMS("Plane address is 0x%lx", planeadr);
+	/* I830 panning flicker work around.
+	** Only for non-LVDS output, only for i830.
+	** Tiling mode is still not supported.
+	*/
+	if (obj->tiling_mode != I915_TILING_NONE) {
+		if ((planeadr & 0x40)) {
+			DRM_DEBUG_KMS("Detected potential flicker");
+			DRM_DEBUG_KMS("No workaround for tiling mode");
+			DRM_DEBUG_KMS("Use a linear frame buffer");
+		}
+	} else {
+		switch (fb->pixel_format) {
+		case DRM_FORMAT_XRGB1555:
+		case DRM_FORMAT_ARGB1555:
+		case DRM_FORMAT_RGB565:
+		case DRM_FORMAT_XRGB8888:
+		case DRM_FORMAT_ARGB8888:
+		case DRM_FORMAT_XBGR8888:
+		case DRM_FORMAT_ABGR8888:
+			oldadr = I915_READ(DSPADDR(plane));
+			if (((oldadr ^ planeadr) & 0x40) &&
+			    (planeadr & 0xc0) == 0xc0) {
+				I915_WRITE(DSPADDR(plane),
+					   planeadr & (~(0x80)));
+				POSTING_READ(reg);
+				intel_wait_for_vblank(dev, intel_crtc->pipe);
+			}
+			break;
+		}
+	}
+	I915_WRITE(DSPADDR(plane), planeadr);
+}
+
 static int i9xx_update_plane(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 			     int x, int y)
 {
@@ -2095,6 +2153,9 @@ static int i9xx_update_plane(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 				     i915_gem_obj_ggtt_offset(obj) + intel_crtc->dspaddr_offset);
 		I915_WRITE(DSPTILEOFF(plane), (y << 16) | x);
 		I915_WRITE(DSPLINOFF(plane), linear_offset);
+	} else if (INTEL_INFO(dev)->gen == 2 && IS_I830(dev) &&
+		   !intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {
+		i830_update_plane_offset(crtc, fb, linear_offset);
 	} else
 		I915_WRITE(DSPADDR(plane), i915_gem_obj_ggtt_offset(obj) + linear_offset);
 	POSTING_READ(reg);
Signature · 1 line
1.7.10.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#1 lore raw
Quoted 8 lines · Thomas Richter

Dear Daniel, dear intel-experts,

please find a patch for the dreadful pipe-A underruns on the i830 attached that works at least in the linear framebuffer mode. As far as the tiling mode is concerned, I'm still scratching my head about it. It as at this time unclear how precisely this works, but I'll keep looking.
At least this is something.

A few bikesheds below. Once the polish is done I'll merge it. -Daniel

Quoted 10 lines

So long,
Thomas

From ba72c1e506bb162f8ac595af94cc6ed850439883 Mon Sep 17 00:00:00 2001
From: Thomas Richter <[email protected]>
Date: Thu, 14 Nov 2013 18:16:14 +0100
Subject: [PATCH 11/11] Added a workaround for pipe-A underruns on i830 with
panning.

A bit commit message summarizing what has been done thus far would be good.

Also please keep here a patch reflog, iirc we're at v3 or so. E.g.

v3: Extract workaround into i830_update_plane_offset function.

Quoted 11 lines
---
 drivers/gpu/drm/i915/intel_display.c |   61 ++++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)
drivers/gpu/drm/i915/intel_display.c +0 -0
index 971f6b9..21895b8 100644

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2011

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +7 -0
@@ -2002,6 +2002,64 @@ unsigned long intel_gen4_compute_page_offset(int *x, int *y,
 	}
 }
 
+/*
+** Update the linear frame pointer for i830 based devices. Due to some
+** unknown hardware feature, updating the frame pointer may cause
+** the unified FIFO of these chips run dry, then causing a flicker
+** and a potential shutdown of the video overlay, or worse.
+** thor, 14.11.2013
+*/

Multi-line comments should be like this:

/*
* text ...
*/

with the text flown to align to 80 chars. vim will do this for you.

Quoted 8 lines
drivers/gpu/drm/i915/intel_display.c +8 -0
+static void i830_update_plane_offset(struct drm_crtc *crtc,
+				     struct drm_framebuffer *fb,
+				     unsigned long linear_offset)
+{
+	struct drm_i915_gem_object *obj;
+	struct intel_framebuffer *intel_fb;
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2029

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
+	unsigned long planeadr, oldadr;
+	int plane = intel_crtc->plane;
+	u32 reg = DSPCNTR(plane);
+
+	intel_fb = to_intel_framebuffer(fb);
+	obj = intel_fb->obj;
+
+	planeadr = i915_gem_obj_ggtt_offset(obj) + linear_offset;
+	DRM_DEBUG_KMS("Plane address is 0x%lx", planeadr);

I'd ditch this debug output due to dmesg spam. I guess it was useful for debugging, but should go away for upstream.

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2035

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+	/* I830 panning flicker work around.
+	** Only for non-LVDS output, only for i830.
+	** Tiling mode is still not supported.
+	*/
+	if (obj->tiling_mode != I915_TILING_NONE) {
+		if ((planeadr & 0x40)) {

We tend to use decimal numbers for limits like these.

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2039

Quoted 4 lines
drivers/gpu/drm/i915/intel_display.c +4 -0
+			DRM_DEBUG_KMS("Detected potential flicker");
+			DRM_DEBUG_KMS("No workaround for tiling mode");
+			DRM_DEBUG_KMS("Use a linear frame buffer");
+		}

I'd just ditch this for now, no point spamming dmesg with that much noise if we don't have a fix.

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2048

Quoted 9 lines
drivers/gpu/drm/i915/intel_display.c +9 -0
+	} else {
+		switch (fb->pixel_format) {
+		case DRM_FORMAT_XRGB1555:
+		case DRM_FORMAT_ARGB1555:
+		case DRM_FORMAT_RGB565:
+		case DRM_FORMAT_XRGB8888:
+		case DRM_FORMAT_ARGB8888:
+		case DRM_FORMAT_XBGR8888:
+		case DRM_FORMAT_ABGR8888:

The switch here can be killed - higher levels already check for valid pixel formats.

Quoted 18 lines
drivers/gpu/drm/i915/intel_display.c +14 -0
+			oldadr = I915_READ(DSPADDR(plane));
+			if (((oldadr ^ planeadr) & 0x40) &&
+			    (planeadr & 0xc0) == 0xc0) {
+				I915_WRITE(DSPADDR(plane),
+					   planeadr & (~(0x80)));
+				POSTING_READ(reg);
+				intel_wait_for_vblank(dev, intel_crtc->pipe);
+			}
+			break;
+		}
+	}
+	I915_WRITE(DSPADDR(plane), planeadr);
+}
+
 static int i9xx_update_plane(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 			     int x, int y)
 {

drivers/gpu/drm/i915/intel_display.c static int i9xx_update_plane(struct drm_crtc *crtc, struct drm_framebuffer *fb, · line 2157

Quoted 5 lines
drivers/gpu/drm/i915/intel_display.c +2 -0
@@ -2095,6 +2153,9 @@ static int i9xx_update_plane(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 				     i915_gem_obj_ggtt_offset(obj) + intel_crtc->dspaddr_offset);
 		I915_WRITE(DSPTILEOFF(plane), (y << 16) | x);
 		I915_WRITE(DSPLINOFF(plane), linear_offset);
+	} else if (INTEL_INFO(dev)->gen == 2 && IS_I830(dev) &&
+		   !intel_pipe_has_type(crtc, INTEL_OUTPUT_LVDS)) {

IS_I830(dev) is good enough - the gen2 check is redundant and i830M doesn't have an LVDS encoder (lvds is only possible with the DVO encoder).

Quoted 7 lines
drivers/gpu/drm/i915/intel_display.c +1 -0
+		i830_update_plane_offset(crtc, fb, linear_offset);
 	} else
 		I915_WRITE(DSPADDR(plane), i915_gem_obj_ggtt_offset(obj) + linear_offset);
 	POSTING_READ(reg);
Signature · 1 line
1.7.10.4
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

A message that is not in the archive (sent off-list or not yet received): 26136_1384455070_52851B9E_26136_18661_1_20131114185141.GJ22741@phenom.ffwll.local

#3 lore raw

Hi Daniel,

a couple of comments on your comments. (-;

Quoted 1 line

with the text flown to align to 80 chars. vim will do this for you.

..which I ain't using. "Using" is not a word that goes along well with "vim". (-; But no, let's not start an editor war here.

Quoted 4 lines

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2035

Quoted 2 lines
drivers/gpu/drm/i915/intel_display.c +2 -0
+	if (obj->tiling_mode != I915_TILING_NONE) {
+		if ((planeadr & 0x40)) {

We tend to use decimal numbers for limits like these.

That's actually a bitmask to check for alignment. (-; In principle, the whole "if" can go because it's currently only a dummy with the real code still missing once I understand what's going on in the tiled mode. Or at least, once I find a way to fiddle around the pipe underflow.

Quoted 11 lines

drivers/gpu/drm/i915/intel_display.c unsigned long intel_gen4_compute_page_offset(int *x, int *y, · line 2048

Quoted 8 lines
drivers/gpu/drm/i915/intel_display.c +8 -0
+		switch (fb->pixel_format) {
+		case DRM_FORMAT_XRGB1555:
+		case DRM_FORMAT_ARGB1555:
+		case DRM_FORMAT_RGB565:
+		case DRM_FORMAT_XRGB8888:
+		case DRM_FORMAT_ARGB8888:
+		case DRM_FORMAT_XBGR8888:
+		case DRM_FORMAT_ABGR8888:

The switch here can be killed - higher levels already check for valid pixel formats.

Validity is not quite the issue here. The issue is that the workaround depends on the bytes per pixel. The formats above are all 24 and 16 bit modes for which the workaround, well, works. 8bpp and 30bpp modes are also currently unsupported. If you prefer, I could also compute a bpp value at the caller and pass it along.

Quoted 2 lines

IS_I830(dev) is good enough - the gen2 check is redundant and i830M doesn't have an LVDS encoder (lvds is only possible with the DVO encoder).

Ok.

Greetings,
Thomas