[PATCH] drm, gma500: Fix Cedarview boot failures in 3.3-rc

* [PATCH] drm, gma500: Fix Cedarview boot failures in 3.3-rc
@ 2012-03-05 14:22 Alan Cox
  2012-03-05 16:50 ` Linus Torvalds
  0 siblings, 1 reply; 6+ messages in thread
From: Alan Cox @ 2012-03-05 14:22 UTC (permalink / raw)
  To: airlied, dri-devel, torvalds

From: Alan Cox <alan@linux.intel.com>

[Resending with correct address for Linus]

Production GMA3600/3650 hardware turns out to be subtly different to the
development platforms. This combined with a minor driver bug is causing
the kernel to hang on these platforms.

This patch does the following

- turn down a couple of messages that were meant to be debug and are
  causing much confusion

- ensure the hotplug interrupt is disabled on Cedartrail systems.

- fix a bug where gtt roll mode called psbfb_sync, which tries to sync
  the 2D engine. On other devices it is harmless as the 2D engine is
  present but not in use when in gtt roll mode, on Cedartrail it causes
  a hang

Without these changes 3.3-rc hangs on boot on Cedartrail based systems.

Signed-off-by: Alan Cox <alan@linux.intel.com>
---

 drivers/gpu/drm/gma500/cdv_device.c  |    2 ++
 drivers/gpu/drm/gma500/framebuffer.c |    1 -
 drivers/gpu/drm/gma500/gtt.c         |    9 ++++-----
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/gma500/cdv_device.c b/drivers/gpu/drm/gma500/cdv_device.c
index 4a5b099..53404af 100644
--- a/drivers/gpu/drm/gma500/cdv_device.c
+++ b/drivers/gpu/drm/gma500/cdv_device.c
@@ -321,6 +321,8 @@ static int cdv_chip_setup(struct drm_device *dev)
 	cdv_get_core_freq(dev);
 	gma_intel_opregion_init(dev);
 	psb_intel_init_bios(dev);
+	REG_WRITE(PORT_HOTPLUG_EN, 0);
+	REG_WRITE(PORT_HOTPLUG_STAT, REG_READ(PORT_HOTPLUG_STAT));
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/gma500/framebuffer.c b/drivers/gpu/drm/gma500/framebuffer.c
index 830dfdd6b..be61673 100644
--- a/drivers/gpu/drm/gma500/framebuffer.c
+++ b/drivers/gpu/drm/gma500/framebuffer.c
@@ -247,7 +247,6 @@ static struct fb_ops psbfb_roll_ops = {
 	.fb_imageblit = cfb_imageblit,
 	.fb_pan_display = psbfb_pan,
 	.fb_mmap = psbfb_mmap,
-	.fb_sync = psbfb_sync,
 	.fb_ioctl = psbfb_ioctl,
 };
 
diff --git a/drivers/gpu/drm/gma500/gtt.c b/drivers/gpu/drm/gma500/gtt.c
index 5d5330f..aff194f 100644
--- a/drivers/gpu/drm/gma500/gtt.c
+++ b/drivers/gpu/drm/gma500/gtt.c
@@ -446,10 +446,9 @@ int psb_gtt_init(struct drm_device *dev, int resume)
 	pg->gtt_start = pci_resource_start(dev->pdev, PSB_GTT_RESOURCE);
 	gtt_pages = pci_resource_len(dev->pdev, PSB_GTT_RESOURCE)
 								>> PAGE_SHIFT;
-	/* Some CDV firmware doesn't report this currently. In which case the
-	   system has 64 gtt pages */
+	/* CDV doesn't report this. In which case the system has 64 gtt pages */
 	if (pg->gtt_start == 0 || gtt_pages == 0) {
-		dev_err(dev->dev, "GTT PCI BAR not initialized.\n");
+		dev_dbg(dev->dev, "GTT PCI BAR not initialized.\n");
 		gtt_pages = 64;
 		pg->gtt_start = dev_priv->pge_ctl;
 	}
@@ -461,10 +460,10 @@ int psb_gtt_init(struct drm_device *dev, int resume)
 
 	if (pg->gatt_pages == 0 || pg->gatt_start == 0) {
 		static struct resource fudge;	/* Preferably peppermint */
-		/* This can occur on CDV SDV systems. Fudge it in this case.
+		/* This can occur on CDV systems. Fudge it in this case.
 		   We really don't care what imaginary space is being allocated
 		   at this point */
-		dev_err(dev->dev, "GATT PCI BAR not initialized.\n");
+		dev_dbg(dev->dev, "GATT PCI BAR not initialized.\n");
 		pg->gatt_start = 0x40000000;
 		pg->gatt_pages = (128 * 1024 * 1024) >> PAGE_SHIFT;
 		/* This is a little confusing but in fact the GTT is providing

^ permalink raw reply related	[flat|nested] 6+ messages in thread