Ver 2 of the patch [was: Re: i915 driver update testing]

[Konstantin Belousov]

On Wed, Oct 08, 2014 at 03:12:08PM -0400, Adam McDougall wrote:
> On 10/08/2014 13:05, Konstantin Belousov wrote:
> > There are more occurences of the bug I fixed once in patch version 2.
> > Also, since pmap changes were committed in modified form, please try
> > the updated patch at https://www.kib.kiev.ua/kib/drm/i915.3.patch .
> > 
> 
> No apparent change:
> http://www.egr.msu.edu/~mcdouga9/i915-patch3-1.txt
cite> (kgdb) p *(struct drm_i915_private *)(dev_private)
cite> No symbol "dev_private" in current context.

This is p *(struct drm_i915_private *)(dev->dev_private)

I regenerated patch after recent merges and changes in KPI on HEAD.
https://www.kib.kiev.ua/kib/drm/i915.4.patch

Please apply it, I think the issue should be there still.  Then apply
the following debugging patch, and set
	kenv drm.debug=0x3
before loading i915kms.ko.  I want to see the same debugging information,
and dmesg from the moment of loading the driver.

diff --git a/sys/dev/drm2/i915/intel_sdvo.c b/sys/dev/drm2/i915/intel_sdvo.c
index 74e479a..e1f1d09 100644
--- a/sys/dev/drm2/i915/intel_sdvo.c
+++ b/sys/dev/drm2/i915/intel_sdvo.c
@@ -1952,8 +1952,10 @@ intel_sdvo_select_i2c_bus(struct drm_i915_private *dev_priv,
 		sdvo->i2c = intel_gmbus_get_adapter(dev_priv, pin);
 		intel_gmbus_set_speed(sdvo->i2c, GMBUS_RATE_1MHZ);
 		intel_gmbus_force_bit(sdvo->i2c, true);
+printf("i915: select i2c pin %d priv %p i2c %p\n", pin, dev_priv, sdvo->i2c);
 	} else {
 		sdvo->i2c = intel_gmbus_get_adapter(dev_priv, GMBUS_PORT_DPB);
+printf("i915: select i2c DPB %d priv %p i2c %p\n", pin, dev_priv, sdvo->i2c);
 	}
 }
 
@@ -2601,6 +2603,7 @@ bool intel_sdvo_init(struct drm_device *dev, uint32_t sdvo_reg, bool is_sdvob)
 
 	intel_sdvo = malloc(sizeof(struct intel_sdvo), DRM_MEM_KMS,
 	    M_WAITOK | M_ZERO);
+printf("i915: intel_sdvo %p\n", intel_sdvo);
 
 	intel_sdvo->sdvo_reg = sdvo_reg;
 	intel_sdvo->is_sdvob = is_sdvob;