[Intel-gfx] [PATCH v2] drm/i915: Fix BUG in i915_gem.c when switch to console

[Xi Ruoyao]

On 03/12/2015 04:28 PM, Jani Nikula wrote:
> On Thu, 12 Mar 2015, Xi Ruoyao <xry111 at outlook.com> wrote:
>> In intel_crtc_page_flip, intel_display.c, the code changed the framebuffer
>> assigned to plane crtc->primary by
>>
>> crtc->primary->fb = fb;
>>
>> However, it forgot to change crtc->primary->state->fb. However, when we
>> switch to console, some kernel code will read crtc->primary->state->fb
>> to get the framebuffer assigned to crtc->primaty. Then a framebuffer
>> object can be unpinned twice and a kernel BUG will be produced in i915_gem.c.
>>
>> So, update crtc->primary->state->fb in intel_display.c using
>> drm_atomic_set_fb_for_plane to fix the BUG.
>>
>> Signed-off-by: Xi Ruoyao <xry111 at outlook.com>
>> Fixed: Bug 93711 <https://bugzilla.kernel.org/show_bug.cgi?id=93711>
> Is this a problem with drm-intel-nightly? In particular see
>
> commit afd65eb4cc0578a9c07d621acdb8a570e2782bf7
> Author: Matt Roper <matthew.d.roper at intel.com>
> Date:   Tue Feb 3 13:10:04 2015 -0800
>
>      drm/i915: Ensure plane->state->fb stays in sync with plane->fb
>
> Matt, do you think this fixes the described issue? Can we backport to
> drm-intel-fixes (and v4.0)?
>
> BR,
> Jani.
>
>
>
>> ---
>>   Sorry, the previous patch is mangled by email client, so I am re-sending it.
>>
>>   drivers/gpu/drm/i915/intel_display.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
>> index e730789..97083fd 100644
>> --- a/drivers/gpu/drm/i915/intel_display.c
>> +++ b/drivers/gpu/drm/i915/intel_display.c
>> @@ -37,6 +37,7 @@
>>   #include <drm/i915_drm.h>
>>   #include "i915_drv.h"
>>   #include "i915_trace.h"
>> +#include <drm/drm_atomic.h>
>>   #include <drm/drm_atomic_helper.h>
>>   #include <drm/drm_dp_helper.h>
>>   #include <drm/drm_crtc_helper.h>
>> @@ -9816,6 +9817,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
>>   	drm_gem_object_reference(&obj->base);
>>   
>>   	crtc->primary->fb = fb;
>> +	drm_atomic_set_fb_for_plane(crtc->primary->state, fb);
>>   
>>   	work->pending_flip_obj = obj;
>>   
>> -- 
>> 1.9.1
>>
Thank you to read my patch.

Matt's patch is better than mine since it fixed more conditions causing the
unsynchronize of plane->fb and plane->fb. But unfortunately, it cannot be
applied to mainline kernel (Linux 4.0.0-rc3) directly.

I'll try to apply Matt's changes to the mainline kernel.