[Pixman] [BUG] SIGSEGV in sse2_fill

[Frédéric Fauberteau]

Le 2018-09-10 12:34, Michel Dänzer a écrit :
> On 2018-09-10 12:01 p.m., Frédéric Fauberteau wrote:
>> Le 2018-09-10 09:26, Michel Dänzer a écrit :
>>> On 2018-09-10 8:22 a.m., Frédéric Fauberteau wrote:
>>>> Le 2018-09-01 15:16, Michel Dänzer a écrit :
>>>>> On 2018-09-01 9:22 a.m., Frédéric Fauberteau wrote:
>>>>>> 
>>>>>> [  5770.134] (EE) RADEON(0): Failed to make prime FD for handle: 
>>>>>> 22
>>>>>> [  5770.134] (EE) RADEON(0): Failed to create textured screen.
>>>>> 
>>>>> This is the problem. Please try current xf86-video-ati Git master,
>>>>> specifically
>>>>> https://gitlab.freedesktop.org/xorg/driver/xf86-video-ati/commit/db28d35ce9fd07a2a4703f3df0633d4c8291ff9b
>>>>> 
>>>>> 
>>>>> could help for this.
>>>> 
>>>> I just tested this commit and xorg server now starts without
>>>> segfaulting.
>>>> 
>>>> I am from a remote host and I don't know if the display is correct. 
>>>> But
>>>> the log file shows an error:
>>>> 
>>>> [688070.735] (II) Initializing extension DRI2
>>>> [688070.737] (II) RADEON(0): Setting screen physical size to 783 x 
>>>> 277
>>>> [688070.923] failed to add FB for modeset
>>>> [688070.923] (WW) RADEON(0): Failed to set mode on CRTC 0
>>>> [688070.954] failed to add FB for modeset
>>>> [688070.954] (WW) RADEON(0): Failed to set mode on CRTC 1
>>>> [688070.955] (EE) RADEON(0): Failed to enable any CRTC
>>> 
>>> Please try the Git master branch, which will soon become the 18.1
>>> release. If there's still a problem with that, please provide the
>>> corresponding full Xorg log file.
>> 
>> Please find the Xorg.0.log produced after upgrading the driver to
>> de88ea2755611bdcb18d91d8234d2ab5be8ff2e9:
> 
> Thanks, the rest of the log file looks fine.
> 
> The above warning and error messages indicate that drmModeAddFB fails.
> Since you got the "Failed to make prime FD for handle" error message
> from glamor with the 18.0.1 driver, there might still be an issue
> related to the GEM handle of the front buffer.
> 
> I've never seen this before, so I suspect it could be an issue specific
> to NetBSD's KMS support. Can you poke around a bit to see why
> drmModeAddFB fails? E.g., what are the parameters passed to it by
> radeon_fb_create, and what error does it return?

My wm runs but I have a blank screen.

The problem seems come from drmModeAddFB () from 
/usr/pkg/lib/libdrm.so.2 but I have no debugging symbol.

Thanks for your feedback. I'll try to follow your trail.