Graphics Support

[Alex Merritt]

The Xorg changes seemed to work! See below.

On Mon, Jul 20, 2015 at 3:43 PM, Matthew Dillon <dillon at backplane.com>
wrote:

> You might be able to disable the nvidia card in the BIOS or change the
> display priority to the internal haswell gpu.  If you can disable the
> nvidia card entirely that might de-confuse the drivers sufficiently.
>

Unfortunately, this BIOS only has "enable", "disable", and "auto" for the
"Hybrid Graphics" mode. Apparently it seems either I can disable the
Haswell GPU, allow both to be visible, or "allow the OS to determine"
(=auto). I've configured it this time as "enable".


>
> Safe mode shouldn't be any different from normal mode unless you are doing
> extra module loads in your/boot/loader.conf or /etc/rc.local.
>

Both of those files are empty, currently.


>
> Unfortunately we're working of bits and pieces of information here.  It
> looks like the kernel sees two vga devices and agp is able to attach to one
> of them.  X should be able to attach to the one agp found.  If it isn't,
> you may be able to force it in your xorg.conf in the "Device" section by
> specifying the BusId of the haswell gpu and forcing the intel driver.  It's
> usually device 0:2:0 on the PCI bus:
>
>     BusId "PCI:0:2:0"
>     Driver "intel"
>
> Try that.  Boot verbose and save the *entire* dmesg output to pastebin.
> Don't preload the i915kms or drm kernel module.  Try starting X directly
> with 'startx' from a root prompt.
>
> ssh in from another box so, presumably it hasn't crashed, you can then
> access the /var/log/Xorg.0.log after starting X and put that entire file on
> pastebin as well.
>

I realize this is a limited-information method of debugging, thank you for
helping me figure this out. I created a new config with Xorg -configure and
edited the file to have the two lines above.

Here is the entire dmesg: http://pastebin.com/VX8M5KcA The consecutive
newlines were added by me to indicate where boot completed, and where
startx was issued.

The screen blinked a few times with garbage, but finally shows a bare
window manager with terminals and a clock, at 1920x1080. Going to any
virtual terminal blanks the screen, and I am unable to return to F9. This
is as root.

startx as non-root also works -- I entered into fluxbox ($HOME/.xinitrc has
startfluxbox). I see these additional messages in dmesg:

error: [drm:pid1161:i915_write32] *ERROR* Unknown unclaimed register before
writing to 70088
error: [drm:pid-1:i915_write32] *ERROR* Unknown unclaimed register before
writing to 4400c
error: [drm:pid-1:i915_write32] *ERROR* Unknown unclaimed register before
writing to 4400c
error: [drm:pid-1:i915_write32] *ERROR* Unknown unclaimed register before
writing to 4400c

The last line continues repeating very slowly (once every 40-60 seconds or
so).


>
> Maybe we can figure out a workaround to at least get the haswell gpu
> operational.  If it were haswell alone it probably would just work.
>
> -Matt
>
>
> On Sun, Jul 19, 2015 at 11:10 PM, Alex Merritt <merritt.alex at gmail.com>
> wrote:
>
>>
>> On Fri, Jul 17, 2015 at 11:36 AM, Matthew Dillon <dillon at backplane.com>
>> wrote:
>>
>>> You can run 'xrandr' (I think it will work with the VESA driver) to see
>>> what your resolution options are.  The VESA driver is limited to what the
>>> BIOS says it can do, and BIOSes are often broken in that regard.
>>>
>>
>> xrandr shows 640x480 anx 1024x768 as the maximum.
>>
>>
>>> In terms of the graphics outputs... what you need to do is carefully
>>> read the specs for your motherboard.  Not all motherboards use the on-die
>>> GPU, many also have a discrete radeon or nvidia chipsets.  If the
>>> motherboard says it can do either you may be able to control which one is
>>> used with a BIOS Setup option.  But it depends on the mobo.
>>>
>>
>> Ah, yes I forgot to check the BIOS for this. In the BIOS there is an
>> option to twiddle the integrated vs NVIDIA card ("hybrid mode"). Now the
>> i915 module states the following:
>>
>> agp0: <Haswell mobile GT2 IG> on vgapci1
>> agp0: aperture size is 256M, detected 524284k stolen memory
>> info: [drm] Initialized drm 1.1.0 20060810
>> link_elf_obj_obj: symbol AcpiGetHandle undefined
>> agp0: detached
>> linker_load_file: Unsupported file type
>>
>> This I am only able to see when starting in "safe mode". Booting as
>> normal, the screen becomes black and I am unable to switch among the
>> virtual terminals. SSH also hangs or shows severe delays (maybe
>> unrelated?).
>>
>> I booted into the more verbose option in DF, and see the following when
>> the module loads:
>>
>> http://pastebin.com/zdk7Kquh
>>
>> info: [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
>> info: [drm] Driver supports precise vblank timestamp query.
>> drm1: taking over the fictitious range 0xa0000000-0xb0000000
>> info: [drm] Initialized i915 1.6.0 20080730
>> info: [drm] Enabling RC6 states: RC6 off, RC6p off, RC6pp off
>>
>> -Alex
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dragonflybsd.org/pipermail/users/attachments/20150720/a2f76acc/attachment-0003.htm>