Daiajo Tibdixious posted on Sat, 08 Oct 2016 20:35:55 +1100 as excerpted:
> This relates to my ATI driver conflict with xorg-server problem.
> Sorry if this comes though in html format I did not realise the last one
> was in html mode.
> I'm trying the solution to stop using fglrx and just use the radeon
> driver,
> however I'm getting weird problems.
>
> I removed fglrx from VIDEO_CARDS. Enabled radeon driver as a module in
> the kernel,
> plus several other options recommended by the wiki.
> First I hacked \etc\X11\xorg.conf to removed the ATI specific code,
> however I get the same problem with X -configure code.
>
> Googling the actual error "Screen 0 deleted because of no matching
> config section"
> shows that its not that, its some kind of driver problem.
> I've tried everything I can think of (except putting the driver in
> kernel with firmware blobs).
>
> Instead of my phone I borrowed another computer and can transfer stuff
> to it via USB.
>
> I have enclosed xorg.conf (from X -configure) and /var/log/Xorg.0.log
>
> Any hints on what else I can try? I don't want to try the other
> alternative (use fglrx with old versions of xorg-server and the
> input-evdev,video-ati drivers) yet.
From the log...
> [ 32385.590] Build Operating System: Linux 4.4.6-gentoo x86_64 Gentoo
> [ 32385.606] (II) [KMS] drm report modesetting isn't supported.
> [ 32385.606] (EE) Screen 0 deleted because of no matching config
> section.
> [ 32385.606] (II) UnloadModule: "radeon"
That's the problem. Reasonably new xf86-video-ati (the radeon driver)
requires KMS, kernel mode setting. Because it can't detect that it
doesn't even get as far as detecting the radeon chipset.
Just to be sure we're looking at the same wiki page:
https://wiki.gentoo.org/wiki/Radeon Most of this appears to be covered in the wiki, which you mentioned, but
just to be sure (I have a monolithic-build kernel here, module loading
disabled, so the below reflects that):
Kernel config, drivers, graphics:
[*] Direct Rendering Manager
(My 4.7 kernel doesn't show the radeon options here under DRM/DRI as the
wiki implies. They're elsewhere. Of course that means no mode-setting
option here either.)
Still directly under graphics:
[*] ATI Radeon
(There's an always enable userptr support sub-option, not checked here,
but the help says it selects mmu_notifier, which a search says is already
on for other reasons. It shouldn't hurt to turn it on, tho, and may
help.)
Still directly under graphics:
Frame buffer Devices
This is hard-enabled here by options I've set elsewhere.
*** WARNING, UNINTUITIVE ***
Do *not* set ATI Radeon display support here, under frame-buffer
devices. On the wiki it's unset, but it's both unintuitive and easy to
miss that it's unset. The radeon-specific framebuffer driver is for UMS,
not KMS. All that's needed is that the main/generic framebuffer support
option be enabled (which as I said it is here, no option to disable it).
I also have enabled, but am not sure it's necessary:
Under graphics, Console display driver support:
[*] Framebuffer console support
The map the console to the primary display device sub-option is hard-
enabled, no way to turn it off.
(I also keep vga text console enabled as a fallback, in case I switch to
an incompatible video card or something, but the framebuffer console is
what's actually used.)
If all those checkout, note that as the wiki says, firmware blobs are
required for r600+. I use the radeon-ucode package, but if you need
other blobs you may prefer the linux-firmware package.
As the wiki says, if you're building the module into the kernel (as I do,
since I have a monolithic kernel with all modules builtin), you'll need
the firmware builtin as well.
The kernel log (dmesg) should confirm whether and what firmware loaded,
or report that it couldn't, if that's the problem.
(Personal note, since I'm known to be rather particular about proprietary
code. My personal line is code loaded on whatever device, vs. code run
on the CPU. While I don't consider firmware blobs like the above that
run on some other device, in this case the GPU, to be ideal, I can and
will, if reluctantly, load them, while I won't run the proprietary
drivers as they run on the CPU. That's the distinction I have made for
some time, tho with GPUs being used for more than graphics these days, I
realize that can be more of a compromise than it was when such things
were single-purpose dedicated. But as I'm not (yet?) running anything
but graphics on the GPU, that's a bridge I've yet to cross.)
--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman
Xorg.0.txt (17.4 KB)
