4.3.x DRM_I915_FBDEV ... no framebuffer!!

[khayyam]

hello ...

I've not been able to build a (3.x) kernel since 3.13.11 (or thereabouts), I'm now able to build 4.3.4 but am completely unable to get a framebuffer. I've built 4.3.4 27 times this morning, all with various combinations of X86_SYSFB, DRM_I915/DRM_FBDEV_EMULATION, FB_SIMPLE, FB_EFI, enabled, disabled, and what-have-you. Nothing but a blackscreen ...

Added to that whoever wrote the "help" for X86_SYSFB and FB_SIMPLE needs shooting, I mean, who can tell me what the "new generic system-framebuffer drivers" are? Apparently such a thing exists, and there are various gotchas with regards to "legacy fbdev drivers [...] not be[ing] able to pick up generic system framebuffers if this option [X86_SYSFB] is selected". So, I should "say Y" and am "highly encouraged" to also enable simplefb.

It doesn't matter however, as no matter what combination of the above I try I get nothing on the display. Previously I had DRM_I915, DRM_I915_KMS, DRM_I915_FBDEV enabled (and nothing under framebuffers =>) and got inteldrmfb, with 4.3.4 there doesn't seem to be any means of getting a famebuffer with a similarly enabled .config ... nor with the "new generic system-framebuffer drivers" (which right now I can only assume is ... singularly ... simplefb).

So, what is it you're supposed to enable to get inteldrmfb ... or *any* framebuffer? Is there some magic combination of options I apparently missed? Can anyone make any sense of the X86_SYSFB "help", and explain what "legacy" and "new generic system framebuffers" mean, or what is being explained?

Thanks in advance ... khay

[charles17]

No explanation, sorry. But from gentoo-sources-4.4.0-r1 I have the following:

[NeddySeagoon]

khayyam,

Simple framebuffer expects to be able to draw in the framebuffer without doing any set up.
It expects that whatever runs before it comes along has done that.

Now, I have no idea what does the setup, or what chipsets it might apply to.

While that may be an interesting factoid, its probably not helping with your problem.
I have an Intel 945 chip on 4.x and its framebuffer works. I can pastebin its .config file if that helps, or you can pastebin yours.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[ennui]

Are you booting using BIOS or UEFI? Can you see anything relevant in the initial kernel output during boot from dmesg? For example, my simple framebuffer shows up as:

[khayyam]

Thanks to all who've replied so far ... I should probably have emphasised that I think I've tried every possible combination of options, with the 27 builds reflecting that fact. The above post was more a case of my having given up and/or the hope I'd overlooked something ... anyhow:

@charles17 ... those options are exactly as I'd had enabled at the outset, and have since replicated, nothing. As an aside, two pattern searches, I'd generally use the following rather than 'grep | grep' ...

[charles17]

khayyam,
There was a similar problem for me with kernel 3.14.
I had to connect to a VGA monitor (which could stay power off) to see anything on the laptop's built-in screen.

And, thanks for the awk advise. I'm not much experienced with sed and awk

[fpemud]

I encountered the same problem. I solved it by selecting DRM_FBDEV_EMULATION.

[s4e8]

build i915 driver as module, and add it to /etc/modprobe.d/blacklist.conf. Boot new kernel and running "dmesg -n 8; modprobe i915".

[genoobish]

I use the same driver, and have the same problem. But my 4.3.3 works, the 4.4.0 is the one that despite thousands of attempts (changing kernel config and recompiling) stiil has no framebuffer.

The drm loads however, so X also loads (I have it autologin).

looking at the dmesg, inteldrmfb says it loads:

[s4e8]

My IVB require CONFIG_DRM_I915_PRELIMINARY_HW_SUPPORT=y since 4.4.0.

[khayyam]

[s4e8]

You means no efifb? Did you enable the CONFIG_FRAMEBUFFER_CONSOLE

[ennui]

[khayyam]

[genoobish]

[khayyam]

[Anon-E-moose]

I run nouveau and radeon but these are what I have set for graphics/fb/console in the .config for 4.3.3

[s4e8]

efifb is very simple, the kernel just take over the framebuffer and writting pixel to it. If you built all fbdev/drm driver as module except efifb, you should get a visible console, unless the kernel panic very earlier, before fbdev/fbcon initialzed. Kernel panic beforefbdev/ fbcon it's very bad situation, there's no good way to debug it unless you can setup serial console. There's kernel parameter earlyprintk=efi would print some very earlier message, after dummy console take over, and before fbcon/fbdev full initialized, you can't see any error messages.
[quote="khayyam"]

[Ant P.]

Is it possible 4.3.x thinks your primary display is different, possibly unconnected?

Might sound insane, but I've seen i915 do insane things in the past: one device (internal LVDS with no external monitor port) thought it had two screens mapped to the same display region for a long time... turns out the kernel driver was at fault there too.

[khayyam]

[genoobish]

Ok, I think I'vie found a workaround.
I'm able to get the framebuffer if the vga monitor is not connected (wtf) on the 4.5.0-rc2 vanilla(that's the only one I tried though). The vga monitor got accidentally disconnected at boot time just as I was testing with the kernel, it may work with others too though. With just the laptop or the hdmi monitor it works fine.

This is the config: https://bpaste.net/show/c907c7d5b448

And the vga monitor still works fine on X, but if I want to switch to the console, I have to unplug it for the frambuffer to work(the normal behaviour would be that it would mirror the lsvd, and the hdmi would stay blank).
and my dmesg: https://bpaste.net/show/2c872ce7093f

Please, don't tell me your monitor is not vga
if it is indeed a vga montior, can you test with a digital monitor?

[khayyam]

[s4e8]

AGP is optional since 3.14:
00fe639 drm/i915: Make AGP support optional

[khayyam]

[genoobish]