MacBook Pro 2012 Retina

mzanetti · 2012-06-30 08:56:09

Hi,

since a few days, I'm trying to install archlinux on the MacBook Pro 2012 Retina. As I guess this will be a long way until everything works as expected, I would like to use this thread as a collection of all issues for this model. Anyways, currently I'm still stuck quite at the beginning:

I managed to boot the the 2012.04 archboot iso by passing "noapic" to the kernel and installed everything.
After the reboot, grub loads fine. But when I select the kernel, the screen just goes blank. I have tried it with "nomodeset", doesn't help. The only thing that helps to boot with having the screen enabled is passing "noefi" to the kernel. This however, breaks lots of other stuff (IRQs not working etc). Adding "fakebios" to grub does not help either.

If you already know how to get around this, please let me know. If you have other problems with this model, feel free to use this thread too. Once we have the most important issues solved, information will be transferred to a wiki page.

litemotiv · 2012-06-30 09:17:22

You may be able to find some suggestions here: problems booting archlinux on a MacBook Air Mid 2012

mzanetti · 2012-06-30 09:54:57

cool, thanks a lot...This helped me to get over the first problem. Passing "noapic" not only fixes the APIC related kernel panic with the live imange, but also the black screen issue with the installed one.

Next stop: current stable kernel (3.4.4) stops with "[drm:intel_dsm_platform_mux_info] *ERROR* MUX INFO call failed"
linux-mainline from AUR (which is 3.5-rc4) boots up fine, however, I need to figure how to install nvidia etc with the linux-mainline.

I'll report back on the progress

litemotiv · 2012-06-30 10:10:32

mzanetti wrote:

Next stop: current stable kernel (3.4.4) stops with "[drm:intel_dsm_platform_mux_info] *ERROR* MUX INFO call failed"

That line itself should be harmless i think, i also get it on my 2011 MBA and that boots alright.

mzanetti · 2012-06-30 10:14:10

litemotiv wrote:

mzanetti wrote:
Next stop: current stable kernel (3.4.4) stops with "[drm:intel_dsm_platform_mux_info] *ERROR* MUX INFO call failed"
That line itself should be harmless i think, i also get it on my 2011 MBA and that boots alright.

hmm... ok. at least its the last thing that I see before everything freezes.

Anyways, I managed to install the nvidia module for the 3.5 kernel. it loads fine, but starting X produces the good old blank screen again

andy753421 · 2012-07-03 05:55:08

For what it's worth, I'm currently using Arch (mostly) on one of these laptops. There's a post on the OpenSuse forums that I found pretty useful, especially the couple of kernel patches that the guy wrote.

I think I had to boot in EFI mode for the NVidia drivers to work, I've been using refined rather than grub since that seemed to be easier to install and configure.

I'm still haven't got Intel graphics or the screen backlight working and the WiFi drivers are kind of flaky, let me know if you've had any luck with those

mzanetti · 2012-07-03 06:51:19

I still haven't managed to get the X server running on this thing... The stable kernel still hangs at booting, the linux-mainline from aur boots fine. However, even after installing nvidia-beta-all the X server just doesn't start up. What combination of linux+nvidia are you running (and which kernel arguments)?

When loading the X server with the intel drivers, it messes up even the terminal here, same with nouveau.

I also see the flakiness of the WiFi card (and only b/g seems to be supported for now) but I didn't really look into that so far.

gbrunoro · 2012-07-03 14:14:14

mzanetti wrote:

I still haven't managed to get the X server running on this thing... The stable kernel still hangs at booting, the linux-mainline from aur boots fine. However, even after installing nvidia-beta-all the X server just doesn't start up. What combination of linux+nvidia are you running (and which kernel arguments)?

Which boot setup do you have for GRUB? BIOS emulation, pure EFI, ...?

andy753421 · 2012-07-04 06:17:50

mzanetti wrote:

I still haven't managed to get the X server running on this thing... The stable kernel still hangs at booting, the linux-mainline from aur boots fine. However, even after installing nvidia-beta-all the X server just doesn't start up. What combination of linux+nvidia are you running (and which kernel arguments)?

I'm using Linus's master from a few days ago, but it should be pretty much the same as to the 3.5-rc5. I think it worked with a few other version as well, so anything from the past few weeks should be fine. I don't have any special kernel arguments, and the ones do have probably don't apply for you (root=/dev/sda4 rootflags=ssd,discard,subvol=/arch init=/sbin/mkinit boot). Here's my .config if you want to look at that, but it may not be very good.

I'm using nvidia driver 295.49 right now, but the 302.17 driver worked fine as well and was easier to install. I downloaded the package from the NVidia website and ran it myself instead of using the arch package since I built my own kernel.

The nouveau driver was messing up my screen as well. I think the i915 drivers didn't do that for me, but they wouldn't load X. The nouveau drivers actually started X but the graphics were just all screwed up. The NVidia driver loads X but then if I Ctlr-Alt-F1 back to terminal everything the screen is black.

litemotiv · 2012-07-04 07:40:42

andy753421, it's a bit offtopic here, but how is that screen ppi working out for you on linux?

andy753421 · 2012-07-04 09:56:30

Not to bad, I mostly use gtk+ and terminal programs (rxvt-unicode), so it was really just a matter of changing my font sizes and then most stuff looks fine. There are a few program that mess it all up by trying to implement their own widgets instead of just sticking with the standad stuff (e.g. google chrome). On the topic of fonts, the dpi is high enough that I think serifs actually look good on screen.

Note: I haven't tried using gnome/unity/kde/whatever so I don't know how well the normal "desktops" will handle it. I think there's also some xorg.conf DPI setting that I'm supposed to be using that changes the fonts automatically or something, but I haven't tried that at all yet.

mzanetti · 2012-07-14 20:33:45

gbrunoro wrote:

mzanetti wrote:
I still haven't managed to get the X server running on this thing... The stable kernel still hangs at booting, the linux-mainline from aur boots fine. However, even after installing nvidia-beta-all the X server just doesn't start up. What combination of linux+nvidia are you running (and which kernel arguments)?
Which boot setup do you have for GRUB? BIOS emulation, pure EFI, ...?

I'm using GRUB in EFI mode. The BIOS emulation messed even more up than the EFI more. Haven't managed to boot a kernel directly through rEFInd.

In the meantime I have upgraded to linux 3.5-rc6 and have tried the current stable and beta nVidia driver. Still nothing...

Last edited by mzanetti (2012-07-14 20:35:54)

89c51 · 2012-07-15 07:04:21

GregKH is trying to install Linux on his machine https://plus.google.com/111049168280159033135

Might be of help.

litemotiv · 2012-07-15 08:43:17

89c51 wrote:

GregKH is trying to install Linux on his machine https://plus.google.com/111049168280159033135
Might be of help.

Greg Kroah-Hartman wrote:

It boots! From the disk into EFI mode directly.
Time for a beer, this has been way more work than expected

That says enough.

frigaut · 2012-07-31 00:24:17

andy753421 wrote:

For what it's worth, I'm currently using Arch (mostly) on one of these laptops. There's a post on the OpenSuse forums that I found pretty useful, especially the couple of kernel patches that the guy wrote.
I think I had to boot in EFI mode for the NVidia drivers to work, I've been using refined rather than grub since that seemed to be easier to install and configure.
I'm still haven't got Intel graphics or the screen backlight working and the WiFi drivers are kind of flaky, let me know if you've had any luck with those

Same here. Arch running pretty smooth on that baby:
- arch install using last iso on a USB stick
- Install to directly boot EFI (requires refind installed on osx to pass boot parameters)
- Using noapic and nomodeset, no kernel patch used (currently using 3.5.0-2 from [testing] but started with similar results with the 3.4.6 from [core])
- intel graphic driver not working (for now), nvidia blob working fine (using Option "UseDPLib" "Off")
- Using xfce4 + compiz, smooth. DPI has to be pushed up quite a bit, and buttons/window title end up being uncomfortably small (as they don't scale with DPI), but it'll do for now. Overall, the display is gorgeous and the fonts renders really nicely.
- Sound is working (using pulseaudio here) with minimum fussing (nothing but enable a few channels in pavucontrol)
- wireless is working using b43. Had to extract the firmware with b43-fwcutter.
- dual monitor kinda works using nvidia-settings (but I had a "black screen of death" after modifying the config a few times)
- wire network works using the network to USB adapter from apple (haven't tried the mini-DP ethernet adaptor yet but I suspect it'll have hot-plug issues)
- [edit] the facetime camera works.

Main issues:
- suspend's not working out of the box (that's a biggie) hum... somehow now it works (installed acpid?)!
- may be related to the same issue, but switching back to a VT, the screen does not lit back (it becomes black and stays that way).

Secondary issues:
- none of the control button (backlight, keyboard backlight, volume) work. I can manually control display backlight and volume though, so it'll just be a matter of assigning the action to the keys
- [edit] can't make the microphone works. kind of annoying for skype.
- didn't test bluetooth
- got a lot of "applesmc: read arg fail" in dmesg (in fact the boot hangs for 30 sec or so and I suspect it's related). I know a patch is around but right now I didn't have to build a new kernel and I'd like to continue so if I can.

Can (and will) provide more details. In summary, it looks promising, need some work to make things work properly. I couldn't yet make suspend work, and this affect usability, but everything in its own time (fixed).

Last edited by frigaut (2012-07-31 03:41:41)

andy753421 · 2012-07-31 04:40:48

What did you have to do to get the display backlight working (which driver), and how are you controlling it? That's been the #1 issue for me so far.

You can control the keyboard backlight by poking values into /sys/class/leds/smc::kbd_backlight/brightness. I'm not sure if that depends on the applesmc patch you mentioned or not. I have that applied and the backlight works. I've been using xbindkeys to control all the media keys, using the media key function as default.

frigaut · 2012-07-31 05:23:19

andy753421
I'm using the nvidia blob. I have the display working with 2 caveats:
- once X has started once, I have a black screen when I switch back to VTs (but recover X and the backlight when switching back to vt7 with fn-ctrl-alt-7)
- contrarily to what I said in my previous post, I can't find the way to control the backlight intensity. I'm sure had this working yesterday using the same nvidia blob, but somehow today I can't find the /sys files corresponding to this function. I don't have the /sys tree you are referring to, as I still don't have the fixed applesmc (building it as I write this).
- xbindkeys is fine. I'm just using the xfce4 keyboard menu which works as well for me.

mzanetti · 2012-07-31 06:19:57

Hi, sorry, forgot to update this post... I have it running since some time now too with more or less the same setup than frigaut with the following exceptions:

- boot works with refit/refind too, but I blessed a standalong-grub so I can boot linux by holding the alt-key without the need of refit/refind
- after suspend to ram the machine fails to resume. (that's the main issue I have so far)
- keyboard backlight works fine after patching the kernel with the patch provided in the SuSE forums. This patch makes also the applesmv warnings and the 30 sec hang at boot go away.
- Blutooth also working fine after applying the patch

I'm using KDE. It looks totally awesome on it. There are some smaller parts where hardcoded sizes don't adapt to DPI settings but I already patched some of them.

For me it's currently only the suspend to ram and the screen backlight which need to be fixed. The rest seems to work fantastic.

frigaut · 2012-07-31 06:44:45

mzanetti,

As I said, suspend/resume are working for me now. It wasn't yesterday. I suspect acpid that I have installed since then has something to do with it.
However you use KDE so I understand it provides the acpi features (not sure?).

I think we should create a archwiki page for this laptop. If there's no volunteer, I'll start a page sometime in the near future :-)

mzanetti · 2012-07-31 07:43:22

frigaut,

Yes, KDE has acpi stuff as a dependency. However, the problem is not that it doesn't go to sleep, the problem is that it never wakes up again. Unfortunately also using pm_trace doesn't give any hints on what's going wrong.

+1 for the wiki page...

89c51 · 2012-07-31 10:24:13

There is also this thread in the intel-gfx mailing list:

http://lists.freedesktop.org/archives/i … 19326.html

frigaut · 2012-07-31 13:55:44

Thanks. Indeed relevant.
As Greg describe in one of his email in this thread: By putting "i915" in the modules in mkinitcpio.conf, regenerating the initramfs (and putting it where it belongs in the efi partition, remember, I boot efi direct from the kernel), I was able to go as far as the black screen using intel + integrated graphics. Better than before (I had the "no screen found" in the Xorg.0.log). Now, with the black screen no error message whatsoever in the Xorg log. and I can see stuff are happening and my xfce session has started. Just nothing beside black on this vt. I can switch back to other vt ok.

Not much progress. I'd be good when we can use the intel graphics card cause right now this laptop is using about 21W of power. I assume some significant fraction is for the nvidia card, TBC.

frigaut · 2012-07-31 22:21:45

After 24 hours and 20-30 cycles without failure, I confirm it suspends and resumes ok with the following quirk:
pm-suspend --quirk-dpms-on

without it, it would fail to resume once in a while (a problem related to the blank screen issue when switching to vt?).
Again, that's using the nvidia drivers.
[edit] note that this is with the stock (arch) 3.5.0 kernel. with the linux-mainline (rc7) + patched applesmc, resumes will take longer with every suspend, and eventually fails. go figure. [/edit]

Last edited by frigaut (2012-07-31 22:55:34)

frigaut · 2012-08-01 09:44:28

I was able to make some progress with the integrated HD4000. I have it working, and have X running with it, with acceleration. It's just horribly corrupted.
The issue was indeed related to the discrete nvidia card being selected by default, thus preventing the intel to access the display.

Using gfxcardstatus in osx (an app, http://codykrieger.com/gfxCardStatus ), one can force osx to use one or the other graphic chip. The nice thing is that this setting seems to stick till the next reboot (i.e. apple doesn't force back to the nvidia before shutdown). Bootting into linux after selecting the integrated card with gfxcardstatus makes the HD4000 the selected card as seen by X and the intel driver works.

Now the display is corrupted. Corruption occurs in the frame buffers as well as in X (in fact, as soon as i915 is loaded: if I put it in the initramfs, it's corrupted from the start, if I don't it gets corrupted from when it's loaded during the boot process).
This might be related to the following line in dmesg:

[ 4.427250] i915 0000:00:02.0: Invalid ROM contents
[ 4.427252] [drm] failed to find VBIOS tables

It's reporting using the full 2880x1800 resolution, but the color are all mixed up (surprisingly, colors are ok in a screenshot I took with xfce4-screenshoter, look at [http://maumae.net/retina/intel_corrupted.png], that might give a hint to the more savvy). I'm no specialist in debugging modelines in X. I tried a few things (video=some_mode in kernel parameters and setting some modeline as reported by cvt when using nvidia in xorg.conf), but it didn't pan out.
Xorg logs and dmesg for both intel and nvidia available on request for the brave soul that would be willing to work out the correct modeline.

Last edited by frigaut (2012-08-01 09:53:48)

BlueDragonX · 2012-08-01 13:32:27

frigaut wrote:

I was able to make some progress with the integrated HD4000. I have it working, and have X running with it, with acceleration. It's just horribly corrupted.
The issue was indeed related to the discrete nvidia card being selected by default, thus preventing the intel to access the display.

I've been working on implementing support for the MBPr in the nouveau (open source NVIDIA) driver. I suspected this was the issue with the Intel card not working. There's no support for the gmux in Linux yet. I'll be working on that once I'm done with nouveau if nobody has beat me to it.

I'd recommend booting straight to a VT instead of to X while experimenting with KMS on the i915 driver. I've been taking that track with my nouveau experimentation.

What type of screen corruption are you experiencing? Check the dmesg output for the mode the i915 driver is selecting. If it's the correct mode then the link training is likely not being done properly. I'm currently working through that issue myself with nouveau. The Retina display requires around 10Gbps of bandwidth which requires the DisplayPort link channels to be configured for 5.4Gbps. Nouveau doesn't appear to support link training above the DP v1.1 channel speed of 2.7Gbps.

You should be able to enable additional debugging on the i915 module by passing appropriate kernel arguments on boot. I'm not sure what they are, you'll have to look into that it yourself. Doing so should hopefully get you debug info on how the eDP link to the Retina panel is begin configured.

Now, if the incorrect modeline is being set that's another story entirely. I had this issue with nouveau and since fixed it. But it was specific to the NVIDIA card so I doubt this issue will be the same. But the GPIO's weren't being initialized and so the driver could not read auxch data off the panel. As a result the driver assumed there was no display plugged in.

Arch Linux

#1 2012-06-30 08:56:09

MacBook Pro 2012 Retina

#2 2012-06-30 09:17:22

Re: MacBook Pro 2012 Retina

#3 2012-06-30 09:54:57

Re: MacBook Pro 2012 Retina

#4 2012-06-30 10:10:32

Re: MacBook Pro 2012 Retina

#5 2012-06-30 10:14:10

Re: MacBook Pro 2012 Retina

#6 2012-07-03 05:55:08

Re: MacBook Pro 2012 Retina

#7 2012-07-03 06:51:19

Re: MacBook Pro 2012 Retina

#8 2012-07-03 14:14:14

Re: MacBook Pro 2012 Retina

#9 2012-07-04 06:17:50

Re: MacBook Pro 2012 Retina

#10 2012-07-04 07:40:42

Re: MacBook Pro 2012 Retina

#11 2012-07-04 09:56:30

Re: MacBook Pro 2012 Retina

#12 2012-07-14 20:33:45

Re: MacBook Pro 2012 Retina

#13 2012-07-15 07:04:21

Re: MacBook Pro 2012 Retina

#14 2012-07-15 08:43:17

Re: MacBook Pro 2012 Retina

#15 2012-07-31 00:24:17

Re: MacBook Pro 2012 Retina

#16 2012-07-31 04:40:48

Re: MacBook Pro 2012 Retina

#17 2012-07-31 05:23:19

Re: MacBook Pro 2012 Retina

#18 2012-07-31 06:19:57

Re: MacBook Pro 2012 Retina

#19 2012-07-31 06:44:45

Re: MacBook Pro 2012 Retina

#20 2012-07-31 07:43:22

Re: MacBook Pro 2012 Retina

#21 2012-07-31 10:24:13

Re: MacBook Pro 2012 Retina

#22 2012-07-31 13:55:44

Re: MacBook Pro 2012 Retina

#23 2012-07-31 22:21:45

Re: MacBook Pro 2012 Retina

#24 2012-08-01 09:44:28

Re: MacBook Pro 2012 Retina

#25 2012-08-01 13:32:27

Re: MacBook Pro 2012 Retina

Board footer