Built-in display doesn't work with AMDGPU on Linux 5.6 Macbook (MacBookPro13,3) #142
Comments
|
Yes, the amdgpu driver apparently got broken in the 5.6 kernel - I noticed because the I haven't had time to file a bug report, but it would be a good idea to do so. |
|
I am facing same problem on fc31 after last dnf upgrade which installed kernel-5.6 |
|
Linux kernel 5.7.0-050700rc3-generic My eDP (builtin display) has no modes for 5.6+ xrandr --query | grep '\bconnected\b' xrandr: Output eDP is not disconnected but has no modes
eDP connected (normal left inverted right x axis y axis)
DisplayPort-0 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 521mm x 293mm |
|
I've also opened a bug for Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=1829174 |
|
I've posted a question at [email protected] and got the following reply: So I'll try to do a bisect in the next days... |
|
I've bisected the problem to: b9f1246df179522bc28fda50b720553c845863db is the first bad commit .../gpu/drm/amd/display/dc/dcn21/dcn21_resource.c | 20 ++++++++++---------- |
|
You guys might have luck with this patch. Its a quick and dirty hack to force all DPCD link rate values to whats expected by Retina panels. Currently, this will prevent externally attached displays from working. |
Thanks I'll try that! Could you please report your findings at: |
|
I have reported my findings with the bisected kernel commits between v5.5.0 and v5.6.0 at https://bugzilla.kernel.org/show_bug.cgi?id=207589 with the related Xorg and dmesg logs. |
|
Hi @aunali1 thanks for posting your patch, that seems to fix it on my system! Updated URL: |
|
Is somebody currently working to get a fix for that upstreamed? What's the status there? |
|
@dmarinuswoodwing Thanks for testing! Great to see that it works for other models aswell. @Dunedan Unfortunately, this patch is not upstream ready since it causes issues with 16inch MBPs. Additional testing is also needed for other models. |
|
@aunali1 is there an overview of models that need to be tested? What kind of issues do you have with 16" MBP's? |
|
It seems that just an hour after I made the last comment, @marcosfad tested it again on his 16inch model and it works without issue! @dmarinuswoodwing Regarding other models, I would prob say the MacBookPro14,x series. |
|
The patch https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=dec9de2ada523b344eb2428abfedf9d6cd0a0029 is included in the upstream as commit 69d3e5a5a66bb59c39f36dcb9cf4e9a4239aa8cd. I am checking out latest commit from Linux kernel master branch and compiling it and I will test it on my Linux Mint distro. |
|
I can confirm the built-in display is working for kernel version 5.7.10 where the patch is included in the commit 69d3e5a5a66bb59c39f36dcb9cf4e9a4239aa8cd from the Linux kernel master branch. |
|
@dmarinuswoodwing time to close the issue? :) |
|
I have verified that the MBP 13,3 built-in display is working for Linux version 5.8.0-rc6+ from Linux master branch with commit 8c26c87b05323a7ccdc632820b85253e0bf47fd9 (HEAD -> master, origin/master, origin/HEAD) https://termbin.com/uby9 |
|
This issue is still open, it should be closed since the fix is in the upstream master branch since kernel v5.7. |
|
It might make sense to atleast give some notice about what kernels are impacted on https://github.com/Dunedan/mbp-2016-linux#amd before closing this issue. |
|
@dmarinuswoodwing after bisecting the main source tree https://bugzilla.kernel.org/show_bug.cgi?id=207589#c18, the built-in display stopped working for me on commit 4a8ca46bae8affba063aabac85a0b1401ba810a3 and they said it was fixed for commit 473e3f7720f652716eb7a5b8f49628d6f66b1e87 https://bugzilla.kernel.org/show_bug.cgi?id=207589#c21
So built-in display stopped working for kernel version v5.6 and started to work again for version v5.7 (since they include the release candidates tags) |
|
Does one of you want to open a PR for that? If not I can do it as well. |
|
I can do it to keep track of history kernel changes that have made impact of the built-in display for MBP model 13,3 unless anyone else is up for it. Just be sure to involve as reviewers before merge. |
|
I will test Linux kernel v5.7 from upstream and report my findings for commit 3d77e6a8804abcc0504c904bd6e5cdf3a5cf8162 (HEAD, tag: v5.7) and if it is working I will create new PR and update the README.md accordingly if no one else has done it by then and verified what commit works. |
|
Ok, after a few testing, the kernel version 5.7.8 was the one where the built-in display started to work again for MBP model 13,3, however the command
I know I tried compiling v5.6 before, the commit hash mentioned above that is included in that specific tag does not seems to fix the problem even if the developers says it does https://bugzilla.kernel.org/show_bug.cgi?id=207589#c21 what am I missing? I just want to be sure we pinpoint the actual proper fix tag versions before mentioning it in the README.md |
|
It seems like commit 639e0db2d70fb84833d96e782cc4a01825e03b13 is actually the commit that fixes the built-in display issue for MBP model 13,3.
v5.8 `commit 639e0db2d70fb84833d96e782cc4a01825e03b13 ` So the kernel version 5.7.10 I am currently running must include that patch somehow. Will try compiling that specific commit and the one before. |
|
@fannullone you should close the bug report at https://bugzilla.kernel.org/show_bug.cgi?id=207589 |
|
I'm on a MacBook Pro 13,3 and can confirm that built-in display works fine on @andyholst quick question, I can no longer run I don't understand what |
|
@cristianmiranda create a new issue of the details you are having problem with the amdgpu driver. I am currently running Linux kernel version 5.7.14 and |
|
@andyholst You're right. I've opened a new issue for this. Thanks. #159 |
|
@Dunedan time to finally close this issue? |
|
Yes 👍 |
When I update to the latest kernel (5.6.6) using Fedora 31 my built-in display shows a (backlighted) black screen.
Could someone please share their experience with a AMDGPU Macbook and linux-5.6?
The text was updated successfully, but these errors were encountered: