You are not logged in.
Hi,
I'm an old Arch user, and usually i try to fix my problems by myself. But, my current situation is not what i expected, and now need help of more advanced users.
I have a (maybe old) Lenovo Laptop AMD A10-7300 with Dual Graphics with R6 Kavieri and R5 M230.
I think these cards belongs to GCN 1.1 and i use amdgpu driver. (Kerner parameters for cik support works pretty well).
A few months ago was pretty excited because amdgpu.dc=1 started to work well in kernel 4.17, because my built-in display (eDP) had several banding issued and amdgpu.dc option fixed them.
However, after kernel 4.18 update until today (updated to 4.19.1), the inclusion of amdgpu.dc=1 (if spite of generating several errors) produced a black screen when booting the system.
The weird thing is, if i boot arch with a HDMI display connected (or if i connect the HDMI display after boot), the eDP leaves their black screen state and works normally. Obviously, this behaviour is unwanted, because i don't have always a HDMI screen available (only at work), i have to use amdgpu.dc=0 option to properly use arch, but the banding issue returns.
I've read a lot of stuff about these amdgpu problems (since kernel 4.15) but now, i pretty confused about all what is happening behind them and i want to fix this stuff. There's some patches for similar cases in Macbooks running ubuntu systems, but i don't know how kernel development works in Arch, and if there is somebody with similar issues to me.
I've read the recent entries or posts on others AMD Graphics Cards (Vega 64, or HD 7850) but no answers to activate mi eDP without the need of an HDMI display for use the amdgpu.dc option.
I don't know how to start, and with what log, because i don't know how to obtain logs from "black screen" boots (i do hard resets). But if somebody can help me, i can provide all information desired as soon as possible.
Last edited by nikofunk (2019-01-01 23:20:41)
Offline
Hi,
I can't help you with your problem, however you can obtain bootlogs from previous boots with "journalctl".
"journalctl -b -1" shows the log of the last boot, for example.
Hope that helps to solve your troubles.
Offline
Hi,
I can't help you with your problem, however you can obtain bootlogs from previous boots with "journalctl".
"journalctl -b -1" shows the log of the last boot, for example.
Hope that helps to solve your troubles.
Thank you, i updated today to 4.19.1. And now, the system hangs completely with amdgpu.dc=1.
No journalctl -b -X entry is found with Kernel 4.19.1 with amdgpu.dc=1.
For sure, there's must be another thing to found in all this stuff.
Offline
I get a kernel panic on AMD Stoney, I have to boot with with amdgpu.dc=0
Also I'm sure that command should be amdgpu.dc=1 and not amdgpu.gc=1 https://wiki.archlinux.org/index.php/AMDGPU#AMD_DC
Offline
Thank you, i updated today to 4.19.1. And now, the system hangs completely with amdgpu.dc=1.
No journalctl -b -X entry is found with Kernel 4.19.1 with amdgpu.dc=1.
For sure, there's must be another thing to found in all this stuff.
There are at least 2 system-breaking blackscreen amdgpu 4.19 bugs. Trying more to get "journalctl" output as knotic said is worth it. I've found that I need to give my system a minute before power cycling, or the appropriate part of the logs aren't written. I'm guessing this is because of running btrfs which has 30 second commit intervals. I don't think you'll see output from either bug unless you look through "journalctl" after being able to boot again.
There's one that starts like this. See: https://bugs.freedesktop.org/show_bug.cgi?id=108781
[drm] Invalid PCC GPIO: 13!
...
[drm:dm_pp_get_static_clocks [amdgpu]] *ERROR* DM_PPLIB: invalid powerlevel state: 0!
[drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!I'm working on bisecting, and there's currently no fix posted upstream. There's hundreds/a thousand commits that won't compile which slowed me down.
If you're lucky, there's another that has a supposedly working patch, that starts like this. See: https://bugs.freedesktop.org/show_bug.cgi?id=108704
[ 2.700664] amdgpu: [powerplay] Failed to retrieve minimum clocks.
[ 2.700665] amdgpu: [powerplay] Error in phm_get_clock_info
[ 2.700922] [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!Offline
Same issue to me when I tried to boot with amdgpu.dc=1
No issues without it, but I got back to radeon driver since it's less problematic and amdgpu has some issues when GPU decoding is active.
Offline
I get a kernel panic on AMD Stoney, I have to boot with with amdgpu.dc=0
Also I'm sure that command should be amdgpu.dc=1 and not amdgpu.gc=1 https://wiki.archlinux.org/index.php/AMDGPU#AMD_DC
indeed, i know its amdgpu.dc , maybe i've misspelled and then i edited.
Well, still in black screen with amdgpu.cik support and amdgpu.dc=1.
Offline
nikofunk wrote:Thank you, i updated today to 4.19.1. And now, the system hangs completely with amdgpu.dc=1.
No journalctl -b -X entry is found with Kernel 4.19.1 with amdgpu.dc=1.
For sure, there's must be another thing to found in all this stuff.
There are at least 2 system-breaking blackscreen amdgpu 4.19 bugs. Trying more to get "journalctl" output as knotic said is worth it. I've found that I need to give my system a minute before power cycling, or the appropriate part of the logs aren't written. I'm guessing this is because of running btrfs which has 30 second commit intervals. I don't think you'll see output from either bug unless you look through "journalctl" after being able to boot again.
There's one that starts like this. See: https://bugs.freedesktop.org/show_bug.cgi?id=108781
[drm] Invalid PCC GPIO: 13! ... [drm:dm_pp_get_static_clocks [amdgpu]] *ERROR* DM_PPLIB: invalid powerlevel state: 0! [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!I'm working on bisecting, and there's currently no fix posted upstream. There's hundreds/a thousand commits that won't compile which slowed me down.
If you're lucky, there's another that has a supposedly working patch, that starts like this. See: https://bugs.freedesktop.org/show_bug.cgi?id=108704
[ 2.700664] amdgpu: [powerplay] Failed to retrieve minimum clocks. [ 2.700665] amdgpu: [powerplay] Error in phm_get_clock_info [ 2.700922] [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4!
thanks! i've also read a lot about the problem, and several linux users gave up because of the huge bisecting work. At least the bugs are being reported quickly!
Offline
The patch from https://bugs.freedesktop.org/show_bug.cgi?id=108704 was applied in 4.19.7
Offline
The patch from https://bugs.freedesktop.org/show_bug.cgi?id=108704 was applied in 4.19.7
In Kernel 4.20 all bugs were fixed. I've marked as Solved this topic.
Offline