Closed exuvo closed 2 years ago
hi @exuvo , thanks for reporting this. Weird indeed, can you please share your raw /sys/class/drm/card0/device/pp_od_clk_voltage
, before and after the changes were applied? You are running this? Does the same thing happen on vanilla Arch 5.13.x?
Not sure if i understand what you want, i already included pp_od_clk_voltage which remains the same before and after as the patch is in amdgpu-clocks not the kernel driver? Yes that is the kernel source i am using. I will get back to you after testing with vanilla kernel.
Not sure if i understand what you want
A raw binary copy of /sys/class/drm/card0/device/pp_od_clk_voltage
Okay but that data is already in the xxd code block. I had to give it a file ending for the upload to work here. pp_od_clk_voltage.txt
@exuvo thanks for sending the raw file, I just wanted to be 200% sure about these nulls @ pp_od_clk_voltage when I do tests on my side. Looks like mapfile
gets totally confused when dealing with these nulls, so some tr prepping in the pipe was added.
The latest commit should had fixed the issue, please verify on your side.
Any specific reason you went with tr instead of the sed line i suggested as a fix?
not really
I am having a wierd minor problem with reading pp_od_clk_voltage and it having a lot of null characters that break parsing. A simple cat does not show it:
But xxd shows it. Before OD_VDDC_CURVE there is a lot of null characters and this seems to make the case for OD_VDDC_CURVE to be missed and it then parses those values as if they were OD_MCLK.
I then get some invalid argument lines as it tries to write a mclk state with voltage which is not allowed on my GPU RX 5700 XT.
I managed to fix it by filtering out the nulls at line 61 in
function_parse_states
:GPU RX 5700 XT Kernel linux-zen 5.15.13 OS Arch