I'm also seeing this on 4.20 with amdgpu.  In my case though it also triggers a
problem with amdgpu, a problem that would not be hit if the options were
honored correctly:

[    7.843957] [drm] Display Core initialized with v3.1.68!
[    7.844448] [drm] forcing DP-3 connector on
[   11.785325] [drm:emulated_link_detect [amdgpu]] *ERROR* Failed to read EDID
[   11.785372] [drm] forcing DP-4 connector on
[   11.805994] [drm] Got external EDID base block and 1 extension from
"edid/edid.bin" for connector "DP-4"
[   11.822809] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[   11.822825] [drm] Driver supports precise vblank timestamp query.
[   11.849368] [drm] UVD and UVD ENC initialized successfully.
[   11.950326] [drm] VCE initialized successfully.
[   11.950636] [drm:amdgpu_dm_connector_mode_valid [amdgpu]] *ERROR* No EDID
firmware found on connector: DP-3 ,forcing to OFF!
[   11.950731] [drm:amdgpu_dm_connector_mode_valid [amdgpu]] *ERROR* No EDID
firmware found on connector: DP-3 ,forcing to OFF!
[   11.950782] ------------[ cut here ]------------
[   11.950802] kernel BUG at mm/slub.c:294

Kernel command line:
amdgpu.dc=1 drm_kms_helper.edid_firmware=DP-3:edid/edid.bin
drm_kms_helper.edid_firmware=DP-4:edid/edid.bin video=DP-3:1920x1080D
video=DP-4:1920x1080D

This is for a video broadcast system that "features" output sinks being
connected and disconnected quite frequently; this should be handled seamlessly
but Linux is not honoring the options here.