[Bug 203325] New: 5.1 amdgpu on polaris11: broken text-fb multihead, works in X

[bugzilla-daemon at bugzilla.kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=203325

            Bug ID: 203325
           Summary: 5.1 amdgpu on polaris11: broken text-fb multihead,
                    works in X
           Product: Drivers
           Version: 2.5
    Kernel Version: 5.1
          Hardware: x86-64
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: Video(DRI - non Intel)
          Assignee: drivers_video-dri at kernel-bugs.osdl.org
          Reporter: 1i5t5.duncan at cox.net
        Regression: Yes

Created attachment 282339
  --> https://bugzilla.kernel.org/attachment.cgi?id=282339&action=edit
kernel config

>From 5.1-rc1, my multi-head amdgpu polaris11 system will often only display on
one head in (non-X) text frame-buffer mode.  The not-working head stays on, but
blank.  It worked fine in 5.0, and is broken at least thru 5.1-rc5.

The initial DRI display will come up on both heads if both TV/monitors are on
at kernel load, but even then, if I start X (with KDE/Plasma), where it
displays correctly on both monitors regardless of what it does outside X, and
ctrl-alt-Fn back to a different VT still in text-fb mode, or quit X, it'll
often (always?) only display the text-fb mode on one instead of both.  Which
one seems somewhat random tho it might be consistent with which one, or both,
were on at amdgpu driver load, as it seems (mostly? entirely?) consistent for a
single boot.

At a guess it might be trying to use the same clock rate and etc settings for
both outputs at the same resolution, but one needs slightly different settings.
 X presumably programs the different outputs correctly, so it works there, and
as I said it worked correctly in text-fb mode in 5.0.

dmesg reports:
[drm] initializing kernel modesetting (POLARIS11 0x1002:0x67EF 0x1458:0x22DD
0xCF).
...
[drm] Display Core initialized with v3.2.17!
[drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[drm] Driver supports precise vblank timestamp query.
...
fbcon: amdgpudrmfb (fb0) is primary device
Console: switching to colour frame buffer device 320x108
amdgpu 0000:01:00.0: fb0: amdgpudrmfb frame buffer device
[drm] Initialized amdgpu 3.30.0 20150101 for 0000:01:00.0 on minor 0


Xorg's log reports:
AMDGPU(0): Chipset: "AMD Radeon (TM) RX 460 Graphics" (ChipID = 0x67ef)
...
AMDGPU(0): glamor X acceleration enabled on AMD Radeon (TM) RX 460 Graphics
(POLARIS11, DRM 3.30.0, 5.1.0-rc5-dirty, LLVM 8.0.0)


I have the following CONFIG_CMDLINE:

CONFIG_CMDLINE="rootwait net.ifnames=0 video=HDMI-A-1:1920x1080
video=DVI-D-1:1920x1080 video=DisplayPort-1:1920x1080"

But only two of the three video ports are actually connected, HDMI-A-1 (kernel,
X says 0 not 1) and DisplayPort-1.

HDMI-A-1 is actually connected to a 4k TV and runs at its native 3840x2160 in
X, but as you can see it's set for 1920x1080 in the text-framebuffer. 
DisplayPort-1 is a full-HD TV, thus native 1920x1080.

Of note, the cable for the full-HD TV is HDMI/DVI-D single-link, TV-side HDMI,
with a DVI-D/DisplayPort adaptor, so I can connect it via DVI-D or DisplayPort.
 As noted, it's currently connected via DisplayPort.

I'm attaching kernel config.  I'll add boot dmesg.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.