| View previous topic :: View next topic |
| Author |
Message |
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
 Posted: Mon Mar 25, 2019 11:24 pm Post subject: |
 |
|
not_a_robot,
As you say ... very interesting. Lets start with the Xorg log.
| Code: | [ 17.937] (==) Matched modesetting as autoconfigured driver 0
[ 17.937] (==) Matched fbdev as autoconfigured driver 1
[ 17.937] (==) Matched vesa as autoconfigured driver 2 |
The intel driver is not even considered as an option for Xorg.
It goes on
| Code: | [ 18.942] (II) modeset(0): glamor X acceleration enabled on Mesa DRI Intel(R) HD Graphics (Coffeelake 3x8 GT2)
[ 18.942] (II) modeset(0): glamor initialized
[ 18.990] (II) modeset(0): Output DP-1 has no monitor section
[ 18.990] (II) modeset(0): Output DP-2 has no monitor section
[ 18.998] (II) modeset(0): Output HDMI-1 has no monitor section
[ 19.046] (II) modeset(0): EDID for output DP-1
[ 19.048] (II) modeset(0): Manufacturer: SAM Model: f13 Serial#: 16780800
[ 19.048] (II) modeset(0): Year: 2018 Week: 1 |
to detect a SAMsung display on the Display Port 1 connector. Check the serial number on the label, if its 16780800, that's your display :)
It was made in the first week of January 2018.
| Code: | | [ 19.048] (II) modeset(0): Max Image Size [cm]: horiz.: 142 vert.: 80 |
That's a better size too. Its cm, not mm this time.
Lastly, there is
| Code: | [ 19.048] (II) modeset(0): Printing probed modes for output DP-1
[ 19.048] (II) modeset(0): Modeline "3840x2160"x30.0 297.00 3840 4016 4104 4400 2160 2168 2178 2250 +hsync +vsync (67.5 kHz eP) |
which is the expected 4k modeline.
In short, Xorg worked as expected. Now dmesg.
| Code: | [ 4.855363] efifb: framebuffer at 0x90000000, using 3072k, total 3072k
[ 4.855364] efifb: mode is 1024x768x32, linelength=4096, pages=1 |
It starts in 1024x768 mode. Ahhh. Thats where your strange EDID comes from. Its the capabilities of the EFI framebuffer.
| Code: | [ 6.122931] Console: switching to colour frame buffer device 480x135
[ 6.164024] i915 0000:00:02.0: fb0: inteldrmfb frame buffer device |
shows that the inteldrmfb is in use for the console.
I don't see the resolution in use by the inteldrmfb frame buffer.
The takeaway is that the shape of the connector that the display is plugged into in not always related to the electronics driving the port.
Your video card sees the display as connected to Display Port 1, even if its an HDMI connector.
This suggests the following approach.
Don't force anything on the kernel as it boots. That worked on Ubuntu. Maybe it works on Gentoo now too.
By setting nomodeset, the autodetection was turmed off, so whatever grub did was carried through to Xorg.
That's why it was 1024x768. Your forcing of HDMI-1 was ignored too.
We know that your display provides correct EDID data. If the above does not work. Try forcing each output in turn but do not provide any EDID data.
Xorg says | Code: | [ 18.990] (II) modeset(0): Output DP-1 has no monitor section
[ 18.990] (II) modeset(0): Output DP-2 has no monitor section
[ 18.998] (II) modeset(0): Output HDMI-1 has no monitor section |
So the electronics knows of three outputs, that's regardless of the number of physical ports on the card.
I suspect that Gentoo will report your display on DP-1 when it works. Its the same codebase as Ubuntu.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
 |
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Tue Mar 26, 2019 10:03 am Post subject: |
|
|
not_a_robot,
| Code: | [ 9.614] (--) intel(0): Output DP1 using initial mode 3840x2160 on pipe 0
[ 9.614] (==) intel(0): TearFree enabled |
It still thinks its using DP1. TearFree enabled means double buffering is turned on.
That's the Intel driver in use too.
What does "Goes to Sleep mean"?
| Code: | | [ 9.618] (II) Initializing extension DPMS |
Is Display Power Management being enabled.
The screen will go blank after a period of inactivity and go to standby sometime later.
Pressing a key or moving the mouse will wake it up. Wakeup from blanking is very fast. Wakeup from sleep takes a lot longer, my system takes about 30 sec.
DPMS can be fine tuned or disabled if you want.
The page says
| Code: | Option "BlankTime" "time"
sets the inactivity timeout for the blank phase of the screen‐
saver. time is in minutes. This is equivalent to the Xorg
server's -s flag, and the value can be changed at run-time with
xset(1). Default: 10 minutes. |
The automatics do a fairly good job oy setting up a single display system but there are lots of knobs to play with.
In days of old, with CRT based displays it was possible to destroy your display. With flat panels, the worst that happens is you get no picture or a "Video out of range" message.
The display may have its own DPMS too.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Tue Mar 26, 2019 1:26 pm Post subject: |
|
|
| NeddySeagoon wrote: |
What does "Goes to Sleep mean"?
|
The screen goes black, and eventually displays "No Signal" and returns to some kind of standby mode where it displays info about the TV (this is also what happens if I turn the TV on with nothing connected).
Not sure what to actually call this, so I've been referring to it as "cutting out" or "going to sleep" throughout the thread. I guess the signal just drops? |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Tue Mar 26, 2019 2:24 pm Post subject: |
|
|
not_a_robot,
Essentially, yes. There are three parts to the signal, this dates from the analogue days.
The Vertical Sync, used to signal the end of a frame.
The Horizontal Sync, used to signal the end of a line.
The Video, which is the picture you see.
When the video is removed (blanked), there is no picture. Everything is operating normally and mouse movement or a keystroke should remove the blanking signal.
In this state, the display can't tell its just the video missing, so you won't get any messages generated by the display itself.
Standby is trigged by sending something other than syncs on the sync lines. The display detects this and should react accordingly.
Now you may get "No Signal", if the TV does not go into standby.
Both of these states are normal reduced power operation. The system should wake up in response to the mouse or keyboard.
Does it?
How long does the screen take to go black?
How long does it take for the "No Signal" message to appear?
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Tue Mar 26, 2019 2:38 pm Post subject: |
|
|
The display goes black instantly - I barely glimpse the "fb0: switching..." message.
The TV's own waiting animation (a black screen with some dots rotating in a circle) starts immedately after, and then goes to a screen that says "No Signal" after a few seconds.
Pressing the keyboard does nothing - though my keystrokes are registered, and I'm able to log in, get dmesg output, etc. by typing blind.
I have a second motherboard, identical to this one, arriving today. I've already replaced the CPU.
When I hook it up to a 1080p projector, there is no tearing in ubuntu. |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Tue Mar 26, 2019 2:59 pm Post subject: |
|
|
not_a_robot,
I'n not worried about the tearing. That will be missing options on mesa and maybe a few other packages, which together provides the video acceleration.
To fix mesa and friends, you need to set the VIDEO_CARDS in make.conf, then rebuild affected packages.
We have been doing the bare minimum to get video drivers installed, without worrying about all the trimmings.
Set your VIDEO_CARDS in make.conf, then run | Code: | | emerge -uDNav @world --keep-going |
-- edit --
Clicked submit too soon.
It sounds like the framebuffer is coming up on the wrong output.
Try forcing each output in turn. One of them must be right.
Xorg thinks its using DP-1 and it claims to be reading EDID data from there too.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Tue Mar 26, 2019 4:09 pm Post subject: |
|
|
Ok, ran emerge, and tried disabling each display in turn.
DP-2 and HDMI have the same behavior as I've seen up to now.
Forcing DP-1 is interesting. The screen freezes at the boot message "fb0: switching to inteldrmfb". The system otherwise behaves as normal, I can log in and perform commands by typing blind. When I change terminals with Alt-Fn, the screen goes black, but does not go to "No Signal".
DP-1 is the only one that seems to get EDID, but DP-2 doesn't even attempt any resolution, while HDMI attempts 1024x768.
xorg logs:
for DP-1: http://dpaste.com/03WAAXD
for DP-2: http://dpaste.com/2SB3CE2
for HDMI: http://dpaste.com/017Z2XZ |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Tue Mar 26, 2019 5:25 pm Post subject: |
|
|
not_a_robot,
What Jaglover said, but the vendor and device ID were listed in the kernel, so the Intel framebuffer console should work, even if Xorg doesn't.
| Quote: | | ="not_a_robot"The screen freezes at the boot message "fb0: switching to inteldrmfb". |
Thats the last thing written to the EFI framebuffer before the switch to inteldrmfb, as if inteldrmfb was drawing on a different output.
As its still on the screen, either inteldrmfb is drawing to a different region of memory, that is not being displayed, or its drawing to a different region of memory that is being displayed on another output.
Nothing ties the memory region being drawn to the one being displayed The two are deliberately separate, or double buffering would be difficult.
As Xorg claims you have three outputs, what I have in mind is to send the image to all three outputs at the same time.
That's called clone mode. It may fix Xorg but won't do anything for the console.
I'll need to read up about it as it will involve writing an xorg.conf file and its been a while since I've done that.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Tue Mar 26, 2019 8:05 pm Post subject: |
|
|
Now this is interesting:
I enabled all 3 outputs this time, and my screen went black as usual, but then I logged in, ran startx, and the screen came back!
Here is xorg log: http://dpaste.com/3HQ428G
I still don't see 4K mode under HDMI! But my TV does confirm (via an "info" button on the remote) that it is at 1024x768@60Hz.
Seems the 4K modes are only under DP-1 and have some kind of timing issue. |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Tue Mar 26, 2019 8:58 pm Post subject: |
|
|
not_a_robot,
This is making my head hurt again.
| Code: | [ 16.904] (--) intel(0): Output DP2 using initial mode 1024x768 on pipe 0
[ 16.904] (--) intel(0): Output HDMI1 using initial mode 1024x768 on pipe 0
...
[ 16.945] (II) intel(0): EDID for output DP1
[ 16.945] (II) intel(0): Manufacturer: SAM Model: f13 Serial#: 16780800
...
[ 16.946] (II) intel(0): Printing DDC gathered Modelines:
[ 16.946] (II) intel(0): Modeline "3840x2160"x0.0 297.00 3840 4016 4104 4400 2160 2168 2178 2250 +hsync +vsync (67.5 kHz eP)
...
[ 17.041] (II) intel(0): Output DP1 connected
[ 17.041] (II) intel(0): Output DP2 connected
[ 17.041] (II) intel(0): Output HDMI1 connected
[ 17.041] (II) intel(0): Output VIRTUAL1 disconnected
[ 17.041] (II) intel(0): Using fuzzy aspect match for initial modes
[ 17.041] (II) intel(0): Output DP1 using initial mode 1024x768 +0+0
[ 17.041] (II) intel(0): Output DP2 using initial mode 1024x768 +0+0
[ 17.041] (II) intel(0): Output HDMI1 using initial mode 1024x768 +0+0
|
so all outputs are using the best resolution they have in common anh your display is connected to one of them but we can't tell which.
I can believe the EDID data is on one and the video is on another though.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Tue Mar 26, 2019 9:03 pm Post subject: |
|
|
Could it be an issue with the onboard video of this particular mobo? I have a replacement that just arrived - I'd rather send it back without opening it but I can give it a try if these problems could be due to a fault in this specific board.
By the way, xrandr --listmonitors says:
| Code: |
Monitors: 2
0: +*DP1 1024/1420x768/800+0+0 DP1
1: +DP2 1024/271x768/203+0+0 DP2 HDMI1
|
DP2 and HDMI are the same output? And DP1 is my primary display?
Also, DP1 is the only one with correct screen size. 1420x800mm instead of that weird 271x203 popping up again for DP2 and HDMI1.
| Code: | | xrandr --ouput HDMI1 --auto |
has no visible effect.
| Code: | | xrandr --ouput DP1 --auto |
causes the display to cut out.
Switching output back to HDMI1, and the display remains off.
Looks like we are outputting to HDMI1, after all, but EDID is being received over DP1? |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Thu Mar 28, 2019 4:07 am Post subject: |
|
|
Another interesting thing - 2nd to last line of Xorg log is (EE) intel(0): failed to set mode: Invalid argument
But this is for when startx works, and switches to 1024x768
Is it possible that the intel driver just doesn't support the 4K@60Hz output mode? |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Thu Mar 28, 2019 7:33 pm Post subject: |
|
|
| What is a live driver? Is that like the nightly build or something? And I would need i915, right? |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Thu Mar 28, 2019 8:59 pm Post subject: |
|
|
not_a_robot,
Live, otherwise -9999 version ebuilds pull directly from the upstream project.
You can get commit by commit access, so live, as it happens.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
not_a_robot
n00b
Joined: 10 Mar 2019
Posts: 36
|
| Posted: Fri Mar 29, 2019 5:52 am Post subject: |
|
|
Nice! That did something. Here's the diff on my last 2 Xorg logs:
| Code: |
2c2
< (WW) intel(0): Unknown chipset
---
> (--) intel(0): gen9 engineering sample
381c381
< (II) intel(0): SNA initialized with disabled backend
---
> (II) intel(0): SNA initialized with Coffeelake (gen9) backend
387,388d386
< (II) intel(0): Textured video not supported on this hardware or backend
< (WW) intel(0): loading DRI2 whilst acceleration is disabled.
|
Sadly, display still goes to No Signal, and HDMI still reports 1024x768 max.
Full xorg log: http://dpaste.com/3QDFFGK |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Fri Mar 29, 2019 11:02 pm Post subject: |
|
|
not_a_robot,
We will try doing it the hard way. I'll write an xorg.conf file but not tonight.
I'll need to do some reading of
Its a while since I've configured Xorg by hand.
What I have in mind is to tell Xorg you have three identical displays connected, even though you don't.
I expect Xorg to draw the same image on all three but we won't know which in the one actually in use.
If there is an image, we can do some trial and error to turn them off one at a time.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54243
Location: 56N 3W
|
| Posted: Sat Mar 30, 2019 1:36 pm Post subject: |
|
|
Jaglover,
Its also possible to use xrandr to fiddle with outputs. Thats a lot easier than a manual xorg.conf
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Jaglover
Watchman
Joined: 29 May 2005
Posts: 8291
Location: Saint Amant, Acadiana
|
|
| Back to top |
|
|
|
Kernel & Hardware
