Opened 7 years ago

Closed 7 years ago

Last modified 7 years ago

#8913 closed bug (fixed)

Monitor turns off booting with HD4870X2

Reported by: jstressman Owned by: kallisti5
Priority: normal Milestone: R1
Component: Drivers/Graphics/radeon_hd Version: R1/Development
Keywords: Cc:
Blocked By: Blocking: #9472
Has a Patch: no Platform: All

Description

This is similar to #8859

hrev44571 gcc2h

Booting with my monitor attached over DVI gets to the rocket and then goes black and switches into power save mode.

Safe mode (up to 1600x1200@32bpp) works fine.

If I attach VGA instead (dvi to vga adapter), then the monitor works correctly, giving me the native resolution of 1920x1200@60hz.

I'm attaching a syslog from the non-working DVI boot (44571log1.txt), and one from the "working" VGA boot (syslog-vga.txt).

Attachments (4)

44571log1.txt (278.3 KB ) - added by jstressman 7 years ago.
non-working DVI connected boot. (goes to blank screen after the rocket)
syslog-vga.txt (290.4 KB ) - added by jstressman 7 years ago.
"working" VGA connected boot. Gives me the correct native resolution etc.
rv770visualaid1.jpg (67.6 KB ) - added by jstressman 7 years ago.
one DVI connector works with Haiku, the other doesn't.
syslog (261.1 KB ) - added by jstressman 7 years ago.
syslog for black screen after rocket with hrev45220

Download all attachments as: .zip

Change History (14)

by jstressman, 7 years ago

Attachment: 44571log1.txt added

non-working DVI connected boot. (goes to blank screen after the rocket)

by jstressman, 7 years ago

Attachment: syslog-vga.txt added

"working" VGA connected boot. Gives me the correct native resolution etc.

comment:1 by kallisti5, 7 years ago

Yeah... having infinite trouble with DVI and this. Here is the issue:

DVI-I Connector --+             EDID digital flag? Digital monitor
| DVI-D (Digital) |            /
| DDC (data)      | --- edid data? monitor attached.
| DVI-A (Analog)  |            \
+-----------------+             Analog load? Analog monitor.    

Since the digital and analog wires share the same edid data line... it is rough figuring out which wires are in use. You can't detect digital load, but you can detect analog load.

It sounds easy to work around, but it seems every combination of code paths I use breaks with at least *one* person :P

comment:2 by jstressman, 7 years ago

Interesting update...

I tried hooking the DVI cable to the other DVI port on the back of my video card, and rebooted... that worked. I switched it back to the one I normally use (which works in linux, am currently using without trouble in Windows, and which worked in Haiku with a VGA adapter) and it didn't work. So one DVI connector works with Haiku, the other DVI connector doesn't.

I'm attaching a picture to make it clear. :)

by jstressman, 7 years ago

Attachment: rv770visualaid1.jpg added

one DVI connector works with Haiku, the other doesn't.

comment:3 by jstressman, 7 years ago

Looks like it broke again with hrev45219 - http://cgit.haiku-os.org/haiku/commit/?id=00bc40ad588f4bd59cd7574788b5dc111b62198a

Attaching a syslog from the failed boot.

(hrev45218 boots fine, hrev45220 goes to black screen after the rocket and the monitor goes into power saving mode)

by jstressman, 7 years ago

Attachment: syslog added

syslog for black screen after rocket with hrev45220

comment:4 by jstressman, 7 years ago

For now I've copied the driver over from hrev45218 and everything seems to work again (at least as much as it did before hrev45219). :)

comment:5 by kallisti5, 7 years ago

Thanks for the syslogs. The largest change is Bill's new duplicate monitor detection. As we were having issues with this kind of thing in the past with your card... i'm guessing it's suspect.

http://cgit.haiku-os.org/haiku/tree/src/add-ons/accelerants/radeon_hd/display.cpp#n244

Bill's code is more graceful... but isn't time tested :)

I'll take a look and compare to what the Linux drm code does.

comment:6 by jstressman, 7 years ago

I'm sad to announce that after hrev45319 ( http://cgit.haiku-os.org/haiku/commit/?id=59dc9ee9552e994240a535467e88d15ee279938b ), it still isn't working.

Still drops to power saving mode after the rocket icon on either DVI connector on hrev45324

Maybe I should setup a build environment on here and we can coordinate on this where I compile and test your changes immediately before committing them to git?

This is a core i7 965EE (hyperthreaded quad 3.2ghz) box with 12GB ram. Should be able to knock out compilation pretty quickly for iterative testing.

comment:7 by kallisti5, 7 years ago

Resolved with hrev45325. Thanks for finding this bug!!!

comment:8 by kallisti5, 7 years ago

Resolution: fixed
Status: newclosed

comment:9 by kallisti5, 7 years ago

Blocking: 9472 added

comment:10 by jstressman, 7 years ago

Just wanted to note that the new driver also works correctly on both DVI connectors now. :)

Thanks again!

Note: See TracTickets for help on using tickets.