Patrick wrote:
Hi all,
I downloaded FC6T2 x86_64 to give it a spin on an Acer 4005WMLi which
has an ATI X700 PCIE card. The ATI card is detected properly but X
selects the external video port (called "CRTC1" in /tmp/X.log) instead
of the internal one (LVDS) which drives the laptop screen. Since there
is no screen attached to the external port I only get a black screen and
no graphical install.
After installing FC6T2 in text mode due to lack of an external screen I
had to add the Option "MonitorLayout" "LVDS,NONE" to the videocard0
device section in /etc/xorg.conf to get X going. Seems the radeon driver
is rather fond of that external video port.
The Radeon driver's display autodetection/selection code is very
broken and has been for a while now. The MonitorLayout directive
exists to force the driver to override what the broken detection
logic uses by default.
Is there any way to force the installer to use the internal LVDS video
port instead of the external CRTC1 video port?
MonitorLayout is the only way.
Or should I bz it? If so under anaconda or xorg-x11-drv-ati, something else?
Not really. The problem is that in order for everything to "just work",
_something_ has to be able to probe the video card and correctly
autodetect what attached displays are present, and also correctly
assign them. That really is the job of the video driver, and it is
not doing that properly right now. The installer does not have any
low level knowledge of the internals of every video card, so this is
totally the responsibility of the video drivers to properly detect
attached video hardware.
If the drivers do not detect the hardware properly, then the hardware
can not easily be set up to work properly with the attached displays.
There is some confusion by people over MonitorLayout as well. Many
people have to use the MonitorLayout directive in order to have working
video at all, and so think that anaconda should use MonitorLayout by
default so they have working video. The problem is that MonitorLayout
is *only* required when the driver does not detect the hardware
properly, and when that is the case, there is no way for anaconda to
know what exact hardware is connected, or how it should specify the
MonitorLayout option as it varies from hardware to hardware.
As a result, there are only 2 options at this time:
1) Users have to manually insert MonitorLayout directives into their
config file themselves, as it is not something that is
currently autodetectable, and if it was, then there would in fact
be no need for the MonitorLayout directive itself. Keep in mind
this directive is a *hack* to work around display detection code
which is broken and does not work properly on all hardware.
2) Wait until someone fixes the radeon driver to properly detect
all attached displays on all Radeon hardware, and it works 100%
of the time on _ALL_ Radeons. In which case the MonitorLayout
directive should essentially become obsolete and no longer required.
Since these flaws in the radeon driver are already very widely known
both to Red Hat and to X.Org upstream, and have been reported by many
people in bugzilla over the last year or so, it's not particularly
useful IMHO to file another bug report which will just get closed as
a duplicate of an existing bug...
Instead, query X.Org bugzilla for an existing report which has various
dupes closed against it, and CC yourself on that bug to track any
progress.
I'm not sure if anyone upstream is currently working on a solution to
this long standing problem or not, but it'll likely be sorted out
eventually. Until then, MonitorLayout specified manually by the user
is the only option, or sticking with a much older (and no longer
supported) OS release that predates the display detection code going
to hell. ;o)
--
Mike A. Harris * Open Source Advocate * http://mharris.ca
Linux fans: Check out Tym Morrison's hit new heavy metal single
"Only Linux" at http://tymmorrison.com - If you would like to support
this great Canadian metal artist and open source fanatic, you can buy
a copy of Tym's Solo Project CD at the "Buy CD" link on his site.
--
fedora-test-list mailing list
fedora-test-list@xxxxxxxxxx
To unsubscribe:
https://www.redhat.com/mailman/listinfo/fedora-test-list