Bug#207543: you patch to restrict PCI device resets in XFree86
having talked with Branden on irc i try to fill in some blanks as
good as i know the content and history of the project ...
* Aivils (aivils@unibanka.lv) [040402 13:08]:
> Here is a possibility add very doubtful thing. Nobody knows how to run
> secondary video adapter and secondary xf86 before tests.
only testing a specific setup will show if it really works
> I have good experience with Nvidia adapters in company with Nvidia closed
> source drivers, which set allow usage of GLX on all local xf86 servers and
> VGA text mode console on 1st xf86. Even this experiece do not allow me write
> "allways success".
propriatary Nvidia drivers work mostly, but not allways, as the
device associated with the inital console.
> My current video set (GeForceMX440, GeForceFX5200PCI, TNT2PCI) will not work
> with this "isolateDevice" option, but does oops inside nvidia kernel module
> during 3rd xf86 start, both 1st and 2nd xf86 stay alive. Similar story can
> write any , which try to start multiple local xf86.
Aivils test-box with three heads does not work with the patch in
question but oopses in the propriatary nvidia driver during
initialisation of it`s third head. This problem occures when
more then one local xfree is run.
(a head is here the monitor + graphics card + keyboard + mouse
combo used when running multi-user-workstations. not the same as
in xinerama).
> Video card driver should correct use libint10.a libvgahw.a for adapter
> initialization, save/restore text mode. Seems condition fulfil for single
> xf86, but fail for multiple local xf86.
graphics device drivers must use libint10a and libvgahw.a
correctly to initialize the hardware and to save/restore text
mode. That it works well for one xf86 does not necessary mean
that it does for several.
> Video driver, libint10.a and vgahw.a
> analyzing is to hard for me and i use this patch as is.
without thoroughly understanding both libint10.a, vgahw.a and the driver
logics, i still use this patch. (?)
> Known are matrox, ati open source drivers which support multiple local xf86,
> whith various troubles to start hardware DRI.
drivers which are known to work relativly well are matrox and
ati. DRI works only in exceptions.
> Succesful xinerama start does not correlate with successful multiple local
> xf86 start, even if hardware set is identic and unchanged.
even if xinerama works, multiple local Xservers with this patch
can or can not work.
Reply to: