Bug#522605: xserver-xorg-video-mga: Fatal server error: xf86MapDomainMem(): mmap() failure
On Sun, 2009-04-05 at 22:26 +1000, Shaddy Baddah wrote:
> First up, contents of /proc/version:
>
> Linux version 2.6.26-1-sparc64 (Debian 2.6.26-13lenny2)
> (dannf@debian.org) (gcc version 4.1.3 20080704 (prerelease) (Debian
> 4.1.2-25)) #1 Fri Mar 13 17:42:29 UTC 2009
>
weird, i would have expected things to work with that version, since the
problematic kernel change was reverted for lenny. maybe it's a
different bug after all?
> In terms of case by case approach, I followed the bug around the time
> that it was raised/addressed. Here is a summary of my understanding:
> * from message #119, Bastian Blank of the kernel team downgraded the
> severity of a let's call it "generic X.org on sparc problem", stating:
>
> 'There is only a small fraction of machines affected, so this is not RC.'
>
> * my reading is that he was persuaded to give the bug a greater
> severity when it was intimated that MACH64 being broken on lenny was
> going to cause a problem for (paraphrasing) a large fraction of Sun
> Ultra 5/10 owners, because that is the onboard video card.
> * but for this exception, it read that X.org brokenness on sparc
> seemed to be not a critical issue to any stable Debian release.
> * my scenario does not fall within the exception. I am trying to (as
> message #126 words it) assemble a unique set of components will not work
> together well.
> * except that I believe that as xserver-xorg-video-mga is provided for
> sparc, I should be able to expect that adding a MGA card is a reasonable
> thing to do. It doesn't have to work first go, I'm not having a whinge
> about that. I am just pre-empting the bug being considered not important
> because of what I read as a blanket policy. That's what I mean by case
> by case consideration.
so here's a summary of my understanding:
- Xorg < 1.5 handles access to the pci bus all by itself
- that code has a bug, which got exposed by a kernel change which was
necessary to make pci-express based sparcs work
- Xorg >= 1.5 let the kernel handle access to the pci bus, so the bug
doesn't apply to newer versions
- the kernel change was reverted for lenny r0 to make X work again, but
will be reinstated for r1 because breaking pci-express machines is not
acceptable, and X will fall back to the fbdev driver instead of going
through pci, in an attempt to start up anyway.
So your case with 2 X servers is probably not going to work with that
workaround, and I'm not sure how to make it work with the X version in
lenny, unfortunately. Patches welcome, though...
Cheers,
Julien
Reply to: