f20, anaconda, net install and video out of range ....

[Adam Jackson]

On Mon, 2014-02-03 at 13:52 -0500, Paul Wouters wrote:
> On Mon, 3 Feb 2014, Adam Jackson wrote:
> 
> > On Sun, 2014-02-02 at 22:02 -0500, Paul Wouters wrote:
> >
> >> ftp://ftp.nohats.ca/Xorg.0.log
> >
> > [    54.323] (II) VESA(0): VESA VBE Total Mem: 2048 kB
> > [    54.323] (II) VESA(0): VESA VBE OEM: Cirrus Logic GD-5480 VGA
> > [    54.324] (II) VESA(0): VESA VBE OEM Software Rev: 1.32
> >
> > Your friend has an actual Cirrus Logic card.  Fortunately, it's not
> > contagious, and it can be cured.
> >
> > You're getting the vesa driver loaded because we don't install cirrus by
> > default anymore (since the most common Cirrus card by far is the
> > emulated one in qemu, which we have a kernel driver for).  You might
> > fare better with xorg-x11-drv-cirrus installed to avoid VBIOS bugs.
> 
> Why would anaconda not load the driver when it sees the physical card?

Anaconda doesn't load X drivers, X does.  And X tries to load the cirrus
driver, it's just not there.

> Why would it not install xorg-x11-drv-cirrus when it sees the physical
> card?

We don't have anything like the kernel's modaliases for X drivers, at
least not exposed in a way anaconda could use.

> Can it not detect the difference between the virtual and real
> card? Should I file a proper bug for this, or do you really deem this
> unsolvable?

It's not unsolvable.  I can think of plenty of solutions, depending on
how much effort one wants to put into it.  We could include cirrus in
the install image, assuming that works.  We could fix wheatever's going
wrong in the vesa emulator.  We could add physical cirrus support to the
kms driver.  Or we could just document the workaround.

My personal level of interest in 17-year-old GPUs, though probably
higher than average, is not quite high enough to work on cirrus.  And
the driver is orphaned; if we really want to fix this it would be nice
to find it a proper maintainer (and ideally get the KMS port done).

- ajax