[gentoo-user] Virtualbox-guest-additions-4.3.24 breaks 3D acceleration in linux guests (?)

[gentoo-user] Virtualbox-guest-additions-4.3.24 breaks 3D acceleration in linux guests (?)

[walt]

I state this as a question because I'm seeing different behavior on
different ~amd64 machines (as usual ;)

After upgrading from vbox-4.3.20-r1 to 4.3.24 yesterday I find that
my linux guests have no 3D acceleration.

Downgrading Virtualbox to 4.3.20-r1 fixes the problem on one ~amd64
machine but, of course, not the other one <sigh>.  I'm still playing
with downgrading mesa, etc, on the still-broken machine to see what
happens.

Anyone else noticing similar problems?

[gentoo-user] Re: Virtualbox-guest-additions-4.3.24 breaks 3D acceleration in linux guests (?)

[walt]

On 03/05/2015 02:25 PM, walt wrote:
> Downgrading Virtualbox to 4.3.20-r1 fixes the problem on one ~amd64
> machine but, of course, not the other one <sigh>.

For reasons I don't know, the gentoo devs split the "guest-additions"
into two parts: they put the vboxvideo kernel module into the gentoo
xf86-video-virtualbox package, and the vboxguest+vboxsf kernel modules
into the virtualbox-guest-additions package.

Anyway, starting with a vbox gentoo snapshot from Mar 3 (before the
update to vbox-4.3.24) I upgraded only the virtualbox-guest-additions
(not the xf86-video-virtualbox package) and found that everything still
works normally.

I then upgraded the xf86-video-virtualbox package and discovered that
3D video acceleration was broken, just as before.

I then downgraded the xf86-video-virtualbox package to 4.3.20 (leaving
the virtualbox-guest-additions at 4.3.24) and found the 3D acceleration
working normally again.

That clearly implicates the vboxvideo.ko kernel module as the cause of
the breakage, I think.

Opinions/comments are most welcome.

Re: [gentoo-user] Re: Virtualbox-guest-additions-4.3.24 breaks 3D acceleration in linux guests (?)

[Alan McKinnon]

On Thu, 05 Mar 2015 16:27:25 -0800
walt <w41ter@gmail.com> wrote:

> On 03/05/2015 02:25 PM, walt wrote:
> > Downgrading Virtualbox to 4.3.20-r1 fixes the problem on one ~amd64
> > machine but, of course, not the other one <sigh>.
> 
> For reasons I don't know, the gentoo devs split the "guest-additions"
> into two parts: they put the vboxvideo kernel module into the gentoo
> xf86-video-virtualbox package, and the vboxguest+vboxsf kernel modules
> into the virtualbox-guest-additions package.
> 
> Anyway, starting with a vbox gentoo snapshot from Mar 3 (before the
> update to vbox-4.3.24) I upgraded only the virtualbox-guest-additions
> (not the xf86-video-virtualbox package) and found that everything
> still works normally.
> 
> I then upgraded the xf86-video-virtualbox package and discovered that
> 3D video acceleration was broken, just as before.
> 
> I then downgraded the xf86-video-virtualbox package to 4.3.20 (leaving
> the virtualbox-guest-additions at 4.3.24) and found the 3D
> acceleration working normally again.
> 
> That clearly implicates the vboxvideo.ko kernel module as the cause of
> the breakage, I think.
> 
> Opinions/comments are most welcome.


File a bug at b.g.o.

what you describe has much potential to trip up many people :-)

Alan

[gentoo-user] Re: Virtualbox-guest-additions-4.3.24 breaks 3D acceleration in linux guests (?)

[James]

Alan McKinnon <alan.mckinnon <at> gmail.com> writes:


> > > Downgrading Virtualbox to 4.3.20-r1 fixes the problem on one ~amd64
> > > machine but, of course, not the other one <sigh>.

> > (not the xf86-*

> File a bug at b.g.o.
> what you describe has much potential to trip up many people 

Well, I have to concur with what I did not prune out of this thread.
My recent experience with xf86* on my amd64 and the 3.18.6 kernel
had me reading quite a few horror stories of others; finally I 
found a solution, that is more a result of "marathonic" efforts
as opposed to finding the correct nuggets of information.

I.E. in this area I'm willing to bet there are ample bugs
and things that have been modified or fixed, but the knowledge
is just not disseminated. I filed no bugs because clarity
never appeared as to the exact problem(s).

For example in /lib/firmware/radeon I have these relevant drivers
some are lower case some are uppercase and my card requires the (V|v)erde*
drivers and at least one "TAHITI" driver..... In fact the inter
relationships on some kernel options are there, but are not, or
are "poorly", defined. I think this is a result of the quickened
turn_around speed of the kernel versions. Quality was sacrificed,
imho.

I found no documentation as to why. It's just a mess, imho. I could
list dozens of things, but the impression I got was (dev)folks are
hacking at this, but it's not cleaned up because there is little
clarity for (kernel) devs or other subsequent development efforts. 
I sure hope I'm wrong. ymmv.


James