Created attachment 1257184 [details]
Running glxgears in Fedora 25 guest after mesa-dri-drivers update
Description of problem:
Updating mesa-dri-drivers to version 13.0.3 makes 3D rendering in guests totally garbled, see attachment for an example. It looks like the wrong buffer is sent to the screen.
Version-Release number of selected component (if applicable):
mesa-dri-drivers 13.0.3-5.fc25 is broken.
mesa-dri-drivers 12.0.3-3.fc25 (the one selected by dnf downgrade) works
How reproducible: Always
Steps to Reproduce:
1. Install a Fedora 25 guest with virtio and 3D acceleration enabled. See this post https://redskincat.wordpress.com/2017/02/16/virtualized-accelerated-3d-is-complicated/#more-1082 for information on how I installed it.
2. Run glxgears -info and verify that
a. You have a Gallium / virtgl rendering
b. Rendering is correct
3. Update the guest with dnf update, or simply mesa-dri-drivers (which brings meas-filesystem with it).
4. Run glxgears -info again, and verify that:
a. There are error messages on the console from MESA-LOADER
b. The window output is garbled
Garbled window, with, most often, remnants of some other part of the screen
Glxgears in its beautiful 3-color glory
Fixed in current mesa master.
Checked out mesa from git, version c0e9e61c9a1.
./configure --prefix=/usr --enable-libglvnd --enable-selinux --enable-gallium-osmesa --with-dri-driverdir=/usr/lib64/dri --enable-gl --disable-gles1 --enable-gles2 --disable-xvmc --with-egl-platforms=drm,x11,surfaceless,wayland --enable-shared-glapi --enable-gbm --enable-glx-tls --enable-texture-float=yes --enable-gallium-llvm --enable-llvm-shared-libs --enable-dri --with-gallium-drivers=i915,nouveau,r300,svga,swrast,virgl --with-dri-drivers=swrast,nouveau
The resulting version works.
Author: Kenneth Graunke <firstname.lastname@example.org>
Date: Tue Feb 28 11:48:51 2017 -0800
mesa: Drop unused STATE_TEXRECT_SCALE program statevars.
Unlikely to be the fixed,
Could you bisect the commit that fixed it (for you)? thanks
(In reply to Marc-Andre Lureau from comment #2)
> commit c0e9e61c9a1eaaee9ca4dfc2f05da85a791c4c5d
> Unlikely to be the fixed,
> Could you bisect the commit that fixed it (for you)? thanks
Yes, bisect is in progress. I got confused because git does not deal correctly with finding the "first good fix", so I had to swap "good" and "bad".
The actual fix seems to be dc2d9b8da14b97376cca9c87e2408763b11bdaa7.
This seems to match the fix for https://bugzilla.redhat.com/show_bug.cgi?id=1417932. I was not sure this was the same thing, because the symptoms were different (garbage on screen instead of black windows). But I now believe the two problems are duplicates.
*** This bug has been marked as a duplicate of bug 1417932 ***