broadwell bugs with fglrx in xserver-xorg-video-intel
Bug #1359105 reported by
Maarten Lankhorst
This bug affects 2 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
HWE Next |
Fix Released
|
Undecided
|
Unassigned | ||
xserver-xorg-video-intel (Ubuntu) |
Fix Released
|
Undecided
|
Unassigned | ||
Precise |
Invalid
|
Undecided
|
Unassigned | ||
Trusty |
Fix Released
|
Undecided
|
Unassigned | ||
xserver-xorg-video-intel-lts-trusty (Ubuntu) |
Invalid
|
Undecided
|
Unassigned | ||
Precise |
Fix Released
|
Undecided
|
Unassigned | ||
Trusty |
Invalid
|
Undecided
|
Unassigned |
Bug Description
When using the fglrx driver gpu-manager will fallback to UXA. UXA is not implemented for broadwell in the 910 driver, but is required when using fglrx for offloading. When using FGLRX in the IGPU (integrated gpu) mode fglrx unloads itself and makes intel drive the screen. This crashes with SNA, because fglrx steals the driver privates.
Upstream has some commits to fix this, but they're missing in the 910 driver from trusty.
Related branches
Changed in hwe-next: | |
status: | New → Fix Released |
tags: |
added: verification-done removed: verification-needed |
Changed in xserver-xorg-video-intel-lts-trusty (Ubuntu Trusty): | |
status: | New → Invalid |
Changed in xserver-xorg-video-intel-lts-trusty (Ubuntu): | |
status: | New → Invalid |
Changed in xserver-xorg-video-intel (Ubuntu Precise): | |
status: | New → Invalid |
To post a comment you must log in.
[Impact]
* Fixes crash when using SNA with fglrx IGPU mode.
* Adds UXA support for broadwell.
[Test Case]
* On an affected system:
* Run aticonfig --px-igpu to force intel driver being used.
* Disable gpu-manager, set accelmethod in /etc/X11/xorg.conf to UXA or SNA
* SNA: Observe a crash in freescreen, while a normal xserver with full acceleration is expected.
* UXA (default): xserver starts normally, but llvmpipe is used and no acceleration is used in intel at all.
[Regression Potential]
* Adding pci-id and names for broadwell and crv should always be safe, no regressions expected there.
* The uxa patches seem to be a noop for < gen8, which means I don't expect any regressions either.
* The sna-freescreen patch can not make things worse, because if sna is NULL the next statement would already crash, so in the worst case intel would crash elsewhere.
* I'll watch for regressions in -proposed, but I don't expect any from them because those patches are in utopic too. :-)