Bug 194697
| Summary: | Xorg freeze on Apple MacBook Pro 8,2 | ||
|---|---|---|---|
| Product: | Drivers | Reporter: | Wilfried Klaebe (linux-kernel) |
| Component: | Video(Other) | Assignee: | drivers_video-other |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | ||
| Priority: | P1 | ||
| Hardware: | Intel | ||
| OS: | Linux | ||
| Kernel Version: | 4.10 and others | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
| Attachments: | Xorg and kernel logs | ||
Created attachment 254909 [details] Xorg and kernel logs Hi all, with Linux 4.10 I still have a problem I already had with at least 4.8: On my MacBook Pro 8.2 (aka "early 2011") with Intel IGD and AMD Radeon discrete GPU, using the Radeon (not switching to IGD with vga_switcheroo before) causes Xorg not to start. Somewhere I had found this mentioned in some discussion, together with a hint to revert commit 4eebd5a4e72697aac25a8a57d3f888a9d5f80370 ("apple-gmux: lock iGP IO to protect from vgaarb changes"), which helps. I seem to remember there was a mention of some kind of deadlock somewhere in the DRM infrastructure that thread, causing not only Xorg to hang, but also making it impossible to access any gpu-driven console. It is possible to do a regular shutdown via ACPI powerbutton though. I had hoped this would somehow get resolved by the people on that thread, but obviously, it didn't. I'm attaching some Xorg and kernel logs of both working and non-working states. The kernel logs both(!) contain a warning about a possible circular locking dependency where vgaswitcheroo tries to acquire a lock it already holds. The "4.10.0" logs are from the non-working, hanging state, while the "4.10.0-00001-gdf6ffb763137" logs are from the working state running 4.10.0 with the mentioned commit reverted.