[Intel-gfx] [PATCH] kthread: finer-grained lockdep/cross-release completion

[Peter Zijlstra]

On Thu, Dec 07, 2017 at 11:08:49AM +0100, Daniel Vetter wrote:
> Since -rc1 we're hitting a bunch of lockdep splats using the new
> cross-release stuff around the 2 kthread completions. In all cases
> they are because totally independent uses of kthread are mixed up by
> lockdep into the same locking class, creating artificial deadlocks.
> 
> Fix this by converting kthread code in the same way as e.g.
> alloc_workqueue already works: Use macros for the public api so we can
> have a callsite specific lockdep key, then pass that through the
> entire callchain. Due to the many entry points this is slightly
> tedious.

Do you have a splat somewhere? I'm having trouble seeing how all this
comes together. That is, I've no real idea what you're actual problem is
and if this is the right solution.