On Wed, May 18, 2016 at 05:53:49PM +0200, Alexander Stein wrote:
On Wednesday 18 May 2016 16:20:28, Mark Brown wrote:
Oh, and the other thing about this which I should have mentioned is that it applies to every single resource acquisition in the kernel - this needs addressing at a global error rather than piecemeal. Raphael was intending to work on avoiding probe deferral in the first place by sorting probe and there were also some proposals to have a helper for printing error messages which would suppress the display of probe deferral errors before the final run at late initcall. Obviously the second mechanism is substantially simpler to implement.
I remember those patches a bit. But working with late initcall sound to me this has no effect on probe deferral in drivers loaded by modules, no?
Correct, once you're using modules you're pretty much relying on userspace sorting out the init ordering.