On Wed, Jun 24, 2015 at 08:44:47AM +0100, Chris Wilson wrote:
On Tue, Jun 23, 2015 at 10:45:51PM +0200, Daniel Vetter wrote:
Hi all,
Dave&I have been discussing connector hotplug and unplugging around DP MST and if there's one thing that's clear it's that we don't even really know where all the problems are. Hence first step is to figure that out. One of the bigger items is walking the encoder/connector lists without appropriate locking to protect against concurrent hotadd/removal of connectors.
This patch series tries to untangle things a bit here. RFC since only lightly tested and missing conversion of the radoen mst code to the new locking scheme. I think rolling out the new macros for i915 should be done as a second step, to avoid hitting too many WARN_ON ;-)
Bah, I was hoping to see scru! :)
This is just to start figuring out how bad the situation is and who's all affected. Later on we need to figure out how to properly protect the connector/encoder list without just grabbing all kinds of locks in the hotadd/remove code - stalling screen updates on the main screen just because you plug something in tends to upset people. -Daniel