<SNIP> On 09/17/2017 05:17 AM, Sagar Arun Kamble wrote:
Teardown of GuC HW/SW state was not properly done in unload path. guc_submission_disable was called as part of intel_uc_fini_hw which happens post gem_unload in the i915_driver_unload path. s/i915_gem_fini/i915_gem_cleanup as it looks more suitable as that function does cleanup. To differentiate the tasks during suspend and unload w.r.t GuC this patch introduces new function i915_gem_fini which in addition to disabling GuC interfaces also disables GuC submission during which communication with GuC is needed for destroying doorbell. i915_gem_fini is copy of i915_gem_suspend with difference w.r.t GuC operations. To achieve this, new helpers i915_gem_context_suspend and i915_gem_suspend_complete are prepared.
Sagar, I just realized this patch would make this comment in guc_client_free superfluous, right?:
/* FIXME: in many cases, by the time we get here the GuC has been * reset, so we cannot destroy the doorbell properly. Ignore the * error message for now */ Can you make sure you remove it in this patch? Thanks! _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
