RFC proposal to get community feedback to handle the lmem mmap memory mappings. We can have two solutions.
1. Avoid rpm for any mmap mapping. 2. Update the page tables while entering to runtime suspend to raise a page fault while device in D3. Solution 2 will incur huge latency versus solution 1 which effectively disable the runtime PM. Caveat with solution 1 that, it doesn't address the case when user directly maps PCI Bar resource i.e. below mentioned resource. "/sys/bus/pci/devices/0000\:03\:00.0/resource2" Anshuman Gupta (1): drm/i915/dgfx: Avoid parent bridge rpm on mmap mappings drivers/gpu/drm/i915/gem/i915_gem_mman.c | 11 ++++++++ drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 8 ++++++ drivers/gpu/drm/i915/intel_runtime_pm.c | 35 ++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_runtime_pm.h | 2 ++ 4 files changed, 56 insertions(+) -- 2.26.2
