2. store_mem_state() still needs a tweak, right? It was only returning -EINVAL by accident, due to the convoluted sequence I listed in the patch.
Yes, absolutely. It returning -EINVAL for "nline" is what is accidently
preserving it's backwards compatibility :) Note that device_online()
returns 1 if already online and memory_subsys_online() returns 0 if online
in this case. So we want store_mem_state() to return -EINVAL if
device_online() returns non-zero (this was in my first email).