Good to see general resume improvements but ...<p>Can we not try to force s2idle? It uses way too much power. I've not yet seen a laptop which is capable of staying "sleeping" in s2idle as long as in s3 sleep. Correct me if I'm wrong though.<p>I also don't really get why: s3 is more than fast enough to be awake when my laptop is open and I'm ready to login, the only "need to be faster" way I can see is when you do not lock you laptop and go straight to desktop but that has ... issues.
It may be a "one-liner", but it's the kind where the engineer had to have broad and detailed knowledge about the system and how it works. Well done!
I just upgraded my Linux kernel and had major suspend issues due to the XHCI driver. I ended up having to disable the ability for XHCI to wake my machine.<p>On fedora cinnamon, I was getting random freezes on resume, and momentary freezes on resume on Linux Mint. Disabling the ability for XHCI to wake my machine fixed it.<p>This sleep() looks like bad code designed to work around a race condition. Perhaps it's related to the issue I had.<p>This issue was the worst Linux issue I've had in ~10+ years.<p>(I'm on ryzen first gen cpu)
I've been thinking there must be a generic approach to waiting for events rather than sleeping and registering actions on events.<p>Like epoll for events
Ah, the most powerful optimization technique: add random sleeps to the code and when people start complaining remove them and collect your bonus<p></joke>