Re: [PATCH] nvme-pci: Shutdown the device if D3Cold is allowed by the user
From: Rafael J. Wysocki
Date: Mon Dec 16 2024 - 14:43:48 EST
On Mon, Dec 16, 2024 at 8:40 PM Keith Busch <kbusch@xxxxxxxxxx> wrote:
>
> On Mon, Dec 16, 2024 at 08:34:24PM +0100, Rafael J. Wysocki wrote:
> > However, there is an additional concern that putting an NVMe device
> > into D3cold every time during system suspend on Android might cause it
> > to wear out more quickly.
> >
> > Is there anything else?
>
> I recall a previous reason for this behavior was because the resume
> latency was significantly faster if we don't prepare the device for D3,
> and the nvme protocol specific power states for some platforms was
> sufficiently low enough. Apparently this choice hasn't been universally
> optimal.
Thanks for chiming in!
I'm recalling some other reasons too (mentioned in one of my previous
messages), but overall this is a driver design choice, it is not done
to adhere to some general rule.