`nvme_disable_ctrl()` takes 411 ms on a Dell XPS 13 with SK hynix PC300 NVMEe

[Keith Busch]

On Wed, May 01, 2024 at 10:58:05PM +0200, Paul Menzel wrote:
> Am 01.05.24 um 09:58 schrieb Keith Busch:
> > Exactly. Unless the device reports a lower D3 entry latency, then this
> > sounds like everything is working-as-designed.
> 
> Maybe according to the spec, but I have a hard time to believe, that disks
> should take longer to shut down than coreboot to initialize a mainboard.

That doesn't seem too hard to believe to me. A safe shutdown can often
take a while time for an SSD. I've seen other implementations orders of
magnitude worse than what you're showing. You could do an unsafe
shutdown instead, but the device will just take even more time to enable
on its next power-on.
 
> In the end, in my opinion, users cannot make an informed decision, if these
> things are hidden. If it would be visible somehow in the logs - maybe not
> warning but info level - then even not so technical users could inform
> themselves and factor this in their buying decision.

What good is it to advertise a shutdown time when vendors are clearly
unreliable at reporting an accurate value? If you need to see the driver
report it from emperical testing, then you've already bought the device,
right?

> > You can check your device's advertised shutdown time (assuming your
> > device is nvme0):
> > 
> >    nvme id-ctrl /dev/nvme0 | grep rtd3e
> > 
> > The value is reported in microseconds. If it shows 0, then the device
> > doesn't report an expected shutdown time.
> 
> Thank you for sharing. It´s 60 ms: