Bear in mind, some spinning rust is only designed to be spun up/down so many times so getting too cute with saving power can prematurely wear down your drives.
The article only looked at SSDs, and mentioned that the findings would likely be very different what you’d see on spinning rust.
Yep, just noting a reason you’d want to avoid doing anything like this on spinning disks. I actually have cronjobs setup to poke/dd my hard drives specifically so they don’t spin down at idle due to their stated spin up/down ratings (100 000 cycles iirc, which sounds like a lot but can add up rather quickly).
Is there something like med_power_with_dipm for FreeBSD?
This “DIPM” word is mentioned nowhere, but pm_level 1 is “device is allowed to initiate PM state change, host is passive”. And the higher PM levels do not mention disabling the device’s ability to change power states, which sort of implies that any non-zero level has “DIPM”?? So, just set hint.ahcich.N.pm_level=5 for all N I guess.
Thankfully none of this is relevant for NVMe :)