Hi,
On 13-11-17 18:53, Laura Abbott wrote:
On 11/13/2017 07:40 AM, Hans de Goede wrote:
> Hi all,
>
> As mentioned before I'm working on trying to improve the OOTB
> power-consumption of Fedora Workstation on laptops
> (I need to create an F28 feature page for this).
>
> On many laptops the btusb device is the only USB device not
> having USB autosuspend enabled, this causes not only the btusb
> device but also the USB controller to stay awake, together using
> aprox. 0.4W of power. Modern ultrabooks idle around 6W
> (at 50% screen brightness), 3.5W for Apollo Lake devices.
> 0.4W is a significant chunk of this (7 / 11%).
>
> So I would like to enable btusb autosuspend by default, to
> make this possible I've submitted the following kernel
> patch upstream which has just been merged (queued for 4.16):
>
https://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next....
>
> If it is ok with the Fedora kernel team I would like to add
> this as a patch to the Fedora kernels for now and set
> the new BT_HCIBTUSB_AUTOSUSPEND option to y. This is not
> without a risk of regressions, so this is something for
> rawhide/F28 only.
>
> Regards,
>
> Hans
I'm concerned about this mostly because it goes through the
USB stack. USB hardware seems to be particularly bad at hardware
power management based on the archives of linux-usb. I don't
have a good alternative though so I'm okay with it going in
as long as we are prepared for a possible influx of bug reports.
Or maybe I'll be wrong and all hardware will just magically
work correctly.
So as I mentioned in my reply to Justin, to clarify this is
NOT about enabling USB autosuspend everywhere. Most bugreports
about USB autosuspend are about mice and keyboards which claim
to implement remote-wakeup (a feature necessary for auto-suspend)
but don't implement it (correctly).
This thread is only about enabling autosuspend for for USB
connected bluetooth receivers.
Anyways as said this was just to clarify and I do share
your worries that even just for btusb this may cause issues,
but there is a significant power-saving to be had here and
only one way to find out if this causes problems. So I will
go ahead and add the patch + the Kconfig option to rawhide.
Regards,
Hans