Re: [PATCH] Revert "USB / PM: Allow USB devices to remain runtime-suspended when sleeping"

From: Alan Stern
Date: Mon May 02 2016 - 11:13:23 EST


On Mon, 2 May 2016, Johan Hovold wrote:

> This reverts commit e3345db85068ddb937fc0ba40dfc39c293dad977, which
> broke system resume for a large class of devices.
>
> Devices that after having been reset during resume need to be rebound
> due to a missing reset_resume callback, are now left in a suspended
> state. This specifically broke resume of common USB-serial devices,
> which are now unusable after system suspend (until disconnected and
> reconnected) when USB persist is enabled.
>
> During resume, usb_resume_interface will set the needs_binding flag for
> such interfaces, but unlike system resume, run-time resume does not
> honour it.
>
> Cc: stable <stable@xxxxxxxxxxxxxxx> # 4.5
> Signed-off-by: Johan Hovold <johan@xxxxxxxxxx>
> ---
>
> Greg, Alan,
>
> This patch for v4.6-rc7 fixes a 4.5-regression that broke system suspend
> for a large class of devices, including USB-serial devices, for example
> when USB persist is enabled.
>
> We may be able to find a way around this, but since it's a user-visible
> regression and late in the rc-cycle, I believe reverting the offending
> commit is the right thing to do.

The description of the problem doesn't sound right to me. For
instance, would it help if usb_runtime_resume() did honor the
needs_binding flag? I doubt it. Things like the wakeup setting would
still be lost before the runtime resume occurred.

I suspect the right answer is always to resume a USB device if it needs
a reset-resume, but otherwise allow it to remain in runtime suspend.

Reverting the patch for now is okay with me. Tomeu may want to work on
a better solution. Part of the difficulty is that the PM core wants to
know before suspending whether skipping resume will be okay, but the
USB stack doesn't know until after the host controller has been
resumed.

In the end, we'll probably have the PM core call usb_resume all the
time, but usb_resume will leave the device in runtime suspend if it
can. This isn't ideal but it may be the best we can do.

Alan Stern