Re: [TESTPATCH v2] xhci: fix usb2 resume timing and races.

[Mathias Nyman <mathias.nyman@linux.intel.com>]

On 01.12.2015 16:32, Felipe Balbi wrote:
>
> Hi,
>
> Mathias Nyman <mathias.nyman@linux.intel.com> writes:
>> usb2 ports need to signal resume for 20ms before moving to U0 state.
>
> at least 20ms ;-) Recently, we decided to drive resume for 40ms to
> support devices with broken FW.
>

True, but specs talk about 20ms, and I'm just trying to give some context for what's
going on. This testpatch doesn't touch the timings.

Daniel is able to trigger a USB2 xhci resume issue which I hope is fixed with this patch.
This is especially made for his setup running a 4.3 kernel

If this works I'll clean up all the "20ms" in the commit message and comments.

Just noticed that xhci USB2 host initiated resume uses a hardcoded msleep(20).
That needs to be changed to USB_RESUME_TIMEOUT at some point.
For now I'm just interested in knowing if this patch works.

>
> this 'v2' note doesn't have to go into commit log, IMO.
>

It's going to be cleaned out as well, just there to explain to Daniel, and the world why
a second version of the testpatch was created.

-Mathias