Re: [PATCH 1/3] arm: dts: imx8mm: Sync with Linux 6.3

[Tim Harvey <tharvey@gateworks.com>]

On Fri, Apr 28, 2023 at 8:32 AM Adam Ford <aford173@gmail.com> wrote:
>
> On Fri, Apr 28, 2023 at 10:27 AM Tim Harvey <tharvey@gateworks.com> wrote:
> >
> > On Fri, Apr 28, 2023 at 4:57 AM Adam Ford <aford173@gmail.com> wrote:
> > >
> > > On Thu, Apr 27, 2023 at 5:25 PM Tim Harvey <tharvey@gateworks.com> wrote:
> > > >
> > > > On Thu, Apr 27, 2023 at 12:49 PM Fabio Estevam <festevam@gmail.com> wrote:
> > > > >
> > > > > On Thu, Apr 27, 2023 at 4:44 PM Tim Harvey <tharvey@gateworks.com> wrote:
> > > > >
> > > > > > Fabio,
> > > > > >
> > > > > > Sorry for the confusion.
> > > > > >
> > > > > > This imx8mm dt sync patch will hang on imx8mm boards that use 'both'
> > > > > > usbotg1 and usbotg2. You can reproduce this hang on your imx8mm-evk by
> > > > > > enabling usbotg2 in the dt (the board has it but it is not enabled due
> > > > > > to the gpio based usb 3.0 mux not being sorted out yet):
> > > > > > +&usbotg2 {
> > > > > > +       dr_mode = "otg";
> > > > > > +       status = "okay";
> > > > > > +};
> > > > > > +
> > > > > >
> > > > > > u-boot=> usb start && usb tree
> > > > > > starting USB...
> > > > > > Bus usb@32e40000: Bus usb@32e50000:
> > > > > > ^^^ imx8mm-evk hangs
> > > > >
> > > > > Yes, I can reproduce the hang, but it happens with or without the
> > > > > imx8mm dt sync.
> > > > >
> > > >
> > > > Fabio,
> > > >
> > > > I do 'not' see a hang on imx8mm-evk on 'usb start && usb tree' on
> > > > master (my other issue was on a 'usb stop' but only with usb
> > > > controllers in host mode).
> > > >
> > > > > This hang is a separate issue, not dt related, as far as I understand.
> > > > >
> > > > > The imx8mm dts sync does solve the issue of running 'ums' after CTRL+C.
> > > >
> > > > I don't agree. The hang 'is' related because all my imx8mm-venice-*
> > > > boards which use 'both' USB controllers hang with this patch on a 'usb
> > > > start' and don't hang without it. While a basic 'review' of the patch
> > > > looks good but actual product testing shows issues. As a maintainer
> > > > for ARM FREESCALE IMX you must have another imx8mm board which uses
> > > > both usbotg devices to test against and verify you see what I see?
> > > >
> > > > Until we know what other fix is needed to go along with this:
> > > > Nacked-by: Tim Harvey <tharvey@gateworks.com>
> > >
> > > What is the harm is sync'ing the device tree with the kernel? I seemed
> > > like you found a solution with the regulator patch.  Did I
> > > misunderstand that?
> > >
> > > adam
> >
> > Adam,
> >
> > No, the regulator patch did 'not' resolve the issue created by syncing
> > the imx8mm dt (I caused confusion by responding to the wrong thread -
> > the regulator patch resolved a different issue).
>
> Ok.
> >
> > Could you please verify my results on a board that uses both usbotg1
> > and usbotg2? What I see is on master + this imx8mm dt sync
> > (specifically the changes from Linux commit 4585c79ff477f ("arm64:
> > dts: imx8mm: correct usb power domains")) the board hangs on usb start
> > when bringing up usbotg2.

Adam,

Sorry to hear that :(

>
> I can, but I am about to board a plane to go visit some sick family,
> but I'll try to do it early next week.
> I have a board with both USB controllers enabled.  My OTG2 is
> host-only, so I think it's similar to your setup.

Yes I think that is similar enough to test. In my experience simply
enabling otg2 via dt on imx8mm-evk shows the issue I see here but
Fabio says he sees a hang on 'usb start' even before this dt sync and
I don't know why my results on an imx8mm-evk differ.

>
> Should I apply the regulator patch when I test?

No, don't apply that as this exposes another issue: Error enabling
VBUS supply (ret=-114)

I'm still looking into that. I'm assuming when the regulaor refcnt
support gets merged it may expose a lot of issues from unbalanced
regulator enable/disable calls. The regulator refcnt series resolved
the hang I see on 'usb stop' for boards where otg2 is in host mode
(internal usb_hub device powers down the power domain before
ehci_shutdown tries to access the registers to disable the ports).

Tim