(EXT) Re: [RFC 1/1] usb: chipidea: ci_hdrc_imx: disable runtime pm for HSIC interface

[Jun Li]

> -----Original Message-----
> From: Peter Chen <peter.chen at kernel.org>
> Sent: Saturday, April 9, 2022 10:20 AM
> To: Alexander Stein <alexander.stein at ew.tq-group.com>
> Cc: Peter Chen <hzpeterchen at gmail.com>; Greg Kroah-Hartman
> <gregkh at linuxfoundation.org>; Shawn Guo <shawnguo at kernel.org>; Sascha
> Hauer <s.hauer at pengutronix.de>; Fabio Estevam <festevam at gmail.com>;
> Pengutronix Kernel Team <kernel at pengutronix.de>; dl-linux-imx
> <linux-imx at nxp.com>; USB list <linux-usb at vger.kernel.org>;
> linux-arm-kernel at lists.infradead.org
> Subject: Re: (EXT) Re: [RFC 1/1] usb: chipidea: ci_hdrc_imx: disable runtime
> pm for HSIC interface
> 
> On 22-03-29 10:14:36, Alexander Stein wrote:
> > Hello Peter,
> >
> > Am Dienstag, 15. März 2022, 02:23:23 CEST schrieb Peter Chen:
> > > On Wed, Mar 2, 2022 at 5:42 PM Alexander Stein
> > >
> > > <alexander.stein at ew.tq-group.com> wrote:
> > > > With the add of power-domain support in commit 02f8eb40ef7b ("ARM:
> dts:
> > > > imx7s: Add power domain for imx7d HSIC") runtime suspend will disable
> > > > the power-domain. This prevents IRQs to occur when a new device is
> > > > attached
> > > > on a downstream hub.
> > > >
> > > > Signed-off-by: Alexander Stein <alexander.stein at ew.tq-group.com>
> > > > ---
> > > > Our board TQMa7x + MBa7x (i.MX7 based) uses a HSIC link to mounted
> USB HUB
> > > > on usbh device. Cold plugging an USB mass storage device is working
> fine.
> > > > But once the last non-HUB device is disconnected the ci_hdrc device
> goes
> > > > into runtime suspend.
> > >
> > > Would you please show the difference between cold boot and runtime
> > > suspend after disconnecting
> > > the last USB device?
> > >
> > > - Power domain on/off status for HUB device
> > > - Runtime suspend status at /sys entry for HUB device
> > > - "/sys/..power/wakeup" /sys entry  for HUB device
> >
> > I hope I got all entries you requested.
> >
> > For reference this is the bus topology:
> > lsusb -t
> > /:  Bus 02.Port 1: Dev 1, Class=root_hub, Driver=ci_hdrc/1p, 480M
> > /:  Bus 01.Port 1: Dev 1, Class=root_hub, Driver=ci_hdrc/1p, 480M
> >     |__ Port 1: Dev 2, If 0, Class=Hub, Driver=hub/4p, 480M
> >         |__ Port 2: Dev 3, If 0, Class=Mass Storage, Driver=usb-storage,
> 480M
> >
> > Bus 2 is a different connector and doesn't matter here. I'm disconnecting
> 'Dev
> > 3' in this scenario.
> >
> > After boot up with the bus as shown above:
> > $ cat /sys/bus/usb/devices/1-1/power/wakeup
> > disabled
> > $ cat /sys/bus/usb/devices/1-1/power/runtime_status
> > active
> > $ cat /sys/kernel/debug/pm_genpd/usb-hsic-phy/current_state
> > on
> >
> > After disconnecting Dev 3 from the bus ('usb 1-1.2: USB disconnect, device
> > number 3' in dmesg) the status changes as follows (without the patch):
> > $ cat /sys/bus/usb/devices/1-1/power/wakeup
> > disabled
> > $ cat /sys/bus/usb/devices/1-1/power/runtime_status
> > suspended
> > $ cat /sys/kernel/debug/pm_genpd/usb-hsic-phy/current_state
> > off-0
> >
> > For the record, when applying the posted patch this changes into:
> > $ cat /sys/bus/usb/devices/1-1/power/wakeup
> > disabled
> > $ cat /sys/bus/usb/devices/1-1/power/runtime_status
> > suspended
> > $ cat /sys/kernel/debug/pm_genpd/usb-hsic-phy/current_state
> > on
> >
> 
> Okay, I think the problem here is the power domain for USB controller is
> off at runtime, but USB controller/PHY needs to detect the USB wakeup
> signal at runtime, so the USB controller/PHY's power domain should be
> not off. The proper change may keep power domain on at runtime, and the
> power domain could be off at system suspend.

Can this "hsic phy power domain off breaks wakeup" be confirmed?
Like with some hack to move hsic phy power domain on some other device
node:

non-usb-node {
        ...
        power-domains = <&pgc_hsic_phy>;
        status = "okay";
};

Just make sure this non-usb-node to be runtime active when
do hsic hub test.

Li Jun 

> 
> The controller/PHY clk could be off at runtime, it does not affect the
> wakeup detecting from controll/PHY side.
> 
> --
> 
> Thanks,
> Peter Chen