[PATCH] usb: gadget: udc: renesas_usb3: Fix host to USB_ROLE_NONE transition

[PATCH] usb: gadget: udc: renesas_usb3: Fix host to USB_ROLE_NONE transition

[Adam Ford]

The support the external role switch a variety of situations were
addressed, but the transition from USB_ROLE_HOST to USB_ROLE_NONE
leaves the host up which can cause some error messages when
switching from host to none, to gadget, to none, and then back
to host again.

 xhci-hcd ee000000.usb: Abort failed to stop command ring: -110
 xhci-hcd ee000000.usb: xHCI host controller not responding, assume dead
 xhci-hcd ee000000.usb: HC died; cleaning up
 usb 4-1: device not accepting address 6, error -108
 usb usb4-port1: couldn't allocate usb_device

After this happens it will not act as a host again.
Fix this by releasing the host mode when transitioning to USB_ROLE_NONE.

Fixes: 0604160d8c0b ("usb: gadget: udc: renesas_usb3: Enhance role switch support")
Signed-off-by: Adam Ford <aford173@gmail.com>

diff --git a/drivers/usb/gadget/udc/renesas_usb3.c b/drivers/usb/gadget/udc/renesas_usb3.c
index 57d417a7c3e0..601829a6b4ba 100644
--- a/drivers/usb/gadget/udc/renesas_usb3.c
+++ b/drivers/usb/gadget/udc/renesas_usb3.c
@@ -2378,6 +2378,8 @@ static void handle_ext_role_switch_states(struct device *dev,
 	switch (role) {
 	case USB_ROLE_NONE:
 		usb3->connection_state = USB_ROLE_NONE;
+		if (cur_role == USB_ROLE_HOST)
+			device_release_driver(host);
 		if (usb3->driver)
 			usb3_disconnect(usb3);
 		usb3_vbus_out(usb3, false);
-- 
2.32.0

RE: [PATCH] usb: gadget: udc: renesas_usb3: Fix host to USB_ROLE_NONE transition

[Yoshihiro Shimoda]

Hi Adam-san,

Thank you for the patch!

> From: Adam Ford, Sent: Saturday, January 29, 2022 7:36 AM
> 
> The support the external role switch a variety of situations were
> addressed, but the transition from USB_ROLE_HOST to USB_ROLE_NONE
> leaves the host up which can cause some error messages when
> switching from host to none, to gadget, to none, and then back
> to host again.

I think so. An external role switch in this driver is possible to
enter the role to USB_ROLE_NONE. (Just a record for me: in other words,
the "internal" role switch didn't enter the role to USB_ROLE_NONE.)

>  xhci-hcd ee000000.usb: Abort failed to stop command ring: -110
>  xhci-hcd ee000000.usb: xHCI host controller not responding, assume dead
>  xhci-hcd ee000000.usb: HC died; cleaning up
>  usb 4-1: device not accepting address 6, error -108
>  usb usb4-port1: couldn't allocate usb_device
> 
> After this happens it will not act as a host again.
> Fix this by releasing the host mode when transitioning to USB_ROLE_NONE.
> 
> Fixes: 0604160d8c0b ("usb: gadget: udc: renesas_usb3: Enhance role switch support")
> Signed-off-by: Adam Ford <aford173@gmail.com>
> 
> diff --git a/drivers/usb/gadget/udc/renesas_usb3.c b/drivers/usb/gadget/udc/renesas_usb3.c
> index 57d417a7c3e0..601829a6b4ba 100644
> --- a/drivers/usb/gadget/udc/renesas_usb3.c
> +++ b/drivers/usb/gadget/udc/renesas_usb3.c
> @@ -2378,6 +2378,8 @@ static void handle_ext_role_switch_states(struct device *dev,
>  	switch (role) {
>  	case USB_ROLE_NONE:
>  		usb3->connection_state = USB_ROLE_NONE;
> +		if (cur_role == USB_ROLE_HOST)
> +			device_release_driver(host);

The handle_ext_role_switch_states() with role = USB_ROLE_NONE seems
to be called multiple times. So, we have to avoid multiple calling of
device_release_driver() somehow.

Best regards,
Yoshihiro Shimoda

>  		if (usb3->driver)
>  			usb3_disconnect(usb3);
>  		usb3_vbus_out(usb3, false);
> --
> 2.32.0

Re: [PATCH] usb: gadget: udc: renesas_usb3: Fix host to USB_ROLE_NONE transition

[Adam Ford]

On Sun, Jan 30, 2022 at 6:23 PM Yoshihiro Shimoda
<yoshihiro.shimoda.uh@renesas.com> wrote:
>
> Hi Adam-san,
>
> Thank you for the patch!
>
> > From: Adam Ford, Sent: Saturday, January 29, 2022 7:36 AM
> >
> > The support the external role switch a variety of situations were
> > addressed, but the transition from USB_ROLE_HOST to USB_ROLE_NONE
> > leaves the host up which can cause some error messages when
> > switching from host to none, to gadget, to none, and then back
> > to host again.
>
> I think so. An external role switch in this driver is possible to
> enter the role to USB_ROLE_NONE. (Just a record for me: in other words,
> the "internal" role switch didn't enter the role to USB_ROLE_NONE.)
>
> >  xhci-hcd ee000000.usb: Abort failed to stop command ring: -110
> >  xhci-hcd ee000000.usb: xHCI host controller not responding, assume dead
> >  xhci-hcd ee000000.usb: HC died; cleaning up
> >  usb 4-1: device not accepting address 6, error -108
> >  usb usb4-port1: couldn't allocate usb_device
> >
> > After this happens it will not act as a host again.
> > Fix this by releasing the host mode when transitioning to USB_ROLE_NONE.
> >
> > Fixes: 0604160d8c0b ("usb: gadget: udc: renesas_usb3: Enhance role switch support")
> > Signed-off-by: Adam Ford <aford173@gmail.com>
> >
> > diff --git a/drivers/usb/gadget/udc/renesas_usb3.c b/drivers/usb/gadget/udc/renesas_usb3.c
> > index 57d417a7c3e0..601829a6b4ba 100644
> > --- a/drivers/usb/gadget/udc/renesas_usb3.c
> > +++ b/drivers/usb/gadget/udc/renesas_usb3.c
> > @@ -2378,6 +2378,8 @@ static void handle_ext_role_switch_states(struct device *dev,
> >       switch (role) {
> >       case USB_ROLE_NONE:
> >               usb3->connection_state = USB_ROLE_NONE;
> > +             if (cur_role == USB_ROLE_HOST)
> > +                     device_release_driver(host);
>
> The handle_ext_role_switch_states() with role = USB_ROLE_NONE seems
> to be called multiple times. So, we have to avoid multiple calling of
> device_release_driver() somehow.

From what I can tell, it appears that device_release_driver ultimately
calls __device_release_driver which checks for a NULL, and during the
release sets it to NULL, so subsequent releases should just get
ignored.  Maybe I am missing something.

adam
>
> Best regards,
> Yoshihiro Shimoda
>
> >               if (usb3->driver)
> >                       usb3_disconnect(usb3);
> >               usb3_vbus_out(usb3, false);
> > --
> > 2.32.0
>

RE: [PATCH] usb: gadget: udc: renesas_usb3: Fix host to USB_ROLE_NONE transition

[Yoshihiro Shimoda]

Hi Adam-san,

> From: Adam Ford, Sent: Monday, January 31, 2022 10:01 AM
> 
> On Sun, Jan 30, 2022 at 6:23 PM Yoshihiro Shimoda
> <yoshihiro.shimoda.uh@renesas.com> wrote:
<snip>
> > > diff --git a/drivers/usb/gadget/udc/renesas_usb3.c b/drivers/usb/gadget/udc/renesas_usb3.c
> > > index 57d417a7c3e0..601829a6b4ba 100644
> > > --- a/drivers/usb/gadget/udc/renesas_usb3.c
> > > +++ b/drivers/usb/gadget/udc/renesas_usb3.c
> > > @@ -2378,6 +2378,8 @@ static void handle_ext_role_switch_states(struct device *dev,
> > >       switch (role) {
> > >       case USB_ROLE_NONE:
> > >               usb3->connection_state = USB_ROLE_NONE;
> > > +             if (cur_role == USB_ROLE_HOST)
> > > +                     device_release_driver(host);
> >
> > The handle_ext_role_switch_states() with role = USB_ROLE_NONE seems
> > to be called multiple times. So, we have to avoid multiple calling of
> > device_release_driver() somehow.
> 
> From what I can tell, it appears that device_release_driver ultimately
> calls __device_release_driver which checks for a NULL, and during the
> release sets it to NULL, so subsequent releases should just get
> ignored.  Maybe I am missing something.

Thank you for your comment. You're correct. I'm sorry, I didn't check
the __device_release_driver() behavior. This patch looks good to me. So,

Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>

Best regards,
Yoshihiro Shimoda