| Message ID | 20230927073027.27952-1-quic_kriskura@quicinc.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 0ea39e030a80be2b1b5f98d6b330a8b97dcf3342 |
| Headers | show |
| Series | [v5] usb: gadget: udc: Handle gadget_connect failure during bind operation | expand |
On Wed, Sep 27, 2023 at 01:00:27PM +0530, Krishna Kurapati wrote: > In the event gadget_connect call (which invokes pullup) fails, > propagate the error to udc bind operation which in turn sends the > error to configfs. The userspace can then retry enumeration if > it chooses to. Will this break userspace that is not expecting error codes to be returned? What userspace code will now be modified to handle this? Where is that work happening? thanks, greg k-h
On 9/27/2023 1:46 PM, Greg Kroah-Hartman wrote: > On Wed, Sep 27, 2023 at 01:00:27PM +0530, Krishna Kurapati wrote: >> In the event gadget_connect call (which invokes pullup) fails, >> propagate the error to udc bind operation which in turn sends the >> error to configfs. The userspace can then retry enumeration if >> it chooses to. > > Will this break userspace that is not expecting error codes to be > returned? What userspace code will now be modified to handle this? > Where is that work happening? > > thanks, Hi Greg, This only handles cases where the pullup failed but the error code wasn't returned to configfs_udc_store approrpiately. In userspace when we do the following: echo "UDC NAME" > /usb_gadget/<>/UDC in the issue I was facing, the core soft reset was failing and we return -110 from dwc3/core.c to udc's bind_to_driver call, but it is not checked any where today and we return 0 to udc_store in configfs. The userspace assumes the UDC write went through (and enum happened) but it actually doesn't. If we propagate the -ETIMEDOUT coming from dwc3-core all the way to configfs_udc_store, it will reach userspace as well indicating that the echo command was not successful which then the user can choose to retry or not. On Android devices, I saw userspace retry happening when we propagate dwc3-core error to userspace. So nothing in userspace breaks. It can stay as it but this time, the echo in userspace will fail if pullup fails for some reason. Regards, Krishna,
On Wed, Sep 27, 2023 at 01:00:27PM +0530, Krishna Kurapati wrote: > In the event gadget_connect call (which invokes pullup) fails, > propagate the error to udc bind operation which in turn sends the > error to configfs. The userspace can then retry enumeration if > it chooses to. > > Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com> > --- > Changes in v5: Addressed proper unlocking of control_lock mutex Reviewed-by: Alan Stern <stern@rowland.harvard.edu> > drivers/usb/gadget/udc/core.c | 19 +++++++++++++++---- > 1 file changed, 15 insertions(+), 4 deletions(-) > > diff --git a/drivers/usb/gadget/udc/core.c b/drivers/usb/gadget/udc/core.c > index 7d49d8a0b00c..824fe64e078a 100644 > --- a/drivers/usb/gadget/udc/core.c > +++ b/drivers/usb/gadget/udc/core.c > @@ -1125,12 +1125,12 @@ EXPORT_SYMBOL_GPL(usb_gadget_set_state); > /* ------------------------------------------------------------------------- */ > > /* Acquire connect_lock before calling this function. */ > -static void usb_udc_connect_control_locked(struct usb_udc *udc) __must_hold(&udc->connect_lock) > +static int usb_udc_connect_control_locked(struct usb_udc *udc) __must_hold(&udc->connect_lock) > { > if (udc->vbus) > - usb_gadget_connect_locked(udc->gadget); > + return usb_gadget_connect_locked(udc->gadget); > else > - usb_gadget_disconnect_locked(udc->gadget); > + return usb_gadget_disconnect_locked(udc->gadget); > } > > static void vbus_event_work(struct work_struct *work) > @@ -1604,12 +1604,23 @@ static int gadget_bind_driver(struct device *dev) > } > usb_gadget_enable_async_callbacks(udc); > udc->allow_connect = true; > - usb_udc_connect_control_locked(udc); > + ret = usb_udc_connect_control_locked(udc); > + if (ret) > + goto err_connect_control; > + > mutex_unlock(&udc->connect_lock); > > kobject_uevent(&udc->dev.kobj, KOBJ_CHANGE); > return 0; > > + err_connect_control: > + udc->allow_connect = false; > + usb_gadget_disable_async_callbacks(udc); > + if (gadget->irq) > + synchronize_irq(gadget->irq); > + usb_gadget_udc_stop_locked(udc); > + mutex_unlock(&udc->connect_lock); > + > err_start: > driver->unbind(udc->gadget); > > -- > 2.42.0 >
diff --git a/drivers/usb/gadget/udc/core.c b/drivers/usb/gadget/udc/core.c index 7d49d8a0b00c..824fe64e078a 100644 --- a/drivers/usb/gadget/udc/core.c +++ b/drivers/usb/gadget/udc/core.c @@ -1125,12 +1125,12 @@ EXPORT_SYMBOL_GPL(usb_gadget_set_state); /* ------------------------------------------------------------------------- */ /* Acquire connect_lock before calling this function. */ -static void usb_udc_connect_control_locked(struct usb_udc *udc) __must_hold(&udc->connect_lock) +static int usb_udc_connect_control_locked(struct usb_udc *udc) __must_hold(&udc->connect_lock) { if (udc->vbus) - usb_gadget_connect_locked(udc->gadget); + return usb_gadget_connect_locked(udc->gadget); else - usb_gadget_disconnect_locked(udc->gadget); + return usb_gadget_disconnect_locked(udc->gadget); } static void vbus_event_work(struct work_struct *work) @@ -1604,12 +1604,23 @@ static int gadget_bind_driver(struct device *dev) } usb_gadget_enable_async_callbacks(udc); udc->allow_connect = true; - usb_udc_connect_control_locked(udc); + ret = usb_udc_connect_control_locked(udc); + if (ret) + goto err_connect_control; + mutex_unlock(&udc->connect_lock); kobject_uevent(&udc->dev.kobj, KOBJ_CHANGE); return 0; + err_connect_control: + udc->allow_connect = false; + usb_gadget_disable_async_callbacks(udc); + if (gadget->irq) + synchronize_irq(gadget->irq); + usb_gadget_udc_stop_locked(udc); + mutex_unlock(&udc->connect_lock); + err_start: driver->unbind(udc->gadget);
In the event gadget_connect call (which invokes pullup) fails, propagate the error to udc bind operation which in turn sends the error to configfs. The userspace can then retry enumeration if it chooses to. Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com> --- Changes in v5: Addressed proper unlocking of control_lock mutex drivers/usb/gadget/udc/core.c | 19 +++++++++++++++---- 1 file changed, 15 insertions(+), 4 deletions(-)