| Message ID | 20230822112455.18957-1-oneukum@suse.com |
|---|---|
| State | New |
| Headers | show |
| Series | USB: dwc2: hande irq on dead controller correctly | expand |
Hi Oliver, On 8/22/23 15:24, Oliver Neukum wrote: > If the controller is dead, the honest answer to the > question whether it has caused an irq is: unknown > As the purpose of the irq return is to trigger switching > off an IRQ, the correct response if you cannot > determine if your device has caused the interrupt is > IRQ_HANDLED > If core is dead then it can not assert interrupt, it's more possible that it's spurious interrupt or interrupt from other device (IRQ is shared) and nothing should be handled. In this case reply by IRQ_HANDLED is incorrect because nothing is handled. Thanks, Minas > Signed-off-by: Oliver Neukum <oneukum@suse.com> > --- > drivers/usb/dwc2/hcd_intr.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/drivers/usb/dwc2/hcd_intr.c b/drivers/usb/dwc2/hcd_intr.c > index c9740caa5974..0144ca8350c3 100644 > --- a/drivers/usb/dwc2/hcd_intr.c > +++ b/drivers/usb/dwc2/hcd_intr.c > @@ -2203,11 +2203,13 @@ static void dwc2_hc_intr(struct dwc2_hsotg *hsotg) > irqreturn_t dwc2_handle_hcd_intr(struct dwc2_hsotg *hsotg) > { > u32 gintsts, dbg_gintsts; > - irqreturn_t retval = IRQ_NONE; > + irqreturn_t retval = IRQ_HANDLED; > > if (!dwc2_is_controller_alive(hsotg)) { > dev_warn(hsotg->dev, "Controller is dead\n"); > return retval; > + } else { > + retval = IRQ_NONE; > } > > spin_lock(&hsotg->lock);
On 28.08.23 12:52, Minas Harutyunyan wrote: Hi, > If core is dead then it can not assert interrupt, it's more possible No, but it may have asserted the interrupt and then died. > that it's spurious interrupt or interrupt from other device (IRQ is > shared) and nothing should be handled. In this case reply by IRQ_HANDLED > is incorrect because nothing is handled. In this case we cannot be sure if it was our interrupt. IRQ_NONE however means that we are sure. Likelihood is not a relevant question here. We must know for sure if we claim IRQ_NONE, because IRQ_NONE has consequences. Regards Oliver
diff --git a/drivers/usb/dwc2/hcd_intr.c b/drivers/usb/dwc2/hcd_intr.c index c9740caa5974..0144ca8350c3 100644 --- a/drivers/usb/dwc2/hcd_intr.c +++ b/drivers/usb/dwc2/hcd_intr.c @@ -2203,11 +2203,13 @@ static void dwc2_hc_intr(struct dwc2_hsotg *hsotg) irqreturn_t dwc2_handle_hcd_intr(struct dwc2_hsotg *hsotg) { u32 gintsts, dbg_gintsts; - irqreturn_t retval = IRQ_NONE; + irqreturn_t retval = IRQ_HANDLED; if (!dwc2_is_controller_alive(hsotg)) { dev_warn(hsotg->dev, "Controller is dead\n"); return retval; + } else { + retval = IRQ_NONE; } spin_lock(&hsotg->lock);
If the controller is dead, the honest answer to the question whether it has caused an irq is: unknown As the purpose of the irq return is to trigger switching off an IRQ, the correct response if you cannot determine if your device has caused the interrupt is IRQ_HANDLED Signed-off-by: Oliver Neukum <oneukum@suse.com> --- drivers/usb/dwc2/hcd_intr.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)