Re: [PATCHv6 6/8] usb: dwc2: gadget: Do not fail probe if there isn't a clock node

[Felipe Balbi <balbi@ti.com>]

[-- Attachment #1: Type: text/plain, Size: 5369 bytes --]

Hi,

On Fri, Oct 31, 2014 at 02:31:31PM -0500, Dinh Nguyen wrote:
> On 10/31/2014 12:42 PM, Felipe Balbi wrote:
> > Hi,
> > 
> > On Fri, Oct 31, 2014 at 10:20:06AM -0500, Dinh Nguyen wrote:
> >>>> @@ -339,7 +339,8 @@ static void dwc2_handle_wakeup_detected_intr(struct dwc2_hsotg *hsotg)
> >>>>  		}
> >>>>  		/* Change to L0 state */
> >>>>  		hsotg->lx_state = DWC2_L0;
> >>>> -		call_gadget(hsotg, resume);
> >>>> +		if (!IS_ERR(hsotg->clk))
> >>>> +			call_gadget(hsotg, resume);
> >>>
> >>> instead of exposing the clock detail to the entire driver, add IS_ERR()
> >>> checks to resume and suspend instead. In fact, NULL is a valid clock, so
> >>> you might as well:
> >>>
> >>> 	clk = clk_get(foo, bar);
> >>> 	if (IS_ERR(clk))
> >>> 		dwc->clk = NULL;
> >>> 	else
> >>> 		dwc->clk = clk;
> >>>
> >>> Then you don't need any IS_ERR() checks sprinkled around the driver.
> >>
> >> But we would still need to check for the clock before accessing gadget
> >> functionality right?
> >>
> >> 	if (dwc2->clk)
> >> 		call_gadget();
> > 
> > Read my comment again. "NULL is a valid clock".  Look at what
> > clk_enable() does when a NULL pointer is passed:
> > 
> > static int __clk_enable(struct clk *clk)
> > {
> > 	int ret = 0;
> > 
> > 	if (!clk)
> > 		return 0;
> > 
> > 	if (WARN_ON(clk->prepare_count == 0))
> > 		return -ESHUTDOWN;
> > 
> > 	if (clk->enable_count == 0) {
> > 		ret = __clk_enable(clk->parent);
> > 
> > 		if (ret)
> > 			return ret;
> > 
> > 		if (clk->ops->enable) {
> > 			ret = clk->ops->enable(clk->hw);
> > 			if (ret) {
> > 				__clk_disable(clk->parent);
> > 				return ret;
> > 			}
> > 		}
> > 	}
> > 
> > 	clk->enable_count++;
> > 	return 0;
> > }
> > 
> > int clk_enable(struct clk *clk)
> > {
> > 	unsigned long flags;
> > 	int ret;
> > 
> > 	flags = clk_enable_lock();
> > 	ret = __clk_enable(clk);
> > 	clk_enable_unlock(flags);
> > 
> > 	return ret;
> > }
> > EXPORT_SYMBOL_GPL(clk_enable);
> 
> Ah yes, thanks for the explanation. So if clk=NULL, it just return 0.
> But what I'm saying is that if the driver is configured for dual-role
> mode, and no clock is specified, then the driver should not be accessing
> any gadget functionality.

why ? Why only for gadget and why can't it work on platforms without
clk? What if Paul wants to run the gadget side on his HAPS-5x platform
configured as a PCIe card ?

You haven't explained why gadget has this hard-dependency on clk and why
*only* gadget has it. This sounds really, really wrong. Why can host
side run without clk but gadget can't ?

Moreover, if you really want to prevent people from using gadget
without clock, fail dwc2_gadget_init() and have the core fallback to
host-only, then print a warning message.

> So as the patch series stands right now, if I swap out an A connector to
> a B-connector, then I get a connect_id_status change interrupt. The
> status would show a device and I would initialize the gadget portion of
> the driver.

that's fine, but why the hard-dependency on clk ?

> diff --git a/drivers/usb/dwc2/hcd.c b/drivers/usb/dwc2/hcd.c
> index 44c609f..96810f7 100644
> --- a/drivers/usb/dwc2/hcd.c
> +++ b/drivers/usb/dwc2/hcd.c
> @@ -1371,7 +1371,8 @@ static void dwc2_conn_id_status_change(struct
> work_struct *work)
>                 hsotg->op_state = OTG_STATE_B_PERIPHERAL;
>                 dwc2_core_init(hsotg, false, -1);
>                 dwc2_enable_global_interrupts(hsotg);
> -               s3c_hsotg_core_init(hsotg);
> +               if (hsotg->clk)
> +                       s3c_hsotg_core_init(hsotg);
> 
> So if I don't have a valid clock, I'll be accessing the peripheral
> portion of the IP.

so what ?

> But I guess not having the check for the valid clock here should be fine
> as I don't see a case where there can be 2 different clocks for host and
> peripheral?

probably the same thing.

> >>>> @@ -400,7 +401,8 @@ static void dwc2_handle_usb_suspend_intr(struct dwc2_hsotg *hsotg)
> >>>>  			"DSTS.Suspend Status=%d HWCFG4.Power Optimize=%d\n",
> >>>>  			!!(dsts & DSTS_SUSPSTS),
> >>>>  			hsotg->hw_params.power_optimized);
> >>>> -		call_gadget(hsotg, suspend);
> >>>> +		if (!IS_ERR(hsotg->clk))
> >>>> +			call_gadget(hsotg, suspend);
> >>>>  	} else {
> >>>>  		if (hsotg->op_state == OTG_STATE_A_PERIPHERAL) {
> >>>>  			dev_dbg(hsotg->dev, "a_peripheral->a_host\n");
> >>>> @@ -477,7 +479,8 @@ irqreturn_t dwc2_handle_common_intr(int irq, void *dev)
> >>>>  	spin_lock(&hsotg->lock);
> >>>>  
> >>>>  	if (dwc2_is_device_mode(hsotg))
> >>>> -		retval = s3c_hsotg_irq(irq, dev);
> >>>> +		if (!IS_ERR(hsotg->clk))
> >>>> +			retval = s3c_hsotg_irq(irq, dev);
> >>>
> >>> wait a minute, if there is no clock we don't call the gadget interrupt
> >>> handler ? Why ? Who will disable the IRQ line ?
> >>
> >> This portion is no static int __clk_enable(struct clk *clk)
> > 
> > huh ? What I mean is that this has the potential of leaving that IRQ
> > line enabled. Imagine you don't have a clock and s3c_hsotg_irq() isn't
> > called, then a peripheral IRQ fires, since the handler isn't called, who
> > will clear the interrupt ?
> > 
> 
> Yes, right. This portion of the code is no longer needed when I switched
> to use a shared IRQ.

k

-- 
balbi

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]