From: Felipe Balbi <felipe.balbi@linux.intel.com>
To: Roger Quadros <rogerq@ti.com>,
        Mathias Nyman <mathias.nyman@linux.intel.com>
Cc: vivek.gautam@codeaurora.org, linux-usb@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 3/4] usb: dwc3: add dual-role support
In-Reply-To: <46440b24-a2cd-93d6-e92e-2d80482f7d32@ti.com>
References: <1487250377-13653-1-git-send-email-rogerq@ti.com> <1487250377-13653-4-git-send-email-rogerq@ti.com> <87r31hhd7g.fsf@linux.intel.com> <46440b24-a2cd-93d6-e92e-2d80482f7d32@ti.com>
Date: Wed, 29 Mar 2017 16:15:37 +0300
Message-ID: <87a884kyva.fsf@linux.intel.com>
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 21036
Lines: 617


Hi,

Roger Quadros <rogerq@ti.com> writes:
>>> @@ -839,6 +841,505 @@ static int dwc3_core_get_phy(struct dwc3 *dwc)
>>>  	return 0;
>>>  }
>>>  
>>> +static int dwc3_drd_start_host(struct dwc3 *dwc, int on, bool skip);
>>> +static int dwc3_drd_start_gadget(struct dwc3 *dwc, int on);
>>> +
>>> +/* dwc->lock must be held */
>>> +static void dwc3_drd_statemachine(struct dwc3 *dwc, int id, int vbus)
>>> +{
>>> +	enum usb_otg_state new_state;
>>> +	int protocol;
>>> +
>>> +	if (id == dwc->otg_fsm.id && vbus == dwc->otg_fsm.b_sess_vld)
>>> +		return;
>>> +
>>> +	dwc->otg_fsm.id = id;
>>> +	dwc->otg_fsm.b_sess_vld = vbus;
>>> +
>>> +	if (!id) {
>>> +		new_state = OTG_STATE_A_HOST;
>>> +	} else{
>>> +		if (vbus)
>>> +			new_state = OTG_STATE_B_PERIPHERAL;
>>> +		else
>>> +			new_state = OTG_STATE_B_IDLE;
>>> +	}
>>> +
>>> +	if (dwc->otg.state == new_state)
>>> +		return;
>>> +
>>> +	protocol = dwc->otg_fsm.protocol;
>>> +	switch (new_state) {
>>> +	case OTG_STATE_B_IDLE:
>>> +		if (protocol == PROTO_GADGET)
>>> +			dwc3_drd_start_gadget(dwc, 0);
>>> +		else if (protocol == PROTO_HOST)
>>> +			dwc3_drd_start_host(dwc, 0, 0);
>>> +		dwc->otg_fsm.protocol = PROTO_UNDEF;
>>> +		break;
>>> +	case OTG_STATE_B_PERIPHERAL:
>>> +		if (protocol == PROTO_HOST)
>>> +			dwc3_drd_start_host(dwc, 0, 0);
>>> +
>>> +		if (protocol != PROTO_GADGET) {
>>> +			dwc->otg_fsm.protocol = PROTO_GADGET;
>>> +			dwc3_drd_start_gadget(dwc, 1);
>>> +		}
>>> +		break;
>>> +	case OTG_STATE_A_HOST:
>>> +		if (protocol == PROTO_GADGET)
>>> +			dwc3_drd_start_gadget(dwc, 0);
>>> +
>>> +		if (protocol != PROTO_HOST) {
>>> +			dwc->otg_fsm.protocol = PROTO_HOST;
>>> +			dwc3_drd_start_host(dwc, 1, 0);
>>> +		}
>>> +		break;
>>> +	default:
>>> +		dev_err(dwc->dev, "drd: invalid usb-drd state: %s\n",
>>> +			usb_otg_state_string(new_state));
>>> +		return;
>>> +	}
>>> +
>>> +	dwc->otg.state = new_state;
>>> +}
>> 
>> I think I've mentioned this before. Why don't we start with the simplest
>> possible implementation? Something that *just* allows us to get ID pin
>> value and set the mode. After that's stable, then we add more pieces to
>> the mix.
>
> That is exactly what I'm doing. Maybe the switch case is making it look complicated.
> dwc3_drd_statemachine() has only 2 inputs VBUS and ID.
>
> I can change it to if/else if you prefer that. I like the way it is cause
> we can define 3 states IDLE, PERIPHERAL and HOST.

Right, I like the three states, but somehow the code looks really
complex :-s

>> For something that simple, we wouldn't even need to use OTG FSM layer
>> because that brings no benefit for such a simple requirement.
>
> no no. I think you got it wrong. I'm not using the OTG FSM layer at all :).

what are all the otg_fsm mentions then? Also you have:

 +	struct usb_otg		otg;
 +	struct otg_fsm		otg_fsm;

>>> +/* dwc->lock must be held */
>>> +static void dwc3_otg_fsm_sync(struct dwc3 *dwc)
>>> +{
>>> +	u32 reg;
>>> +	int id, vbus;
>>> +
>>> +	/*
>>> +	 * calling dwc3_otg_fsm_sync() during resume breaks host
>>> +	 * if adapter was removed during suspend as xhci driver
>>> +	 * is not prepared to see hcd removal before xhci_resume.
>>> +	 */
>>> +	if (dwc->otg_prevent_sync)
>>> +		return;
>>> +
>>> +	reg = dwc3_readl(dwc->regs, DWC3_OSTS);
>>> +	id = !!(reg & DWC3_OSTS_CONIDSTS);
>>> +	vbus = !!(reg & DWC3_OSTS_BSESVLD);
>>> +	dwc3_drd_statemachine(dwc, id, vbus);
>>> +}
>> 
>> Just consider this for a moment. Consider the steps taken to get here.
>> 
>> 	- User plugs cable
>>         - Hardirq handler run
>>        	- read register
>>         - if (reg) return IRQ_WAKE_THREAD;
>>         - schedule bottom half handler to sometime in the future
>>         - scheduler runs our threaded handler
>>         - lock dwc3
>>         - if (host)
>>         	- configure register
>>                 - add xHCI device
>> 	- else
>>         	- otg_fsm_sync()
>>                 - drd_statemachine()
>> 		- configure registers
>>                 - reimplement gadget initialization (same thing we do
>> 		    when registering UDC
>> 
>> I mean, just looking at this we can already see that it's really overly
>> complex. Right now we need simple, dead simple. This will limit the
>> possibility of errors.
>
> OK. I can probably get rid of the state machine model and just use if/else?
> Anything else you want me to get rid of?

The workqueue, unless it's really, really necessary and it appears like
it shouldn't be.

We _can_ keep the switch statement. The problem is not the switch
statement, it's the reimplementation of code we already have.

All you do with adding and removing UDC/HCD is already available
somewhere. Perhaps it just needs to be extracted to a function you can
call, but the code is already there, since we need it for
loading/unloading dwc3 itself.

>>> +static void dwc3_drd_work(struct work_struct *work)
>>> +{
>>> +	struct dwc3 *dwc = container_of(work, struct dwc3,
>>> +					otg_work);
>>> +
>>> +	spin_lock(&dwc->lock);
>>> +	dwc3_otg_fsm_sync(dwc);
>>> +	spin_unlock(&dwc->lock);
>>> +}
>> 
>> so this is only called from ->complete(). You mentioned your commit log
>> that calling dwc3_otg_fsm_sync() directly from ->complete() can lock up
>> the system. Why? I have a feeling your locking isn't proper and that's
>> why sometimes it locks up. You introduced a deadlock and to work around
>> it, the "solution" was to add a workqueue.
>> 
>> Can you either confirm or refute the statement above?
>
> The real problem was that if host adapter was removed during a system suspend
> then while resuming XHCI was locking up like below. This is probably because
> we're trying to remove/Halt the HCD in the otg_irq_handler before XHCI driver has resumed?
>
> How can we ensure that we call dwc3_host_exit() only *after* XHCI driver has resumed?

Well, xHCI is our child, so driver model should *already* be
guaranteeing that, no? Specially since you're calling this from
->complete() which happens after ->resume() has been called for the
entire tree. It's true, however, that dwc3's ->complete() will be called
before xhci's ->complete(). Is this the problem you're pointing at? Or
do you mean xHCI is runtime suspended (or runtime resuming) while you
call dwc3_host_exit()? If that's the case, then there's a bug in xHCI
itself.

We should be able to remove a device from platform/pci bus at any time.

> [ 1057.565573] PM: Syncing filesystems ... done.
> [ 1057.573986] PM: Preparing system for sleep (mem)
> [ 1057.580282] Freezing user space processes ... (elapsed 0.001 seconds) done.
> [ 1057.589626] Freezing remaining freezable tasks ... (elapsed 0.001 seconds) done.
> [ 1057.598931] PM: Suspending system (mem)
> [ 1057.617852] PM: suspend of devices complete after 13.163 msecs
> [ 1057.628279] PM: late suspend of devices complete after 4.296 msecs
> [ 1057.635858] PM: noirq resume of devices complete after 0.178 msecs
> [ 1057.642783] PM: noirq suspend of devices failed
> [ 1057.649703] PM: early resume of devices complete after 2.134 msecs
> [ 1057.658046] net eth0: initializing cpsw version 1.15 (0)
> [ 1057.663634] cpsw 48484000.ethernet: initialized cpsw ale version 1.4
> [ 1057.670325] cpsw 48484000.ethernet: ALE Table size 1024
> [ 1057.683322] Generic PHY 48485000.mdio:02: attached PHY driver [Generic PHY] (mii_bus:phy_addr=48485000.mdio:02, irq=-1)
> [ 1057.706453] usb usb1: root hub lost power or was reset
> [ 1057.711847] usb usb2: root hub lost power or was reset
> [ 1057.987078] ata1: SATA link down (SStatus 0 SControl 300)
> [ 1059.762288] cpsw 48484000.ethernet eth0: Link is Up - 100Mbps/Full - flow control rx/tx
>
> [ 1061.846695] PM: resume of devices complete after 4190.473 msecs
> [ 1061.853294] xhci-hcd xhci-hcd.0.auto: remove, state 1
> [ 1061.858644] usb usb2: USB disconnect, device number 1
> [ 1061.890701] xhci-hcd xhci-hcd.0.auto: Host halt failed, -110
> [ 1061.896640] xhci-hcd xhci-hcd.0.auto: Host controller not halted, aborting reset.
> [ 1061.904535] xhci-hcd xhci-hcd.0.auto: USB bus 2 deregistered
> [ 1061.910514] xhci-hcd xhci-hcd.0.auto: remove, state 1
> [ 1061.915848] usb usb1: USB disconnect, device number 1
> [ 1061.921146] usb 1-1: USB disconnect, device number 2
>
>> 
>>> +static void dwc3_otg_disable_events(struct dwc3 *dwc, u32 disable_mask)
>>> +{
>>> +	dwc->oevten &= ~(disable_mask);
>>> +	dwc3_writel(dwc->regs, DWC3_OEVTEN, dwc->oevten);
>>> +}
>> 
>> we should disable OTG events from our top half handler, otherwise top
>> and bottom half can race with each other.
>
> If we disable OTG events then there is a chance that we miss the events that
> happen while they were disabled.

no, they'll be in the register. Once we reenable them, then the IRQ line
will be raised once more and our handler will get scheduled.

> We need a way to mask the OTG events without loosing them while they are masked.
> Do you know how that could be achieved?

masking doesn't clear events. It just masks them. Look at gadget.c for
how we do it. Basically, the code we have here is racy, really racy and
will cause hard-to-debug lockups. Your code can only work with
IRQF_ONESHOT, which we don't want to add back.

In any case, to mask events, you set BIT 31 in GEVNTSIZ register. Just
copy it from gadget.c ;-)

>>> +	spin_unlock(&dwc->lock);
>>> +
>>> +	return IRQ_HANDLED;
>>> +}
>>> +
>>> +static irqreturn_t dwc3_otg_irq(int irq, void *_dwc)
>>> +{
>>> +	u32 reg;
>>> +	struct dwc3 *dwc = _dwc;
>>> +	irqreturn_t ret = IRQ_NONE;
>>> +
>>> +	reg = dwc3_readl(dwc->regs, DWC3_OEVT);
>>> +	if (reg) {
>>> +		if ((dwc->otg_fsm.protocol == PROTO_HOST) &&
>>> +		    !(reg & DWC3_OEVT_DEVICEMODE))
>>> +			dwc->otg_needs_host_start = 1;
>>> +		dwc3_writel(dwc->regs, DWC3_OEVT, reg);
>>> +		ret = IRQ_WAKE_THREAD;
>> 
>> disable_events();
>
> We can't disable events if we don't want to miss any events while they were disabled.

right, disable events is not the best thing, sorry. We should set bit 31
in GEVNTSIZ.

> But I agree that we need to prevent them from firing another hard IRQ.
> I couldn't figure out how that can be done.

gadget.c ;-)

>>> +
>>> +	return ret;
>>> +}
>>> +
>>> +/* --------------------- Dual-Role management ------------------------------- */
>>> +static void dwc3_otgregs_init(struct dwc3 *dwc)
>>> +{
>>> +	u32 reg;
>>> +
>>> +	/*
>>> +	 * Prevent host/device reset from resetting OTG core.
>>> +	 * If we don't do this then xhci_reset (USBCMD.HCRST) will reset
>>> +	 * the signal outputs sent to the PHY, the OTG FSM logic of the
>>> +	 * core and also the resets to the VBUS filters inside the core.
>>> +	 */
>>> +	reg = dwc3_readl(dwc->regs, DWC3_OCFG);
>>> +	reg |= DWC3_OCFG_SFTRSTMASK;
>>> +	dwc3_writel(dwc->regs, DWC3_OCFG, reg);
>>> +
>>> +	/* Disable hibernation for simplicity */
>>> +	reg = dwc3_readl(dwc->regs, DWC3_GCTL);
>>> +	reg &= ~DWC3_GCTL_GBLHIBERNATIONEN;
>>> +	dwc3_writel(dwc->regs, DWC3_GCTL, reg);
>> 
>> no, don't do that. We support hibernation on some Intel devices. You'd
>> be regressing them, most likely.
>
> Do they support dual-role as well? Either ways I can omit this step.

At least for now, let's skip it.

>>> +	/*
>>> +	 * Initialize OTG registers as per
>>> +	 * Figure 11-4 OTG Driver Overall Programming Flow
>>> +	 */
>>> +	/* OCFG.SRPCap = 0, OCFG.HNPCap = 0 */
>>> +	reg = dwc3_readl(dwc->regs, DWC3_OCFG);
>>> +	reg &= ~(DWC3_OCFG_SRPCAP | DWC3_OCFG_HNPCAP);
>> 
>> are you sure *NO* DWC3 implementation out there is SRP capable and HNP
>> capable?
>> 
>> HNP I understand that you want to disable because we're not support full
>> OTG, but SRP should be easy to support and it's also rather handy. In
>> any case, perhaps add a slightly longer comment explaining why you're
>> disabling these?
>
> This is done according to Fig 11.4 in the TRM. IMO it needs to be done
> even if the controller supports SRP and HNP.

I see, fair enough.

>>> +	dwc3_writel(dwc->regs, DWC3_OCFG, reg);
>>> +	/* OEVT = FFFF */
>>> +	dwc3_writel(dwc->regs, DWC3_OEVT, ~0);
>> 
>> hmmm, flushing pending events. Are you sure you can even have them at
>> this point? This should be called after we reset the controller.
>
> This is again as per Fig 11.4 in TRM.

documentation might have made assumptions which don't apply to us :-)

>>> +	/* OEVTEN = 0 */
>>> +	dwc3_otg_disable_events(dwc, DWC3_OTG_ALL_EVENTS);
>>> +	/* OEVTEN.ConIDStsChngEn = 1. Instead we enable all events */
>>> +	dwc3_otg_enable_events(dwc, DWC3_OTG_ALL_EVENTS);
>> 
>> oh, disable everything and enable everything right after. What gives?
>
> I did this following Fig 11.4. But there there don't enable all events,
> so it was a good idea to be on a clean slate by disabling all events first
> and then only enabling selected events.
>
> In any case I think it is good practice. i.e. clear before OR operation?
> FYI. dwc3_otg_enable_events doesn't clear the events that are not enabled so
> if some old event not part of enable_mask was enabled it will stay enabled.

can't this result in IRQ triggering forever and us not handling it? ;-)

>>> +	/*
>>> +	 * OCTL.PeriMode = 1, OCTL.DevSetHNPEn = 0, OCTL.HstSetHNPEn = 0,
>>> +	 * OCTL.HNPReq = 0
>>> +	 */
>>> +	reg = dwc3_readl(dwc->regs, DWC3_OCTL);
>>> +	reg |= DWC3_OCTL_PERIMODE;
>>> +	reg &= ~(DWC3_OCTL_DEVSETHNPEN | DWC3_OCTL_HSTSETHNPEN |
>>> +		 DWC3_OCTL_HNPREQ);
>>> +	dwc3_writel(dwc->regs, DWC3_OCTL, reg);
>>> +}
>>> +
>>> +/* dwc->lock must be held */
>>> +static int dwc3_drd_start_host(struct dwc3 *dwc, int on, bool skip)
>>> +{
>>> +	u32 reg;
>>> +
>>> +	/* switch OTG core */
>>> +	if (on) {
>>> +		/* As per Figure 11-10 A-Device Flow Diagram */
>>> +		/* OCFG.HNPCap = 0, OCFG.SRPCap = 0 */
>>> +		reg = dwc3_readl(dwc->regs, DWC3_OCFG);
>>> +		reg &= ~(DWC3_OCFG_SRPCAP | DWC3_OCFG_HNPCAP);
>> 
>> didn't you do this already? Why do you need to do this again?
>> 
>
> Was blindly following Fig 11-10. Might be necessary whenever we support HNP/SRP.
> I can get rid of it though if you prefer.

please do, minimal support for now ;-)

>>> +		dwc3_writel(dwc->regs, DWC3_OCFG, reg);
>>> +
>>> +		/*
>>> +		 * OCTL.PeriMode=0, OCTL.TermSelDLPulse = 0,
>>> +		 * OCTL.DevSetHNPEn = 0, OCTL.HstSetHNPEn = 0
>>> +		 */
>>> +		reg = dwc3_readl(dwc->regs, DWC3_OCTL);
>>> +		reg &= ~(DWC3_OCTL_PERIMODE | DWC3_OCTL_TERMSELIDPULSE |
>>> +			 DWC3_OCTL_DEVSETHNPEN | DWC3_OCTL_HSTSETHNPEN);
>> 
>> HNP already disabled elsewhere. Why disable it again?
>> 
>
> Strictly following TRM. nothing else. What do you want me to do here?

assume your register writes actually stick :-)

>>> +		dwc3_writel(dwc->regs, DWC3_OCTL, reg);
>>> +
>>> +		/*
>>> +		 * OCFG.DisPrtPwrCutoff = 0/1
>>> +		 */
>>> +		reg = dwc3_readl(dwc->regs, DWC3_OCFG);
>>> +		reg &= ~DWC3_OCFG_DISPWRCUTTOFF;
>>                                           ^^
>>                                           one T enough?
>> 
>>> +		dwc3_writel(dwc->regs, DWC3_OCFG, reg);
>> 
>> should you really always disable port power cutoff ?
>
> If I remember right this should be disabled for non OTG cases else
> core will turn off VBUS after A_WAIT_BCON timeout when port is
> disconnected.

aha, good point :-)

>>> +		/* start the xHCI host driver */
>>> +		if (!skip) {
>> 
>> when would skip be true?
>> 
>
> only during system resume.

hmmm, is there a reason for that? I mean, could we live without it for
the time being? Seems like all this achieves is avoiding reenumeration
of some devices during resume. Do we care from a starting
implementation?

>>> +	} else {
>>> +		/*
>>> +		 * Exit from A-device flow as per
>>> +		 * Figure 11-4 OTG Driver Overall Programming Flow
>>> +		 */
>>> +		/* stop the HCD */
>>> +		if (!skip) {
>>> +			spin_unlock(&dwc->lock);
>>> +			dwc3_host_exit(dwc);
>>> +			spin_lock(&dwc->lock);
>>> +		}
>>> +
>>> +		/*
>>> +		 * OEVTEN.OTGADevBHostEndEvntEn=0, OEVTEN.OTGADevHNPChngEvntEn=0
>>> +		 * OEVTEN.OTGADevSessEndDetEvntEn=0,
>>> +		 * OEVTEN.OTGADevHostEvntEn = 0
>>> +		 * But we don't disable any OTG events
>> 
>> why not?
>
> because we kept all of them enabled based on your suggestion last year
> (unlike what TRM says)

Hmm, I see. I, clearly, forgot what I said. :-p Sorry

>>> +		 */
>>> +
>>> +		/* OCTL.HstSetHNPEn = 0, OCTL.PrtPwrCtl=0 */
>>> +		reg = dwc3_readl(dwc->regs, DWC3_OCTL);
>>> +		reg &= ~(DWC3_OCTL_HSTSETHNPEN | DWC3_OCTL_PRTPWRCTL);
>> 
>> disabled elsewhere. Why do it again?
>
> I can optimize it away if you prefer.

we gotta start with an assumption that the HW works. If it doesn't, we
quirk it out.

>>> +		dwc3_writel(dwc->regs, DWC3_OCTL, reg);
>>> +
>>> +		/* Initialize OTG registers */
>>> +		dwc3_otgregs_init(dwc);
>> 
>> again you reinitialize everything. Why so many reinitializations? Seems
>> like you were having issues getting this to work and ended up with silly
>> reinitializations and workqueues in an effort to get it working.
>
> Felipe, last year you told me to strictly follow the TRM programming model.
> This is what it says to do. Please refer Fig 11.4
>
> I know some things are silly but I deliberately didn't optimize them.
> If you want to now not strictly follow the TRM I'm fine with that as well.

I see what you're doing now.

>> This patch gives me the impression that the feature hasn't been tested
>> properly. :-s
>
> It is currently undergoing testing for TI release. So far there haven't been
> any surprises.

good to know

>>> +		dwc3_writel(dwc->regs, DWC3_GCTL, reg);
>>> +
>>> +		/* start the Peripheral driver  */
>>> +		if (dwc->gadget_driver) {
>>> +			__dwc3_gadget_start(dwc);
>>> +			if (dwc->gadget_pullup)
>>> +				dwc3_gadget_run_stop(dwc, true, false);
>> 
>> why don't you add/remove the UDC just like you do for the HCD? (you
>> wouldn't add/remove a device, but rather call
>> usb_del_gadget_udc()/usb_add_gadget_udc() directly. Would that clean up
>> some of this?
>
> It causes more problems than solving anything.
> e.g. An already loaded gadget driver will have to be manually removed and re-loaded to
> work after a peripheral to host to peripheral mode switch.

is that really still true? When we remove the UDC, the currently-loaded
gadget driver will be moved to the pending list. Once a UDC is added
back, udc-core will bind it again to the UDC.

>>> +	dwc3_writel(dwc->regs, DWC3_GUSB2PHYCFG(0), reg);
>>> +
>>> +	/* Initialize OTG registers */
>>> +	dwc3_otgregs_init(dwc);
>>> +
>>> +	/* force drd state machine update the first time */
>>> +	dwc->otg_fsm.b_sess_vld = -1;
>>> +	dwc->otg_fsm.id = -1;
>> 
>> Does this work if you boot with cable already attached? Both host and
>> peripheral cables?
>
> Yes.

fair enough

>>> +static int dwc3_drd_init(struct dwc3 *dwc)
>>> +{
>>> +	int ret, irq;
>>> +	unsigned long flags;
>>> +
>>> +	INIT_WORK(&dwc->otg_work, dwc3_drd_work);
>>> +
>>> +	irq = dwc3_otg_get_irq(dwc);
>>> +	if (irq < 0)
>>> +		return irq;
>>> +
>>> +	dwc->otg_irq = irq;
>>> +
>>> +	/* disable all otg irqs */
>>> +	dwc3_otg_disable_events(dwc, DWC3_OTG_ALL_EVENTS);
>>> +	/* clear all events */
>>> +	dwc3_writel(dwc->regs, DWC3_OEVT, ~0);
>> 
>> 
>> this is really odd. You have a bunch of these duplicated chunks of code
>> all over the place...
>> 
>>> +	irq_set_status_flags(dwc->otg_irq, IRQ_NOAUTOEN);
>> 
>> why?
>
> I don't know how to fix this. I have to do this because dwc3_omap is doing it
> on the shared IRQ line and the flags must match if we need to share it.

hmmm... Then why does dwc_omap IRQ have IRQ_NOAUTOEN and otg_irq doesn't?

>>> +	ret = request_threaded_irq(dwc->otg_irq, dwc3_otg_irq,
>>> +				   dwc3_otg_thread_irq,
>>> +				   IRQF_SHARED, "dwc3-otg", dwc);
>>> +	if (ret) {
>>> +		dev_err(dwc->dev, "failed to request irq #%d --> %d\n",
>>> +			dwc->otg_irq, ret);
>>> +		ret = -ENODEV;
>>> +		return ret;
>>> +	}
>>> +
>>> +	ret = dwc3_gadget_init(dwc);
>> 
>> unconditionally? What if I booted with a micro-A plugged to my port and
>> a USB-stick attached to it?
>
> We are not starting the gadget controller though and we want UDC to be initialized
> anyways so users can load a gadget driver before hand.

Users will be able to load gadget driver and that will be kept in the
pending list until a UDC is loaded.

> This is another point against using usb_del_gadget_udc(). The gadget controller
> is really there and user wants to have a persistant gadget driver loaded
> between host/peripheral mode switches.

see above.

>>> @@ -1827,7 +1835,7 @@ static int dwc3_gadget_start(struct usb_gadget *g,
>>>  	return ret;
>>>  }
>>>  
>>> -static void __dwc3_gadget_stop(struct dwc3 *dwc)
>>> +void __dwc3_gadget_stop(struct dwc3 *dwc)
>> 
>> shouldn't have to. Just rely on usb_add/del_gadget_udc()
>> 
> Please let's not use usb_add/del_gadget_udc(). It causes more trouble
> for user :)

I can't see why it would :-s

> gadget_start/stop has been working beautifully with the benefit of
> user being able to load gadget driver at any time (even when booted
> with host mode) and not worrying about re-loading it between
> host/peripheral role swithces.

If that's still necessary, we have a bug in udc-core. That needs to be
fixed :-)

-- 
balbi