Re: [RFC][PATCH] usb: dwc2: Make sure we disconnect the gadget state on reset

From: Felipe Balbi
Date: Mon Oct 31 2016 - 07:19:40 EST



Hi,

John Stultz <john.stultz@xxxxxxxxxx> writes:
> I had seen some odd behavior with HiKey's usb-gadget interface
> that I finally seemed to have chased down. Basically every other
> time I pluged in the OTG port, the gadget interface would
> properly initialize. The other times, I'd get a big WARN_ON
> in dwc2_hsotg_init_fifo() about the fifo_map not being clear.
>
> Ends up If we don't disconnect the gadget state on reset, the
> fifo-map doesn't get cleared properly, which causes WARN_ON
> messages and also results in the device not properly being
> setup as a gadget every other time the OTG port is connected.
>
> So this patch adds a call to dwc2_hsotg_disconnect() in the
> reset path so the state is properly cleared.
>
> With it, the gadget interface initializes properly on every
> plug in.
>
> I don't know if this is actually the right fix, but it seems
> to work well. Feedback would be greatly appreciated!
>
> Cc: Wei Xu <xuwei5@xxxxxxxxxxxxx>
> Cc: Guodong Xu <guodong.xu@xxxxxxxxxx>
> Cc: Chen Yu <chenyu56@xxxxxxxxxx>
> Cc: Amit Pundir <amit.pundir@xxxxxxxxxx>
> Cc: Rob Herring <robh+dt@xxxxxxxxxx>
> Cc: Mark Rutland <mark.rutland@xxxxxxx>
> Cc: John Youn <johnyoun@xxxxxxxxxxxx>
> Cc: Douglas Anderson <dianders@xxxxxxxxxxxx>
> Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> Cc: linux-usb@xxxxxxxxxxxxxxx
> Signed-off-by: John Stultz <john.stultz@xxxxxxxxxx>
> ---
> drivers/usb/dwc2/gadget.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/usb/dwc2/gadget.c b/drivers/usb/dwc2/gadget.c
> index 24fbebc..5505001 100644
> --- a/drivers/usb/dwc2/gadget.c
> +++ b/drivers/usb/dwc2/gadget.c
> @@ -2519,6 +2519,8 @@ void dwc2_hsotg_core_init_disconnected(struct dwc2_hsotg *hsotg,
>
> /* Kill any ep0 requests as controller will be reinitialized */
> kill_all_requests(hsotg, hsotg->eps_out[0], -ECONNRESET);
> + /* Make sure everything is disconnected */
> + dwc2_hsotg_disconnect(hsotg);

Dunno, seems like you're actually working around a different
bug. Looking at USB Reset handler we have:

if (gintsts & (GINTSTS_USBRST | GINTSTS_RESETDET)) {

u32 usb_status = dwc2_readl(hsotg->regs + GOTGCTL);
u32 connected = hsotg->connected;

dev_dbg(hsotg->dev, "%s: USBRst\n", __func__);
dev_dbg(hsotg->dev, "GNPTXSTS=%08x\n",
dwc2_readl(hsotg->regs + GNPTXSTS));

dwc2_writel(GINTSTS_USBRST, hsotg->regs + GINTSTS);

/* Report disconnection if it is not already done. */
dwc2_hsotg_disconnect(hsotg);

if (usb_status & GOTGCTL_BSESVLD && connected)
dwc2_hsotg_core_init_disconnected(hsotg, true);
}

so, dwc2_hsotg_disconnect() is already called from Reset path. What
you're doing here is that you could, potentially, call
driver->disconnect() twice.

The real problem could be your implementation for ->pullup() which
duplicates part of what ->udc_start() does:

static int dwc2_hsotg_pullup(struct usb_gadget *gadget, int is_on)
{
struct dwc2_hsotg *hsotg = to_hsotg(gadget);
unsigned long flags = 0;

dev_dbg(hsotg->dev, "%s: is_on: %d op_state: %d\n", __func__, is_on,
hsotg->op_state);

/* Don't modify pullup state while in host mode */
if (hsotg->op_state != OTG_STATE_B_PERIPHERAL) {
hsotg->enabled = is_on;
return 0;
}

spin_lock_irqsave(&hsotg->lock, flags);
if (is_on) {
hsotg->enabled = 1;
dwc2_hsotg_core_init_disconnected(hsotg, false);
dwc2_hsotg_core_connect(hsotg);
} else {
dwc2_hsotg_core_disconnect(hsotg);
dwc2_hsotg_disconnect(hsotg);
hsotg->enabled = 0;
}

hsotg->gadget.speed = USB_SPEED_UNKNOWN;
spin_unlock_irqrestore(&hsotg->lock, flags);

return 0;
}

Here's what I think dwc2_hsotg_pullup() and dwc2_hsotg_udc_start()
should contain:


diff --git a/drivers/usb/dwc2/gadget.c b/drivers/usb/dwc2/gadget.c
index 9dc6c482b89e..dbe28947d3a9 100644
--- a/drivers/usb/dwc2/gadget.c
+++ b/drivers/usb/dwc2/gadget.c
@@ -3388,6 +3388,7 @@ static void dwc2_hsotg_init(struct dwc2_hsotg *hsotg)
dwc2_writel(0, hsotg->regs + DAINTMSK);

/* Be in disconnected state until gadget is registered */
+ /* REVISIT!!!! This should be done in probe()!!!! */
__orr32(hsotg->regs + DCTL, DCTL_SFTDISCON);

/* setup fifos */
@@ -3428,26 +3429,6 @@ static int dwc2_hsotg_udc_start(struct usb_gadget *gadget,
unsigned long flags;
int ret;

- if (!hsotg) {
- pr_err("%s: called with no device\n", __func__);
- return -ENODEV;
- }
-
- if (!driver) {
- dev_err(hsotg->dev, "%s: no driver\n", __func__);
- return -EINVAL;
- }
-
- if (driver->max_speed < USB_SPEED_FULL)
- dev_err(hsotg->dev, "%s: bad speed\n", __func__);
-
- if (!driver->setup) {
- dev_err(hsotg->dev, "%s: missing entry points\n", __func__);
- return -EINVAL;
- }
-
- WARN_ON(hsotg->driver);
-
driver->driver.bus = NULL;
hsotg->driver = driver;
hsotg->gadget.dev.of_node = hsotg->dev->of_node;
@@ -3550,17 +3531,15 @@ static int dwc2_hsotg_pullup(struct usb_gadget *gadget, int is_on)
/* Don't modify pullup state while in host mode */
if (hsotg->op_state != OTG_STATE_B_PERIPHERAL) {
hsotg->enabled = is_on;
- return 0;
+ return -EINVAL;
}

spin_lock_irqsave(&hsotg->lock, flags);
if (is_on) {
hsotg->enabled = 1;
- dwc2_hsotg_core_init_disconnected(hsotg, false);
dwc2_hsotg_core_connect(hsotg);
} else {
dwc2_hsotg_core_disconnect(hsotg);
- dwc2_hsotg_disconnect(hsotg);
hsotg->enabled = 0;
}


And BTW, your usb_gadget_ops had indentation problems, here's a fix for
that:

@@ -3617,10 +3596,10 @@ static int dwc2_hsotg_vbus_draw(struct usb_gadget *gadget, unsigned mA)
}

static const struct usb_gadget_ops dwc2_hsotg_gadget_ops = {
- .get_frame = dwc2_hsotg_gadget_getframe,
+ .get_frame = dwc2_hsotg_gadget_getframe,
.udc_start = dwc2_hsotg_udc_start,
.udc_stop = dwc2_hsotg_udc_stop,
- .pullup = dwc2_hsotg_pullup,
+ .pullup = dwc2_hsotg_pullup,
.vbus_session = dwc2_hsotg_vbus_session,
.vbus_draw = dwc2_hsotg_vbus_draw,
};

John Y, it sees like dwc2 needs quite a bit of work when it comes to its
linkage to the Gadget API, please try to schedule some time to revisit
all of that. Also, the way debugging is done with dwc2 is rather bad,
with a few #ifdef DEBUG in the driver. Some of that information could be
exposed via tracepoints and some would make more sense with debugfs.

Anyway, I don't think $subject is the right fix for the problem
described so I'm not taking it, at least not at this moment or without a
better explanation of how we got to the conclusion that $subject is the
right fix ;-)

cheers

--
balbi

Attachment: signature.asc
Description: PGP signature