agree your findings.

so it's better move the synchronize_irq() after the spin_unlock_irqrestore().
static int dwc3_suspend_common(struct dwc3 *dwc)
{
	unsigned long	flags;

	switch (dwc->dr_mode) {
	case USB_DR_MODE_PERIPHERAL:
	case USB_DR_MODE_OTG:
		spin_lock_irqsave(&dwc->lock, flags);
		dwc3_gadget_suspend(dwc);
		spin_unlock_irqrestore(&dwc->lock, flags);
		synchronize_irq()

-----Original Message-----
From: Marek Szyprowski <m.szyprowski@samsung.com> 
Sent: Monday, January 28, 2019 10:53 PM
To: Felipe Balbi <redacted>; Linux USB <redacted>
Cc: He, Bo <redacted>; Wang, Yu Y <redacted>; 'Linux Samsung SOC' <redacted>
Subject: Re: [PATCH] usb: dwc3: gadget: synchronize_irq dwc irq in suspend

Hi Felipe,

On 2019-01-28 14:30, Felipe Balbi wrote:

Marek Szyprowski [off-list ref] writes:

quoted

On 2019-01-14 09:30, Felipe Balbi wrote:

quoted

From: Bo He <redacted>

We see dwc3 endpoint stopped by unwanted irq during suspend resume 
test, which is caused dwc3 ep can't be started with error "No 
Resource".

Here, add synchronize_irq before suspend to sync the pending IRQ 
handlers complete.

Signed-off-by: Bo He <redacted>
Signed-off-by: Yu Wang <redacted>
Signed-off-by: Felipe Balbi <redacted>

This patch causes following kernel BUG on Samsung Exynos based 
platforms during system suspend/resume cycle:

are you calling ->suspend() from ->suspend_noirq() time? Are we not 
allowed to call synchronize_irq() during ->suspend()?

dwc3_suspend_common() calls dwc3_gadget_suspend() with dwc->lock spinlock held. This is not the proper context for calling sleeping functions like synchronize_irq().

Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland