Re: [PATCH] usb: gadget: lpc32xx_udc: Fix a race in lpc32xx_get_frame()
From: Greg KH
Date: Fri Nov 20 2026 - 09:21:03 EST
On Tue, Sep 22, 2026 at 02:53:57PM +0800, Ginger Li wrote:
> lpc32xx_get_frame() tests udc->clocked before it takes udc->lock, but
> udc->clocked is cleared by udc_clk_set() with the lock held, for example from
> lpc32xx_vbus_session() when the session is disconnected, which also gates the
> UDC clock.
>
> When the test passes on a stale value, udc_get_current_frame() then reads UDC
> registers whose clock has already been turned off. Check udc->clocked inside
> the critical section so that the state can not change between the check and
> the register access.
>
> Fixes: 24a28e428351 ("USB: gadget driver for LPC32xx")
> Signed-off-by: Ginger Li <ginger.jzllee@xxxxxxxxx>
> ---
> drivers/usb/gadget/udc/lpc32xx_udc.c | 8 +++++---
> 1 file changed, 5 insertions(+), 3 deletions(-)
How was this found and tested?
>
> diff --git a/drivers/usb/gadget/udc/lpc32xx_udc.c b/drivers/usb/gadget/udc/lpc32xx_udc.c
> --- a/drivers/usb/gadget/udc/lpc32xx_udc.c
> +++ b/drivers/usb/gadget/udc/lpc32xx_udc.c
> @@ -2403,10 +2403,12 @@ static int lpc32xx_get_frame(struct usb_gadget *gadget
> unsigned long flags;
> struct lpc32xx_udc *udc = to_udc(gadget);
>
> - if (!udc->clocked)
> - return -EINVAL;
> + spin_lock_irqsave(&udc->lock, flags);
guard()?