"RTOS unfriendly" in chipidea udc code

"RTOS unfriendly" in chipidea udc code

[Jeaho Hwang]

Hi RT experts.

We are reviewing device drivers of our product board and found qualmy
codes in chipidea usb device, 4.19.0 applied RT patch(still exists in
current)

The function hw_usb_reset, called by isr_reset_handler, is using
udelay to wait hw reset and commented "RTOS unfriendly". Can I be sure
RT-safe is it? Could you suggest any better alternative?

Thanks. Regards

drivers/usb/chipidea/udc.c:325
/**
 * hw_usb_reset: restart device after a bus reset (execute without
 *               interruption)
 * @ci: the controller
 *
 * This function returns an error code
 */
static int hw_usb_reset(struct ci_hdrc *ci)
{
hw_usb_set_address(ci, 0);

/* ESS flushes only at end?!? */
hw_write(ci, OP_ENDPTFLUSH,    ~0, ~0);

/* clear setup token semaphores */
hw_write(ci, OP_ENDPTSETUPSTAT, 0,  0);

/* clear complete status */
hw_write(ci, OP_ENDPTCOMPLETE,  0,  0);

/* wait until all bits cleared */
while (hw_read(ci, OP_ENDPTPRIME, ~0))
udelay(10);             /* not RTOS friendly */
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/* reset all endpoints ? */

/* reset internal status and wait for further instructions
  no need to verify the port reset status (ESS does it) */

return 0;
}

-- 
황재호, Jay Hwang, linux team manager of RTst
010-7242-1593

Re: "RTOS unfriendly" in chipidea udc code

[Sebastian Andrzej Siewior]

On 2020-09-14 10:01:46 [+0900], Jeaho Hwang wrote:
> The function hw_usb_reset, called by isr_reset_handler, is using
> udelay to wait hw reset and commented "RTOS unfriendly". Can I be sure
> RT-safe is it? Could you suggest any better alternative?

If I traced it right, that function is used as an interrupt handler
which should run threaded on PREEMPT_RT (it should be requested via
request_irq() or so). That means that the udelay() can be preempted by a
task with higher priority.

> Thanks. Regards

Sebastian