[PATCH 3/4] ARM: Xilinx: Adding timer support to the platform

[Jamie Iles]

On Mon, Feb 07, 2011 at 07:16:01AM -0700, John Linn wrote:
> > -----Original Message-----
> > From: Jamie Iles [mailto:jamie at jamieiles.com]
> > Sent: Saturday, February 05, 2011 6:04 PM
> > To: John Linn
> > Cc: linux-arm-kernel at lists.infradead.org;
> linux-kernel at vger.kernel.org; linux at arm.linux.org.uk;
> > catalin.marinas at arm.com; glikely at secretlab.ca; Kiran Sutariya
> > Subject: Re: [PATCH 3/4] ARM: Xilinx: Adding timer support to the
> platform
> > 
> > Hi John,
> > 
> > A couple more nitpicks.
> > 
> > Jamie
> > 
> > On Sat, Feb 05, 2011 at 09:17:16AM -0700, John Linn wrote:
> > > +static irqreturn_t xttcpss_clock_event_interrupt(int irq, void
> *dev_id)
> > > +{
> > > +	struct clock_event_device *evt = &xttcpss_clockevent;
> > > +	struct xttcpss_timer *timer = dev_id;
> > > +	u32 ctrl_reg;
> > > +
> > > +	/* Acknowledge the interrupt and call event handler */
> > > +	xttcpss_write(timer->base_addr + XTTCPSS_ISR_OFFSET,
> > > +		xttcpss_read(timer->base_addr + XTTCPSS_ISR_OFFSET));
> > > +
> > > +	if (timer->mode == CLOCK_EVT_MODE_ONESHOT) {
> > > +
> > > +		/* Disable the counter as it would keep running. */
> > > +		ctrl_reg = xttcpss_read(timer->base_addr +
> > > +					XTTCPSS_CNT_CNTRL_OFFSET);
> > > +		ctrl_reg |= ~(XTTCPSS_CNT_CNTRL_ENABLE_MASK);
> > 
> > The clock events framework should reprogram the next event so you
> don't
> > actually need to disable the timer here.  Once the event handler has
> > been called the timer will be reenabled with a new period, and as
> we're
> > running with interrupts disabled here we don't need to stop the timer.
> > 
> 
> What about if there's not another event to be started?

Then the clockevents code can set the mode to CLOCK_EVT_MODE_SHUTDOWN 
and this should disable the timer.  For NOHZ, AFAICT the 
clockevents/tick code will make sure that the next tick always falls 
within your max_delta_ns.

Jamie