On Tue, Jul 03, 2018 at 11:39:05AM +0200, Thomas Gleixner wrote:
> -EEMPTYCHANGELOG
Ok
> > +// SPDX-License-Identifier: GPL-2.0
> > +// Copyright (C) 2018 Hangzhou NationalChip Science & Technology Co.,Ltd.
>
> newline please
Ok
> > +#define BITS_CSKY_TIMER 56
> > +
> > +DECLARE_PER_CPU(struct timer_of, csky_to);
>
> static?
Ok.
>
> > +
> > +static int csky_timer_irq;
> > +static int csky_timer_rate;
> > +
> > +static inline u64 get_ccvr(void)
> > +{
> > + u32 lo, hi, t;
> > +
> > + do {
> > + hi = mfcr(PTIM_CCVR_HI);
> > + lo = mfcr(PTIM_CCVR_LO);
> > + t = mfcr(PTIM_CCVR_HI);
> > + } while(t != hi);
>
> No idea which frequency this timer ticks at, but if the 32 bit wrap does
> not come too fast, then you really should avoid that loop. That function is
> called very frequently.
0000006c <clksrc_read>:
hi = mfcr(PTIM_CCVR_HI);
6c: c1c26023 mfcr r3, cr<2, 14>
lo = mfcr(PTIM_CCVR_LO);
70: c1c36021 mfcr r1, cr<3, 14>
t = mfcr(PTIM_CCVR_HI);
74: c1c26022 mfcr r2, cr<2, 14>
} while(t != hi);
78: 648e cmpne r3, r2
7a: 0bf9 bt 0x6c // 6c <clksrc_read>
When two read cr<2, 14> is not equal, we'll retry. So only when
CCVR_LO is at 0xffffffff between the two read of CCVR_HI. That's very
very small probability event for "bt 0x6c".
Don't worry about the "do {...} whie(t != hi)", it's no performance issue.
> > +DEFINE_PER_CPU(struct timer_of, csky_to) = {
>
> static
Ok.
> > + .flags = TIMER_OF_CLOCK | TIMER_OF_IRQ,
> > +
> > + .clkevt = {
> > + .name = "C-SKY SMP Timer V1",
> > + .rating = 300,
> > + .features = CLOCK_EVT_FEAT_PERCPU | CLOCK_EVT_FEAT_ONESHOT,
> > + .set_state_shutdown = csky_timer_shutdown,
> > + .set_state_oneshot = csky_timer_oneshot,
> > + .set_state_oneshot_stopped = csky_timer_oneshot_stopped,
> > + .set_next_event = csky_timer_set_next_event,
> > + },
> > +
> > + .of_irq = {
> > + .handler = timer_interrupt,
> > + .flags = IRQF_TIMER,
> > + .percpu = 1,
>
> This is inconsistent. You made it half tabular and half not. Please use
> tabular style consistently.
Ok.
.clkevt = {
.name = "C-SKY SMP Timer V1",
.features = CLOCK_EVT_FEAT_PERCPU |
CLOCK_EVT_FEAT_ONESHOT,
.rating = 300,
.set_state_shutdown = csky_timer_shutdown,
.set_state_oneshot = csky_timer_oneshot,
.set_state_oneshot_stopped = csky_timer_oneshot_stopped,
.set_next_event = csky_timer_set_next_event,
},
.of_irq = {
.handler = timer_interrupt,
.flags = IRQF_TIMER,
.percpu = 1,
> > +/*** clock event for percpu ***/
>
> Please refrain from inventing new horrible comment styles.
Ok.
/* clock event for percpu */
> > +struct clocksource csky_clocksource = {
> > + .name = "csky_timer_v1_clksrc",
> > + .rating = 400,
> > + .mask = CLOCKSOURCE_MASK(BITS_CSKY_TIMER),
> > + .flags = CLOCK_SOURCE_IS_CONTINUOUS,
> > + .read = clksrc_read,
>
> tabular style please
Ok.
> > + ret = cpuhp_setup_state(CPUHP_AP_DUMMY_TIMER_STARTING,
> > + "clockevents/csky/timer:starting",
> > + csky_timer_starting_cpu,
> > + csky_timer_dying_cpu);
>
> Oh no. Just picking a random hotplug event is not how it works. Add your
> own please and make sure it's at the proper place.
like this?
include/linux/cpuhotplug.h:
CPUHP_AP_KVM_ARM_TIMER_STARTING,
+ CPUHP_AP_CSKY_TIMER_STARTING,
/* Must be the last timer callback */
CPUHP_AP_DUMMY_TIMER_STARTING,
> > + struct clock_event_device *ce = (struct clock_event_device *) dev;
>
> Pointless type cast.
Ok.
struct clock_event_device *ce = dev;
> > + .flags = IRQF_TIMER | IRQF_IRQPOLL,
> > + },
>
> See above
Ok, tabular
Guo Ren
