Skip to contentSkip to conversation

Trivial clocksource driver

DISCUSSION Mason 3 participants 7 messages Sep 29, 2015

lore
Share & tools
lore

Mason

author
#0 lore raw

Hello everyone,

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Did I understand correctly? Is this the right place to submit code as provided below?

Regards.

#include <linux/delay.h> /* register_current_timer_delay */
#include <linux/clocksource.h> /* clocksource_register_hz */
#include <linux/sched_clock.h> /* sched_clock_register */
#include <linux/of_address.h> /* of_iomap */
#include <linux/clk.h> /* of_clk_get, clk_get_rate */

static void __iomem *xtal_in_cnt;
static struct delay_timer delay_timer;

static unsigned long read_xtal_counter(void)

{
	return readl_relaxed(xtal_in_cnt);
}

static u64 read_sched_clock(void)

{
	return read_xtal_counter();
}

static cycle_t read_clocksource(struct clocksource *cs)

{
	return read_xtal_counter();
}
static struct clocksource tango_xtal = {
	.name	= "tango-xtal",
	.rating	= 350,
	.read	= read_clocksource,
	.mask	= CLOCKSOURCE_MASK(32),
	.flags	= CLOCK_SOURCE_IS_CONTINUOUS,
};

static void __init tango_clksrc_init(struct device_node *np)

{
	struct clk *clk = of_clk_get(np, 0);
	unsigned int xtal_freq = clk_get_rate(clk);
	xtal_in_cnt = of_iomap(np, 0);

	delay_timer.freq = xtal_freq;
	delay_timer.read_current_timer = read_xtal_counter;
	register_current_timer_delay(&delay_timer);
	sched_clock_register(read_sched_clock, 32, xtal_freq);
	clocksource_register_hz(&tango_xtal, xtal_freq);
}

CLOCKSOURCE_OF_DECLARE(tango, "sigma,xtal_in_cnt", tango_clksrc_init);

#1 lore raw
Quoted 8 lines · Mason

Hello everyone,

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Did I understand correctly? Is this the right place to submit code as provided below?

Yes, drivers/clocksource is the right place. You just need to submit a formal patch, which includes a proper subject line, changelog, plus the necessary Makefile and Kconfig modifications.

Quoted 1 line

#include <linux/delay.h> /* register_current_timer_delay */

Please get rid of these silly tail comments. They provide absolutely no value.

Other than that this looks reasonable.

Thanks,

tglx

author
#2 lore raw
Quoted 6 lines · Thomas Gleixner
Quoted 4 lines · Mason

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Quoted 6 lines
Quoted 2 lines

Did I understand correctly? Is this the right place to submit code as provided below?

Yes, drivers/clocksource is the right place. You just need to submit a formal patch, which includes a proper subject line, changelog, plus the necessary Makefile and Kconfig modifications.

OK, I'll send a formal patch tomorrow.
There are no Kconfig modifications, is that OK?

Also, that patch is part of a larger patch-set (most of the patches intended for arch/arm). I should send you only the clocksource patch, or the whole patch-set?

Quoted 4 lines
Quoted 1 line

#include <linux/delay.h> /* register_current_timer_delay */

Please get rid of these silly tail comments. They provide absolutely no value.

I will remove them, since you asked.

In my opinion, they serve one purpose: if code is refactored, and the function call is removed, the comment is a reminder to also remove the relevant include directive.

Do you disagree?

Quoted 1 line

Other than that this looks reasonable.

Just wanted to ask:
Can register_current_timer_delay, sched_clock_register, and clocksource_register_hz be called in any order?

Regards.

#3 lore raw
Quoted 11 lines · Mason
Quoted 9 lines · Thomas Gleixner
Quoted 6 lines · Mason

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Did I understand correctly? Is this the right place to submit code as provided below?

Quoted 6 lines
Quoted 3 lines

Yes, drivers/clocksource is the right place. You just need to submit a formal patch, which includes a proper subject line, changelog, plus the necessary Makefile and Kconfig modifications.

OK, I'll send a formal patch tomorrow.
There are no Kconfig modifications, is that OK?

Why don't you use my driver? It's even simpler, and it works just fine on the 87xx chip.

https://github.com/mansr/linux-tangox/blob/master/drivers/clocksource/clksrc-tangox.c

Signature · 2 lines
Måns Rullgård
[email protected]
#4 lore raw
Quoted 17 lines · Måns Rullgård
Quoted 15 lines · Mason
Quoted 12 lines · Thomas Gleixner
Quoted 6 lines · Mason

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Did I understand correctly? Is this the right place to submit code as provided below?

Yes, drivers/clocksource is the right place. You just need to submit a formal patch, which includes a proper subject line, changelog, plus the necessary Makefile and Kconfig modifications.

Quoted 7 lines
Quoted 2 lines

OK, I'll send a formal patch tomorrow.
There are no Kconfig modifications, is that OK?

Why don't you use my driver? It's even simpler, and it works just fine on the 87xx chip.

https://github.com/mansr/linux-tangox/blob/master/drivers/clocksource/clksrc-tangox.c

Perhaps because that driver is not upstream either?

Thanks,

tglx

author
#5 lore raw
In reply to Måns Rullgård

Hello Mans,

Quoted 17 lines · Måns Rullgård
Quoted 15 lines · Mason
Quoted 12 lines · Thomas Gleixner
Quoted 6 lines · Mason

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Did I understand correctly? Is this the right place to submit code as provided below?

Yes, drivers/clocksource is the right place. You just need to submit a formal patch, which includes a proper subject line, changelog, plus the necessary Makefile and Kconfig modifications.

Quoted 7 lines
Quoted 2 lines

OK, I'll send a formal patch tomorrow.
There are no Kconfig modifications, is that OK?

Why don't you use my driver? It's even simpler, and it works just fine on the 87xx chip.

https://github.com/mansr/linux-tangox/blob/master/drivers/clocksource/clksrc-tangox.c

As you know, I am using two of your complex drivers, namely ethernet and interrupt controller (which would have taken me several weeks to write). I will be submitting them upstream shortly, is that OK with you?

I'm not using this particular driver of yours because I had already written the code, and porting it to DT was a good learning process. This is probably my only opportunity to actually write any kind of code for this port.

Regards.

#6 lore raw
Quoted 27 lines · Mason

Hello Mans,

Quoted 23 lines · Måns Rullgård
Quoted 11 lines · Mason
Quoted 9 lines · Thomas Gleixner
Quoted 6 lines · Mason

I am trying to submit a new ARM port, and Arnd pointed out that the clocksource code could not live in arch/arm/$PLATFORM, but had to move to drivers/clocksource (and it had to support DT).

Did I understand correctly? Is this the right place to submit code as provided below?

Quoted 6 lines
Quoted 3 lines

Yes, drivers/clocksource is the right place. You just need to submit a formal patch, which includes a proper subject line, changelog, plus the necessary Makefile and Kconfig modifications.

OK, I'll send a formal patch tomorrow.
There are no Kconfig modifications, is that OK?

Why don't you use my driver? It's even simpler, and it works just fine on the 87xx chip.

Quoted 6 lines
Quoted 1 line

https://github.com/mansr/linux-tangox/blob/master/drivers/clocksource/clksrc-tangox.c

As you know, I am using two of your complex drivers, namely ethernet and interrupt controller (which would have taken me several weeks to write). I will be submitting them upstream shortly, is that OK with you?

I want to test them a bit more on the 87xx first. It's booting, but there are a couple of niggles that should be looked into. Getting some documentation would expedite that process.

Quoted 4 lines

I'm not using this particular driver of yours because I had already written the code, and porting it to DT was a good learning process. This is probably my only opportunity to actually write any kind of code for this port.

You should still be using the existing clocksource_mmio helper. In fact, that interface could be exposed using a generic DT binding.

Signature · 2 lines
Måns Rullgård
[email protected]