Re: [PATCH 1/5] clocksource: st_lpc: Add LPC timer as a clocksource.
Hi Thomas,
On Tue, 21 Apr 2015, Thomas Gleixner wrote:
> On Fri, 17 Apr 2015, Peter Griffin wrote:
> > +/* Low Power Timer */
> > +#define LPC_LPT_LSB_OFF0x400
> > +#define LPC_LPT_MSB_OFF0x404
> > +#define LPC_LPT_START_OFF 0x408
> > +
> > +struct st_lpc {
> > + struct clk *clk;
> > + void __iomem *iomem_cs;
> > +};
> > +
> > +static struct st_lpc *st_lpc;
> > +
> > +static u64 notrace st_lpc_counter_read(void)
> > +{
> > + u64 counter;
> > + u32 lower;
> > + u32 upper, old_upper;
> > +
> > + upper = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_MSB_OFF);
> > + do {
> > + old_upper = upper;
> > + lower = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_LSB_OFF);
> > + upper = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_MSB_OFF);
> > + } while (upper != old_upper);
> > +
> > + counter = upper;
> > + counter <<= 32;
> > + counter |= lower;
> > + return counter;
>
>
> What's the point of this exercise? The kernel can handle 32bit
> clocksources nicely. So why do you want to artificially expand them to
> 64bit by adding useless loops and hoops to a hotpath?
Thanks for reviewing, yes your correct, we don't really need to be
doing this. We will fix this in the v2 version.
Lee Jones has also pointed out something which I was not aware of when
submitting this, and that is this IP has shared registers with the RTC
driver.
He has been working on some patches here https://lkml.org/lkml/2015/4/9/609
which deal with configuring between watchdog and rtc and ensuring only one
can be configured at a time, and this driver should also be making use of
that mechanism.
As he is already working on this, we decided it would be best if he takes
over for the V2 submission, and aligns it with the work he has already done
for this IP.
regards,
Peter.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 1/5] clocksource: st_lpc: Add LPC timer as a clocksource.
Hello Thomas,
On 04/21/2015 11:56 AM, Thomas Gleixner wrote:
On Fri, 17 Apr 2015, Peter Griffin wrote:
+/* Low Power Timer */
+#define LPC_LPT_LSB_OFF0x400
+#define LPC_LPT_MSB_OFF0x404
+#define LPC_LPT_START_OFF 0x408
+
+struct st_lpc {
+ struct clk *clk;
+ void __iomem *iomem_cs;
+};
+
+static struct st_lpc *st_lpc;
+
+static u64 notrace st_lpc_counter_read(void)
+{
+ u64 counter;
+ u32 lower;
+ u32 upper, old_upper;
+
+ upper = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_MSB_OFF);
+ do {
+ old_upper = upper;
+ lower = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_LSB_OFF);
+ upper = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_MSB_OFF);
+ } while (upper != old_upper);
+
+ counter = upper;
+ counter <<= 32;
+ counter |= lower;
+ return counter;
What's the point of this exercise? The kernel can handle 32bit
clocksources nicely. So why do you want to artificially expand them to
64bit by adding useless loops and hoops to a hotpath?
I agree we should use only the lower 32 bits.
This timer is moreover clocked at a slow rate (30MHz), so we won't have
too much interrupts.
Peter, doing that, you could use something like this directly:
clocksource_mmio_init(st_lpc->iomem_cs + LPC_LPT_LSB_OFF,
"st_lpc_timer", rate,
200, 24, clocksource_mmio_readl_down);
Thanks,
Maxime
Thanks,
tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 1/5] clocksource: st_lpc: Add LPC timer as a clocksource.
On Fri, 17 Apr 2015, Peter Griffin wrote:
> +/* Low Power Timer */
> +#define LPC_LPT_LSB_OFF 0x400
> +#define LPC_LPT_MSB_OFF 0x404
> +#define LPC_LPT_START_OFF0x408
> +
> +struct st_lpc {
> + struct clk *clk;
> + void __iomem *iomem_cs;
> +};
> +
> +static struct st_lpc *st_lpc;
> +
> +static u64 notrace st_lpc_counter_read(void)
> +{
> + u64 counter;
> + u32 lower;
> + u32 upper, old_upper;
> +
> + upper = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_MSB_OFF);
> + do {
> + old_upper = upper;
> + lower = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_LSB_OFF);
> + upper = readl_relaxed(st_lpc->iomem_cs + LPC_LPT_MSB_OFF);
> + } while (upper != old_upper);
> +
> + counter = upper;
> + counter <<= 32;
> + counter |= lower;
> + return counter;
What's the point of this exercise? The kernel can handle 32bit
clocksources nicely. So why do you want to artificially expand them to
64bit by adding useless loops and hoops to a hotpath?
Thanks,
tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 1/5] clocksource: st_lpc: Add LPC timer as a clocksource.
Hi Paul,
On Fri, 17 Apr 2015, Paul Bolle wrote:
> On Fri, 2015-04-17 at 11:50 +0100, Peter Griffin wrote:
> > --- a/drivers/clocksource/Kconfig
> > +++ b/drivers/clocksource/Kconfig
>
> > +config CLKSRC_ST_LPC_CLOCK
> > + bool
> > + depends on ARCH_STI
> > + select CLKSRC_OF if OF
> > + help
> > + Enable this option to use the Low Power controller timer
> > + as clock source.
> > +
> > +config CLKSRC_ST_LPC_TIMER_SCHED_CLOCK
> > + bool
> > + depends on ST_LPC_CLOCK
>
> It looks like you meant
> depends on CLKSRC_ST_LPC_CLOCK
Yes your correct, will fix in v2. I did a last minute change to append CLKSRC_,
and seem to have messed it up.
> > --- /dev/null
> > +++ b/drivers/clocksource/st_lpc.c
>
> > +#ifdef CONFIG_CLKSRC_LPC_TIMER_SCHED_CLOCK
>
> #ifdef CONFIG_CLKSRC_ST_LPC_TIMER_SCHED_CLOCK here?
Yes, will fix in v2
>
> > +static u64 notrace st_lpc_sched_clock_read(void)
> > +{
> > + return st_lpc_counter_read();
> > +}
> > +#endif
>
> > +#ifdef CONFIG_CLKSRC_LPC_TIMER_SCHED_CLOCK
>
> Again, #ifdef CONFIG_CLKSRC_ST_LPC_TIMER_SCHED_CLOCK here?
Yes, will fix in v2
>
> > + sched_clock_register(st_lpc_sched_clock_read, 64, rate);
> > +#endif
>
> Assuming the above suggestions are correct: checkkconfigsymbols.py, as
> shipped in linux-next, helps detect stuff like this. See
> scripts/checkkconfigsymbols.py --help.
Thanks for the pointer to the script, I'd not heard of that before, so
will take a look.
Thanks for reviewing,
regards,
Peter.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 1/5] clocksource: st_lpc: Add LPC timer as a clocksource.
On Fri, 2015-04-17 at 11:50 +0100, Peter Griffin wrote:
> --- a/drivers/clocksource/Kconfig
> +++ b/drivers/clocksource/Kconfig
> +config CLKSRC_ST_LPC_CLOCK
> + bool
> + depends on ARCH_STI
> + select CLKSRC_OF if OF
> + help
> + Enable this option to use the Low Power controller timer
> + as clock source.
> +
> +config CLKSRC_ST_LPC_TIMER_SCHED_CLOCK
> + bool
> + depends on ST_LPC_CLOCK
It looks like you meant
depends on CLKSRC_ST_LPC_CLOCK
> + default y
> + help
> + Use Low Power controller timer clock source as sched_clock
> --- a/drivers/clocksource/Makefile
> +++ b/drivers/clocksource/Makefile
> +obj-$(CONFIG_CLKSRC_ST_LPC_CLOCK)+= st_lpc.o
> --- /dev/null
> +++ b/drivers/clocksource/st_lpc.c
> +#ifdef CONFIG_CLKSRC_LPC_TIMER_SCHED_CLOCK
#ifdef CONFIG_CLKSRC_ST_LPC_TIMER_SCHED_CLOCK here?
> +static u64 notrace st_lpc_sched_clock_read(void)
> +{
> + return st_lpc_counter_read();
> +}
> +#endif
> +#ifdef CONFIG_CLKSRC_LPC_TIMER_SCHED_CLOCK
Again, #ifdef CONFIG_CLKSRC_ST_LPC_TIMER_SCHED_CLOCK here?
> + sched_clock_register(st_lpc_sched_clock_read, 64, rate);
> +#endif
Assuming the above suggestions are correct: checkkconfigsymbols.py, as
shipped in linux-next, helps detect stuff like this. See
scripts/checkkconfigsymbols.py --help.
Thanks,
Paul Bolle
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
