| Message ID | 1346344678-18248-1-git-send-email-linus.walleij@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, 30 Aug 2012 18:37:58 +0200 Linus Walleij <linus.walleij@linaro.org> wrote: > This converts the ageing clps711x timer driver to the modern > clocksource/clockevent API. The patch does not account for the changes that I sent to the mailing list recently, where I did support different clock frequencies for the processor. > This was blind-coded using the datasheet, it'd be great if someone > who has this machine could test it. ... > + /* Set both timers to 512 kHz */ The patch uses a second timer, which does not allow its use for other purposes, I can not test the functionality of the device in this case, unfortunately. I have also tried to do something like this without using additional processor resources, but not yet brought to the finish line yet.
On Thu, Aug 30, 2012 at 10:02 AM, Alexander Shiyan <shc_work@mail.ru> wrote: > On Thu, 30 Aug 2012 18:37:58 +0200 > Linus Walleij <linus.walleij@linaro.org> wrote: > >> This converts the ageing clps711x timer driver to the modern >> clocksource/clockevent API. > > The patch does not account for the changes that I sent to the mailing list > recently, where I did support different clock frequencies for the processor. No problem, I can rebase on top of your patches as soon as you submit them to ARM SoC so I can see them in linux-next. >> This was blind-coded using the datasheet, it'd be great if someone >> who has this machine could test it. > ... >> + /* Set both timers to 512 kHz */ > > The patch uses a second timer, which does not allow its use for other purposes, What does this mean? Reading the datasheet it didn't say anything about this timer being dedicated to something else. I am using TC1D as clocksource and TC2D as clockevent, repurposing it from the previous use as singe time source. Doing "git grep TC1D" I cannot see any code in the kernel making use of this timer? I could think of things like the boot loader or firmware setting it up to run in free-running mode ... in that case we use it as clocksource without initialization as long as we know what frequency it's running on. As far as I could understand from the datasheet there is really no way to actually disable these timers, so they always run anyway? > I have also tried to do something like this without using additional processor > resources, but not yet brought to the finish line yet. OK do you think you can switch the clps71x over to generic time and clockevents inspired by this patch or so? Yours, Linus Walleij
On Thu, 30 Aug 2012 10:21:44 -0700 Linus Walleij <linus.walleij@linaro.org> wrote: > >> This converts the ageing clps711x timer driver to the modern > >> clocksource/clockevent API. > > The patch does not account for the changes that I sent to the mailing list > > recently, where I did support different clock frequencies for the processor. > No problem, I can rebase on top of your patches as soon as you > submit them to ARM SoC so I can see them in linux-next. Patches have already been sent to the mailing list, and as far I understand, these patches are awaiting approval from Arnd. ... > > The patch uses a second timer, which does not allow its use for other purposes, > What does this mean? Reading the datasheet it didn't say anything about > this timer being dedicated to something else. > I am using TC1D as clocksource and TC2D as clockevent, repurposing > it from the previous use as singe time source. > Doing "git grep TC1D" I cannot see any code in the kernel making use > of this timer? These timers are only two in the processor and I just want to say that the second timer CAN be used for other purposes, such as I use. In the kernel of the second timer is really not used. > I could think of things like the boot loader or firmware setting it up > to run in free-running mode ... in that case we use > it as clocksource without initialization as long as we know what > frequency it's running on. Not understand this... > As far as I could understand from the datasheet there is really > no way to actually disable these timers, so they always run > anyway? Yes. We can only enable/disable IRQ for the timers, switch clock source and mode. > > I have also tried to do something like this without using additional processor > > resources, but not yet brought to the finish line yet. > OK do you think you can switch the clps71x over to generic time and > clockevents inspired by this patch or so? Unfortunately, I'm leaving tomorrow for a little while so I can not check anything or try to do better, but the idea of ??switching to a new API I like. Upon arrival back, I'll try to make a similar using only one timer or use the CLOCKSOURCE/CLOCKEVENTS resources to generate intervals necessary for me.
diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 181b697..8a3da9e 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -383,8 +383,9 @@ config ARCH_HIGHBANK config ARCH_CLPS711X bool "Cirrus Logic CLPS711x/EP721x/EP731x-based" select CPU_ARM720T - select ARCH_USES_GETTIMEOFFSET select NEED_MACH_MEMORY_H + select GENERIC_CLOCKEVENTS + select CLKSRC_MMIO help Support for Cirrus Logic 711x/721x/731x based boards. diff --git a/arch/arm/mach-clps711x/common.c b/arch/arm/mach-clps711x/common.c index f15293b..850697f 100644 --- a/arch/arm/mach-clps711x/common.c +++ b/arch/arm/mach-clps711x/common.c @@ -26,6 +26,8 @@ #include <linux/io.h> #include <linux/irq.h> #include <linux/sched.h> +#include <linux/clockchips.h> +#include <linux/clocksource.h> #include <asm/sizes.h> #include <mach/hardware.h> @@ -157,25 +159,42 @@ void __init clps711x_init_irq(void) clps_writel(0, KBDEOI); } -/* - * gettimeoffset() returns time since last timer tick, in usecs. - * - * 'LATCH' is hwclock ticks (see CLOCK_TICK_RATE in timex.h) per jiffy. - * 'tick' is usecs per jiffy. - */ -static unsigned long clps711x_gettimeoffset(void) +static void clps711x_set_mode(enum clock_event_mode mode, + struct clock_event_device *evt) { - unsigned long hwticks; - hwticks = LATCH - (clps_readl(TC2D) & 0xffff); /* since last underflow */ - return (hwticks * (tick_nsec / 1000)) / LATCH; + u32 syscon; + u32 period = DIV_ROUND_CLOSEST(512000, HZ); + + switch (mode) { + case CLOCK_EVT_MODE_PERIODIC: + case CLOCK_EVT_MODE_RESUME: + /* Go into prescaler mode, program frequency */ + syscon = clps_readl(SYSCON1); + syscon |= SYSCON1_TC2M; + clps_writel(syscon, SYSCON1); + clps_writel(period, TC2D); + break; + default: + break; + } } +static struct clock_event_device clockevent_clps711x = { + .name = "clps711x_tc2d", + .rating = 300, + /* Only support periodic mode */ + .features = CLOCK_EVT_FEAT_PERIODIC, + .set_mode = clps711x_set_mode, +}; + /* * IRQ handler for the timer */ static irqreturn_t p720t_timer_interrupt(int irq, void *dev_id) { - timer_tick(); + struct clock_event_device *evt = &clockevent_clps711x; + + evt->event_handler(evt); return IRQ_HANDLED; } @@ -190,17 +209,29 @@ static void __init clps711x_timer_init(void) unsigned int syscon; syscon = clps_readl(SYSCON1); - syscon |= SYSCON1_TC2S | SYSCON1_TC2M; + /* Timer 1 in free running mode */ + syscon &= ~(SYSCON1_TC1M); + /* Timer 2 in prescale mode */ + syscon |= SYSCON1_TC2M; + /* Set both timers to 512 kHz */ + syscon |= (SYSCON1_TC1S | SYSCON1_TC2S); clps_writel(syscon, SYSCON1); - clps_writel(LATCH-1, TC2D); /* 512kHz / 100Hz - 1 */ + /* Start timer 1 in free running mode */ + clps_writel(0xFFFFU, TC1D); + if (clocksource_mmio_init(CLPS711X_VIRT_BASE + TC1D, + "clps711x_tc1d", 512000, 300, 16, + clocksource_mmio_readl_down)) + pr_err("timer: failed to initialize clps711x clock source\n"); + /* Configure and register the clockevent */ + clockevents_config_and_register(&clockevent_clps711x, 512000, + 1, 0xFFFFU); setup_irq(IRQ_TC2OI, &clps711x_timer_irq); } struct sys_timer clps711x_timer = { .init = clps711x_timer_init, - .offset = clps711x_gettimeoffset, }; void clps711x_restart(char mode, const char *cmd)
This converts the ageing clps711x timer driver to the modern clocksource/clockevent API. This was blind-coded using the datasheet, it'd be great if someone who has this machine could test it. Signed-off-by: Linus Walleij <linus.walleij@linaro.org> --- arch/arm/Kconfig | 3 +- arch/arm/mach-clps711x/common.c | 59 +++++++++++++++++++++++++++++--------- 2 files changed, 47 insertions(+), 15 deletions(-)