Re: [PATCH] timer: mchp-pit64b: add support for pit64b
On 07.10.2020 20:49, Sean Anderson wrote: > EXTERNAL EMAIL: Do not click links or open attachments unless you know the > content is safe > > On 9/7/20 11:36 AM, Claudiu Beznea wrote: >> Add support for Microchip PIT64B timer. The timer is 64 bit length and >> is used as a free running counter (in continuous mode with highest values >> for period registers). The clock feeding the timer would be no more >> than 12.5MHz. >> >> Signed-off-by: Claudiu Beznea >> --- >> drivers/timer/Kconfig | 7 +++ >> drivers/timer/Makefile| 1 + >> drivers/timer/mchp-pit64b-timer.c | 109 >> ++ >> 3 files changed, 117 insertions(+) >> create mode 100644 drivers/timer/mchp-pit64b-timer.c >> >> diff --git a/drivers/timer/Kconfig b/drivers/timer/Kconfig >> index 5f4bc6edb67b..13a98be4ab92 100644 >> --- a/drivers/timer/Kconfig >> +++ b/drivers/timer/Kconfig >> @@ -181,4 +181,11 @@ config MTK_TIMER >> Select this to enable support for the timer found on >> MediaTek devices. >> >> +config MCHP_PIT64B_TIMER >> + bool "Microchip 64-bit periodic interval timer support" >> + depends on TIMER >> + help >> + Select this to enable support for Microchip 64-bit periodic >> + interval timer. >> + >> endmenu >> diff --git a/drivers/timer/Makefile b/drivers/timer/Makefile >> index fa35bea6c5b2..4744a8d9c93c 100644 >> --- a/drivers/timer/Makefile >> +++ b/drivers/timer/Makefile >> @@ -21,3 +21,4 @@ obj-$(CONFIG_STI_TIMER) += sti-timer.o >> obj-$(CONFIG_STM32_TIMER)+= stm32_timer.o >> obj-$(CONFIG_X86_TSC_TIMER) += tsc_timer.o >> obj-$(CONFIG_MTK_TIMER) += mtk_timer.o >> +obj-$(CONFIG_MCHP_PIT64B_TIMER) += mchp-pit64b-timer.o >> diff --git a/drivers/timer/mchp-pit64b-timer.c >> b/drivers/timer/mchp-pit64b-timer.c >> new file mode 100644 >> index ..ead8c9b84ad5 >> --- /dev/null >> +++ b/drivers/timer/mchp-pit64b-timer.c >> @@ -0,0 +1,109 @@ >> +// SPDX-License-Identifier: GPL-2.0+ >> +/* >> + * 64-bit Periodic Interval Timer driver >> + * >> + * Copyright (C) 2020 Microchip Technology Inc. and its subsidiaries >> + * >> + * Author: Claudiu Beznea >> + */ >> + >> +#include >> +#include >> +#include >> +#include >> +#include >> + >> +#define MCHP_PIT64B_CR 0x00/* Control Register */ >> +#define MCHP_PIT64B_CR_STARTBIT(0) >> +#define MCHP_PIT64B_CR_SWRSTBIT(8) >> +#define MCHP_PIT64B_MR 0x04/* Mode Register */ >> +#define MCHP_PIT64B_MR_CONT BIT(0) >> +#define MCHP_PIT64B_LSB_PR 0x08/* LSB Period Register */ >> +#define MCHP_PIT64B_MSB_PR 0x0C/* MSB Period Register */ >> +#define MCHP_PIT64B_TLSBR0x20/* Timer LSB Register */ >> +#define MCHP_PIT64B_TMSBR0x24/* Timer MSB Register */ >> + >> +struct mchp_pit64b_priv { >> + void __iomem *base; >> +}; >> + >> +static int mchp_pit64b_get_count(struct udevice *dev, u64 *count) >> +{ >> + struct mchp_pit64b_priv *priv = dev_get_priv(dev); >> + >> + u32 lsb = readl(priv->base + MCHP_PIT64B_TLSBR); >> + u32 msb = readl(priv->base + MCHP_PIT64B_TMSBR); > > Is there a chance of rollover here? E.g. lets say the 64-bit counter is > 0x___ when we read lsb, but it increments to > 0x_0001__ when we read msb. Then the next time we read the > timer, it will look like we jumped backward in time. This should not happen as the hardware takes care of it as follows: when using a 64 bit period TLSB must be read first, followed by the read of TMSB. This sequence generates an atomic read of the 64 bit timer value whatever the lapse of time between the accesses. Thank you, Claudiu Beznea > > One way to get around this is by doing something like > > do { > msb = readl(priv->base + MCHP_PIT64B_TMSBR); > lsb = readl(priv->base + MCHP_PIT64B_TLSBR); > } while (msb != readl(priv->base + MCHP_PIT64B_TMSBR)); > > That way, if we ever roll-over between the two reads, we just redo the > lsb read. This is the method used by drivers/timer/riscv_timer.c on > 32-bit systems. > > --Sean > >> + >> + *count = ((u64)msb << 32) | lsb; >> + >> + return 0; >> +} >> + >> +static int mchp_pit64b_probe(struct udevice *dev) >> +{ >> + struct timer_dev_priv *uc_priv = dev_get_uclass_priv(dev); >> + struct mchp_pit64b_priv *priv = dev_get_priv(dev); >> + struct clk clk; >> + ulong rate; >> + int ret; >> + >> + priv->base = dev_read_addr_ptr(dev); >> + if (IS_ERR(priv->base)) >> + return PTR_ERR(priv->base); >> + >> + ret = clk_get_by_index(dev, 0, &clk); >> + if (ret) >> + return ret; >> + >> + ret = clk_enable(&clk); >> + if (ret) >> + return ret; >> + >> + rate = clk_get_rate(&clk); >> + if (!rate) { >> + clk_disable(&
Re: [PATCH] timer: mchp-pit64b: add support for pit64b
On 9/7/20 11:36 AM, Claudiu Beznea wrote: > Add support for Microchip PIT64B timer. The timer is 64 bit length and > is used as a free running counter (in continuous mode with highest values > for period registers). The clock feeding the timer would be no more > than 12.5MHz. > > Signed-off-by: Claudiu Beznea > --- > drivers/timer/Kconfig | 7 +++ > drivers/timer/Makefile| 1 + > drivers/timer/mchp-pit64b-timer.c | 109 > ++ > 3 files changed, 117 insertions(+) > create mode 100644 drivers/timer/mchp-pit64b-timer.c > > diff --git a/drivers/timer/Kconfig b/drivers/timer/Kconfig > index 5f4bc6edb67b..13a98be4ab92 100644 > --- a/drivers/timer/Kconfig > +++ b/drivers/timer/Kconfig > @@ -181,4 +181,11 @@ config MTK_TIMER > Select this to enable support for the timer found on > MediaTek devices. > > +config MCHP_PIT64B_TIMER > + bool "Microchip 64-bit periodic interval timer support" > + depends on TIMER > + help > + Select this to enable support for Microchip 64-bit periodic > + interval timer. > + > endmenu > diff --git a/drivers/timer/Makefile b/drivers/timer/Makefile > index fa35bea6c5b2..4744a8d9c93c 100644 > --- a/drivers/timer/Makefile > +++ b/drivers/timer/Makefile > @@ -21,3 +21,4 @@ obj-$(CONFIG_STI_TIMER) += sti-timer.o > obj-$(CONFIG_STM32_TIMER)+= stm32_timer.o > obj-$(CONFIG_X86_TSC_TIMER) += tsc_timer.o > obj-$(CONFIG_MTK_TIMER) += mtk_timer.o > +obj-$(CONFIG_MCHP_PIT64B_TIMER) += mchp-pit64b-timer.o > diff --git a/drivers/timer/mchp-pit64b-timer.c > b/drivers/timer/mchp-pit64b-timer.c > new file mode 100644 > index ..ead8c9b84ad5 > --- /dev/null > +++ b/drivers/timer/mchp-pit64b-timer.c > @@ -0,0 +1,109 @@ > +// SPDX-License-Identifier: GPL-2.0+ > +/* > + * 64-bit Periodic Interval Timer driver > + * > + * Copyright (C) 2020 Microchip Technology Inc. and its subsidiaries > + * > + * Author: Claudiu Beznea > + */ > + > +#include > +#include > +#include > +#include > +#include > + > +#define MCHP_PIT64B_CR 0x00/* Control Register */ > +#define MCHP_PIT64B_CR_STARTBIT(0) > +#define MCHP_PIT64B_CR_SWRSTBIT(8) > +#define MCHP_PIT64B_MR 0x04/* Mode Register */ > +#define MCHP_PIT64B_MR_CONT BIT(0) > +#define MCHP_PIT64B_LSB_PR 0x08/* LSB Period Register */ > +#define MCHP_PIT64B_MSB_PR 0x0C/* MSB Period Register */ > +#define MCHP_PIT64B_TLSBR0x20/* Timer LSB Register */ > +#define MCHP_PIT64B_TMSBR0x24/* Timer MSB Register */ > + > +struct mchp_pit64b_priv { > + void __iomem *base; > +}; > + > +static int mchp_pit64b_get_count(struct udevice *dev, u64 *count) > +{ > + struct mchp_pit64b_priv *priv = dev_get_priv(dev); > + > + u32 lsb = readl(priv->base + MCHP_PIT64B_TLSBR); > + u32 msb = readl(priv->base + MCHP_PIT64B_TMSBR); Is there a chance of rollover here? E.g. lets say the 64-bit counter is 0x___ when we read lsb, but it increments to 0x_0001__ when we read msb. Then the next time we read the timer, it will look like we jumped backward in time. One way to get around this is by doing something like do { msb = readl(priv->base + MCHP_PIT64B_TMSBR); lsb = readl(priv->base + MCHP_PIT64B_TLSBR); } while (msb != readl(priv->base + MCHP_PIT64B_TMSBR)); That way, if we ever roll-over between the two reads, we just redo the lsb read. This is the method used by drivers/timer/riscv_timer.c on 32-bit systems. --Sean > + > + *count = ((u64)msb << 32) | lsb; > + > + return 0; > +} > + > +static int mchp_pit64b_probe(struct udevice *dev) > +{ > + struct timer_dev_priv *uc_priv = dev_get_uclass_priv(dev); > + struct mchp_pit64b_priv *priv = dev_get_priv(dev); > + struct clk clk; > + ulong rate; > + int ret; > + > + priv->base = dev_read_addr_ptr(dev); > + if (IS_ERR(priv->base)) > + return PTR_ERR(priv->base); > + > + ret = clk_get_by_index(dev, 0, &clk); > + if (ret) > + return ret; > + > + ret = clk_enable(&clk); > + if (ret) > + return ret; > + > + rate = clk_get_rate(&clk); > + if (!rate) { > + clk_disable(&clk); > + return -ENOTSUPP; > + } > + > + /* Reset the timer in case it was used by previous bootloaders. */ > + writel(MCHP_PIT64B_CR_SWRST, priv->base + MCHP_PIT64B_CR); > + > + /* > + * Use highest prescaller (for a peripheral clock running at 200MHz > + * this will lead to the timer running at 12.5MHz) and continuous mode. > + */ > + writel((15 << 8) | MCHP_PIT64B_MR_CONT, priv->base + MCHP_PIT64B_MR); > + uc_priv->clock_rate = rate / 16; > + > + /* > + * Simulate free running counter by setting max values to per
Re: [PATCH] timer: mchp-pit64b: add support for pit64b
On 22.09.2020 13:49, Claudiu Beznea - M18063 wrote: > Hi Eugen, > > On 22.09.2020 11:32, Eugen Hristev - M18282 wrote: >> On 07.09.2020 18:36, Claudiu Beznea wrote: >>> Add support for Microchip PIT64B timer. The timer is 64 bit length and >>> is used as a free running counter (in continuous mode with highest values >>> for period registers). The clock feeding the timer would be no more >>> than 12.5MHz. >>> >>> Signed-off-by: Claudiu Beznea >>> --- >> >> Hi Claudiu, >> >> To make CI/CD scripts happy, this new driver needs a MAINTAINERS entry > > Would you like me to create a new entry or add the file under the > ARM MICROCHIP/ATMEL AT91 hood? > > Thank you, > Claudiu Beznea > >> >> Thanks ! >> Eugen Applied to u-boot-atmel/next , thanks !
Re: [PATCH] timer: mchp-pit64b: add support for pit64b
On 22.09.2020 13:49, Claudiu Beznea - M18063 wrote: > Hi Eugen, > > On 22.09.2020 11:32, Eugen Hristev - M18282 wrote: >> On 07.09.2020 18:36, Claudiu Beznea wrote: >>> Add support for Microchip PIT64B timer. The timer is 64 bit length and >>> is used as a free running counter (in continuous mode with highest values >>> for period registers). The clock feeding the timer would be no more >>> than 12.5MHz. >>> >>> Signed-off-by: Claudiu Beznea >>> --- >> >> Hi Claudiu, >> >> To make CI/CD scripts happy, this new driver needs a MAINTAINERS entry > > Would you like me to create a new entry or add the file under the > ARM MICROCHIP/ATMEL AT91 hood? I am fine with both solutions, but I think one separate entry will be better. Eugen > > Thank you, > Claudiu Beznea > >> >> Thanks ! >> Eugen
Re: [PATCH] timer: mchp-pit64b: add support for pit64b
Hi Eugen, On 22.09.2020 11:32, Eugen Hristev - M18282 wrote: > On 07.09.2020 18:36, Claudiu Beznea wrote: >> Add support for Microchip PIT64B timer. The timer is 64 bit length and >> is used as a free running counter (in continuous mode with highest values >> for period registers). The clock feeding the timer would be no more >> than 12.5MHz. >> >> Signed-off-by: Claudiu Beznea >> --- > > Hi Claudiu, > > To make CI/CD scripts happy, this new driver needs a MAINTAINERS entry Would you like me to create a new entry or add the file under the ARM MICROCHIP/ATMEL AT91 hood? Thank you, Claudiu Beznea > > Thanks ! > Eugen >
Re: [PATCH] timer: mchp-pit64b: add support for pit64b
On 07.09.2020 18:36, Claudiu Beznea wrote: > Add support for Microchip PIT64B timer. The timer is 64 bit length and > is used as a free running counter (in continuous mode with highest values > for period registers). The clock feeding the timer would be no more > than 12.5MHz. > > Signed-off-by: Claudiu Beznea > --- Hi Claudiu, To make CI/CD scripts happy, this new driver needs a MAINTAINERS entry Thanks ! Eugen