arm: zynq: Use standard timer binding
Subsystems (5)
PATCH 1/10 arm: zynq: Use standard timer binding
Use standard dt description for zynq timer
and also prepare for moving to clocksource
initialization.
---
Josh: We have talked about this change at ELC.
Using standard dt binding as other timers.
I have also discussed this with Rob some time ago.
https://patchwork.kernel.org/patch/2112791/
--- arch/arm/boot/dts/zynq-7000.dtsi | 41 +----- arch/arm/boot/dts/zynq-zc702.dts | 10 -- arch/arm/mach-zynq/common.c | 1 + arch/arm/mach-zynq/timer.c | 261 +++++++++++++++++++++++++++----------- 4 files changed, 194 insertions(+), 119 deletions(-)
arch/arm/boot/dts/zynq-7000.dtsi +4 -37
@@ -111,56 +111,23 @@ }; ttc0: ttc0 at f8001000 { - #address-cells = <1>; - #size-cells = <0>; + interrupt-parent = <&intc>; + interrupts = < 0 10 4 0 11 4 0 12 4 >; compatible = "xlnx,ttc"; reg = <0xF8001000 0x1000>; clocks = <&cpu_clk 3>; clock-names = "cpu_1x"; clock-ranges; - - ttc0_0: ttc0.0 { - status = "disabled"; - reg = <0>; - interrupts = <0 10 4>; - }; - ttc0_1: ttc0.1 { - status = "disabled"; - reg = <1>; - interrupts = <0 11 4>; - }; - ttc0_2: ttc0.2 { - status = "disabled"; - reg = <2>; - interrupts = <0 12 4>; - }; }; ttc1: ttc1 at f8002000 { - #interrupt-parent = <&intc>; - #address-cells = <1>; - #size-cells = <0>; + interrupt-parent = <&intc>; + interrupts = < 0 37 4 0 38 4 0 39 4 >; compatible = "xlnx,ttc"; reg = <0xF8002000 0x1000>; clocks = <&cpu_clk 3>; clock-names = "cpu_1x"; clock-ranges; - - ttc1_0: ttc1.0 { - status = "disabled"; - reg = <0>; - interrupts = <0 37 4>; - }; - ttc1_1: ttc1.1 { - status = "disabled"; - reg = <1>; - interrupts = <0 38 4>; - }; - ttc1_2: ttc1.2 { - status = "disabled"; - reg = <2>; - interrupts = <0 39 4>; - }; }; }; };
arch/arm/boot/dts/zynq-zc702.dts +0 -10
@@ -32,13 +32,3 @@ &ps_clk { clock-frequency = <33333330>; }; - -&ttc0_0 { - status = "ok"; - compatible = "xlnx,ttc-counter-clocksource"; -}; - -&ttc0_1 { - status = "ok"; - compatible = "xlnx,ttc-counter-clockevent"; -};
arch/arm/mach-zynq/common.c +1 -0
@@ -20,6 +20,7 @@#include <linux/platform_device.h> #include <linux/clk.h> #include <linux/clk/zynq.h> +#include <linux/clocksource.h> #include <linux/of_address.h> #include <linux/of_irq.h> #include <linux/of_platform.h>
arch/arm/mach-zynq/timer.c +189 -72
368 lines not on this page · raw message
Signature · 1 line
1.7.9.7
PATCH 2/10 arm: zynq: Move timer to clocksource interface
Use clocksource timer initialization.
--- arch/arm/mach-zynq/common.c | 2 +- arch/arm/mach-zynq/common.h | 2 -- arch/arm/mach-zynq/timer.c | 27 ++++++++------------------- 3 files changed, 9 insertions(+), 22 deletions(-)
arch/arm/mach-zynq/common.c +1 -1
@@ -78,7 +78,7 @@ static void __init xilinx_zynq_timer_init(void)xilinx_zynq_clocks_init (slcr); -xttcps_timer_init (); +clocksource_of_init (); }/**
arch/arm/mach-zynq/common.h +0 -2
@@ -17,6 +17,4 @@#ifndef __MACH_ZYNQ_COMMON_H__#define __MACH_ZYNQ_COMMON_H__ -void __init xttcps_timer_init (void ); -#endif
arch/arm/mach-zynq/timer.c +8 -19
@@ -22,7 +22,6 @@#include <linux/of_irq.h> #include <linux/slab.h> #include <linux/clk-provider.h> -#include "common.h" /* * This driver configures the 2 16-bit count-up timers as follows:
@@ -387,11 +386,17 @@ static void __init zynq_ttc_setup_clockevent(struct clk *clk,* Initializes the timer hardware and register the clock source and clock event * timers with Linux kernal timer framework */ -static void __init xttcps_timer_init_of (struct device_node *timer) +static void __init xttcps_timer_init (struct device_node *timer) {unsigned int irq;void __iomem *timer_baseaddr;struct clk *clk; +static int initialized; + +if (initialized) +return ; + + initialized =1 ;/* * Get the 1st Triple Timer Counter (TTC) block from the device tree
@@ -422,20 +427,4 @@ static void __init xttcps_timer_init_of(struct device_node *timer)pr_info ("%s #0 at %p, irq=%d\n" , timer->name, timer_baseaddr, irq); } - -void __init xttcps_timer_init (void ) -{ -const char *const timer_list[] = { -"xlnx,ttc" , -NULL - }; -struct device_node *timer; - - timer =of_find_compatible_node (NULL ,NULL , timer_list[0 ]); -if (!timer) { -pr_err ("ERROR: no compatible timer found\n" ); -BUG (); - } - -xttcps_timer_init_of (timer); -} +CLOCKSOURCE_OF_DECLARE (zynq,"xlnx,ttc" , xttcps_timer_init);
Signature · 1 line
1.7.9.7
PATCH 3/10 arm: zynq: Move timer to generic location
In reply to Michal SimekMove zynq timer out of mach folder to generic location.
---
Arnd: This is the last patch as we discuss about moving this timer to generic location.
It depends on ARCH_ZYNQ as Tegra and others.
It should be ok to use it or we can use different name if you like.
--- arch/arm/mach-zynq/Makefile | 2 +- arch/arm/mach-zynq/timer.c | 430 -------------------------------------- drivers/clocksource/Makefile | 1 + drivers/clocksource/zynq_timer.c | 430 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 432 insertions(+), 431 deletions(-) delete mode 100644 arch/arm/mach-zynq/timer.c create mode 100644 drivers/clocksource/zynq_timer.c
arch/arm/mach-zynq/Makefile +1 -1
@@ -3,4 +3,4 @@ # # Common support -obj-y := common.o timer.o +obj-y := common.o
arch/arm/mach-zynq/timer.c deleted +0 -430
430 lines not on this page · raw message
drivers/clocksource/Makefile +1 -0
@@ -19,6 +19,7 @@ obj-$(CONFIG_ARCH_BCM2835) += bcm2835_timer.o obj-$(CONFIG_SUNXI_TIMER) += sunxi_timer.o obj-$(CONFIG_ARCH_TEGRA) += tegra20_timer.o obj-$(CONFIG_VT8500_TIMER) += vt8500_timer.o +obj-$(CONFIG_ARCH_ZYNQ) += zynq_timer.o obj-$(CONFIG_ARM_ARCH_TIMER) += arm_arch_timer.o obj-$(CONFIG_CLKSRC_METAG_GENERIC) += metag_generic.o
drivers/clocksource/zynq_timer.c new file +430 -0
430 lines not on this page · raw message
Signature · 1 line
1.7.9.7
PATCH 3/10 arm: zynq: Move timer to generic location
Hi Michal!
Quoted 10 lines · Michal Simek
Move zynq timer out of mach folder to generic location.
Signed-off-by Michal Simek <[email protected]>---
Arnd: This is the last patch as we discuss about moving this timer to generic location.It depends on
ARCH_ZYNQas Tegra and others.
It should be ok to use it or we can use different name if you like.
---
Moving out of mach-zynq is good, because this is a
Cadence Rev06 Triple Timer Counter (see Zynq-7000 TRM page 25)
Which leads me to a proposed rename to: cadence_timer.c or something like that. And the fn-names should be changed accordingly while at it.
Quoted 686 lines
arch/arm/mach-zynq/Makefile | 2 +- arch/arm/mach-zynq/timer.c | 430 -------------------------------------- drivers/clocksource/Makefile | 1 + drivers/clocksource/zynq_timer.c | 430 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 432 insertions(+), 431 deletions(-) delete mode 100644 arch/arm/mach-zynq/timer.c create mode 100644 drivers/clocksource/zynq_timer.carch/arm/mach-zynq/Makefile +1
-1@@ -3,4 +3,4 @@ # # Common support -obj-y := common.o timer.o +obj-y := common.oarch/arm/mach-zynq/timer.c deleted +0
-430@@ -1,430 +0,0 @@ -/* -* This file contains driver for the Xilinx PS Timer Counter IP. -* -* Copyright (C) 2011 Xilinx -* -* based on arch/mips/kernel/time.c timer driver -* -* This software is licensed under the terms of the GNU General Public -* License version 2, as published by the Free Software Foundation, and -* may be copied, distributed, and modified under those terms. -* -* This program is distributed in the hope that it will be useful, -* but WITHOUT ANY WARRANTY; without even the implied warranty of -* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -* GNU General Public License for more details. -*/ - -#include <linux/clk.h> -#include <linux/interrupt.h> -#include <linux/clockchips.h> -#include <linux/of_address.h> -#include <linux/of_irq.h> -#include <linux/slab.h> -#include <linux/clk-provider.h> - -/* -* This driver configures the 2 16-bit count-up timers as follows: -* -* T1: Timer 1, clocksource for generic timekeeping -* T2: Timer 2, clockevent source for hrtimers -* T3: Timer 3, <unused> -* -* The input frequency to the timer module for emulation is 2.5MHz which is -* common to all the timer channels (T1, T2, and T3). With a pre-scaler of 32, -* the timers are clocked at 78.125KHz (12.8 us resolution). - -* The input frequency to the timer module in silicon is configurable and -* obtained from device tree. The pre-scaler of 32 is used. -*/ - -/* -* Timer Register Offset Definitions of Timer 1, Increment base address by 4 -* and use same offsets for Timer 2 -*/ -#define XTTCPS_CLK_CNTRL_OFFSET 0x00/* Clock Control Reg, RW */ -#define XTTCPS_CNT_CNTRL_OFFSET 0x0C/* Counter Control Reg, RW */ -#define XTTCPS_COUNT_VAL_OFFSET 0x18/* Counter Value Reg, RO */ -#define XTTCPS_INTR_VAL_OFFSET 0x24/* Interval Count Reg, RW */ -#define XTTCPS_ISR_OFFSET 0x54/* Interrupt Status Reg, RO */ -#define XTTCPS_IER_OFFSET 0x60/* Interrupt Enable Reg, RW */ - -#define XTTCPS_CNT_CNTRL_DISABLE_MASK 0x1 - -/* -* Setup the timers to use pre-scaling, using a fixed value for now that will -* work across most input frequency, but it may need to be more dynamic -*/ -#define PRESCALE_EXPONENT 11/* 2 ^ PRESCALE_EXPONENT = PRESCALE */ -#define PRESCALE 2048/* The exponent must match this */ -#define CLK_CNTRL_PRESCALE ((PRESCALE_EXPONENT - 1) << 1) -#define CLK_CNTRL_PRESCALE_EN 1 -#define CNT_CNTRL_RESET (1 << 4) - -/** -* struct xttcps_timer - This definition defines local timer structure -* -* @base_addr: Base address of timer -* @clk: Associated clock source -* @clk_rate_change_nb Notifier block for clock rate changes -*/ -struct xttcps_timer { -void __iomem *base_addr; -struct clk *clk; -struct notifier_block clk_rate_change_nb; -}; - -#define to_xttcps_timer (x) \ - container_of(x, struct xttcps_timer, clk_rate_change_nb) - -struct xttcps_timer_clocksource { -struct xttcps_timer xttc; -struct clocksource cs; -}; - -#define to_xttcps_timer_clksrc (x) \ - container_of(x, struct xttcps_timer_clocksource, cs) - -struct xttcps_timer_clockevent { -struct xttcps_timer xttc; -struct clock_event_device ce; -}; - -#define to_xttcps_timer_clkevent (x) \ - container_of(x, struct xttcps_timer_clockevent, ce) - -/** -* xttcps_set_interval - Set the timer interval value -* -* @timer: Pointer to the timer instance -* @cycles: Timer interval ticks -**/ -static void xttcps_set_interval (struct xttcps_timer *timer, -unsigned long cycles) -{ -u32 ctrl_reg; - -/* Disable the counter, set the counter value and re-enable counter */ - ctrl_reg =__raw_readl (timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); - ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); - -__raw_writel (cycles, timer->base_addr +XTTCPS_INTR_VAL_OFFSET ); - -/* -* Reset the counter (0x10) so that it starts from 0, one-shot -* mode makes this needed for timing to be right. -*/ - ctrl_reg |=CNT_CNTRL_RESET ; - ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -} - -/** -* xttcps_clock_event_interrupt - Clock event timer interrupt handler -* -* @irq: IRQ number of the Timer -* @dev_id: void pointer to the xttcps_timer instance -* -* returns: Always IRQ_HANDLED - success -**/ -static irqreturn_t xttcps_clock_event_interrupt (int irq,void *dev_id) -{ -struct xttcps_timer_clockevent *xttce = dev_id; -struct xttcps_timer *timer = &xttce->xttc; - -/* Acknowledge the interrupt and call event handler */ -__raw_readl (timer->base_addr +XTTCPS_ISR_OFFSET ); - - xttce->ce.event_handler (&xttce->ce); - -return IRQ_HANDLED ; -} - -/** -* __xttc_clocksource_read - Reads the timer counter register -* -* returns: Current timer counter register value -**/ -static cycle_t __xttc_clocksource_read (struct clocksource *cs) -{ -struct xttcps_timer *timer = &to_xttcps_timer_clksrc (cs)->xttc; - -return (cycle_t )__raw_readl (timer->base_addr + -XTTCPS_COUNT_VAL_OFFSET ); -} - -/** -* xttcps_set_next_event - Sets the time interval for next event -* -* @cycles: Timer interval ticks -* @evt: Address of clock event instance -* -* returns: Always 0 - success -**/ -static int xttcps_set_next_event (unsigned long cycles, -struct clock_event_device *evt) -{ -struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); -struct xttcps_timer *timer = &xttce->xttc; - -xttcps_set_interval (timer, cycles); -return 0 ; -} - -/** -* xttcps_set_mode - Sets the mode of timer -* -* @mode: Mode to be set -* @evt: Address of clock event instance -**/ -static void xttcps_set_mode (enum clock_event_mode mode, -struct clock_event_device *evt) -{ -struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); -struct xttcps_timer *timer = &xttce->xttc; -u32 ctrl_reg; - -switch (mode) { -case CLOCK_EVT_MODE_PERIODIC : -xttcps_set_interval (timer, -DIV_ROUND_CLOSEST (clk_get_rate (xttce->xttc.clk), -PRESCALE *HZ )); -break ; -case CLOCK_EVT_MODE_ONESHOT : -case CLOCK_EVT_MODE_UNUSED : -case CLOCK_EVT_MODE_SHUTDOWN : - ctrl_reg =__raw_readl (timer->base_addr + -XTTCPS_CNT_CNTRL_OFFSET ); - ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, - timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -break ; -case CLOCK_EVT_MODE_RESUME : - ctrl_reg =__raw_readl (timer->base_addr + -XTTCPS_CNT_CNTRL_OFFSET ); - ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, - timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -break ; - } -} - -static int xttcps_rate_change_clocksource_cb (struct notifier_block *nb, -unsigned long event,void *data) -{ -struct clk_notifier_data *ndata = data; -struct xttcps_timer *xttcps =to_xttcps_timer (nb); -struct xttcps_timer_clocksource *xttccs =container_of (xttcps, - struct xttcps_timer_clocksource, xttc); - -switch (event) { -case POST_RATE_CHANGE : -/* -* Do whatever is necessare to maintain a proper time base -* -* I cannot find a way to adjust the currently used clocksource -* to the new frequency. __clocksource_updatefreq_hz() sounds -* good, but does not work. Not sure what's that missing. -* -* This approach works, but triggers two clocksource switches. -* The first after unregister to clocksource jiffies. And -* another one after the register to the newly registered timer. -* -* Alternatively we could 'waste' another HW timer to ping pong -* between clock sources. That would also use one register and -* one unregister call, but only trigger one clocksource switch -* for the cost of another HW timer used by the OS. -*/ -clocksource_unregister (&xttccs->cs); -clocksource_register_hz (&xttccs->cs, - ndata->new_rate /PRESCALE ); -/* fall through */ -case PRE_RATE_CHANGE : -case ABORT_RATE_CHANGE : -default : -return NOTIFY_DONE ; - } -} - -static void __init zynq_ttc_setup_clocksource (struct clk *clk, -void __iomem *base) -{ -struct xttcps_timer_clocksource *ttccs; -int err; - - ttccs =kzalloc (sizeof (*ttccs),GFP_KERNEL ); -if (WARN_ON (!ttccs)) -return ; - - ttccs->xttc.clk = clk; - - err =clk_prepare_enable (ttccs->xttc.clk); -if (WARN_ON (err)) -return ; - - ttccs->xttc.clk_rate_change_nb.notifier_call = - xttcps_rate_change_clocksource_cb; - ttccs->xttc.clk_rate_change_nb.next =NULL ; -if (clk_notifier_register (ttccs->xttc.clk, - &ttccs->xttc.clk_rate_change_nb)) -pr_warn ("Unable to register clock notifier.\n" ); - - ttccs->xttc.base_addr = base; - ttccs->cs.name ="xttcps_clocksource" ; - ttccs->cs.rating =200 ; - ttccs->cs.read = __xttc_clocksource_read; - ttccs->cs.mask =CLOCKSOURCE_MASK (16 ); - ttccs->cs.flags =CLOCK_SOURCE_IS_CONTINUOUS ; - -/* -* Setup the clock source counter to be an incrementing counter -* with no interrupt and it rolls over at 0xFFFF. Pre-scale -* it by 32 also. Let it start running now. -*/ -__raw_writel (0x0 , ttccs->xttc.base_addr +XTTCPS_IER_OFFSET ); -__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , - ttccs->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); -__raw_writel (CNT_CNTRL_RESET , - ttccs->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); - - err =clocksource_register_hz (&ttccs->cs, -clk_get_rate (ttccs->xttc.clk) /PRESCALE ); -if (WARN_ON (err)) -return ; - -} - -static int xttcps_rate_change_clockevent_cb (struct notifier_block *nb, -unsigned long event,void *data) -{ -struct clk_notifier_data *ndata = data; -struct xttcps_timer *xttcps =to_xttcps_timer (nb); -struct xttcps_timer_clockevent *xttcce =container_of (xttcps, - struct xttcps_timer_clockevent, xttc); - -switch (event) { -case POST_RATE_CHANGE : - { -unsigned long flags; - -/* -* clockevents_update_freq should be called with IRQ disabled on -* the CPU the timer provides events for. The timer we use is -* common to both CPUs, not sure if we need to run on both -* cores. -*/ -local_irq_save (flags); -clockevents_update_freq (&xttcce->ce, - ndata->new_rate /PRESCALE ); -local_irq_restore (flags); - -/* fall through */ - } -case PRE_RATE_CHANGE : -case ABORT_RATE_CHANGE : -default : -return NOTIFY_DONE ; - } -} - -static void __init zynq_ttc_setup_clockevent (struct clk *clk, -void __iomem *base,u32 irq) -{ -struct xttcps_timer_clockevent *ttcce; -int err; - - ttcce =kzalloc (sizeof (*ttcce),GFP_KERNEL ); -if (WARN_ON (!ttcce)) -return ; - - ttcce->xttc.clk = clk; - - err =clk_prepare_enable (ttcce->xttc.clk); -if (WARN_ON (err)) -return ; - - ttcce->xttc.clk_rate_change_nb.notifier_call = - xttcps_rate_change_clockevent_cb; - ttcce->xttc.clk_rate_change_nb.next =NULL ; -if (clk_notifier_register (ttcce->xttc.clk, - &ttcce->xttc.clk_rate_change_nb)) -pr_warn ("Unable to register clock notifier.\n" ); - - ttcce->xttc.base_addr = base; - ttcce->ce.name ="xttcps_clockevent" ; - ttcce->ce.features =CLOCK_EVT_FEAT_PERIODIC |CLOCK_EVT_FEAT_ONESHOT ; - ttcce->ce.set_next_event = xttcps_set_next_event; - ttcce->ce.set_mode = xttcps_set_mode; - ttcce->ce.rating =200 ; - ttcce->ce.irq = irq; - ttcce->ce.cpumask = cpu_possible_mask; - -/* -* Setup the clock event timer to be an interval timer which -* is prescaled by 32 using the interval interrupt. Leave it -* disabled for now. -*/ -__raw_writel (0x23 , ttcce->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , - ttcce->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); -__raw_writel (0x1 , ttcce->xttc.base_addr +XTTCPS_IER_OFFSET ); - - err =request_irq (irq, xttcps_clock_event_interrupt, -IRQF_DISABLED |IRQF_TIMER , - ttcce->ce.name, ttcce); -if (WARN_ON (err)) -return ; - -clockevents_config_and_register (&ttcce->ce, -clk_get_rate (ttcce->xttc.clk) /PRESCALE ,1 ,0xfffe ); -} - -/** -* xttcps_timer_init - Initialize the timer -* -* Initializes the timer hardware and register the clock source and clock event -* timers with Linux kernal timer framework -*/ -static void __init xttcps_timer_init (struct device_node *timer) -{ -unsigned int irq; -void __iomem *timer_baseaddr; -struct clk *clk; -static int initialized; - -if (initialized) -return ; - - initialized =1 ; - -/* -* Get the 1st Triple Timer Counter (TTC) block from the device tree -* and use it. Note that the event timer uses the interrupt and it's the -* 2nd TTC hence the irq_of_parse_and_map(,1) -*/ - timer_baseaddr =of_iomap (timer,0 ); -if (!timer_baseaddr) { -pr_err ("ERROR: invalid timer base address\n" ); -BUG (); - } - - irq =irq_of_parse_and_map (timer,1 ); -if (irq <=0 ) { -pr_err ("ERROR: invalid interrupt number\n" ); -BUG (); - } - - clk =of_clk_get_by_name (timer,"cpu_1x" ); -if (IS_ERR (clk)) { -pr_err ("ERROR: timer input clock not found\n" ); -BUG (); - } - -zynq_ttc_setup_clocksource (clk, timer_baseaddr); -zynq_ttc_setup_clockevent (clk, timer_baseaddr +4 , irq); - -pr_info ("%s #0 at %p, irq=%d\n" , timer->name, timer_baseaddr, irq); -} - -CLOCKSOURCE_OF_DECLARE (zynq,"xlnx,ttc" , xttcps_timer_init);drivers/clocksource/Makefile +1
-0@@ -19,6 +19,7 @@ obj-$(CONFIG_ARCH_BCM2835) += bcm2835_timer.o obj-$(CONFIG_SUNXI_TIMER) += sunxi_timer.o obj-$(CONFIG_ARCH_TEGRA) += tegra20_timer.o obj-$(CONFIG_VT8500_TIMER) += vt8500_timer.o +obj-$(CONFIG_ARCH_ZYNQ) += zynq_timer.o obj-$(CONFIG_ARM_ARCH_TIMER) += arm_arch_timer.o obj-$(CONFIG_CLKSRC_METAG_GENERIC) += metag_generic.odrivers/clocksource/zynq_timer.c new file +214
-0@@ -0,0 +1,430 @@ +/* +* This file contains driver for the Xilinx PS Timer Counter IP. +* +* Copyright (C) 2011 Xilinx +* +* based on arch/mips/kernel/time.c timer driver +* +* This software is licensed under the terms of the GNU General Public +* License version 2, as published by the Free Software Foundation, and +* may be copied, distributed, and modified under those terms. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +*/ + +#include <linux/clk.h> +#include <linux/interrupt.h> +#include <linux/clockchips.h> +#include <linux/of_address.h> +#include <linux/of_irq.h> +#include <linux/slab.h> +#include <linux/clk-provider.h> + +/* +* This driver configures the 2 16-bit count-up timers as follows: +* +* T1: Timer 1, clocksource for generic timekeeping +* T2: Timer 2, clockevent source for hrtimers +* T3: Timer 3, <unused> +* +* The input frequency to the timer module for emulation is 2.5MHz which is +* common to all the timer channels (T1, T2, and T3). With a pre-scaler of 32, +* the timers are clocked at 78.125KHz (12.8 us resolution). + +* The input frequency to the timer module in silicon is configurable and +* obtained from device tree. The pre-scaler of 32 is used. +*/ + +/* +* Timer Register Offset Definitions of Timer 1, Increment base address by 4 +* and use same offsets for Timer 2 +*/ +#define XTTCPS_CLK_CNTRL_OFFSET 0x00/* Clock Control Reg, RW */ +#define XTTCPS_CNT_CNTRL_OFFSET 0x0C/* Counter Control Reg, RW */ +#define XTTCPS_COUNT_VAL_OFFSET 0x18/* Counter Value Reg, RO */ +#define XTTCPS_INTR_VAL_OFFSET 0x24/* Interval Count Reg, RW */ +#define XTTCPS_ISR_OFFSET 0x54/* Interrupt Status Reg, RO */ +#define XTTCPS_IER_OFFSET 0x60/* Interrupt Enable Reg, RW */ + +#define XTTCPS_CNT_CNTRL_DISABLE_MASK 0x1 + +/* +* Setup the timers to use pre-scaling, using a fixed value for now that will +* work across most input frequency, but it may need to be more dynamic +*/ +#define PRESCALE_EXPONENT 11/* 2 ^ PRESCALE_EXPONENT = PRESCALE */ +#define PRESCALE 2048/* The exponent must match this */ +#define CLK_CNTRL_PRESCALE ((PRESCALE_EXPONENT - 1) << 1) +#define CLK_CNTRL_PRESCALE_EN 1 +#define CNT_CNTRL_RESET (1 << 4) + +/** +* struct xttcps_timer - This definition defines local timer structure +* +* @base_addr: Base address of timer +* @clk: Associated clock source +* @clk_rate_change_nb Notifier block for clock rate changes +*/ +struct xttcps_timer { +void __iomem *base_addr; +struct clk *clk; +struct notifier_block clk_rate_change_nb; +}; + +#define to_xttcps_timer (x) \ + container_of(x, struct xttcps_timer, clk_rate_change_nb) + +struct xttcps_timer_clocksource { +struct xttcps_timer xttc; +struct clocksource cs; +}; + +#define to_xttcps_timer_clksrc (x) \ + container_of(x, struct xttcps_timer_clocksource, cs) + +struct xttcps_timer_clockevent { +struct xttcps_timer xttc; +struct clock_event_device ce; +}; + +#define to_xttcps_timer_clkevent (x) \ + container_of(x, struct xttcps_timer_clockevent, ce) + +/** +* xttcps_set_interval - Set the timer interval value +* +* @timer: Pointer to the timer instance +* @cycles: Timer interval ticks +**/ +static void xttcps_set_interval (struct xttcps_timer *timer, +unsigned long cycles) +{ +u32 ctrl_reg; + +/* Disable the counter, set the counter value and re-enable counter */ + ctrl_reg =__raw_readl (timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); + ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); + +__raw_writel (cycles, timer->base_addr +XTTCPS_INTR_VAL_OFFSET ); + +/* +* Reset the counter (0x10) so that it starts from 0, one-shot +* mode makes this needed for timing to be right. +*/ + ctrl_reg |=CNT_CNTRL_RESET ; + ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +} + +/** +* xttcps_clock_event_interrupt - Clock event timer interrupt handler +* +* @irq: IRQ number of the Timer +* @dev_id: void pointer to the xttcps_timer instance +* +* returns: Always IRQ_HANDLED - success +**/ +static irqreturn_t xttcps_clock_event_interrupt (int irq,void *dev_id) +{ +struct xttcps_timer_clockevent *xttce = dev_id; +struct xttcps_timer *timer = &xttce->xttc; + +/* Acknowledge the interrupt and call event handler */ +__raw_readl (timer->base_addr +XTTCPS_ISR_OFFSET ); + + xttce->ce.event_handler (&xttce->ce); + +return IRQ_HANDLED ; +} + +/** +* __xttc_clocksource_read - Reads the timer counter register +* +* returns: Current timer counter register value +**/ +static cycle_t __xttc_clocksource_read (struct clocksource *cs) +{ +struct xttcps_timer *timer = &to_xttcps_timer_clksrc (cs)->xttc; + +return (cycle_t )__raw_readl (timer->base_addr + +XTTCPS_COUNT_VAL_OFFSET ); +} + +/** +* xttcps_set_next_event - Sets the time interval for next event +* +* @cycles: Timer interval ticks +* @evt: Address of clock event instance +* +* returns: Always 0 - success +**/ +static int xttcps_set_next_event (unsigned long cycles, +struct clock_event_device *evt) +{ +struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); +struct xttcps_timer *timer = &xttce->xttc; + +xttcps_set_interval (timer, cycles); +return 0 ; +} + +/** +* xttcps_set_mode - Sets the mode of timer +* +* @mode: Mode to be set +* @evt: Address of clock event instance +**/ +static void xttcps_set_mode (enum clock_event_mode mode, +struct clock_event_device *evt) +{ +struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); +struct xttcps_timer *timer = &xttce->xttc; +u32 ctrl_reg; + +switch (mode) { +case CLOCK_EVT_MODE_PERIODIC : +xttcps_set_interval (timer, +DIV_ROUND_CLOSEST (clk_get_rate (xttce->xttc.clk), +PRESCALE *HZ )); +break ; +case CLOCK_EVT_MODE_ONESHOT : +case CLOCK_EVT_MODE_UNUSED : +case CLOCK_EVT_MODE_SHUTDOWN : + ctrl_reg =__raw_readl (timer->base_addr + +XTTCPS_CNT_CNTRL_OFFSET ); + ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, + timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +break ; +case CLOCK_EVT_MODE_RESUME : + ctrl_reg =__raw_readl (timer->base_addr + +XTTCPS_CNT_CNTRL_OFFSET ); + ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, + timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +break ; + } +} + +static int xttcps_rate_change_clocksource_cb (struct notifier_block *nb, +unsigned long event,void *data)
drivers/clocksource/zynq_timer.c line 224
Quoted 10 lines
drivers/clocksource/zynq_timer.c new file +10
-0+{ +struct clk_notifier_data *ndata = data; +struct xttcps_timer *xttcps =to_xttcps_timer (nb); +struct xttcps_timer_clocksource *xttccs =container_of (xttcps, + struct xttcps_timer_clocksource, xttc); + +switch (event) { +case POST_RATE_CHANGE : +/* +* Do whatever is necessare to maintain a proper time base
------------------------------------------^ little typo
Quoted 30 lines
drivers/clocksource/zynq_timer.c new file +30
-0+* +* I cannot find a way to adjust the currently used clocksource +* to the new frequency. __clocksource_updatefreq_hz() sounds +* good, but does not work. Not sure what's that missing. +* +* This approach works, but triggers two clocksource switches. +* The first after unregister to clocksource jiffies. And +* another one after the register to the newly registered timer. +* +* Alternatively we could 'waste' another HW timer to ping pong +* between clock sources. That would also use one register and +* one unregister call, but only trigger one clocksource switch +* for the cost of another HW timer used by the OS. +*/ +clocksource_unregister (&xttccs->cs); +clocksource_register_hz (&xttccs->cs, + ndata->new_rate /PRESCALE ); +/* fall through */ +case PRE_RATE_CHANGE : +case ABORT_RATE_CHANGE : +default : +return NOTIFY_DONE ; + } +} + +static void __init zynq_ttc_setup_clocksource (struct clk *clk, +void __iomem *base) +{ +struct xttcps_timer_clocksource *ttccs; +int err;
drivers/clocksource/zynq_timer.c line 264
Quoted 10 lines
drivers/clocksource/zynq_timer.c new file +10
-0+ + ttccs =kzalloc (sizeof (*ttccs),GFP_KERNEL ); +if (WARN_ON (!ttccs)) +return ; + + ttccs->xttc.clk = clk; + + err =clk_prepare_enable (ttccs->xttc.clk); +if (WARN_ON (err)) +return ;
I think you are losing memory here...
Quoted 20 lines
drivers/clocksource/zynq_timer.c new file +20
-0+ + ttccs->xttc.clk_rate_change_nb.notifier_call = + xttcps_rate_change_clocksource_cb; + ttccs->xttc.clk_rate_change_nb.next =NULL ; +if (clk_notifier_register (ttccs->xttc.clk, + &ttccs->xttc.clk_rate_change_nb)) +pr_warn ("Unable to register clock notifier.\n" ); + + ttccs->xttc.base_addr = base; + ttccs->cs.name ="xttcps_clocksource" ; + ttccs->cs.rating =200 ; + ttccs->cs.read = __xttc_clocksource_read; + ttccs->cs.mask =CLOCKSOURCE_MASK (16 ); + ttccs->cs.flags =CLOCK_SOURCE_IS_CONTINUOUS ; + +/* +* Setup the clock source counter to be an incrementing counter +* with no interrupt and it rolls over at 0xFFFF. Pre-scale +* it by 32 also. Let it start running now. +*/
drivers/clocksource/zynq_timer.c line 294
Quoted 10 lines
drivers/clocksource/zynq_timer.c new file +10
-0+__raw_writel (0x0 , ttccs->xttc.base_addr +XTTCPS_IER_OFFSET ); +__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , + ttccs->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); +__raw_writel (CNT_CNTRL_RESET , + ttccs->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); + + err =clocksource_register_hz (&ttccs->cs, +clk_get_rate (ttccs->xttc.clk) /PRESCALE ); +if (WARN_ON (err)) +return ;
... and here and below in zynq_ttc_setup_clockevent the same.
Quoted 96 lines
drivers/clocksource/zynq_timer.c new file +96
-0+ +} + +static int xttcps_rate_change_clockevent_cb (struct notifier_block *nb, +unsigned long event,void *data) +{ +struct clk_notifier_data *ndata = data; +struct xttcps_timer *xttcps =to_xttcps_timer (nb); +struct xttcps_timer_clockevent *xttcce =container_of (xttcps, + struct xttcps_timer_clockevent, xttc); + +switch (event) { +case POST_RATE_CHANGE : + { +unsigned long flags; + +/* +* clockevents_update_freq should be called with IRQ disabled on +* the CPU the timer provides events for. The timer we use is +* common to both CPUs, not sure if we need to run on both +* cores. +*/ +local_irq_save (flags); +clockevents_update_freq (&xttcce->ce, + ndata->new_rate /PRESCALE ); +local_irq_restore (flags); + +/* fall through */ + } +case PRE_RATE_CHANGE : +case ABORT_RATE_CHANGE : +default : +return NOTIFY_DONE ; + } +} + +static void __init zynq_ttc_setup_clockevent (struct clk *clk, +void __iomem *base,u32 irq) +{ +struct xttcps_timer_clockevent *ttcce; +int err; + + ttcce =kzalloc (sizeof (*ttcce),GFP_KERNEL ); +if (WARN_ON (!ttcce)) +return ; + + ttcce->xttc.clk = clk; + + err =clk_prepare_enable (ttcce->xttc.clk); +if (WARN_ON (err)) +return ; + + ttcce->xttc.clk_rate_change_nb.notifier_call = + xttcps_rate_change_clockevent_cb; + ttcce->xttc.clk_rate_change_nb.next =NULL ; +if (clk_notifier_register (ttcce->xttc.clk, + &ttcce->xttc.clk_rate_change_nb)) +pr_warn ("Unable to register clock notifier.\n" ); + + ttcce->xttc.base_addr = base; + ttcce->ce.name ="xttcps_clockevent" ; + ttcce->ce.features =CLOCK_EVT_FEAT_PERIODIC |CLOCK_EVT_FEAT_ONESHOT ; + ttcce->ce.set_next_event = xttcps_set_next_event; + ttcce->ce.set_mode = xttcps_set_mode; + ttcce->ce.rating =200 ; + ttcce->ce.irq = irq; + ttcce->ce.cpumask = cpu_possible_mask; + +/* +* Setup the clock event timer to be an interval timer which +* is prescaled by 32 using the interval interrupt. Leave it +* disabled for now. +*/ +__raw_writel (0x23 , ttcce->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , + ttcce->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); +__raw_writel (0x1 , ttcce->xttc.base_addr +XTTCPS_IER_OFFSET ); + + err =request_irq (irq, xttcps_clock_event_interrupt, +IRQF_DISABLED |IRQF_TIMER , + ttcce->ce.name, ttcce); +if (WARN_ON (err)) +return ; + +clockevents_config_and_register (&ttcce->ce, +clk_get_rate (ttcce->xttc.clk) /PRESCALE ,1 ,0xfffe ); +} + +/** +* xttcps_timer_init - Initialize the timer +* +* Initializes the timer hardware and register the clock source and clock event +* timers with Linux kernal timer framework +*/ +static void __init xttcps_timer_init (struct device_node *timer) +{
drivers/clocksource/zynq_timer.c line 400
Quoted 10 lines
drivers/clocksource/zynq_timer.c new file +10
-0+unsigned int irq; +void __iomem *timer_baseaddr; +struct clk *clk; +static int initialized; + +if (initialized) +return ; + + initialized =1 ; +
??? This looks very strange ???
I think this should be handled in a different manner.
Quoted 20 lines
drivers/clocksource/zynq_timer.c new file +20
-0+/* +* Get the 1st Triple Timer Counter (TTC) block from the device tree +* and use it. Note that the event timer uses the interrupt and it's the +* 2nd TTC hence the irq_of_parse_and_map(,1) +*/ + timer_baseaddr =of_iomap (timer,0 ); +if (!timer_baseaddr) { +pr_err ("ERROR: invalid timer base address\n" ); +BUG (); + } + + irq =irq_of_parse_and_map (timer,1 ); +if (irq <=0 ) { +pr_err ("ERROR: invalid interrupt number\n" ); +BUG (); + } + + clk =of_clk_get_by_name (timer,"cpu_1x" ); +if (IS_ERR (clk)) { +pr_err ("ERROR: timer input clock not found\n" );
drivers/clocksource/zynq_timer.c line 430
Quoted 10 lines
drivers/clocksource/zynq_timer.c new file +10
-0+BUG (); + } + +zynq_ttc_setup_clocksource (clk, timer_baseaddr); +zynq_ttc_setup_clockevent (clk, timer_baseaddr +4 , irq); + +pr_info ("%s #0 at %p, irq=%d\n" , timer->name, timer_baseaddr, irq); +} + +CLOCKSOURCE_OF_DECLARE (zynq,"xlnx,ttc" , xttcps_timer_init);
What if do not want to use the ttc as clocksource and therefore don't want it enabled? I still have the smp_twd which I might prefere.
Regards,
Steffen
Signature · 4 lines
Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
PATCH 3/10 arm: zynq: Move timer to generic location
2013/3/25 Steffen Trumtrar <[email protected]>:
Quoted 10 lines
Hi Michal!
Quoted 7 lines · Michal Simek
Move zynq timer out of mach folder to generic location.
Signed-off-by Michal Simek <[email protected]>---
Arnd: This is the last patch as we discuss about moving this timer to generic location.
Quoted 8 lines
Quoted 3 lines
It depends on
ARCH_ZYNQas Tegra and others.
It should be ok to use it or we can use different name if you like.
---Moving out of mach-zynq is good, because this is a
Cadence Rev06 Triple Timer Counter (see Zynq-7000 TRM page 25)
Which leads me to a proposed rename to:cadence_timer.cor something like that. And the fn-names should be changed accordingly while at it.
Agree. Let me do one middle patch which will just remove all zynq reference before we can move it to generic location.
Quoted 732 lines
Quoted 686 lines
arch/arm/mach-zynq/Makefile | 2 +- arch/arm/mach-zynq/timer.c | 430 -------------------------------------- drivers/clocksource/Makefile | 1 + drivers/clocksource/zynq_timer.c | 430 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 432 insertions(+), 431 deletions(-) delete mode 100644 arch/arm/mach-zynq/timer.c create mode 100644 drivers/clocksource/zynq_timer.carch/arm/mach-zynq/Makefile +1
-1@@ -3,4 +3,4 @@ # # Common support -obj-y := common.o timer.o +obj-y := common.oarch/arm/mach-zynq/timer.c deleted +0
-430@@ -1,430 +0,0 @@ -/* -* This file contains driver for the Xilinx PS Timer Counter IP. -* -* Copyright (C) 2011 Xilinx -* -* based on arch/mips/kernel/time.c timer driver -* -* This software is licensed under the terms of the GNU General Public -* License version 2, as published by the Free Software Foundation, and -* may be copied, distributed, and modified under those terms. -* -* This program is distributed in the hope that it will be useful, -* but WITHOUT ANY WARRANTY; without even the implied warranty of -* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -* GNU General Public License for more details. -*/ - -#include <linux/clk.h> -#include <linux/interrupt.h> -#include <linux/clockchips.h> -#include <linux/of_address.h> -#include <linux/of_irq.h> -#include <linux/slab.h> -#include <linux/clk-provider.h> - -/* -* This driver configures the 2 16-bit count-up timers as follows: -* -* T1: Timer 1, clocksource for generic timekeeping -* T2: Timer 2, clockevent source for hrtimers -* T3: Timer 3, <unused> -* -* The input frequency to the timer module for emulation is 2.5MHz which is -* common to all the timer channels (T1, T2, and T3). With a pre-scaler of 32, -* the timers are clocked at 78.125KHz (12.8 us resolution). - -* The input frequency to the timer module in silicon is configurable and -* obtained from device tree. The pre-scaler of 32 is used. -*/ - -/* -* Timer Register Offset Definitions of Timer 1, Increment base address by 4 -* and use same offsets for Timer 2 -*/ -#define XTTCPS_CLK_CNTRL_OFFSET 0x00/* Clock Control Reg, RW */ -#define XTTCPS_CNT_CNTRL_OFFSET 0x0C/* Counter Control Reg, RW */ -#define XTTCPS_COUNT_VAL_OFFSET 0x18/* Counter Value Reg, RO */ -#define XTTCPS_INTR_VAL_OFFSET 0x24/* Interval Count Reg, RW */ -#define XTTCPS_ISR_OFFSET 0x54/* Interrupt Status Reg, RO */ -#define XTTCPS_IER_OFFSET 0x60/* Interrupt Enable Reg, RW */ - -#define XTTCPS_CNT_CNTRL_DISABLE_MASK 0x1 - -/* -* Setup the timers to use pre-scaling, using a fixed value for now that will -* work across most input frequency, but it may need to be more dynamic -*/ -#define PRESCALE_EXPONENT 11/* 2 ^ PRESCALE_EXPONENT = PRESCALE */ -#define PRESCALE 2048/* The exponent must match this */ -#define CLK_CNTRL_PRESCALE ((PRESCALE_EXPONENT - 1) << 1) -#define CLK_CNTRL_PRESCALE_EN 1 -#define CNT_CNTRL_RESET (1 << 4) - -/** -* struct xttcps_timer - This definition defines local timer structure -* -* @base_addr: Base address of timer -* @clk: Associated clock source -* @clk_rate_change_nb Notifier block for clock rate changes -*/ -struct xttcps_timer { -void __iomem *base_addr; -struct clk *clk; -struct notifier_block clk_rate_change_nb; -}; - -#define to_xttcps_timer (x) \ - container_of(x, struct xttcps_timer, clk_rate_change_nb) - -struct xttcps_timer_clocksource { -struct xttcps_timer xttc; -struct clocksource cs; -}; - -#define to_xttcps_timer_clksrc (x) \ - container_of(x, struct xttcps_timer_clocksource, cs) - -struct xttcps_timer_clockevent { -struct xttcps_timer xttc; -struct clock_event_device ce; -}; - -#define to_xttcps_timer_clkevent (x) \ - container_of(x, struct xttcps_timer_clockevent, ce) - -/** -* xttcps_set_interval - Set the timer interval value -* -* @timer: Pointer to the timer instance -* @cycles: Timer interval ticks -**/ -static void xttcps_set_interval (struct xttcps_timer *timer, -unsigned long cycles) -{ -u32 ctrl_reg; - -/* Disable the counter, set the counter value and re-enable counter */ - ctrl_reg =__raw_readl (timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); - ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); - -__raw_writel (cycles, timer->base_addr +XTTCPS_INTR_VAL_OFFSET ); - -/* -* Reset the counter (0x10) so that it starts from 0, one-shot -* mode makes this needed for timing to be right. -*/ - ctrl_reg |=CNT_CNTRL_RESET ; - ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -} - -/** -* xttcps_clock_event_interrupt - Clock event timer interrupt handler -* -* @irq: IRQ number of the Timer -* @dev_id: void pointer to the xttcps_timer instance -* -* returns: Always IRQ_HANDLED - success -**/ -static irqreturn_t xttcps_clock_event_interrupt (int irq,void *dev_id) -{ -struct xttcps_timer_clockevent *xttce = dev_id; -struct xttcps_timer *timer = &xttce->xttc; - -/* Acknowledge the interrupt and call event handler */ -__raw_readl (timer->base_addr +XTTCPS_ISR_OFFSET ); - - xttce->ce.event_handler (&xttce->ce); - -return IRQ_HANDLED ; -} - -/** -* __xttc_clocksource_read - Reads the timer counter register -* -* returns: Current timer counter register value -**/ -static cycle_t __xttc_clocksource_read (struct clocksource *cs) -{ -struct xttcps_timer *timer = &to_xttcps_timer_clksrc (cs)->xttc; - -return (cycle_t )__raw_readl (timer->base_addr + -XTTCPS_COUNT_VAL_OFFSET ); -} - -/** -* xttcps_set_next_event - Sets the time interval for next event -* -* @cycles: Timer interval ticks -* @evt: Address of clock event instance -* -* returns: Always 0 - success -**/ -static int xttcps_set_next_event (unsigned long cycles, -struct clock_event_device *evt) -{ -struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); -struct xttcps_timer *timer = &xttce->xttc; - -xttcps_set_interval (timer, cycles); -return 0 ; -} - -/** -* xttcps_set_mode - Sets the mode of timer -* -* @mode: Mode to be set -* @evt: Address of clock event instance -**/ -static void xttcps_set_mode (enum clock_event_mode mode, -struct clock_event_device *evt) -{ -struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); -struct xttcps_timer *timer = &xttce->xttc; -u32 ctrl_reg; - -switch (mode) { -case CLOCK_EVT_MODE_PERIODIC : -xttcps_set_interval (timer, -DIV_ROUND_CLOSEST (clk_get_rate (xttce->xttc.clk), -PRESCALE *HZ )); -break ; -case CLOCK_EVT_MODE_ONESHOT : -case CLOCK_EVT_MODE_UNUSED : -case CLOCK_EVT_MODE_SHUTDOWN : - ctrl_reg =__raw_readl (timer->base_addr + -XTTCPS_CNT_CNTRL_OFFSET ); - ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, - timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -break ; -case CLOCK_EVT_MODE_RESUME : - ctrl_reg =__raw_readl (timer->base_addr + -XTTCPS_CNT_CNTRL_OFFSET ); - ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; -__raw_writel (ctrl_reg, - timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -break ; - } -} - -static int xttcps_rate_change_clocksource_cb (struct notifier_block *nb, -unsigned long event,void *data) -{ -struct clk_notifier_data *ndata = data; -struct xttcps_timer *xttcps =to_xttcps_timer (nb); -struct xttcps_timer_clocksource *xttccs =container_of (xttcps, - struct xttcps_timer_clocksource, xttc); - -switch (event) { -case POST_RATE_CHANGE : -/* -* Do whatever is necessare to maintain a proper time base -* -* I cannot find a way to adjust the currently used clocksource -* to the new frequency. __clocksource_updatefreq_hz() sounds -* good, but does not work. Not sure what's that missing. -* -* This approach works, but triggers two clocksource switches. -* The first after unregister to clocksource jiffies. And -* another one after the register to the newly registered timer. -* -* Alternatively we could 'waste' another HW timer to ping pong -* between clock sources. That would also use one register and -* one unregister call, but only trigger one clocksource switch -* for the cost of another HW timer used by the OS. -*/ -clocksource_unregister (&xttccs->cs); -clocksource_register_hz (&xttccs->cs, - ndata->new_rate /PRESCALE ); -/* fall through */ -case PRE_RATE_CHANGE : -case ABORT_RATE_CHANGE : -default : -return NOTIFY_DONE ; - } -} - -static void __init zynq_ttc_setup_clocksource (struct clk *clk, -void __iomem *base) -{ -struct xttcps_timer_clocksource *ttccs; -int err; - - ttccs =kzalloc (sizeof (*ttccs),GFP_KERNEL ); -if (WARN_ON (!ttccs)) -return ; - - ttccs->xttc.clk = clk; - - err =clk_prepare_enable (ttccs->xttc.clk); -if (WARN_ON (err)) -return ; - - ttccs->xttc.clk_rate_change_nb.notifier_call = - xttcps_rate_change_clocksource_cb; - ttccs->xttc.clk_rate_change_nb.next =NULL ; -if (clk_notifier_register (ttccs->xttc.clk, - &ttccs->xttc.clk_rate_change_nb)) -pr_warn ("Unable to register clock notifier.\n" ); - - ttccs->xttc.base_addr = base; - ttccs->cs.name ="xttcps_clocksource" ; - ttccs->cs.rating =200 ; - ttccs->cs.read = __xttc_clocksource_read; - ttccs->cs.mask =CLOCKSOURCE_MASK (16 ); - ttccs->cs.flags =CLOCK_SOURCE_IS_CONTINUOUS ; - -/* -* Setup the clock source counter to be an incrementing counter -* with no interrupt and it rolls over at 0xFFFF. Pre-scale -* it by 32 also. Let it start running now. -*/ -__raw_writel (0x0 , ttccs->xttc.base_addr +XTTCPS_IER_OFFSET ); -__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , - ttccs->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); -__raw_writel (CNT_CNTRL_RESET , - ttccs->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); - - err =clocksource_register_hz (&ttccs->cs, -clk_get_rate (ttccs->xttc.clk) /PRESCALE ); -if (WARN_ON (err)) -return ; - -} - -static int xttcps_rate_change_clockevent_cb (struct notifier_block *nb, -unsigned long event,void *data) -{ -struct clk_notifier_data *ndata = data; -struct xttcps_timer *xttcps =to_xttcps_timer (nb); -struct xttcps_timer_clockevent *xttcce =container_of (xttcps, - struct xttcps_timer_clockevent, xttc); - -switch (event) { -case POST_RATE_CHANGE : - { -unsigned long flags; - -/* -* clockevents_update_freq should be called with IRQ disabled on -* the CPU the timer provides events for. The timer we use is -* common to both CPUs, not sure if we need to run on both -* cores. -*/ -local_irq_save (flags); -clockevents_update_freq (&xttcce->ce, - ndata->new_rate /PRESCALE ); -local_irq_restore (flags); - -/* fall through */ - } -case PRE_RATE_CHANGE : -case ABORT_RATE_CHANGE : -default : -return NOTIFY_DONE ; - } -} - -static void __init zynq_ttc_setup_clockevent (struct clk *clk, -void __iomem *base,u32 irq) -{ -struct xttcps_timer_clockevent *ttcce; -int err; - - ttcce =kzalloc (sizeof (*ttcce),GFP_KERNEL ); -if (WARN_ON (!ttcce)) -return ; - - ttcce->xttc.clk = clk; - - err =clk_prepare_enable (ttcce->xttc.clk); -if (WARN_ON (err)) -return ; - - ttcce->xttc.clk_rate_change_nb.notifier_call = - xttcps_rate_change_clockevent_cb; - ttcce->xttc.clk_rate_change_nb.next =NULL ; -if (clk_notifier_register (ttcce->xttc.clk, - &ttcce->xttc.clk_rate_change_nb)) -pr_warn ("Unable to register clock notifier.\n" ); - - ttcce->xttc.base_addr = base; - ttcce->ce.name ="xttcps_clockevent" ; - ttcce->ce.features =CLOCK_EVT_FEAT_PERIODIC |CLOCK_EVT_FEAT_ONESHOT ; - ttcce->ce.set_next_event = xttcps_set_next_event; - ttcce->ce.set_mode = xttcps_set_mode; - ttcce->ce.rating =200 ; - ttcce->ce.irq = irq; - ttcce->ce.cpumask = cpu_possible_mask; - -/* -* Setup the clock event timer to be an interval timer which -* is prescaled by 32 using the interval interrupt. Leave it -* disabled for now. -*/ -__raw_writel (0x23 , ttcce->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); -__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , - ttcce->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); -__raw_writel (0x1 , ttcce->xttc.base_addr +XTTCPS_IER_OFFSET ); - - err =request_irq (irq, xttcps_clock_event_interrupt, -IRQF_DISABLED |IRQF_TIMER , - ttcce->ce.name, ttcce); -if (WARN_ON (err)) -return ; - -clockevents_config_and_register (&ttcce->ce, -clk_get_rate (ttcce->xttc.clk) /PRESCALE ,1 ,0xfffe ); -} - -/** -* xttcps_timer_init - Initialize the timer -* -* Initializes the timer hardware and register the clock source and clock event -* timers with Linux kernal timer framework -*/ -static void __init xttcps_timer_init (struct device_node *timer) -{ -unsigned int irq; -void __iomem *timer_baseaddr; -struct clk *clk; -static int initialized; - -if (initialized) -return ; - - initialized =1 ; - -/* -* Get the 1st Triple Timer Counter (TTC) block from the device tree -* and use it. Note that the event timer uses the interrupt and it's the -* 2nd TTC hence the irq_of_parse_and_map(,1) -*/ - timer_baseaddr =of_iomap (timer,0 ); -if (!timer_baseaddr) { -pr_err ("ERROR: invalid timer base address\n" ); -BUG (); - } - - irq =irq_of_parse_and_map (timer,1 ); -if (irq <=0 ) { -pr_err ("ERROR: invalid interrupt number\n" ); -BUG (); - } - - clk =of_clk_get_by_name (timer,"cpu_1x" ); -if (IS_ERR (clk)) { -pr_err ("ERROR: timer input clock not found\n" ); -BUG (); - } - -zynq_ttc_setup_clocksource (clk, timer_baseaddr); -zynq_ttc_setup_clockevent (clk, timer_baseaddr +4 , irq); - -pr_info ("%s #0 at %p, irq=%d\n" , timer->name, timer_baseaddr, irq); -} - -CLOCKSOURCE_OF_DECLARE (zynq,"xlnx,ttc" , xttcps_timer_init);drivers/clocksource/Makefile +1
-0@@ -19,6 +19,7 @@ obj-$(CONFIG_ARCH_BCM2835) += bcm2835_timer.o obj-$(CONFIG_SUNXI_TIMER) += sunxi_timer.o obj-$(CONFIG_ARCH_TEGRA) += tegra20_timer.o obj-$(CONFIG_VT8500_TIMER) += vt8500_timer.o +obj-$(CONFIG_ARCH_ZYNQ) += zynq_timer.o obj-$(CONFIG_ARM_ARCH_TIMER) += arm_arch_timer.o obj-$(CONFIG_CLKSRC_METAG_GENERIC) += metag_generic.odrivers/clocksource/zynq_timer.c new file +214
-0@@ -0,0 +1,430 @@ +/* +* This file contains driver for the Xilinx PS Timer Counter IP. +* +* Copyright (C) 2011 Xilinx +* +* based on arch/mips/kernel/time.c timer driver +* +* This software is licensed under the terms of the GNU General Public +* License version 2, as published by the Free Software Foundation, and +* may be copied, distributed, and modified under those terms. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +*/ + +#include <linux/clk.h> +#include <linux/interrupt.h> +#include <linux/clockchips.h> +#include <linux/of_address.h> +#include <linux/of_irq.h> +#include <linux/slab.h> +#include <linux/clk-provider.h> + +/* +* This driver configures the 2 16-bit count-up timers as follows: +* +* T1: Timer 1, clocksource for generic timekeeping +* T2: Timer 2, clockevent source for hrtimers +* T3: Timer 3, <unused> +* +* The input frequency to the timer module for emulation is 2.5MHz which is +* common to all the timer channels (T1, T2, and T3). With a pre-scaler of 32, +* the timers are clocked at 78.125KHz (12.8 us resolution). + +* The input frequency to the timer module in silicon is configurable and +* obtained from device tree. The pre-scaler of 32 is used. +*/ + +/* +* Timer Register Offset Definitions of Timer 1, Increment base address by 4 +* and use same offsets for Timer 2 +*/ +#define XTTCPS_CLK_CNTRL_OFFSET 0x00/* Clock Control Reg, RW */ +#define XTTCPS_CNT_CNTRL_OFFSET 0x0C/* Counter Control Reg, RW */ +#define XTTCPS_COUNT_VAL_OFFSET 0x18/* Counter Value Reg, RO */ +#define XTTCPS_INTR_VAL_OFFSET 0x24/* Interval Count Reg, RW */ +#define XTTCPS_ISR_OFFSET 0x54/* Interrupt Status Reg, RO */ +#define XTTCPS_IER_OFFSET 0x60/* Interrupt Enable Reg, RW */ + +#define XTTCPS_CNT_CNTRL_DISABLE_MASK 0x1 + +/* +* Setup the timers to use pre-scaling, using a fixed value for now that will +* work across most input frequency, but it may need to be more dynamic +*/ +#define PRESCALE_EXPONENT 11/* 2 ^ PRESCALE_EXPONENT = PRESCALE */ +#define PRESCALE 2048/* The exponent must match this */ +#define CLK_CNTRL_PRESCALE ((PRESCALE_EXPONENT - 1) << 1) +#define CLK_CNTRL_PRESCALE_EN 1 +#define CNT_CNTRL_RESET (1 << 4) + +/** +* struct xttcps_timer - This definition defines local timer structure +* +* @base_addr: Base address of timer +* @clk: Associated clock source +* @clk_rate_change_nb Notifier block for clock rate changes +*/ +struct xttcps_timer { +void __iomem *base_addr; +struct clk *clk; +struct notifier_block clk_rate_change_nb; +}; + +#define to_xttcps_timer (x) \ + container_of(x, struct xttcps_timer, clk_rate_change_nb) + +struct xttcps_timer_clocksource { +struct xttcps_timer xttc; +struct clocksource cs; +}; + +#define to_xttcps_timer_clksrc (x) \ + container_of(x, struct xttcps_timer_clocksource, cs) + +struct xttcps_timer_clockevent { +struct xttcps_timer xttc; +struct clock_event_device ce; +}; + +#define to_xttcps_timer_clkevent (x) \ + container_of(x, struct xttcps_timer_clockevent, ce) + +/** +* xttcps_set_interval - Set the timer interval value +* +* @timer: Pointer to the timer instance +* @cycles: Timer interval ticks +**/ +static void xttcps_set_interval (struct xttcps_timer *timer, +unsigned long cycles) +{ +u32 ctrl_reg; + +/* Disable the counter, set the counter value and re-enable counter */ + ctrl_reg =__raw_readl (timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); + ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); + +__raw_writel (cycles, timer->base_addr +XTTCPS_INTR_VAL_OFFSET ); + +/* +* Reset the counter (0x10) so that it starts from 0, one-shot +* mode makes this needed for timing to be right. +*/ + ctrl_reg |=CNT_CNTRL_RESET ; + ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +} + +/** +* xttcps_clock_event_interrupt - Clock event timer interrupt handler +* +* @irq: IRQ number of the Timer +* @dev_id: void pointer to the xttcps_timer instance +* +* returns: Always IRQ_HANDLED - success +**/ +static irqreturn_t xttcps_clock_event_interrupt (int irq,void *dev_id) +{ +struct xttcps_timer_clockevent *xttce = dev_id; +struct xttcps_timer *timer = &xttce->xttc; + +/* Acknowledge the interrupt and call event handler */ +__raw_readl (timer->base_addr +XTTCPS_ISR_OFFSET ); + + xttce->ce.event_handler (&xttce->ce); + +return IRQ_HANDLED ; +} + +/** +* __xttc_clocksource_read - Reads the timer counter register +* +* returns: Current timer counter register value +**/ +static cycle_t __xttc_clocksource_read (struct clocksource *cs) +{ +struct xttcps_timer *timer = &to_xttcps_timer_clksrc (cs)->xttc; + +return (cycle_t )__raw_readl (timer->base_addr + +XTTCPS_COUNT_VAL_OFFSET ); +} + +/** +* xttcps_set_next_event - Sets the time interval for next event +* +* @cycles: Timer interval ticks +* @evt: Address of clock event instance +* +* returns: Always 0 - success +**/ +static int xttcps_set_next_event (unsigned long cycles, +struct clock_event_device *evt) +{ +struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); +struct xttcps_timer *timer = &xttce->xttc; + +xttcps_set_interval (timer, cycles); +return 0 ; +} + +/** +* xttcps_set_mode - Sets the mode of timer +* +* @mode: Mode to be set +* @evt: Address of clock event instance +**/ +static void xttcps_set_mode (enum clock_event_mode mode, +struct clock_event_device *evt) +{ +struct xttcps_timer_clockevent *xttce =to_xttcps_timer_clkevent (evt); +struct xttcps_timer *timer = &xttce->xttc; +u32 ctrl_reg; + +switch (mode) { +case CLOCK_EVT_MODE_PERIODIC : +xttcps_set_interval (timer, +DIV_ROUND_CLOSEST (clk_get_rate (xttce->xttc.clk), +PRESCALE *HZ )); +break ; +case CLOCK_EVT_MODE_ONESHOT : +case CLOCK_EVT_MODE_UNUSED : +case CLOCK_EVT_MODE_SHUTDOWN : + ctrl_reg =__raw_readl (timer->base_addr + +XTTCPS_CNT_CNTRL_OFFSET ); + ctrl_reg |=XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, + timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +break ; +case CLOCK_EVT_MODE_RESUME : + ctrl_reg =__raw_readl (timer->base_addr + +XTTCPS_CNT_CNTRL_OFFSET ); + ctrl_reg &= ~XTTCPS_CNT_CNTRL_DISABLE_MASK ; +__raw_writel (ctrl_reg, + timer->base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +break ; + } +} + +static int xttcps_rate_change_clocksource_cb (struct notifier_block *nb, +unsigned long event,void *data)drivers/clocksource/zynq_timer.c line 224
Quoted 10 lines
drivers/clocksource/zynq_timer.c new file +10
-0+{ +struct clk_notifier_data *ndata = data; +struct xttcps_timer *xttcps =to_xttcps_timer (nb); +struct xttcps_timer_clocksource *xttccs =container_of (xttcps, + struct xttcps_timer_clocksource, xttc); + +switch (event) { +case POST_RATE_CHANGE : +/* +* Do whatever is necessare to maintain a proper time base ------------------------------------------^ little typo
Quoted 34 lines
drivers/clocksource/zynq_timer.c new file +34
-0+* +* I cannot find a way to adjust the currently used clocksource +* to the new frequency. __clocksource_updatefreq_hz() sounds +* good, but does not work. Not sure what's that missing. +* +* This approach works, but triggers two clocksource switches. +* The first after unregister to clocksource jiffies. And +* another one after the register to the newly registered timer. +* +* Alternatively we could 'waste' another HW timer to ping pong +* between clock sources. That would also use one register and +* one unregister call, but only trigger one clocksource switch +* for the cost of another HW timer used by the OS. +*/ +clocksource_unregister (&xttccs->cs); +clocksource_register_hz (&xttccs->cs, + ndata->new_rate /PRESCALE ); +/* fall through */ +case PRE_RATE_CHANGE : +case ABORT_RATE_CHANGE : +default : +return NOTIFY_DONE ; + } +} + +static void __init zynq_ttc_setup_clocksource (struct clk *clk, +void __iomem *base) +{ +struct xttcps_timer_clocksource *ttccs; +int err; + + ttccs =kzalloc (sizeof (*ttccs),GFP_KERNEL ); +if (WARN_ON (!ttccs)) +return ;
Quoted 8 lines
drivers/clocksource/zynq_timer.c line 264
Quoted 6 lines
drivers/clocksource/zynq_timer.c new file +6
-0+ + ttccs->xttc.clk = clk; + + err =clk_prepare_enable (ttccs->xttc.clk); +if (WARN_ON (err)) +return ;I think you are losing memory here...
right. Let's free that memory. I have pinged Soren to look if we can handle clock better.
Till now this is only one timer in the system we have used. Look below.
Quoted 24 lines
Quoted 24 lines
drivers/clocksource/zynq_timer.c new file +23
-0+ + ttccs->xttc.clk_rate_change_nb.notifier_call = + xttcps_rate_change_clocksource_cb; + ttccs->xttc.clk_rate_change_nb.next =NULL ; +if (clk_notifier_register (ttccs->xttc.clk, + &ttccs->xttc.clk_rate_change_nb)) +pr_warn ("Unable to register clock notifier.\n" ); + + ttccs->xttc.base_addr = base; + ttccs->cs.name ="xttcps_clocksource" ; + ttccs->cs.rating =200 ; + ttccs->cs.read = __xttc_clocksource_read; + ttccs->cs.mask =CLOCKSOURCE_MASK (16 ); + ttccs->cs.flags =CLOCK_SOURCE_IS_CONTINUOUS ; + +/* +* Setup the clock source counter to be an incrementing counter +* with no interrupt and it rolls over at 0xFFFF. Pre-scale +* it by 32 also. Let it start running now. +*/ +__raw_writel (0x0 , ttccs->xttc.base_addr +XTTCPS_IER_OFFSET ); +__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , + ttccs->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET );
Quoted 8 lines
drivers/clocksource/zynq_timer.c line 294
Quoted 7 lines
drivers/clocksource/zynq_timer.c new file +7
-0+__raw_writel (CNT_CNTRL_RESET , + ttccs->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); + + err =clocksource_register_hz (&ttccs->cs, +clk_get_rate (ttccs->xttc.clk) /PRESCALE ); +if (WARN_ON (err)) +return ;... and here and below in
zynq_ttc_setup_clockeventthe same.
yep
Quoted 100 lines
Quoted 100 lines
drivers/clocksource/zynq_timer.c new file +100
-0+ +} + +static int xttcps_rate_change_clockevent_cb (struct notifier_block *nb, +unsigned long event,void *data) +{ +struct clk_notifier_data *ndata = data; +struct xttcps_timer *xttcps =to_xttcps_timer (nb); +struct xttcps_timer_clockevent *xttcce =container_of (xttcps, + struct xttcps_timer_clockevent, xttc); + +switch (event) { +case POST_RATE_CHANGE : + { +unsigned long flags; + +/* +* clockevents_update_freq should be called with IRQ disabled on +* the CPU the timer provides events for. The timer we use is +* common to both CPUs, not sure if we need to run on both +* cores. +*/ +local_irq_save (flags); +clockevents_update_freq (&xttcce->ce, + ndata->new_rate /PRESCALE ); +local_irq_restore (flags); + +/* fall through */ + } +case PRE_RATE_CHANGE : +case ABORT_RATE_CHANGE : +default : +return NOTIFY_DONE ; + } +} + +static void __init zynq_ttc_setup_clockevent (struct clk *clk, +void __iomem *base,u32 irq) +{ +struct xttcps_timer_clockevent *ttcce; +int err; + + ttcce =kzalloc (sizeof (*ttcce),GFP_KERNEL ); +if (WARN_ON (!ttcce)) +return ; + + ttcce->xttc.clk = clk; + + err =clk_prepare_enable (ttcce->xttc.clk); +if (WARN_ON (err)) +return ; + + ttcce->xttc.clk_rate_change_nb.notifier_call = + xttcps_rate_change_clockevent_cb; + ttcce->xttc.clk_rate_change_nb.next =NULL ; +if (clk_notifier_register (ttcce->xttc.clk, + &ttcce->xttc.clk_rate_change_nb)) +pr_warn ("Unable to register clock notifier.\n" ); + + ttcce->xttc.base_addr = base; + ttcce->ce.name ="xttcps_clockevent" ; + ttcce->ce.features =CLOCK_EVT_FEAT_PERIODIC |CLOCK_EVT_FEAT_ONESHOT ; + ttcce->ce.set_next_event = xttcps_set_next_event; + ttcce->ce.set_mode = xttcps_set_mode; + ttcce->ce.rating =200 ; + ttcce->ce.irq = irq; + ttcce->ce.cpumask = cpu_possible_mask; + +/* +* Setup the clock event timer to be an interval timer which +* is prescaled by 32 using the interval interrupt. Leave it +* disabled for now. +*/ +__raw_writel (0x23 , ttcce->xttc.base_addr +XTTCPS_CNT_CNTRL_OFFSET ); +__raw_writel (CLK_CNTRL_PRESCALE |CLK_CNTRL_PRESCALE_EN , + ttcce->xttc.base_addr +XTTCPS_CLK_CNTRL_OFFSET ); +__raw_writel (0x1 , ttcce->xttc.base_addr +XTTCPS_IER_OFFSET ); + + err =request_irq (irq, xttcps_clock_event_interrupt, +IRQF_DISABLED |IRQF_TIMER , + ttcce->ce.name, ttcce); +if (WARN_ON (err)) +return ; + +clockevents_config_and_register (&ttcce->ce, +clk_get_rate (ttcce->xttc.clk) /PRESCALE ,1 ,0xfffe ); +} + +/** +* xttcps_timer_init - Initialize the timer +* +* Initializes the timer hardware and register the clock source and clock event +* timers with Linux kernal timer framework +*/ +static void __init xttcps_timer_init (struct device_node *timer) +{ +unsigned int irq; +void __iomem *timer_baseaddr; +struct clk *clk; +static int initialized;
Quoted 8 lines
drivers/clocksource/zynq_timer.c line 400
Quoted 6 lines
drivers/clocksource/zynq_timer.c new file +6
-0+ +if (initialized) +return ; + + initialized =1 ; +??? This looks very strange ???
I think this should be handled in a different manner.
Feel free to propose better solution. Because there are two ttc in the system and the first
is taken, second shouldn't be initialized as clocksource or clockevent.
Quoted 25 lines
Quoted 25 lines
drivers/clocksource/zynq_timer.c new file +25
-0+/* +* Get the 1st Triple Timer Counter (TTC) block from the device tree +* and use it. Note that the event timer uses the interrupt and it's the +* 2nd TTC hence the irq_of_parse_and_map(,1) +*/ + timer_baseaddr =of_iomap (timer,0 ); +if (!timer_baseaddr) { +pr_err ("ERROR: invalid timer base address\n" ); +BUG (); + } + + irq =irq_of_parse_and_map (timer,1 ); +if (irq <=0 ) { +pr_err ("ERROR: invalid interrupt number\n" ); +BUG (); + } + + clk =of_clk_get_by_name (timer,"cpu_1x" ); +if (IS_ERR (clk)) { +pr_err ("ERROR: timer input clock not found\n" ); +BUG (); + } + +zynq_ttc_setup_clocksource (clk, timer_baseaddr); +zynq_ttc_setup_clockevent (clk, timer_baseaddr +4 , irq);
Quoted 8 lines
drivers/clocksource/zynq_timer.c line 430
Quoted 5 lines
drivers/clocksource/zynq_timer.c new file +5
-0+ +pr_info ("%s #0 at %p, irq=%d\n" , timer->name, timer_baseaddr, irq); +} + +CLOCKSOURCE_OF_DECLARE (zynq,"xlnx,ttc" , xttcps_timer_init);What if do not want to use the ttc as clocksource and therefore don't want it enabled? I still have the
smp_twdwhich I might prefere.
Renaming this to cadence name we will have to change dt name to more generic one cdns,ttc? Also we should probably add it to Kconfig.
IRC: If there is not compatible string kernel will try to find out different one.
If you find some of them the more precise one is used. It is quite long time I was playing with it.
Thanks,
Michal
Signature · 5 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel - Microblaze cpu - http://www.monstr.eu/fdt/ Maintainer of Linux kernel - Xilinx Zynq ARM architecture Microblaze U-BOOT custodian and responsible for u-boot arm zynq platform
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
In reply to Michal SimekUse Cortex a9 cp15 to read scu baseaddress.
--- arch/arm/boot/dts/zynq-7000.dtsi | 5 +++++ arch/arm/mach-zynq/common.c | 33 +++++++++++++++++++++------------ arch/arm/mach-zynq/common.h | 2 ++ 3 files changed, 28 insertions(+), 12 deletions(-)
arch/arm/boot/dts/zynq-7000.dtsi +5 -0
@@ -22,6 +22,11 @@ interrupt-parent = <&intc>; ranges; + scu: scu at f8f000000 { + compatible = "arm,cortex-a9-scu"; + reg = <0xf8f00000 0x58>; + }; + intc: interrupt-controller at f8f01000 { compatible = "arm,cortex-a9-gic"; #interrupt-cells = <3>;
arch/arm/mach-zynq/common.c +21 -12
@@ -33,10 +33,13 @@#include <asm/mach-types.h> #include <asm/page.h> #include <asm/pgtable.h> +#include <asm/smp_scu.h> #include <asm/hardware/cache-l2x0.h> #include "common.h" +void __iomem *scu_base; +static struct of_device_id zynq_of_bus_ids[]__initdata = { { .compatible ="simple-bus" , }, {}
@@ -56,17 +59,6 @@ static void __init xilinx_init_machine(void)of_platform_bus_probe (NULL , zynq_of_bus_ids,NULL ); } -#define SCU_PERIPH_PHYS 0xF8F00000 -#define SCU_PERIPH_SIZE SZ_8K -#define SCU_PERIPH_VIRT (VMALLOC_END - SCU_PERIPH_SIZE) - -static struct map_desc scu_desc__initdata = { - .virtual =SCU_PERIPH_VIRT , - .pfn =__phys_to_pfn (SCU_PERIPH_PHYS ), - .length =SCU_PERIPH_SIZE , - .type =MT_DEVICE , -}; -static void __init xilinx_zynq_timer_init (void ) {struct device_node *np;
@@ -81,13 +73,30 @@ static void __init xilinx_zynq_timer_init(void)clocksource_of_init (); } +static struct map_desc zynq_cortex_a9_scu_map__initdata = { + .length =SZ_256 , + .type =MT_DEVICE , +}; + +static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base; +iotable_init (&zynq_cortex_a9_scu_map,1 ); + scu_base =ioremap (base, zynq_cortex_a9_scu_map.length); +BUG_ON (!scu_base); +} +/** * xilinx_map_io() - Create memory mappings needed for early I/O. */ static void __init xilinx_map_io (void ) {debug_ll_io_init (); -iotable_init (&scu_desc,1 ); +scu_init (); }static const char *xilinx_dt_match[] = {
arch/arm/mach-zynq/common.h +2 -0
@@ -17,4 +17,6 @@#ifndef __MACH_ZYNQ_COMMON_H__#define __MACH_ZYNQ_COMMON_H__ +extern void __iomem *scu_base; +#endif
Signature · 1 line
1.7.9.7
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 14 lines · Michal Simek
Use Cortex a9 cp15 to read scu baseaddress.
Signed-off-by Michal Simek <[email protected]>--- arch/arm/boot/dts/zynq-7000.dtsi | 5 +++++ arch/arm/mach-zynq/common.c | 33 +++++++++++++++++++++------------ arch/arm/mach-zynq/common.h | 2 ++ 3 files changed, 28 insertions(+), 12 deletions(-)arch/arm/boot/dts/zynq-7000.dtsi +0
-0
arch/arm/boot/dts/zynq-7000.dtsi line 29
Quoted 8 lines
arch/arm/boot/dts/zynq-7000.dtsi +5
-0@@ -22,6 +22,11 @@ interrupt-parent = <&intc>; ranges; + scu: scu at f8f000000 { + compatible = "arm,cortex-a9-scu"; + reg = <0xf8f00000 0x58>; + }; +
It's fine to add this, but you don't really need it for this patch.
Quoted 45 lines
arch/arm/boot/dts/zynq-7000.dtsi +0
-0intc: interrupt-controller at f8f01000 { compatible = "arm,cortex-a9-gic"; #interrupt-cells = <3>;arch/arm/mach-zynq/common.c +5
-11@@ -33,10 +33,13 @@#include <asm/mach-types.h> #include <asm/page.h> #include <asm/pgtable.h> +#include <asm/smp_scu.h> #include <asm/hardware/cache-l2x0.h> #include "common.h" +void __iomem *scu_base; +static struct of_device_id zynq_of_bus_ids[]__initdata = { { .compatible ="simple-bus" , }, {}@@ -56,17 +59,6 @@ static void __init xilinx_init_machine(void)of_platform_bus_probe (NULL , zynq_of_bus_ids,NULL ); } -#define SCU_PERIPH_PHYS 0xF8F00000 -#define SCU_PERIPH_SIZE SZ_8K -#define SCU_PERIPH_VIRT (VMALLOC_END - SCU_PERIPH_SIZE) - -static struct map_desc scu_desc__initdata = { - .virtual =SCU_PERIPH_VIRT , - .pfn =__phys_to_pfn (SCU_PERIPH_PHYS ), - .length =SCU_PERIPH_SIZE , - .type =MT_DEVICE , -}; -static void __init xilinx_zynq_timer_init (void ) {struct device_node *np;@@ -81,13 +73,30 @@ static void __init xilinx_zynq_timer_init(void)clocksource_of_init (); } +static struct map_desc zynq_cortex_a9_scu_map__initdata = { + .length =SZ_256 ,
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 87
Quoted 10 lines
arch/arm/mach-zynq/common.c +10
-0+ .type = MT_DEVICE, +}; + +static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base;
You are setting the virtual address to the physical base?
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 88
Quoted 1 line
arch/arm/mach-zynq/common.c +1
-0+iotable_init (&zynq_cortex_a9_scu_map,1 );
Then creating a static mapping...
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 89
Quoted 1 line
arch/arm/mach-zynq/common.c +1
-0+ scu_base =ioremap (base, zynq_cortex_a9_scu_map.length);
And also a dynamic mapping?
Rob
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 18 lines · Michal Simek
Quoted 17 lines
Use Cortex a9 cp15 to read scu baseaddress.
Signed-off-by Michal Simek <[email protected]>--- arch/arm/boot/dts/zynq-7000.dtsi | 5 +++++ arch/arm/mach-zynq/common.c | 33 +++++++++++++++++++++------------ arch/arm/mach-zynq/common.h | 2 ++ 3 files changed, 28 insertions(+), 12 deletions(-)arch/arm/boot/dts/zynq-7000.dtsi +0
-0@@ -22,6 +22,11 @@ interrupt-parent = <&intc>; ranges;
Quoted 7 lines
arch/arm/boot/dts/zynq-7000.dtsi line 29
Quoted 5 lines
arch/arm/boot/dts/zynq-7000.dtsi +5
-0+ scu: scu at f8f000000 { + compatible = "arm,cortex-a9-scu"; + reg = <0xf8f00000 0x58>; + }; +It's fine to add this, but you don't really need it for this patch.
yes truth.
Quoted 59 lines
Quoted 46 lines
arch/arm/boot/dts/zynq-7000.dtsi +0
-0intc: interrupt-controller at f8f01000 { compatible = "arm,cortex-a9-gic"; #interrupt-cells = <3>;arch/arm/mach-zynq/common.c +5
-11@@ -33,10 +33,13 @@#include <asm/mach-types.h> #include <asm/page.h> #include <asm/pgtable.h> +#include <asm/smp_scu.h> #include <asm/hardware/cache-l2x0.h> #include "common.h" +void __iomem *scu_base; +static struct of_device_id zynq_of_bus_ids[]__initdata = { { .compatible ="simple-bus" , }, {}@@ -56,17 +59,6 @@ static void __init xilinx_init_machine(void)of_platform_bus_probe (NULL , zynq_of_bus_ids,NULL ); } -#define SCU_PERIPH_PHYS 0xF8F00000 -#define SCU_PERIPH_SIZE SZ_8K -#define SCU_PERIPH_VIRT (VMALLOC_END - SCU_PERIPH_SIZE) - -static struct map_desc scu_desc__initdata = { - .virtual =SCU_PERIPH_VIRT , - .pfn =__phys_to_pfn (SCU_PERIPH_PHYS ), - .length =SCU_PERIPH_SIZE , - .type =MT_DEVICE , -}; -static void __init xilinx_zynq_timer_init (void ) {struct device_node *np;@@ -81,13 +73,30 @@ static void __init xilinx_zynq_timer_init(void)clocksource_of_init (); } +static struct map_desc zynq_cortex_a9_scu_map__initdata = { + .length =SZ_256 ,arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 87
Quoted 10 lines
arch/arm/mach-zynq/common.c +10
-0+ .type = MT_DEVICE, +}; + +static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 7 lines
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 88
Quoted 1 line
arch/arm/mach-zynq/common.c +1
-0+iotable_init (&zynq_cortex_a9_scu_map,1 );Then creating a static mapping...
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 89
Quoted 1 line
arch/arm/mach-zynq/common.c +1
-0+ scu_base =ioremap (base, zynq_cortex_a9_scu_map.length);And also a dynamic mapping?
Yes - exactly.
I was talking to Olof about this code at ELC and he mentioned that someone else might know better way how to do it.
It is quite a long time I played with this code.
I found this solution in vexpress platform (mach-vexpress/platsmp.c)
IRC: Static mapping is necessary to be able to access device so early. On the other hand you can't use this static mapping when kernel runs for that you need to use dynamic allocation.
Calling ioremap so early caused that the return address is the same and you can just use it later. and it is also in the correct vmalloc area.
We are using scu_base for power management code.
Let me check if dynamic mapping is also required.
Thanks,
Michal
Signature · 5 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel - Microblaze cpu - http://www.monstr.eu/fdt/ Maintainer of Linux kernel - Xilinx Zynq ARM architecture Microblaze U-BOOT custodian and responsible for u-boot arm zynq platform
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 5 lines · Michal Simek
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 2 lines · Michal Simek
Quoted 1 line
Use Cortex a9 cp15 to read scu baseaddress.
[...]
Quoted 12 lines
Quoted 12 lines
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 87
Quoted 7 lines
arch/arm/mach-zynq/common.c +7
-0+static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 1 line
+
iotable_init(&zynq_cortex_a9_scu_map, 1);
Quoted 7 lines
Quoted 5 lines
Then creating a static mapping...
Quoted 1 line
+
scu_base=ioremap(base, zynq_cortex_a9_scu_map.length);And also a dynamic mapping?
Yes - exactly.
You are simply getting lucky that it works. If the physical address did not happen to be in the vmalloc address region, it would not work. You should not do this because you have an implicit requirement and the code will look broken to anyone that reads it.
Quoted 7 lines
I was talking to Olof about this code at ELC and he mentioned that someone else might know better way how to do it.
It is quite a long time I played with this code.
I found this solution in vexpress platform (mach-vexpress/platsmp.c)IRC: Static mapping is necessary to be able to access device so early.
Only to read the number of cores. That's not really worth an early mapping. I would move to using DT to count the number of cores. Support for that is already in place.
Quoted 7 lines
On the other hand you can't use this static mapping when kernel runs for that you need to use dynamic allocation.
Calling ioremap so early caused that the return address is the same and you can just use it later. and it is also in the correct vmalloc area.We are using
scu_basefor power management code.
Let me check if dynamic mapping is also required.
The should be no reason you need both.
Rob
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
2013/3/25 Rob Herring <[email protected]>:
Quoted 25 lines · Michal Simek
Quoted 5 lines
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 2 lines · Michal Simek
Quoted 1 line
Use Cortex a9 cp15 to read scu baseaddress.
[...]
Quoted 16 lines
Quoted 16 lines
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 87
Quoted 7 lines
arch/arm/mach-zynq/common.c +7
-0+static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 1 line
+
iotable_init(&zynq_cortex_a9_scu_map, 1);Then creating a static mapping...
Quoted 1 line
+
scu_base=ioremap(base, zynq_cortex_a9_scu_map.length);
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
And also a dynamic mapping?
Yes - exactly.
You are simply getting lucky that it works. If the physical address did not happen to be in the vmalloc address region, it would not work. You should not do this because you have an implicit requirement and the code will look broken to anyone that reads it.
yeah correct. I will add there a comment to mentioned that.
Quoted 11 lines
Quoted 7 lines
I was talking to Olof about this code at ELC and he mentioned that someone else might know better way how to do it.
It is quite a long time I played with this code.
I found this solution in vexpress platform (mach-vexpress/platsmp.c)IRC: Static mapping is necessary to be able to access device so early.
Only to read the number of cores. That's not really worth an early mapping. I would move to using DT to count the number of cores. Support for that is already in place.
What's the functions for that?
Quoted 1 line
From my point of view is better to read it directly on SoC and not to read
dts. It should be also faster.
Quoted 9 lines
Quoted 7 lines
On the other hand you can't use this static mapping when kernel runs for that you need to use dynamic allocation.
Calling ioremap so early caused that the return address is the same and you can just use it later. and it is also in the correct vmalloc area.We are using
scu_basefor power management code.
Let me check if dynamic mapping is also required.The should be no reason you need both.
I have done some tests and that ioremap could be probably removed.
Pawel: You are the author of this code in vexpress. Is there any particular reason
to have there that ioremap?
Thanks,
Michal
Signature · 5 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel - Microblaze cpu - http://www.monstr.eu/fdt/ Maintainer of Linux kernel - Xilinx Zynq ARM architecture Microblaze U-BOOT custodian and responsible for u-boot arm zynq platform
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 28 lines · Michal Simek
2013/3/25 Rob Herring <[email protected]>:
Quoted 27 lines · Michal Simek
Quoted 5 lines
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 2 lines · Michal Simek
Quoted 1 line
Use Cortex a9 cp15 to read scu baseaddress.
[...]
Quoted 18 lines
Quoted 17 lines
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 87
Quoted 7 lines
arch/arm/mach-zynq/common.c +7
-0+static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 1 line
+
iotable_init(&zynq_cortex_a9_scu_map, 1);Then creating a static mapping...
Quoted 1 line
+
scu_base=ioremap(base, zynq_cortex_a9_scu_map.length);And also a dynamic mapping?
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Yes - exactly.
You are simply getting lucky that it works. If the physical address did not happen to be in the vmalloc address region, it would not work. You should not do this because you have an implicit requirement and the code will look broken to anyone that reads it.
yeah correct. I will add there a comment to mentioned that.
I think leaving the define would be better, but either way is fine.
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
I was talking to Olof about this code at ELC and he mentioned that someone else might know better way how to do it.
It is quite a long time I played with this code.
I found this solution in vexpress platform (mach-vexpress/platsmp.c)IRC: Static mapping is necessary to be able to access device so early.
Quoted 7 lines
Quoted 3 lines
Only to read the number of cores. That's not really worth an early mapping. I would move to using DT to count the number of cores. Support for that is already in place.
What's the functions for that?
From my point of view is better to read it directly on SoC and not to read dts. It should be also faster.
Actually, I think you just remove all the code related to set_cpu_possible().
The relevant function is arm_dt_init_cpu_maps.
Rob
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 31 lines · Rob Herring
Quoted 30 lines · Michal Simek
2013/3/25 Rob Herring <[email protected]>:
Quoted 29 lines · Michal Simek
Quoted 5 lines
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 2 lines · Michal Simek
Quoted 1 line
Use Cortex a9 cp15 to read scu baseaddress.
[...]
Quoted 19 lines
Quoted 10 lines
arch/arm/mach-zynq/common.c static void __init xilinx_zynq_timer_init(void) · line 87
Quoted 7 lines
arch/arm/mach-zynq/common.c +7
-0+static void __init scu_init (void ) +{ +unsigned long base; + + base =scu_a9_get_base (); + zynq_cortex_a9_scu_map.pfn =__phys_to_pfn (base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 7 lines
Quoted 1 line
+
iotable_init(&zynq_cortex_a9_scu_map, 1);Then creating a static mapping...
Quoted 1 line
+
scu_base=ioremap(base, zynq_cortex_a9_scu_map.length);And also a dynamic mapping?
Yes - exactly.
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
You are simply getting lucky that it works. If the physical address did not happen to be in the vmalloc address region, it would not work. You should not do this because you have an implicit requirement and the code will look broken to anyone that reads it.
yeah correct. I will add there a comment to mentioned that.
I think leaving the define would be better, but either way is fine.
ok. From my point of view is better to keep scu where it is and any early vmalloc allocator will be really helpful. Because there is also early map for early console where virtual addresses are hardcoded.
It will be just easier to ask generic code to return base if if you request for size.
static unsigned long iomapstart = VMALLOC_END;
unsigned long get_iomap_virtbase(unsigned long size)
{
iomapstart -= size;
return iomapstart;
}Something like this will be helpful and then instead of setup virtual addresses for every static io mapping just to call this function and you will get mapping correct all the time.
Quoted 12 lines
Quoted 12 lines
Quoted 11 lines
Quoted 7 lines
I was talking to Olof about this code at ELC and he mentioned that someone else might know better way how to do it.
It is quite a long time I played with this code.
I found this solution in vexpress platform (mach-vexpress/platsmp.c)IRC: Static mapping is necessary to be able to access device so early.
Only to read the number of cores. That's not really worth an early mapping. I would move to using DT to count the number of cores. Support for that is already in place.
Quoted 8 lines
Quoted 3 lines
What's the functions for that?
From my point of view is better to read it directly on SoC and not to read
dts. It should be also faster.Actually, I think you just remove all the code related to
set_cpu_possible().The relevant function is
arm_dt_init_cpu_maps.
Ok. Thanks.
I will look at it but currently I prefer to cound number of cores via scu because we are using too many dtses.
Thanks,
Michal
Signature · 4 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/ Microblaze U-BOOT custodian
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 49 lines · Michal Simek
Quoted 37 lines · Rob Herring
Quoted 36 lines · Michal Simek
2013/3/25 Rob Herring <[email protected]>:
Quoted 34 lines · Michal Simek
Quoted 5 lines
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 2 lines · Michal Simek
Quoted 1 line
Use Cortex a9 cp15 to read scu baseaddress.
[...]
Quoted 12 lines
Quoted 12 lines
Review
Quoted 7 lines
(quoted diff) +7
-0+static void __init scu_init(void) +{ + unsigned long base; + + base = scu_a9_get_base(); + zynq_cortex_a9_scu_map.pfn = __phys_to_pfn(base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 1 line
+
iotable_init(&zynq_cortex_a9_scu_map, 1);Quoted 7 lines
Quoted 5 lines
Then creating a static mapping...
Quoted 1 line
+
scu_base=ioremap(base, zynq_cortex_a9_scu_map.length);And also a dynamic mapping?
Yes - exactly.
You are simply getting lucky that it works. If the physical address did not happen to be in the vmalloc address region, it would not work. You should not do this because you have an implicit requirement and the code
will look broken to anyone that reads it.Quoted 3 lines
Quoted 1 line
yeah correct. I will add there a comment to mentioned that.
I think leaving the define would be better, but either way is fine.
ok. From my point of view is better to keep scu where it is and any early vmalloc
allocator will be really helpful. Because there is also early map for early console where virtual addresses are hardcoded.
It will be just easier to ask generic code to return base if if you request for size.
Quoted 7 lines
static unsigned long iomapstart =
VMALLOC_END;unsigned long
get_iomap_virtbase(unsigned long size){ iomapstart -= size; return iomapstart; }
It wouldn't be so simple with page/section alignment requirements and such.
Quoted 3 lines
Something like this will be helpful and then instead of setup virtual addresses for every static io mapping just to call this function and you will get mapping correct all the time.
If you use DT for core count, then you don't need the static mapping and can use ioremap (or of_iomap).
Rob
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
2013/3/26 Rob Herring <[email protected]>:
Quoted 52 lines · Michal Simek
Quoted 51 lines · Rob Herring
Quoted 37 lines · Michal Simek
Quoted 36 lines
2013/3/25 Rob Herring <[email protected]>:
Quoted 34 lines · Michal Simek
Quoted 5 lines
Hi Rob,
2013/3/25 Rob Herring <[email protected]>:
Quoted 2 lines · Michal Simek
Quoted 1 line
Use Cortex a9 cp15 to read scu baseaddress.
[...]
Quoted 12 lines
Quoted 12 lines
Review
Quoted 7 lines
(quoted diff) +7
-0+static void __init scu_init(void) +{ + unsigned long base; + + base = scu_a9_get_base(); + zynq_cortex_a9_scu_map.pfn = __phys_to_pfn(base); + zynq_cortex_a9_scu_map.virtual = base;You are setting the virtual address to the physical base?
Quoted 1 line
+
iotable_init(&zynq_cortex_a9_scu_map, 1);Quoted 7 lines
Quoted 5 lines
Then creating a static mapping...
Quoted 1 line
+
scu_base=ioremap(base, zynq_cortex_a9_scu_map.length);And also a dynamic mapping?
Yes - exactly.
You are simply getting lucky that it works. If the physical address did not happen to be in the vmalloc address region, it would not work. You should not do this because you have an implicit requirement and the code
will look broken to anyone that reads it.Quoted 3 lines
Quoted 1 line
yeah correct. I will add there a comment to mentioned that.
I think leaving the define would be better, but either way is fine.
ok. From my point of view is better to keep scu where it is and any early vmalloc
allocator will be really helpful. Because there is also early map for early console where virtual addresses are hardcoded.
It will be just easier to ask generic code to return base if if you request for size.static unsigned long iomapstart =
VMALLOC_END;
Quoted 7 lines
Quoted 5 lines
unsigned long
get_iomap_virtbase(unsigned long size){ iomapstart -= size; return iomapstart; }It wouldn't be so simple with page/section alignment requirements and such.
Adding alignment to it is not so hard.
Quoted 6 lines
Quoted 3 lines
Something like this will be helpful and then instead of setup virtual addresses for every static io mapping just to call this function and you will get mapping correct all the time.
If you use DT for core count, then you don't need the static mapping and can use ioremap (or
of_iomap).
yep but on the other hand I have to be sure that all users are using proper DTS files. And based on my experience with users I would rather relate on scu count instead of DT count.
Thanks,
Michal
Signature · 5 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel - Microblaze cpu - http://www.monstr.eu/fdt/ Maintainer of Linux kernel - Xilinx Zynq ARM architecture Microblaze U-BOOT custodian and responsible for u-boot arm zynq platform
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
In reply to Michal SimekGreetings,
Apologies about "slightly" delayed response.
Quoted 3 lines · Michal Simek
Pawel: You are the author of this code in vexpress. Is there any particular reason
to have there that ioremap?
At the time there was no standard DT->platsmp solution and the traditional way was using the SCU to work out the CPUs.
If I was doing this today (and I'll re-do this code at some point) I'd rely on the DT in the first place and maybe then have a SCU "micro-driver" double-checking the DT data against the SCU information. That way I could reduce the dependency on static mappings. And this is what I want to get - if possible I don't want any static mapping at all.
Pawe?
A message that is not in the archive (sent off-list or not yet received): CAHTX3dKD4G0E8qoxTR2HnJVdagoeOerM+TiZzkJUPjcGwYdX_Q@mail.gmail.com
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 23 lines · Michal Simek
Hi
2013/4/2 Pawel Moll <[email protected]>
Quoted 9 lines
Greetings,
Apologies about "slightly" delayed response.
Quoted 3 lines · Michal Simek
Pawel: You are the author of this code in vexpress. Is there any particular reason
to have there that ioremap?Quoted 9 lines
At the time there was no standard
DT->platsmpsolution and the traditional way was using the SCU to work out the CPUs.If I was doing this today (and I'll re-do this code at some point) I'd rely on the DT in the first place and maybe then have a SCU "micro-driver" double-checking the DT data against the SCU information. That way I could reduce the dependency on static mappings. And this is what I want to get - if possible I don't want any static mapping at all.
Quoted 3 lines
Sure. Currently I want to use that static mapping because I know that user DTSes are just wrong and relating on HW is better for me right now.
Wouldn't we risk getting more faulty DTSes instead of stopping them? If you stop the support starting with kernel 3.xx everyone has to fix their *.dts and be done with it.
Regards,
Steffen
Signature · 4 lines
Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
In reply to a message not in the archiveQuoted 6 lines · Michal Simek
The point is different. Was there any reason why you used this static mapping
and then dynamic?
This code:iotable_init(&vexpress_dt_cortex_a9_scu_map, 1); vexpress_dt_cortex_a9_scu_base = ioremap(phys_addr, SZ_256);
The SCU code is being used very early - too early for "normal" ioremap() (one problem I'm aware of is lack of kmalloc, but there may be more). So the static mapping is required. Now, traditionally there was some (more or less cryptic) pointer arithmetic was used (eg. "base = map_virt_addr
+ 0x100"). I'm sure some would argue it's fine, but I personally dislike it. Then Nico Pitre's patches made the ioremap() aware of the static mappings so it will reuse them when possible. Of course it would be better to be able to do normal ioremap() without any tricks, but it's another story and I don't feel like discussing it.
So, to summarize, the static mapping is there to make the ioremap() possible in the first place. If your ioremap() works without it (which means that you're doing it late enough) you can ignore the iotable_init() completely.
Does this answer the question?
Pawe?
A message that is not in the archive (sent off-list or not yet received): CAHTX3dJMpp+E2u-cAeYbqtxC1WAYWpCeRx6W7G=dWDcgzUz5DA@mail.gmail.com
PATCH 4/10 arm: zynq: Load scu baseaddress at run time
Quoted 3 lines · Michal Simek
Then my question is if you can just remove that ioremap, as I have done for zynq,
without any related problem.
Well, the vexpress platsmp functions are going away soon, replaced by the generic DT code. When this happen I won't even try to access the SCU, so both its static mapping and the ioremap will go as well :-)
Pawe?
PATCH 5/10 arm: zynq: Move slcr initialization to separate file
In reply to Michal SimekCreate separate slcr driver instead of pollute common code.
--- arch/arm/mach-zynq/Makefile | 2 +- arch/arm/mach-zynq/common.c | 10 +------ arch/arm/mach-zynq/common.h | 3 ++ arch/arm/mach-zynq/slcr.c | 69 +++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 74 insertions(+), 10 deletions(-) create mode 100644 arch/arm/mach-zynq/slcr.c
arch/arm/mach-zynq/Makefile +1 -1
@@ -3,4 +3,4 @@ # # Common support -obj-y := common.o +obj-y := common.o slcr.o
arch/arm/mach-zynq/common.c +1 -9
@@ -61,15 +61,7 @@ static void __init xilinx_init_machine(void)static void __init xilinx_zynq_timer_init (void ) { -struct device_node *np; -void __iomem *slcr; - - np =of_find_compatible_node (NULL ,NULL ,"xlnx,zynq-slcr" ); - slcr =of_iomap (np,0 ); -WARN_ON (!slcr); - -xilinx_zynq_clocks_init (slcr); - +slcr_init ();clocksource_of_init (); }
arch/arm/mach-zynq/common.h +3 -0
@@ -17,6 +17,9 @@#ifndef __MACH_ZYNQ_COMMON_H__#define __MACH_ZYNQ_COMMON_H__ +extern int slcr_init (void ); + +extern void __iomem *zynq_slcr_base;extern void __iomem *scu_base;#endif
arch/arm/mach-zynq/slcr.c new file +69 -0
69 lines not on this page · raw message
Signature · 1 line
1.7.9.7
PATCH 5/10 arm: zynq: Move slcr initialization to separate file
Quoted 85 lines · Michal Simek
Create separate slcr driver instead of pollute common code.
Signed-off-by Michal Simek <[email protected]>--- arch/arm/mach-zynq/Makefile | 2 +- arch/arm/mach-zynq/common.c | 10 +------ arch/arm/mach-zynq/common.h | 3 ++ arch/arm/mach-zynq/slcr.c | 69 +++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 74 insertions(+), 10 deletions(-) create mode 100644 arch/arm/mach-zynq/slcr.carch/arm/mach-zynq/Makefile +1
-1@@ -3,4 +3,4 @@ # # Common support -obj-y := common.o +obj-y := common.o slcr.oarch/arm/mach-zynq/common.c +1
-9@@ -61,15 +61,7 @@ static void __init xilinx_init_machine(void)static void __init xilinx_zynq_timer_init (void ) { -struct device_node *np; -void __iomem *slcr; - - np =of_find_compatible_node (NULL ,NULL ,"xlnx,zynq-slcr" ); - slcr =of_iomap (np,0 ); -WARN_ON (!slcr); - -xilinx_zynq_clocks_init (slcr); - +slcr_init ();clocksource_of_init (); }arch/arm/mach-zynq/common.h +3
-0@@ -17,6 +17,9 @@#ifndef __MACH_ZYNQ_COMMON_H__#define __MACH_ZYNQ_COMMON_H__ +extern int slcr_init (void ); + +extern void __iomem *zynq_slcr_base;extern void __iomem *scu_base;#endif arch/arm/mach-zynq/slcr.c new file +23
-0@@ -0,0 +1,69 @@ +/* +* Xilinx SLCR driver +* +* Copyright (c) 2011-2013 Xilinx Inc. +* +* This program is free software; you can redistribute it and/or +* modify it under the terms of the GNU General Public License +* as published by the Free Software Foundation; either version +* 2 of the License, or (at your option) any later version. +* +* You should have received a copy of the GNU General Public +* License along with this program; if not, write to the Free +* Software Foundation, Inc., 675 Mass Ave, Cambridge, MA +* 02139, USA. +*/ + +#include <linux/export.h> +#include <linux/io.h> +#include <linux/fs.h> +#include <linux/interrupt.h> +#include <linux/init.h> +#include <linux/kernel.h> +#include <linux/module.h>
arch/arm/mach-zynq/slcr.c line 33
Quoted 10 lines
arch/arm/mach-zynq/slcr.c new file +10
-0+#include <linux/of_address.h> +#include <linux/uaccess.h> +#include <linux/platform_device.h> +#include <linux/slab.h> +#include <linux/string.h> +#include <linux/clk/zynq.h> +#include "common.h" + +#define SLCR_UNLOCK_MAGIC 0xDF0D +#define SLCR_UNLOCK 0x8/* SCLR unlock register */
--------------------------------------------------^ small typo
Quoted 26 lines
arch/arm/mach-zynq/slcr.c new file +26
-0+ +void __iomem *zynq_slcr_base; + +/** +* xslcr_init() +* Returns 0 on success, negative errno otherwise. +* +* Called early during boot from platform code to remap SLCR area. +*/ +int __init slcr_init (void ) +{ +struct device_node *np; + + np =of_find_compatible_node (NULL ,NULL ,"xlnx,zynq-slcr" ); +if (!np) { +pr_err ("%s: no slcr node found\n" , __func__); +BUG (); + } + + zynq_slcr_base =of_iomap (np,0 ); +if (!zynq_slcr_base) { +pr_err ("%s: Unable to map I/O memory\n" , __func__); +BUG (); + } + +/* unlock the SLCR so that registers can be changed */
arch/arm/mach-zynq/slcr.c line 69
Quoted 10 lines
arch/arm/mach-zynq/slcr.c new file +10
-0+writel (SLCR_UNLOCK_MAGIC , zynq_slcr_base +SLCR_UNLOCK ); + +pr_info ("%s mapped to %p\n" , np->name, zynq_slcr_base); + +xilinx_zynq_clocks_init (zynq_slcr_base); + +of_node_put (np); + +return 0 ; +}
The slcr should definitely get it's own driver. I like that.
But wouldn't it be better to call this in the map_io call and then, for the time being, assume that the slcr is unlocked and put the clocks_init-call in zynq_timer_init?
Regards,
Steffen
Signature · 4 lines
Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
PATCH 5/10 arm: zynq: Move slcr initialization to separate file
Quoted 130 lines · Steffen Trumtrar
Quoted 85 lines · Michal Simek
Create separate slcr driver instead of pollute common code.
Signed-off-by Michal Simek <[email protected]>--- arch/arm/mach-zynq/Makefile | 2 +- arch/arm/mach-zynq/common.c | 10 +------ arch/arm/mach-zynq/common.h | 3 ++ arch/arm/mach-zynq/slcr.c | 69 +++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 74 insertions(+), 10 deletions(-)create mode 100644
arch/arm/mach-zynq/slcr.carch/arm/mach-zynq/Makefile +1
-1@@ -3,4 +3,4 @@ # # Common support -obj-y := common.o +obj-y := common.o slcr.oarch/arm/mach-zynq/common.c +1
-9@@ -61,15 +61,7 @@ static void __init xilinx_init_machine(void)static void __init xilinx_zynq_timer_init (void ) { -struct device_node *np; -void __iomem *slcr; - - np =of_find_compatible_node (NULL ,NULL ,"xlnx,zynq-slcr" ); - slcr =of_iomap (np,0 ); -WARN_ON (!slcr); - -xilinx_zynq_clocks_init (slcr); - +slcr_init ();clocksource_of_init (); }arch/arm/mach-zynq/common.h +3
-0@@ -17,6 +17,9 @@#ifndef __MACH_ZYNQ_COMMON_H__#define __MACH_ZYNQ_COMMON_H__ +extern int slcr_init (void ); + +extern void __iomem *zynq_slcr_base;extern void __iomem *scu_base;#endif arch/arm/mach-zynq/slcr.c new file +23
-0@@ -0,0 +1,69 @@ +/* +* Xilinx SLCR driver +* +* Copyright (c) 2011-2013 Xilinx Inc. +* +* This program is free software; you can redistribute it and/or +* modify it under the terms of the GNU General Public License +* as published by the Free Software Foundation; either version +* 2 of the License, or (at your option) any later version. +* +* You should have received a copy of the GNU General Public +* License along with this program; if not, write to the Free +* Software Foundation, Inc., 675 Mass Ave, Cambridge, MA +* 02139, USA. +*/ + +#include <linux/export.h> +#include <linux/io.h> +#include <linux/fs.h> +#include <linux/interrupt.h> +#include <linux/init.h> +#include <linux/kernel.h> +#include <linux/module.h> arch/arm/mach-zynq/slcr.c line 33
Quoted 10 lines
arch/arm/mach-zynq/slcr.c new file +10
-0+#include <linux/of_address.h> +#include <linux/uaccess.h> +#include <linux/platform_device.h> +#include <linux/slab.h> +#include <linux/string.h> +#include <linux/clk/zynq.h> +#include "common.h" + +#define SLCR_UNLOCK_MAGIC 0xDF0D +#define SLCR_UNLOCK 0x8/* SCLR unlock register */ --------------------------------------------------^ small typo
Quoted 33 lines
arch/arm/mach-zynq/slcr.c new file +33
-0+ +void __iomem *zynq_slcr_base; + +/** +* xslcr_init() +* Returns 0 on success, negative errno otherwise. +* +* Called early during boot from platform code to remap SLCR area. +*/ +int __init slcr_init (void ) +{ +struct device_node *np; + + np =of_find_compatible_node (NULL ,NULL ,"xlnx,zynq-slcr" ); +if (!np) { +pr_err ("%s: no slcr node found\n" , __func__); +BUG (); + } + + zynq_slcr_base =of_iomap (np,0 ); +if (!zynq_slcr_base) { +pr_err ("%s: Unable to map I/O memory\n" , __func__); +BUG (); + } + +/* unlock the SLCR so that registers can be changed */ +writel (SLCR_UNLOCK_MAGIC , zynq_slcr_base +SLCR_UNLOCK ); + +pr_info ("%s mapped to %p\n" , np->name, zynq_slcr_base); + +xilinx_zynq_clocks_init (zynq_slcr_base); + +of_node_put (np);
Quoted 8 lines
arch/arm/mach-zynq/slcr.c line 69
Quoted 3 lines
arch/arm/mach-zynq/slcr.c new file +3
-0+ +return 0 ; +}The slcr should definitely get it's own driver. I like that.
But wouldn't it be better to call this in themap_iocall and then, for the time being, assume that the slcr is unlocked and put the clocks_init-call inzynq_timer_init?
map_io can work with static mapping. Dynamic mapping only doesn't work in that time. That's why it must be done later.
The reason why I call xilinx_zynq_clocks_init() from this code is
that slcr IP contains almost everything and clocks are logical subnode of this device.
Thanks,
Michal
Signature · 4 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/ Microblaze U-BOOT custodian
PATCH 6/10 arm: zynq: Add support for system reset
In reply to Michal SimekDo system reset via slcr registers.
--- arch/arm/mach-zynq/common.c | 6 ++++++ arch/arm/mach-zynq/common.h | 1 + arch/arm/mach-zynq/slcr.c | 26 ++++++++++++++++++++++++++ 3 files changed, 33 insertions(+)
arch/arm/mach-zynq/common.c +6 -0
@@ -91,6 +91,11 @@ static void __init xilinx_map_io(void)scu_init (); } +static void xilinx_system_reset (char mode,const char *cmd) +{ +slcr_system_reset (); +} +static const char *xilinx_dt_match[] = {"xlnx,zynq-zc702" ,"xlnx,zynq-7000" ,
@@ -103,4 +108,5 @@ MACHINE_START(XILINX_EP107, "Xilinx Zynq Platform") .init_machine = xilinx_init_machine, .init_time = xilinx_zynq_timer_init, .dt_compat = xilinx_dt_match, + .restart = xilinx_system_reset,MACHINE_END
arch/arm/mach-zynq/common.h +1 -0
@@ -18,6 +18,7 @@#define __MACH_ZYNQ_COMMON_H__extern int slcr_init (void ); +extern void slcr_system_reset (void );extern void __iomem *zynq_slcr_base;extern void __iomem *scu_base;
arch/arm/mach-zynq/slcr.c +26 -0
@@ -32,9 +32,35 @@#define SLCR_UNLOCK_MAGIC 0xDF0D#define SLCR_UNLOCK 0x8/* SCLR unlock register */ +#define SLCR_PSS_RST_CTRL_OFFSET 0x200/* PS Software Reset Control */ +#define SLCR_REBOOT_STATUS 0x258/* PS Reboot Status */ +void __iomem *zynq_slcr_base;/** +* xslcr_system_reset - Reset the entire system. +* +*/ +void slcr_system_reset (void ) +{ +u32 reboot; + +/* Unlock the SLCR then reset the system. +* Note that this seems to require raw i/o +* functions or there's a lockup? +*/ +__raw_writel (SLCR_UNLOCK_MAGIC , zynq_slcr_base +SLCR_UNLOCK ); + +/* Clear 0x0F000000 bits of reboot status register to workaround +* the FSBL not loading the bitstream after soft-reboot +* This is a temporary solution until we know more. +*/ + reboot =readl (zynq_slcr_base +SLCR_REBOOT_STATUS ); +writel (reboot &0xF0FFFFFF , zynq_slcr_base +SLCR_REBOOT_STATUS ); +writel (1 , zynq_slcr_base +SLCR_PSS_RST_CTRL_OFFSET ); +} + +/** * xslcr_init() * Returns 0 on success, negative errno otherwise. *
Signature · 1 line
1.7.9.7
PATCH 7/10 arm: zynq: Add support for pmu
In reply to Michal SimekZynq is standard PMU with 2 interrupt per core. There is also access via register which is not used right now.
--- arch/arm/boot/dts/zynq-7000.dtsi | 7 +++++++ 1 file changed, 7 insertions(+)
arch/arm/boot/dts/zynq-7000.dtsi +7 -0
@@ -15,6 +15,13 @@ / { compatible = "xlnx,zynq-7000"; + pmu { + compatible = "arm,cortex-a9-pmu"; + interrupts = <0 5 4>, <0 6 4>; + interrupt-parent = <&intc>; + reg = < 0xf8891000 0x1000 0xf8893000 0x1000 >; + }; + amba { compatible = "simple-bus"; #address-cells = <1>;
Signature · 1 line
1.7.9.7
PATCH 8/10 arm: zynq: Add smp support
In reply to Michal SimekZynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
to secondary_startup code.
--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/common.c | 1 + arch/arm/mach-zynq/common.h | 7 ++ arch/arm/mach-zynq/platsmp.c | 160 ++++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-zynq/slcr.c | 29 ++++++++ 5 files changed, 198 insertions(+) create mode 100644 arch/arm/mach-zynq/platsmp.c
arch/arm/mach-zynq/Makefile +1 -0
@@ -4,3 +4,4 @@ # Common support obj-y := common.o slcr.o +obj-$(CONFIG_SMP) += platsmp.o
arch/arm/mach-zynq/common.c +1 -0
@@ -103,6 +103,7 @@ static const char *xilinx_dt_match[] = { };MACHINE_START (XILINX_EP107 ,"Xilinx Zynq Platform" ) + .smp = smp_ops(zynq_smp_ops), .map_io = xilinx_map_io, .init_irq = irqchip_init, .init_machine = xilinx_init_machine,
arch/arm/mach-zynq/common.h +7 -0
@@ -19,6 +19,13 @@extern int slcr_init (void );extern void slcr_system_reset (void ); +extern void slcr_cpu_stop (int cpu); +extern void slcr_cpu_start (int cpu); + +#ifdef CONFIG_SMP +extern void secondary_startup (void ); +#endif +extern struct smp_operations zynq_smp_ops__initdata ;extern void __iomem *zynq_slcr_base;extern void __iomem *scu_base;
arch/arm/mach-zynq/platsmp.c new file +160 -0
160 lines not on this page · raw message
arch/arm/mach-zynq/slcr.c +29 -0
@@ -33,6 +33,11 @@#define SLCR_UNLOCK 0x8/* SCLR unlock register */ #define SLCR_PSS_RST_CTRL_OFFSET 0x200/* PS Software Reset Control */ + +#define SLCR_A9_CPU_CLKSTOP 0x10 +#define SLCR_A9_CPU_RST 0x1 + +#define SLCR_A9_CPU_RST_CTRL 0x244/* CPU Software Reset Control */ #define SLCR_REBOOT_STATUS 0x258/* PS Reboot Status */ void __iomem *zynq_slcr_base;
@@ -61,6 +66,30 @@ void slcr_system_reset(void) }/** +* slcr_cpu_start - Start cpu +* @cpu: cpu number +*/ +void slcr_cpu_start (int cpu) +{ +/* enable CPUn */ +writel (SLCR_A9_CPU_CLKSTOP << cpu, + zynq_slcr_base +SLCR_A9_CPU_RST_CTRL ); +/* enable CLK for CPUn */ +writel (0x0 << cpu, zynq_slcr_base +SLCR_A9_CPU_RST_CTRL ); +} + +/** +* slcr_cpu_stop - Stop cpu +* @cpu: cpu number +*/ +void slcr_cpu_stop (int cpu) +{ +/* stop CLK and reset CPUn */ +writel ((SLCR_A9_CPU_CLKSTOP |SLCR_A9_CPU_RST ) << cpu, + zynq_slcr_base +SLCR_A9_CPU_RST_CTRL ); +} + +/** * xslcr_init() * Returns 0 on success, negative errno otherwise. *
Signature · 1 line
1.7.9.7
PATCH 8/10 arm: zynq: Add smp support
Quoted 13 lines · Michal Simek
Zynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
tosecondary_startupcode.Signed-off-by Michal Simek <[email protected]>--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/common.c | 1 + arch/arm/mach-zynq/common.h | 7 ++ arch/arm/mach-zynq/platsmp.c | 160 ++++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-zynq/slcr.c | 29 ++++++++ 5 files changed, 198 insertions(+) create mode 100644 arch/arm/mach-zynq/platsmp.c
[...]
Quoted 9 lines
arch/arm/mach-zynq/platsmp.c new file +9
-0+/* Secondary CPU kernel startup is a 2 step process. The primary CPU +* starts the secondary CPU by giving it the address of the kernel and +* then sending it an event to wake it up. The secondary CPU then +* starts the kernel and tells the primary CPU it's up and running. +*/ +static void __cpuinit zynq_secondary_init (unsigned int cpu) +{ +/* +* if any interrupts are already enabled for the primary
arch/arm/mach-zynq/platsmp.c line 57
Quoted 10 lines
arch/arm/mach-zynq/platsmp.c new file +10
-0+* core (e.g. timer irq), then they will not have been enabled +* for us: do so +*/ +gic_secondary_init (0 ); + +/* +* Synchronise with the boot thread. +*/ +spin_lock (&boot_lock); +spin_unlock (&boot_lock);
Why do you think this is needed? Platforms that need this synchronization are only the ones that just do wfi for hotplug rather than properly reseting the core. You appear to do the latter and should not need this.
arch/arm/mach-zynq/platsmp.c line 69
Quoted 12 lines
arch/arm/mach-zynq/platsmp.c new file +12
-0+} + +int __cpuinit zynq_cpun_start (u32 address,int cpu) +{ +if (cpu > ncores) { +pr_warn ("CPU No. is not available in the system\n" ); +return -1 ; + } + +/* MS: Expectation that SLCR are directly map and accessible */ +/* Not possible to jump to non aligned address */ +if (!(address &3 ) && (!address || (address >=0xC ))) {
What about Thumb2 kernel entry?
arch/arm/mach-zynq/platsmp.c line 70
Quoted 1 line
arch/arm/mach-zynq/platsmp.c new file +1
-0+slcr_cpu_stop (cpu);
Isn't a secondary cpu already stopped?
Quoted 7 lines
arch/arm/mach-zynq/platsmp.c new file +7
-0+ +/* +* This is elegant way how to jump to any address +* 0x0: Load address at 0x8 to r0 +* 0x4: Jump by mov instruction +* 0x8: Jumping address +*/
arch/arm/mach-zynq/platsmp.c line 87
Quoted 10 lines
arch/arm/mach-zynq/platsmp.c new file +10
-0+if (address) { +/* 0: ldr r0, [8] */ +__raw_writel (0xe59f0000 ,phys_to_virt (0x0 )); +/* 4: mov pc, r0 */ +__raw_writel (0xe1a0f000 ,phys_to_virt (0x4 )); +__raw_writel (address,phys_to_virt (0x8 )); + } + +flush_cache_all (); +outer_flush_all ();
You should only need to flush range on address 0-4 here.
Quoted 121 lines
arch/arm/mach-zynq/platsmp.c new file +73
-0+wmb (); + +slcr_cpu_start (cpu); + +return 0 ; + } + +pr_warn ("Can't start CPU%d: Wrong starting address %x\n" , cpu, address); + +return -1 ; +} +EXPORT_SYMBOL (zynq_cpun_start); + +static int __cpuinit zynq_boot_secondary (unsigned int cpu, +struct task_struct *idle) +{ +int ret; + +/* +* set synchronisation state between this boot processor +* and the secondary one +*/ +spin_lock (&boot_lock); + + ret =zynq_cpun_start (virt_to_phys (secondary_startup), cpu); +if (ret) { +spin_unlock (&boot_lock); +return -1 ; + } + +/* +* now the secondary core is starting up let it run its +* calibrations, then wait for it to finish +*/ +spin_unlock (&boot_lock); + +return 0 ; +} + +/* +* Initialise the CPU possible map early - this describes the CPUs +* which may be present or become present in the system. +*/ +static void __init zynq_smp_init_cpus (void ) +{ +int i; + + ncores =scu_get_core_count (scu_base); + +for (i =0 ; i < ncores && i <CONFIG_NR_CPUS ; i++) +set_cpu_possible (i, true); +} + +static void __init zynq_smp_prepare_cpus (unsigned int max_cpus) +{ +int i; + +/* +* Initialise the present map, which describes the set of CPUs +* actually populated at the present time. +*/ +for (i =0 ; i < max_cpus; i++) +set_cpu_present (i, true); + +scu_enable (scu_base); +} + +struct smp_operations zynq_smp_ops__initdata = { + .smp_init_cpus = zynq_smp_init_cpus, + .smp_prepare_cpus = zynq_smp_prepare_cpus, + .smp_secondary_init = zynq_secondary_init, + .smp_boot_secondary = zynq_boot_secondary, +};arch/arm/mach-zynq/slcr.c +29
-0@@ -33,6 +33,11 @@#define SLCR_UNLOCK 0x8/* SCLR unlock register */ #define SLCR_PSS_RST_CTRL_OFFSET 0x200/* PS Software Reset Control */ + +#define SLCR_A9_CPU_CLKSTOP 0x10 +#define SLCR_A9_CPU_RST 0x1 + +#define SLCR_A9_CPU_RST_CTRL 0x244/* CPU Software Reset Control */ #define SLCR_REBOOT_STATUS 0x258/* PS Reboot Status */ void __iomem *zynq_slcr_base;@@ -61,6 +66,30 @@ void slcr_system_reset(void) }/** +* slcr_cpu_start - Start cpu +* @cpu: cpu number +*/ +void slcr_cpu_start (int cpu) +{ +/* enable CPUn */ +writel (SLCR_A9_CPU_CLKSTOP << cpu, + zynq_slcr_base +SLCR_A9_CPU_RST_CTRL ); +/* enable CLK for CPUn */ +writel (0x0 << cpu, zynq_slcr_base +SLCR_A9_CPU_RST_CTRL ); +} + +/** +* slcr_cpu_stop - Stop cpu +* @cpu: cpu number +*/ +void slcr_cpu_stop (int cpu) +{ +/* stop CLK and reset CPUn */ +writel ((SLCR_A9_CPU_CLKSTOP |SLCR_A9_CPU_RST ) << cpu, + zynq_slcr_base +SLCR_A9_CPU_RST_CTRL ); +} + +/** * xslcr_init() * Returns 0 on success, negative errno otherwise. *
PATCH 8/10 arm: zynq: Add smp support
Quoted 33 lines · Rob Herring
Quoted 13 lines · Michal Simek
Zynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
tosecondary_startupcode.Signed-off-by Michal Simek <[email protected]>--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/common.c | 1 + arch/arm/mach-zynq/common.h | 7 ++ arch/arm/mach-zynq/platsmp.c | 160 ++++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-zynq/slcr.c | 29 ++++++++ 5 files changed, 198 insertions(+)create mode 100644
arch/arm/mach-zynq/platsmp.c[...]
Quoted 16 lines
arch/arm/mach-zynq/platsmp.c new file +16
-0+/* Secondary CPU kernel startup is a 2 step process. The primary CPU +* starts the secondary CPU by giving it the address of the kernel and +* then sending it an event to wake it up. The secondary CPU then +* starts the kernel and tells the primary CPU it's up and running. +*/ +static void __cpuinit zynq_secondary_init (unsigned int cpu) +{ +/* +* if any interrupts are already enabled for the primary +* core (e.g. timer irq), then they will not have been enabled +* for us: do so +*/ +gic_secondary_init (0 ); + +/* +* Synchronise with the boot thread.
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c line 57
Quoted 3 lines
arch/arm/mach-zynq/platsmp.c new file +3
-0+*/ +spin_lock (&boot_lock); +spin_unlock (&boot_lock);Why do you think this is needed? Platforms that need this synchronization are only the ones that just do wfi for hotplug rather than properly reseting the core. You appear to do the latter and should not need this.
ok.
Quoted 6 lines
Quoted 6 lines
arch/arm/mach-zynq/platsmp.c new file +6
-0+} + +int __cpuinit zynq_cpun_start (u32 address,int cpu) +{ +if (cpu > ncores) { +pr_warn ("CPU No. is not available in the system\n" );
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c line 69
Quoted 6 lines
arch/arm/mach-zynq/platsmp.c new file +6
-0+return -1 ; + } + +/* MS: Expectation that SLCR are directly map and accessible */ +/* Not possible to jump to non aligned address */ +if (!(address &3 ) && (!address || (address >=0xC ))) {What about Thumb2 kernel entry?
I have no idea what's that.
Still more microblaze guy than Arm one.
Quoted 4 lines
arch/arm/mach-zynq/platsmp.c line 70
Quoted 2 lines
arch/arm/mach-zynq/platsmp.c new file +1
-0+slcr_cpu_stop (cpu);Isn't a secondary cpu already stopped?
On the normal boot this is really necessary because first stage bootloader doesn't stop cpu just keep it in loop and without stopping cpu and starting it again it doesn't work.
Quoted 11 lines
Quoted 11 lines
arch/arm/mach-zynq/platsmp.c new file +11
-0+ +/* +* This is elegant way how to jump to any address +* 0x0: Load address at 0x8 to r0 +* 0x4: Jump by mov instruction +* 0x8: Jumping address +*/ +if (address) { +/* 0: ldr r0, [8] */ +__raw_writel (0xe59f0000 ,phys_to_virt (0x0 )); +/* 4: mov pc, r0 */
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c line 87
Quoted 6 lines
arch/arm/mach-zynq/platsmp.c new file +6
-0+__raw_writel (0xe1a0f000 ,phys_to_virt (0x4 )); +__raw_writel (address,phys_to_virt (0x8 )); + } + +flush_cache_all (); +outer_flush_all ();You should only need to flush range on address 0-4 here.
ok.
Thanks,
Michal
Signature · 4 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/ Microblaze U-BOOT custodian
PATCH 8/10 arm: zynq: Add smp support
Quoted 16 lines · Michal Simek
Quoted 15 lines · Rob Herring
Quoted 14 lines · Michal Simek
Zynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
tosecondary_startupcode.Signed-off-by Michal Simek <[email protected]>--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/common.c | 1 + arch/arm/mach-zynq/common.h | 7 ++ arch/arm/mach-zynq/platsmp.c | 160 ++++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-zynq/slcr.c | 29 ++++++++ 5 files changed, 198 insertions(+)create mode 100644
arch/arm/mach-zynq/platsmp.c
[...]
Quoted 15 lines
Quoted 6 lines
Quoted 6 lines
arch/arm/mach-zynq/platsmp.c new file +6
-0+} + +int __cpuinit zynq_cpun_start (u32 address,int cpu) +{ +if (cpu > ncores) { +pr_warn ("CPU No. is not available in the system\n" );Quoted 8 lines
arch/arm/mach-zynq/platsmp.c line 69
Quoted 6 lines
arch/arm/mach-zynq/platsmp.c new file +6
-0+return -1 ; + } + +/* MS: Expectation that SLCR are directly map and accessible */ +/* Not possible to jump to non aligned address */ +if (!(address &3 ) && (!address || (address >=0xC ))) {What about Thumb2 kernel entry?
Quoted 2 lines
I have no idea what's that.
Still more microblaze guy than Arm one.
It's the 16-bit (mostly) instruction mode. Why it matters here is bit 0 being set in an address will trigger a switch to Thumb mode in a bx/blx instruction. So you can't really check for alignment as only 0x2 would not be allowed. More below...
Quoted 9 lines
Quoted 5 lines
Quoted 3 lines
+
slcr_cpu_stop(cpu);Isn't a secondary cpu already stopped?
On the normal boot this is really necessary because first stage bootloader doesn't stop cpu just keep it in loop and without stopping cpu and starting it again it doesn't work.
And there is no way to exit the loop other than a reset?
So for for hotplug this would not be needed. Perhaps .smp_prepare_cpus is a better spot for this.
If you can change the bootloader, then you should look at doing PSCI support. Here's some information:
http://lca-13.zerista.com/files_user/attachments/9311/psci_update.pdf
I've also submitted highbank patches which add support for PSCI.
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
arch/arm/mach-zynq/platsmp.c new file +4
-0+ +/* +* This is elegant way how to jump to any address +* 0x0: Load address at 0x8 to r0
Quoted 8 lines
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c line 82
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c new file +8
-0+* 0x4: Jump by mov instruction +* 0x8: Jumping address +*/ +if (address) { +/* 0: ldr r0, [8] */ +__raw_writel (0xe59f0000 ,phys_to_virt (0x0 )); +/* 4: mov pc, r0 */ +__raw_writel (0xe1a0f000 ,phys_to_virt (0x4 ));
This should be a "bx r0" to work with Thumb2 entry address.
Also, this part of the setup could be one time rather than every hotplug.
Quoted 1 line
Quoted 1 line
Quoted 1 line
+
__raw_writel(address, phys_to_virt(0x8));
This should be a per core address including core 0 if you ever want to do something like cpuidle powergating on one core and hotplug on another.
Rob
PATCH 8/10 arm: zynq: Add smp support
2013/3/25 Rob Herring <[email protected]>:
Quoted 35 lines · Michal Simek
Quoted 16 lines · Rob Herring
Quoted 15 lines · Michal Simek
Quoted 14 lines
Zynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
tosecondary_startupcode.Signed-off-by Michal Simek <[email protected]>--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/common.c | 1 + arch/arm/mach-zynq/common.h | 7 ++ arch/arm/mach-zynq/platsmp.c | 160 ++++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-zynq/slcr.c | 29 ++++++++ 5 files changed, 198 insertions(+)create mode 100644
arch/arm/mach-zynq/platsmp.c[...]
Quoted 15 lines
Quoted 14 lines
arch/arm/mach-zynq/platsmp.c line 69
Quoted 12 lines
arch/arm/mach-zynq/platsmp.c new file +12
-0+} + +int __cpuinit zynq_cpun_start (u32 address,int cpu) +{ +if (cpu > ncores) { +pr_warn ("CPU No. is not available in the system\n" ); +return -1 ; + } + +/* MS: Expectation that SLCR are directly map and accessible */ +/* Not possible to jump to non aligned address */ +if (!(address &3 ) && (!address || (address >=0xC ))) {What about Thumb2 kernel entry?
Quoted 7 lines
Quoted 2 lines
I have no idea what's that.
Still more microblaze guy than Arm one.It's the 16-bit (mostly) instruction mode. Why it matters here is bit 0 being set in an address will trigger a switch to Thumb mode in a bx/blx instruction. So you can't really check for alignment as only
0x2would not be allowed. More below...
ok.
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
+
slcr_cpu_stop(cpu);
Quoted 7 lines
Quoted 5 lines
Quoted 1 line
Isn't a secondary cpu already stopped?
On the normal boot this is really necessary because first stage bootloader doesn't stop cpu just keep it in loop and without stopping cpu and starting it again it doesn't work.
And there is no way to exit the loop other than a reset?
You can exit the loop by writing jump address to one location where bootloader expect it. Then it jumps to proper function and it was done like that before. But this is not suitable for cpu hotplug because that loop is placed in OCM (on chip memory)
and it can be used for different purpose.
Also there is no way how to return cpu to this mode.
Quoted 2 lines
So for for hotplug this would not be needed. Perhaps .smp_prepare_cpus is a better spot for this.
I don't think so because smp_prepare_cpus is called by the kernel just once in bootup time.
For cpu hotplug I see that only smp_boot_secondary is called.
Quoted 2 lines
If you can change the bootloader, then you should look at doing PSCI support. Here's some information:
Unfortunately no.
Quoted 4 lines
http://lca-13.zerista.com/files_user/attachments/9311/psci_update.pdf
I've also submitted highbank patches which add support for PSCI.
Interesting. I will look at it. Thanks.
Quoted 13 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
arch/arm/mach-zynq/platsmp.c new file +4
-0+ +/* +* This is elegant way how to jump to any address +* 0x0: Load address at 0x8 to r0 Quoted 8 lines
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c line 82
Quoted 8 lines
arch/arm/mach-zynq/platsmp.c new file +8
-0+* 0x4: Jump by mov instruction +* 0x8: Jumping address +*/ +if (address) { +/* 0: ldr r0, [8] */ +__raw_writel (0xe59f0000 ,phys_to_virt (0x0 )); +/* 4: mov pc, r0 */ +__raw_writel (0xe1a0f000 ,phys_to_virt (0x4 ));
Quoted 3 lines
This should be a "bx r0" to work with Thumb2 entry address.
Also, this part of the setup could be one time rather than every hotplug.
It is better to copy it there all the time because we have one AMP demo where rtos runs from 0x0 because there are reset vectors for it. That's why it is necessary to copy this trampoline code all the time.
Quoted 4 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
+
__raw_writel(address, phys_to_virt(0x8));This should be a per core address including core 0 if you ever want to do something like cpuidle powergating on one core and hotplug on another.
That's interesting idea.
Please correct me if I am wrong, I didn't play with cpuidle.
Zynq is dual core and hotplug can be done only on cpu1. (not sure if in general cpu0
can be unplugged too. If yes, are you doing that?).
I didn't play with cpuidle code but I am not quite sure if you can use hotplug if cpu0 is in idle because this code is for >cpu0.
I can imagine to be more flexible on quad core where your comment make definitely sense.
Thanks,
Michal
Signature · 5 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel - Microblaze cpu - http://www.monstr.eu/fdt/ Maintainer of Linux kernel - Xilinx Zynq ARM architecture Microblaze U-BOOT custodian and responsible for u-boot arm zynq platform
PATCH 8/10 arm: zynq: Add smp support
Quoted 7 lines · Michal Simek
2013/3/25 Rob Herring <[email protected]>:
Quoted 6 lines · Michal Simek
Quoted 5 lines · Rob Herring
Quoted 4 lines · Michal Simek
Quoted 3 lines
Zynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
tosecondary_startupcode.
[...]
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
+
slcr_cpu_stop(cpu);Isn't a secondary cpu already stopped?
On the normal boot this is really necessary because first stage bootloader doesn't stop cpu just keep it in loop and without stopping cpu and starting it again it doesn't work.
Quoted 8 lines
Quoted 1 line
And there is no way to exit the loop other than a reset?
You can exit the loop by writing jump address to one location where bootloader expect it. Then it jumps to proper function and it was done like that before. But this is not suitable for cpu hotplug because that loop is placed in OCM (on chip memory)
and it can be used for different purpose.
Also there is no way how to return cpu to this mode.
I was asking about the cold boot case, not hotplug. If you are spinning in the bootloader waiting for a jump address, then why do you need the slcr_cpu_stop for cold boot? In the hotplug case, you have already called slcr_cpu_stop in the unplug path, so this shouldn't be needed there either.
[...]
Quoted 11 lines
Quoted 4 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
+
__raw_writel(address, phys_to_virt(0x8));This should be a per core address including core 0 if you ever want to do something like cpuidle powergating on one core and hotplug on another.
That's interesting idea.
Please correct me if I am wrong, I didn't play with cpuidle.Zynq is dual core and hotplug can be done only on cpu1. (not sure if in general cpu0
can be unplugged too. If yes, are you doing that?).
You cannot do hotplug on cpu 0, but that is a current Linux limitation. I believe support to hotplug cpu 0 was recently added for x86. So this may change at some point.
What the h/w can support is another issue. Some chips have independent power domains per core and some have combined domains. I can and do powergate individual cores including core 0 in cpuidle for highbank. The main difference to hotplug is whether I set the wake-up address when I go down (cpuidle) or when I wake the core (hotplug). For hotplug, you don't want the core to come back before the kernel is ready for it.
Quoted 5 lines
I didn't play with cpuidle code but I am not quite sure if you can use hotplug if cpu0 is in idle because this code is for >cpu0.
I can imagine to be more flexible on quad core where your comment make definitely sense.
The next part could be quad core...
Rob
PATCH 8/10 arm: zynq: Add smp support
2013/4/2 Rob Herring <[email protected]>:
Quoted 28 lines · Michal Simek
Quoted 7 lines
2013/3/25 Rob Herring <[email protected]>:
Quoted 6 lines · Michal Simek
Quoted 5 lines · Rob Herring
Quoted 4 lines · Michal Simek
Quoted 3 lines
Zynq is dual core Cortex A9 which starts always
at zero. Using simple trampoline ensure long jump
tosecondary_startupcode.[...]
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
+
slcr_cpu_stop(cpu);Isn't a secondary cpu already stopped?
On the normal boot this is really necessary because first stage bootloader doesn't stop cpu just keep it in loop and without stopping cpu and starting it again it doesn't work.
Quoted 8 lines
Quoted 1 line
And there is no way to exit the loop other than a reset?
You can exit the loop by writing jump address to one location where bootloader expect it. Then it jumps to proper function and it was done like that before. But this is not suitable for cpu hotplug because that loop is placed in OCM (on chip memory)
and it can be used for different purpose.
Also there is no way how to return cpu to this mode.
Quoted 5 lines
I was asking about the cold boot case, not hotplug. If you are spinning in the bootloader waiting for a jump address, then why do you need the
slcr_cpu_stopfor cold boot? In the hotplug case, you have already calledslcr_cpu_stopin the unplug path, so this shouldn't be needed there either.
For cold boot case there is an option to write jump addr
to specific address which fsbl (xilinx proprietary bootloader) expects or just do cpu reset.
Waiting loop is placed in OCM. Then u-boot runs and user can change OCM mapping for whatever reason.
My point is that reset is cleaner way because I am exactly sure that cpu will be in the correct state and previous bootloader/code just hasn't done any weird setting/changes.
It means that my requirement is that any bootloader will just ensure that cpu1 is
in any waiting loop/sleep/whatever and I will do proper reset when I need to start it.
I understand your concern about moving slcr_cpu_stop to different location or even just stop cpu in bootloader but that's even more problematic.
Quoted 16 lines
Quoted 11 lines
Quoted 4 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
+
__raw_writel(address, phys_to_virt(0x8));This should be a per core address including core 0 if you ever want to do something like cpuidle powergating on one core and hotplug on another.
That's interesting idea.
Please correct me if I am wrong, I didn't play with cpuidle.Zynq is dual core and hotplug can be done only on cpu1. (not sure if in general cpu0
can be unplugged too. If yes, are you doing that?).You cannot do hotplug on cpu 0, but that is a current Linux limitation. I believe support to hotplug cpu 0 was recently added for x86. So this may change at some point.
Quoted 6 lines
What the h/w can support is another issue. Some chips have independent power domains per core and some have combined domains. I can and do powergate individual cores including core 0 in cpuidle for highbank. The main difference to hotplug is whether I set the wake-up address when I go down (cpuidle) or when I wake the core (hotplug). For hotplug, you don't want the core to come back before the kernel is ready for it.
ok
Quoted 7 lines
Quoted 5 lines
I didn't play with cpuidle code but I am not quite sure if you can use hotplug if cpu0 is in idle because this code is for >cpu0.
I can imagine to be more flexible on quad core where your comment make definitely sense.
The next part could be quad core...
I will look at changing this boot up code for future.
Thanks,
Michal
Signature · 5 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel - Microblaze cpu - http://www.monstr.eu/fdt/ Maintainer of Linux kernel - Xilinx Zynq ARM architecture Microblaze U-BOOT custodian and responsible for u-boot arm zynq platform
PATCH 9/10 arm: zynq: Add hotplug support
In reply to Michal Simek--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/common.h | 3 ++ arch/arm/mach-zynq/hotplug.c | 102 ++++++++++++++++++++++++++++++++++++++++++ arch/arm/mach-zynq/platsmp.c | 3 ++ 4 files changed, 109 insertions(+) create mode 100644 arch/arm/mach-zynq/hotplug.c
arch/arm/mach-zynq/Makefile +1 -0
@@ -4,4 +4,5 @@ # Common support obj-y := common.o slcr.o +obj-$(CONFIG_HOTPLUG_CPU) += hotplug.o obj-$(CONFIG_SMP) += platsmp.o
arch/arm/mach-zynq/common.h +3 -0
@@ -30,4 +30,7 @@ extern struct smp_operations zynq_smp_ops __initdata;extern void __iomem *zynq_slcr_base;extern void __iomem *scu_base; +/* Hotplug */ +extern void platform_cpu_die (unsigned int cpu); +#endif
arch/arm/mach-zynq/hotplug.c new file +102 -0
102 lines not on this page · raw message
arch/arm/mach-zynq/platsmp.c +3 -0
@@ -157,4 +157,7 @@ struct smp_operations zynq_smp_ops __initdata = { .smp_prepare_cpus = zynq_smp_prepare_cpus, .smp_secondary_init = zynq_secondary_init, .smp_boot_secondary = zynq_boot_secondary, +#ifdef CONFIG_HOTPLUG_CPU + .cpu_die = platform_cpu_die, +#endif };
Signature · 1 line
1.7.9.7
PATCH 10/10 arm: zynq: Add cpuidle support
In reply to Michal SimekAdd support for cpuidle.
--- arch/arm/mach-zynq/Makefile | 1 + arch/arm/mach-zynq/cpuidle.c | 131 ++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 132 insertions(+) create mode 100644 arch/arm/mach-zynq/cpuidle.c
arch/arm/mach-zynq/Makefile +1 -0
@@ -4,5 +4,6 @@ # Common support obj-y := common.o slcr.o +obj-$(CONFIG_CPU_IDLE) += cpuidle.o obj-$(CONFIG_HOTPLUG_CPU) += hotplug.o obj-$(CONFIG_SMP) += platsmp.o
arch/arm/mach-zynq/cpuidle.c new file +131 -0
131 lines not on this page · raw message
Signature · 1 line
1.7.9.7