* about system time incorrect after changing cpu frequency
@ 2015-08-31 14:03 ` vichy
0 siblings, 0 replies; 16+ messages in thread
From: vichy @ 2015-08-31 14:03 UTC (permalink / raw)
To: linux-arm-kernel
hi all:
My platform is like below:
1. single core Cortex A9
2. use global timer for system timer
after I porting cpu frequency driver based on snow ball, the cpu
frequency did change as I expected.
But the system time is incorrect( since pherial clk is got from cpu frequency)
for example:
a) cpu 1G (pherial clk = 250M) --> sleep 1 sec (OK)
b) cpu 500M (pherial clk = 125M) --> sleep 1 sec (will be measured as 2 sec)
I try to call below 2 functions to change the frequency of clocksource
and clockevent, but the above b) sleep time is still incorrect when
cpu runs in 500Mhz.
clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate);
__clocksource_updatefreq_hz(>_clocksource, gt_clk_rate);
in Arm cortex A9 single core system with Global timer as system timer,
is there any kernel api to change system timer period when cpu/pherial
frequency change?
appreciate your kind help in advance,
^ permalink raw reply [flat|nested] 16+ messages in thread* about system time incorrect after changing cpu frequency @ 2015-08-31 14:03 ` vichy 0 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-08-31 14:03 UTC (permalink / raw) To: cpufreq, linux-arm-kernel hi all: My platform is like below: 1. single core Cortex A9 2. use global timer for system timer after I porting cpu frequency driver based on snow ball, the cpu frequency did change as I expected. But the system time is incorrect( since pherial clk is got from cpu frequency) for example: a) cpu 1G (pherial clk = 250M) --> sleep 1 sec (OK) b) cpu 500M (pherial clk = 125M) --> sleep 1 sec (will be measured as 2 sec) I try to call below 2 functions to change the frequency of clocksource and clockevent, but the above b) sleep time is still incorrect when cpu runs in 500Mhz. clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); in Arm cortex A9 single core system with Global timer as system timer, is there any kernel api to change system timer period when cpu/pherial frequency change? appreciate your kind help in advance, ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-08-31 14:03 ` vichy @ 2015-09-01 3:57 ` Viresh Kumar -1 siblings, 0 replies; 16+ messages in thread From: Viresh Kumar @ 2015-09-01 3:57 UTC (permalink / raw) To: vichy; +Cc: cpufreq, linux-arm-kernel, Linux PM list On Mon, Aug 31, 2015 at 7:33 PM, vichy <vichy.kuo@gmail.com> wrote: > hi all: > My platform is like below: > 1. single core Cortex A9 > 2. use global timer for system timer > > after I porting cpu frequency driver based on snow ball, the cpu > frequency did change as I expected. > But the system time is incorrect( since pherial clk is got from cpu frequency) > > for example: > a) cpu 1G (pherial clk = 250M) --> sleep 1 sec (OK) > b) cpu 500M (pherial clk = 125M) --> sleep 1 sec (will be measured as 2 sec) > > I try to call below 2 functions to change the frequency of clocksource > and clockevent, but the above b) sleep time is still incorrect when > cpu runs in 500Mhz. > clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); > __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); > > in Arm cortex A9 single core system with Global timer as system timer, > is there any kernel api to change system timer period when cpu/pherial > frequency change? > > appreciate your kind help in advance, The list cpufreq@vger.kernel.org is the wrong list for posting cpufreq queries as we have moved to Linux PM list <linux-pm@vger.kernel.org> list now. Try unsetting CPUFREQ_CONST_LOOPS flag in your driver, if you have it set. ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 3:57 ` Viresh Kumar 0 siblings, 0 replies; 16+ messages in thread From: Viresh Kumar @ 2015-09-01 3:57 UTC (permalink / raw) To: linux-arm-kernel On Mon, Aug 31, 2015 at 7:33 PM, vichy <vichy.kuo@gmail.com> wrote: > hi all: > My platform is like below: > 1. single core Cortex A9 > 2. use global timer for system timer > > after I porting cpu frequency driver based on snow ball, the cpu > frequency did change as I expected. > But the system time is incorrect( since pherial clk is got from cpu frequency) > > for example: > a) cpu 1G (pherial clk = 250M) --> sleep 1 sec (OK) > b) cpu 500M (pherial clk = 125M) --> sleep 1 sec (will be measured as 2 sec) > > I try to call below 2 functions to change the frequency of clocksource > and clockevent, but the above b) sleep time is still incorrect when > cpu runs in 500Mhz. > clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); > __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); > > in Arm cortex A9 single core system with Global timer as system timer, > is there any kernel api to change system timer period when cpu/pherial > frequency change? > > appreciate your kind help in advance, The list cpufreq at vger.kernel.org is the wrong list for posting cpufreq queries as we have moved to Linux PM list <linux-pm@vger.kernel.org> list now. Try unsetting CPUFREQ_CONST_LOOPS flag in your driver, if you have it set. ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-09-01 3:57 ` Viresh Kumar @ 2015-09-01 5:36 ` vichy -1 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-09-01 5:36 UTC (permalink / raw) To: Viresh Kumar; +Cc: cpufreq, linux-arm-kernel, Linux PM list hi Viresh: 2015-09-01 11:57 GMT+08:00 Viresh Kumar <viresh.kumar@linaro.org>: > On Mon, Aug 31, 2015 at 7:33 PM, vichy <vichy.kuo@gmail.com> wrote: >> hi all: >> My platform is like below: >> 1. single core Cortex A9 >> 2. use global timer for system timer >> >> after I porting cpu frequency driver based on snow ball, the cpu >> frequency did change as I expected. >> But the system time is incorrect( since pherial clk is got from cpu frequency) >> >> for example: >> a) cpu 1G (pherial clk = 250M) --> sleep 1 sec (OK) >> b) cpu 500M (pherial clk = 125M) --> sleep 1 sec (will be measured as 2 sec) >> >> I try to call below 2 functions to change the frequency of clocksource >> and clockevent, but the above b) sleep time is still incorrect when >> cpu runs in 500Mhz. >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); >> >> in Arm cortex A9 single core system with Global timer as system timer, >> is there any kernel api to change system timer period when cpu/pherial >> frequency change? >> >> appreciate your kind help in advance, > > The list cpufreq@vger.kernel.org is the wrong list for posting cpufreq queries > as we have moved to Linux PM list <linux-pm@vger.kernel.org> list now. > > Try unsetting CPUFREQ_CONST_LOOPS flag in your driver, if you have it > set. I DIDN'T set the CPUFREQ_CONST_LOOPS when I register my cpufreq driver I pasted my cpufreq driver declariation as below: static struct cpufreq_driver plat_cpufreq_driver = { .flags = CPUFREQ_STICKY, .verify = plat_cpufreq_verify_speed, .target = plat_cpufreq_target, .get = plat_cpufreq_getspeed, .init = plat_cpufreq_init, .name = "plat-cpufreq", .attr = plat_cpufreq_attr, }; I have traced the kernel code if I guess correctly, the sleep accurate is based on jiffies and tick_handle_periodic will periodically update the next event interval void tick_handle_periodic --> for (;;) { if (!clockevents_program_event(dev, next, false)) return; /* * Have to be careful here. If we're in oneshot mode, * before we call tick_periodic() in a loop, we need * to be sure we're using a real hardware clocksource. * Otherwise we could get trapped in an infinite * loop, as the tick_periodic() increments jiffies, * when then will increment time, posibly causing * the loop to trigger again and again. */ if (timekeeping_valid_for_hres()) tick_periodic(cpu); next = ktime_add(next, tick_period); } and in clockevents_program_event, we will use mult, shift to calculate the cycles need for global timer triggering next interrupt event. clc = ((unsigned long long) delta * dev->mult) >> dev->shift; rc = dev->set_next_event((unsigned long) clc, dev); belwo is my tick device information in /proc/timer_list and multi did change to 1/2 when I change cpu freq from 1G to 500Mhz. when cpu run 1GHz except multi and shift, is there any place I need to take care for system timer accurate? appreciate your kind help, Tick Device: mode: 1 Per CPU device: 0 Clock Event Device: arm_global_timer max_delta_ns: 17043521021 min_delta_ns: 1000 mult: 541165879 shift: 31 mode: 3 next_event: 2176344000000 nsecs set_next_event: gt_clockevent_set_next_event set_mode: gt_clockevent_set_mode event_handler: hrtimer_interrupt retries: 0 when cpu run in 500Mhz Tick Device: mode: 1 Per CPU device: 0 Clock Event Device: arm_global_timer max_delta_ns: 34087041979 min_delta_ns: 1000 mult: 270582940 shift: 31 mode: 3 next_event: 2230100000000 nsecs set_next_event: gt_clockevent_set_next_event set_mode: gt_clockevent_set_mode event_handler: hrtimer_interrupt retries: 0 ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 5:36 ` vichy 0 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-09-01 5:36 UTC (permalink / raw) To: linux-arm-kernel hi Viresh: 2015-09-01 11:57 GMT+08:00 Viresh Kumar <viresh.kumar@linaro.org>: > On Mon, Aug 31, 2015 at 7:33 PM, vichy <vichy.kuo@gmail.com> wrote: >> hi all: >> My platform is like below: >> 1. single core Cortex A9 >> 2. use global timer for system timer >> >> after I porting cpu frequency driver based on snow ball, the cpu >> frequency did change as I expected. >> But the system time is incorrect( since pherial clk is got from cpu frequency) >> >> for example: >> a) cpu 1G (pherial clk = 250M) --> sleep 1 sec (OK) >> b) cpu 500M (pherial clk = 125M) --> sleep 1 sec (will be measured as 2 sec) >> >> I try to call below 2 functions to change the frequency of clocksource >> and clockevent, but the above b) sleep time is still incorrect when >> cpu runs in 500Mhz. >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); >> >> in Arm cortex A9 single core system with Global timer as system timer, >> is there any kernel api to change system timer period when cpu/pherial >> frequency change? >> >> appreciate your kind help in advance, > > The list cpufreq at vger.kernel.org is the wrong list for posting cpufreq queries > as we have moved to Linux PM list <linux-pm@vger.kernel.org> list now. > > Try unsetting CPUFREQ_CONST_LOOPS flag in your driver, if you have it > set. I DIDN'T set the CPUFREQ_CONST_LOOPS when I register my cpufreq driver I pasted my cpufreq driver declariation as below: static struct cpufreq_driver plat_cpufreq_driver = { .flags = CPUFREQ_STICKY, .verify = plat_cpufreq_verify_speed, .target = plat_cpufreq_target, .get = plat_cpufreq_getspeed, .init = plat_cpufreq_init, .name = "plat-cpufreq", .attr = plat_cpufreq_attr, }; I have traced the kernel code if I guess correctly, the sleep accurate is based on jiffies and tick_handle_periodic will periodically update the next event interval void tick_handle_periodic --> for (;;) { if (!clockevents_program_event(dev, next, false)) return; /* * Have to be careful here. If we're in oneshot mode, * before we call tick_periodic() in a loop, we need * to be sure we're using a real hardware clocksource. * Otherwise we could get trapped in an infinite * loop, as the tick_periodic() increments jiffies, * when then will increment time, posibly causing * the loop to trigger again and again. */ if (timekeeping_valid_for_hres()) tick_periodic(cpu); next = ktime_add(next, tick_period); } and in clockevents_program_event, we will use mult, shift to calculate the cycles need for global timer triggering next interrupt event. clc = ((unsigned long long) delta * dev->mult) >> dev->shift; rc = dev->set_next_event((unsigned long) clc, dev); belwo is my tick device information in /proc/timer_list and multi did change to 1/2 when I change cpu freq from 1G to 500Mhz. when cpu run 1GHz except multi and shift, is there any place I need to take care for system timer accurate? appreciate your kind help, Tick Device: mode: 1 Per CPU device: 0 Clock Event Device: arm_global_timer max_delta_ns: 17043521021 min_delta_ns: 1000 mult: 541165879 shift: 31 mode: 3 next_event: 2176344000000 nsecs set_next_event: gt_clockevent_set_next_event set_mode: gt_clockevent_set_mode event_handler: hrtimer_interrupt retries: 0 when cpu run in 500Mhz Tick Device: mode: 1 Per CPU device: 0 Clock Event Device: arm_global_timer max_delta_ns: 34087041979 min_delta_ns: 1000 mult: 270582940 shift: 31 mode: 3 next_event: 2230100000000 nsecs set_next_event: gt_clockevent_set_next_event set_mode: gt_clockevent_set_mode event_handler: hrtimer_interrupt retries: 0 ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-09-01 5:36 ` vichy @ 2015-09-01 6:02 ` Viresh Kumar -1 siblings, 0 replies; 16+ messages in thread From: Viresh Kumar @ 2015-09-01 6:02 UTC (permalink / raw) To: vichy; +Cc: cpufreq, linux-arm-kernel, Linux PM list On 01-09-15, 13:36, vichy wrote: > >> I try to call below 2 functions to change the frequency of clocksource > >> and clockevent, but the above b) sleep time is still incorrect when > >> cpu runs in 500Mhz. > >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); > >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); How and when were you calling them? What kernel version is it ? > Clock Event Device: arm_global_timer This driver doesn't have support to update clkevt device's freq. You may need to modify that based on how arch/arm/kernel/smp_twd.c is updated. look for: clockevents_update_freq(). -- viresh ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 6:02 ` Viresh Kumar 0 siblings, 0 replies; 16+ messages in thread From: Viresh Kumar @ 2015-09-01 6:02 UTC (permalink / raw) To: linux-arm-kernel On 01-09-15, 13:36, vichy wrote: > >> I try to call below 2 functions to change the frequency of clocksource > >> and clockevent, but the above b) sleep time is still incorrect when > >> cpu runs in 500Mhz. > >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); > >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); How and when were you calling them? What kernel version is it ? > Clock Event Device: arm_global_timer This driver doesn't have support to update clkevt device's freq. You may need to modify that based on how arch/arm/kernel/smp_twd.c is updated. look for: clockevents_update_freq(). -- viresh ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-09-01 6:02 ` Viresh Kumar @ 2015-09-01 6:39 ` Jisheng Zhang -1 siblings, 0 replies; 16+ messages in thread From: Jisheng Zhang @ 2015-09-01 6:39 UTC (permalink / raw) To: Viresh Kumar; +Cc: vichy, linux-arm-kernel, cpufreq, Linux PM list On Tue, 1 Sep 2015 11:32:36 +0530 Viresh Kumar <viresh.kumar@linaro.org> wrote: > On 01-09-15, 13:36, vichy wrote: > > >> I try to call below 2 functions to change the frequency of clocksource > > >> and clockevent, but the above b) sleep time is still incorrect when > > >> cpu runs in 500Mhz. > > >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); > > >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); > > How and when were you calling them? What kernel version is it ? > > > Clock Event Device: arm_global_timer > > This driver doesn't have support to update clkevt device's freq. You > may need to modify that based on how arch/arm/kernel/smp_twd.c is > updated. look for: clockevents_update_freq(). > This can only help the clockevent. Even with this cpufreq notifier update, the globaltimer clocksource still can't hold correct/accurate time. I think the arm global timer is not suitable for clocksource if cpufreq is enabled. ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 6:39 ` Jisheng Zhang 0 siblings, 0 replies; 16+ messages in thread From: Jisheng Zhang @ 2015-09-01 6:39 UTC (permalink / raw) To: linux-arm-kernel On Tue, 1 Sep 2015 11:32:36 +0530 Viresh Kumar <viresh.kumar@linaro.org> wrote: > On 01-09-15, 13:36, vichy wrote: > > >> I try to call below 2 functions to change the frequency of clocksource > > >> and clockevent, but the above b) sleep time is still incorrect when > > >> cpu runs in 500Mhz. > > >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); > > >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); > > How and when were you calling them? What kernel version is it ? > > > Clock Event Device: arm_global_timer > > This driver doesn't have support to update clkevt device's freq. You > may need to modify that based on how arch/arm/kernel/smp_twd.c is > updated. look for: clockevents_update_freq(). > This can only help the clockevent. Even with this cpufreq notifier update, the globaltimer clocksource still can't hold correct/accurate time. I think the arm global timer is not suitable for clocksource if cpufreq is enabled. ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-09-01 6:39 ` Jisheng Zhang @ 2015-09-01 7:37 ` vichy -1 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-09-01 7:37 UTC (permalink / raw) To: Jisheng Zhang; +Cc: Viresh Kumar, linux-arm-kernel, cpufreq, Linux PM list hi Jisheng: 2015-09-01 14:39 GMT+08:00 Jisheng Zhang <jszhang@marvell.com>: > On Tue, 1 Sep 2015 11:32:36 +0530 > Viresh Kumar <viresh.kumar@linaro.org> wrote: > >> On 01-09-15, 13:36, vichy wrote: >> > >> I try to call below 2 functions to change the frequency of clocksource >> > >> and clockevent, but the above b) sleep time is still incorrect when >> > >> cpu runs in 500Mhz. >> > >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); >> > >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); >> >> How and when were you calling them? What kernel version is it ? >> >> > Clock Event Device: arm_global_timer >> >> This driver doesn't have support to update clkevt device's freq. You >> may need to modify that based on how arch/arm/kernel/smp_twd.c is >> updated. look for: clockevents_update_freq(). >> > > This can only help the clockevent. Even with this cpufreq notifier > update, the globaltimer clocksource still can't hold correct/accurate time. Why even global timer add clockevents_update_freq, it stilll CANNOT hold currect/accurate time? When I trace the kernel, clockevents_program_event will calculate the cycles with mult/shift then put in comparator. if we put the write mult/shift, the global timer should hold correct/accurate time under different phepherial clock, right? Appreciate all your kind help, ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 7:37 ` vichy 0 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-09-01 7:37 UTC (permalink / raw) To: linux-arm-kernel hi Jisheng: 2015-09-01 14:39 GMT+08:00 Jisheng Zhang <jszhang@marvell.com>: > On Tue, 1 Sep 2015 11:32:36 +0530 > Viresh Kumar <viresh.kumar@linaro.org> wrote: > >> On 01-09-15, 13:36, vichy wrote: >> > >> I try to call below 2 functions to change the frequency of clocksource >> > >> and clockevent, but the above b) sleep time is still incorrect when >> > >> cpu runs in 500Mhz. >> > >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); >> > >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); >> >> How and when were you calling them? What kernel version is it ? >> >> > Clock Event Device: arm_global_timer >> >> This driver doesn't have support to update clkevt device's freq. You >> may need to modify that based on how arch/arm/kernel/smp_twd.c is >> updated. look for: clockevents_update_freq(). >> > > This can only help the clockevent. Even with this cpufreq notifier > update, the globaltimer clocksource still can't hold correct/accurate time. Why even global timer add clockevents_update_freq, it stilll CANNOT hold currect/accurate time? When I trace the kernel, clockevents_program_event will calculate the cycles with mult/shift then put in comparator. if we put the write mult/shift, the global timer should hold correct/accurate time under different phepherial clock, right? Appreciate all your kind help, ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-09-01 6:02 ` Viresh Kumar @ 2015-09-01 7:21 ` vichy -1 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-09-01 7:21 UTC (permalink / raw) To: Viresh Kumar; +Cc: cpufreq, linux-arm-kernel, Linux PM list hi Viresh: 2015-09-01 14:02 GMT+08:00 Viresh Kumar <viresh.kumar@linaro.org>: > On 01-09-15, 13:36, vichy wrote: >> >> I try to call below 2 functions to change the frequency of clocksource >> >> and clockevent, but the above b) sleep time is still incorrect when >> >> cpu runs in 500Mhz. >> >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); >> >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); > > How and when were you calling them? What kernel version is it ? Belwo is what I did in my cpufreq driver and my kernel version is 3.0.33. plat_cpufreq_target() { ........ cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); PLAT_Mpll(freqs.new / 1000); PLAT_gt_change_freq(); /* post change notification */ cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); .... } and I add PLAT_gt_change_freq in arm_global_timer.c like below: PLAT_gt_change_freq calling() { gt_clk_rate = get_periph_clk(); clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); } I also do another experiment that DIDN'T change any REAL HW setting but change mult/shift to the half value like below: (But I measure out the "sleep 1" --> still 1secs, even I modify mult/shift) (in my opinion, the "seep 1" I measured should be 500ms, since HW NO change but mult/shift change) plat_cpufreq_target() { ........ cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); // PLAT_Mpll(freqs.new / 1000); //NO HW FREQ change PLAT_gt_change_freq(); /* post change notification */ cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); .... } PLAT_gt_change_freq in arm_global_timer.c like below: PLAT_gt_change_freq calling() { gt_clk_rate = get_periph_clk()/2; //purposely chage 1/2 clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); } > >> Clock Event Device: arm_global_timer > > This driver doesn't have support to update clkevt device's freq. You > may need to modify that based on how arch/arm/kernel/smp_twd.c is > updated. look for: clockevents_update_freq(). I have check the latest kernel about arch/arm/kernel/smp_twd.c it almost do the same thing like I did, except it use cpufreq notifier for calling clockevents_update_freq like below: (the final target of them is the same --> clling clockevents_update_freq to change mult/shift. static void twd_update_frequency(void *data) { twd_timer_rate = clk_get_rate(twd_clk); clockevents_update_freq(raw_cpu_ptr(twd_evt), twd_timer_rate); } I know so far arm global timer seems not support cpu frequcy chage flow. But like my experiment above to fake change gt_clk_rate, the sleep should be 1/2 of the original. if smp_twd call twd_update_frequency to change multi/shift can work, why purposely adding clockevents_update_freq in global timer not working? or there is something else except Mult/shift need to be modify when gt_clk_rate changed? appreciate your help in advance, > > -- > viresh ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 7:21 ` vichy 0 siblings, 0 replies; 16+ messages in thread From: vichy @ 2015-09-01 7:21 UTC (permalink / raw) To: linux-arm-kernel hi Viresh: 2015-09-01 14:02 GMT+08:00 Viresh Kumar <viresh.kumar@linaro.org>: > On 01-09-15, 13:36, vichy wrote: >> >> I try to call below 2 functions to change the frequency of clocksource >> >> and clockevent, but the above b) sleep time is still incorrect when >> >> cpu runs in 500Mhz. >> >> clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); >> >> __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); > > How and when were you calling them? What kernel version is it ? Belwo is what I did in my cpufreq driver and my kernel version is 3.0.33. plat_cpufreq_target() { ........ cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); PLAT_Mpll(freqs.new / 1000); PLAT_gt_change_freq(); /* post change notification */ cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); .... } and I add PLAT_gt_change_freq in arm_global_timer.c like below: PLAT_gt_change_freq calling() { gt_clk_rate = get_periph_clk(); clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); } I also do another experiment that DIDN'T change any REAL HW setting but change mult/shift to the half value like below: (But I measure out the "sleep 1" --> still 1secs, even I modify mult/shift) (in my opinion, the "seep 1" I measured should be 500ms, since HW NO change but mult/shift change) plat_cpufreq_target() { ........ cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); // PLAT_Mpll(freqs.new / 1000); //NO HW FREQ change PLAT_gt_change_freq(); /* post change notification */ cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); .... } PLAT_gt_change_freq in arm_global_timer.c like below: PLAT_gt_change_freq calling() { gt_clk_rate = get_periph_clk()/2; //purposely chage 1/2 clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); } > >> Clock Event Device: arm_global_timer > > This driver doesn't have support to update clkevt device's freq. You > may need to modify that based on how arch/arm/kernel/smp_twd.c is > updated. look for: clockevents_update_freq(). I have check the latest kernel about arch/arm/kernel/smp_twd.c it almost do the same thing like I did, except it use cpufreq notifier for calling clockevents_update_freq like below: (the final target of them is the same --> clling clockevents_update_freq to change mult/shift. static void twd_update_frequency(void *data) { twd_timer_rate = clk_get_rate(twd_clk); clockevents_update_freq(raw_cpu_ptr(twd_evt), twd_timer_rate); } I know so far arm global timer seems not support cpu frequcy chage flow. But like my experiment above to fake change gt_clk_rate, the sleep should be 1/2 of the original. if smp_twd call twd_update_frequency to change multi/shift can work, why purposely adding clockevents_update_freq in global timer not working? or there is something else except Mult/shift need to be modify when gt_clk_rate changed? appreciate your help in advance, > > -- > viresh ^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: about system time incorrect after changing cpu frequency 2015-09-01 7:21 ` vichy @ 2015-09-01 7:42 ` Viresh Kumar -1 siblings, 0 replies; 16+ messages in thread From: Viresh Kumar @ 2015-09-01 7:42 UTC (permalink / raw) To: vichy; +Cc: cpufreq, linux-arm-kernel, Linux PM list On 01-09-15, 15:21, vichy wrote: > and I add PLAT_gt_change_freq in arm_global_timer.c like below: > PLAT_gt_change_freq calling() { > gt_clk_rate = get_periph_clk(); > clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); In earlier versions of kernel, this had a problem that it works only for ONESHOT mode and simply returns in PERIODIC mode. I do see that your's mode is ONESHOT only, and so it should have worked. Anyway, few prints into that routine to see where we are reaching might help. > __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); AFAIR, there is a difference between how different mdelay/udelay are implemented. One of them is using the loops_per_jiffie variable to calculate the time, other uses clkevt device. I am not 100% sure about the theory, but that's what I remember. So, have a look from that perspective as well.. -- viresh ^ permalink raw reply [flat|nested] 16+ messages in thread
* about system time incorrect after changing cpu frequency @ 2015-09-01 7:42 ` Viresh Kumar 0 siblings, 0 replies; 16+ messages in thread From: Viresh Kumar @ 2015-09-01 7:42 UTC (permalink / raw) To: linux-arm-kernel On 01-09-15, 15:21, vichy wrote: > and I add PLAT_gt_change_freq in arm_global_timer.c like below: > PLAT_gt_change_freq calling() { > gt_clk_rate = get_periph_clk(); > clockevents_update_freq(this_cpu_ptr(gt_evt), gt_clk_rate); In earlier versions of kernel, this had a problem that it works only for ONESHOT mode and simply returns in PERIODIC mode. I do see that your's mode is ONESHOT only, and so it should have worked. Anyway, few prints into that routine to see where we are reaching might help. > __clocksource_updatefreq_hz(>_clocksource, gt_clk_rate); AFAIR, there is a difference between how different mdelay/udelay are implemented. One of them is using the loops_per_jiffie variable to calculate the time, other uses clkevt device. I am not 100% sure about the theory, but that's what I remember. So, have a look from that perspective as well.. -- viresh ^ permalink raw reply [flat|nested] 16+ messages in thread
end of thread, other threads:[~2015-09-01 7:42 UTC | newest] Thread overview: 16+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2015-08-31 14:03 about system time incorrect after changing cpu frequency vichy 2015-08-31 14:03 ` vichy 2015-09-01 3:57 ` Viresh Kumar 2015-09-01 3:57 ` Viresh Kumar 2015-09-01 5:36 ` vichy 2015-09-01 5:36 ` vichy 2015-09-01 6:02 ` Viresh Kumar 2015-09-01 6:02 ` Viresh Kumar 2015-09-01 6:39 ` Jisheng Zhang 2015-09-01 6:39 ` Jisheng Zhang 2015-09-01 7:37 ` vichy 2015-09-01 7:37 ` vichy 2015-09-01 7:21 ` vichy 2015-09-01 7:21 ` vichy 2015-09-01 7:42 ` Viresh Kumar 2015-09-01 7:42 ` Viresh Kumar
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.