[PATCH 1/3] ARM: rockchip: fix the CPU soft reset

[heiko@sntech.de (Heiko Stübner)]

Am Freitag, 5. Juni 2015, 17:00:02 schrieb Caesar Wang:
> Heiko,
> 
> ? 2015?06?05? 16:45, Heiko St?bner ??:
> > Hi Caesar,
> > 
> > 
> > thanks for investigating this.
> > 
> > Am Freitag, 5. Juni 2015, 12:47:55 schrieb Caesar Wang:
> >> In general, the correct flow is:
> >> 
> >> cpu off:
> >>      reset_control_assert
> >>      regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), BIT(pd))
> >> 
> >> cpu on:
> >>      regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), 0)
> >>      reset_control_deassert
> >> 
> >> You can repro it with bringing CPU up and down.
> >> Says:(test scripts)
> >> 
> >>      cd /sys/devices/system/cpu/
> >>      for i in $(seq 1000); do
> >>      echo "================= $i ============"
> >>      
> >>          for j in $(seq 100); do
> >>          
> >>              while [[ "$(cat cpu1/online)$(cat cpu2/online)$(cat
> >> 
> >> cpu3/online)" != "000" ]]; do echo 0 > cpu1/online
> >> 
> >>                  echo 0 > cpu2/online
> >>                  echo 0 > cpu3/online
> >>              
> >>              done
> >>              while [[ "$(cat cpu1/online)$(cat cpu2/online)$(cat
> >> 
> >> cpu3/online)" != "111" ]]; do echo 1 > cpu1/online
> >> 
> >>                  echo 1 > cpu2/online
> >>                  echo 1 > cpu3/online
> >>              
> >>              done
> >>          
> >>          done
> >>      
> >>      done
> >> 
> >> The following is reproducile log:
> >> [34466.186812] PM: noirq suspend of devices complete after 0.669 msecs
> >> [34466.186824] Disabling non-boot CPUs ...
> >> [34466.187509] CPU1: shutdown
> >> [34466.188672] CPU2: shutdown
> >> [34473.736627] Kernel panic - not syncing: Watchdog detected hard LOCKUP
> >> on
> >> cpu 0 [34473.736646] CPU: 3 PID: 0 Comm: swapper/3 Not tainted 3.14.0 #1
> >> [34473.736687] [<c010dfb0>] (unwind_backtrace) from [<c010a460>]
> >> (show_stack+0x20/0x24) [34473.736711] [<c010a460>] (show_stack) from
> >> [<c0630670>] (dump_stack+0x70/0x8c) [34473.736731] [<c0630670>]
> >> (dump_stack) from [<c062fa0c>] (panic+0xa8/0x1fc) [34473.736754]
> >> [<c062fa0c>] (panic) from [<c0191efc>] (watchdog_timer_fn+0x234/0x26c)
> >> [34473.736777] [<c0191efc>] (watchdog_timer_fn) from [<c014103c>]
> >> (__run_hrtimer+0x118/0x1e0) [34473.736797] [<c014103c>] (__run_hrtimer)
> >> from [<c0141c04>] (hrtimer_interrupt+0x148/0x2a0) [34473.736820]
> >> [<c0141c04>] (hrtimer_interrupt) from [<c04d9994>]
> >> (arch_timer_handler_phys+0x38/0x48) [34473.736844] [<c04d9994>]
> >> (arch_timer_handler_phys) from [<c016acc4>]
> >> (handle_percpu_devid_irq+0xb8/0x124) [34473.736867] [<c016acc4>]
> >> (handle_percpu_devid_irq) from [<c0166a30>]
> >> (generic_handle_irq+0x30/0x40)
> >> [34473.736887] [<c0166a30>] (generic_handle_irq) from [<c0166d28>]
> >> (__handle_domain_irq+0x8c/0xb0) [34473.736905] [<c0166d28>]
> >> (__handle_domain_irq) from [<c01003a0>] (gic_handle_irq+0x48/0x6c)
> >> [34473.736922] [<c01003a0>] (gic_handle_irq) from [<c010b040>]
> >> (__irq_svc+0x40/0x50) [34473.736936] Exception stack(0xee127f70 to
> >> 0xee127fb8)
> >> [34473.736948] 7f60:                                     ffffffed
> >> 00000000
> >> 2dd6d000 00000000 [34473.736964] 7f80: ee126000 00000015 c0b46bac
> >> c0b46bac
> >> 0000406a 410fc0d1 00000000 ee127fc4 [34473.736979] 7fa0: ee127fb8
> >> ee127fb8
> >> c0107038 c010703c 600f0013 ffffffff [34473.736995] [<c010b040>]
> >> (__irq_svc)
> >> from [<c010703c>] (arch_cpu_idle+0x40/0x48) [34473.737013] [<c010703c>]
> >> (arch_cpu_idle) from [<c0166978>] (cpu_startup_entry+0x170/0x1d0)
> >> [34473.737031] [<c0166978>] (cpu_startup_entry) from [<c010c254>]
> >> (secondary_start_kernel+0x138/0x160) [34473.737059] [<c010c254>]
> >> (secondary_start_kernel) from [<00100464>] (0x100464) [34474.903740] SMP:
> >> failed to stop secondary CPUs
> >> [34476.099964] SMP: failed to stop secondary CPUs
> >> ...
> >> 
> >> Signed-off-by: Caesar Wang <wxt@rock-chips.com>
> >> ---
> >> 
> >>   arch/arm/mach-rockchip/platsmp.c | 13 +++++--------
> >>   1 file changed, 5 insertions(+), 8 deletions(-)
> >> 
> >> diff --git a/arch/arm/mach-rockchip/platsmp.c
> >> b/arch/arm/mach-rockchip/platsmp.c index 5b4ca3c..1230d3d 100644
> >> --- a/arch/arm/mach-rockchip/platsmp.c
> >> +++ b/arch/arm/mach-rockchip/platsmp.c
> >> @@ -88,20 +88,17 @@ static int pmu_set_power_domain(int pd, bool on)
> >> 
> >>   			return PTR_ERR(rstc);
> >>   		
> >>   		}
> >> 
> >> -		if (on)
> >> +		if (on) {
> >> +			regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), val);
> >> 
> >>   			reset_control_deassert(rstc);
> >> 
> >> -		else
> >> +		} else {
> >> 
> >>   			reset_control_assert(rstc);
> >> 
> >> +			regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), val);
> >> +		}
> > 
> > you're loosing the return value of regmap_update_bits here, I guess it
> > should look like below?
> > 
> > 		if (on) {
> > 		
> > 			ret = regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), val);
> 
> I search the kernel, Nobody do that return value.

counter-example:
http://lxr.free-electrons.com/source/drivers/phy/phy-exynos5250-sata.c#L76
;-)

> Need we are that?
> I will fix it on next patch if it's need.

yes, while not very probable, it is nevertheless still good practice to handle 
it. But when chaning the logic as we observed below, the regmap update part 
can already stay how it is now, as you can simply do something like:

if (!on)
	reset_control_assert(rstc);

regmap_update_bits
wait_for_pd

if (on)
	reset_control_deassert(rstc);


Heiko

> > 			if (ret < 0) {
> > 			
> > 				pr_err("%s: could not update power domain\n", __func__);
> > 				
> > 		  		reset_control_put(rstc);
> > 				
> > 				return ret;
> > 			
> > 			}
> > 			
> > 			reset_control_deassert(rstc);
> > 		
> > 		} else {
> > 		
> > 			reset_control_assert(rstc);
> > 			
> > 			ret = regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), val);
> > 			if (ret < 0) {
> > 			
> > 				pr_err("%s: could not update power domain\n", __func__);
> > 				
> > 		  		reset_control_put(rstc);
> > 				
> > 				return ret;
> > 			
> > 			}
> > 		
> > 		}
> > 		
> >>   		reset_control_put(rstc);
> >>   	
> >>   	}
> >> 
> >> -	ret = regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), val);
> >> -	if (ret < 0) {
> >> -		pr_err("%s: could not update power domain\n", __func__);
> >> -		return ret;
> >> -	}
> >> -
> >> 
> >>   	ret = -1;
> >>   	while (ret != on) {
> >>   	
> >>   		ret = pmu_power_domain_is_on(pd);
> > 
> > second question - with this patch, what happens actually is
> > 
> >   cpu off:
> >       reset_control_assert
> >       regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), BIT(pd))
> >       wait_for_power_domain_to_turn_off
> >   
> >   cpu on:
> >       regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), 0)
> >       reset_control_deassert
> >       wait_for_power_domain_to_turn_on
> > 
> > So shouldn't the deassertion of the reset happen after the powerdomain
> > sucessfull turned on? Like
> > 
> >   cpu on:
> >       regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), 0)
> >       wait_for_power_domain_to_turn_on
> >       reset_control_deassert
> 
> You are right.