Re: [PATCH] ARM: OMAP5/DRA7: realtime_counter: Configure CNTFRQ register

[Santosh Shilimkar <santosh.shilimkar@ti.com>]

On Wednesday 18 September 2013 09:44 AM, Sricharan R wrote:
> On Wednesday 18 September 2013 07:01 PM, Santosh Shilimkar wrote:
>> On Wednesday 18 September 2013 07:23 AM, Sricharan R wrote:
>>> The realtime counter called master counter, produces the count
>>> used by the private timer peripherals in the MPU_CLUSTER. The
>>> CNTFRQ per cpu register is used to denote the frequency of the counter.
>>> Currently the frequency value is passed from the
>>> DT file, but this is not scalable when we have other non-DT guest
>>> OS. This register must be set to the right value by the
>>> host OS, as this will be propagated to the guests as well.
>>>
>> Its not host OS but ROM code/secure code. Host OS is not
>> ideal place to set it up.
>  ok, so you are suggesting to re word this right ?
>>> More discussions and the reason for adding this in a non-DT
>>> way can be seen from below.
>>> http://www.mail-archive.com/linux-omap@vger.kernel.org/msg93832.html
>>>
>>> So configuring this secure register for all the cpus here.
>>>
>>> Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
>>> Cc: Nishanth Menon <nm@ti.com>
>>> Cc: Rajendra Nayak <rnayak@ti.com>
>>> Cc: Marc Zyngier <marc.zyngier@arm.com>
>>> Cc: Mark Rutland <mark.rutland@arm.com>
>>> Cc: Tony Lindgren <tony@atomide.com>
>>> Signed-off-by: Sricharan R <r.sricharan@ti.com>
>>> ---
>>>  arch/arm/mach-omap2/omap-secure.h |    2 ++
>>>  arch/arm/mach-omap2/omap-smp.c    |    9 +++++++++
>>>  arch/arm/mach-omap2/timer.c       |    6 ++++++
>>>  3 files changed, 17 insertions(+)
>>>
>>> diff --git a/arch/arm/mach-omap2/omap-secure.h b/arch/arm/mach-omap2/omap-secure.h
>>> index a171a5a..4de9a00 100644
>>> --- a/arch/arm/mach-omap2/omap-secure.h
>>> +++ b/arch/arm/mach-omap2/omap-secure.h
>>> @@ -51,6 +51,8 @@
>>>  
>>>  #define OMAP5_MON_AMBA_IF_INDEX		0x108
>>>  
>>> +#define OMAP5_DRA7_MON_SET_CNTFRQ_INDEX	0x109
>>> +
>> Nice. Looks like OMAP5 ES2.0 addition which I missed.
>>
>>>  /* Secure PPA(Primary Protected Application) APIs */
>>>  #define OMAP4_PPA_L2_POR_INDEX		0x23
>>>  #define OMAP4_PPA_CPU_ACTRL_SMP_INDEX	0x25
>>> diff --git a/arch/arm/mach-omap2/omap-smp.c b/arch/arm/mach-omap2/omap-smp.c
>>> index 98a1146..00873b4 100644
>>> --- a/arch/arm/mach-omap2/omap-smp.c
>>> +++ b/arch/arm/mach-omap2/omap-smp.c
>>> @@ -41,6 +41,8 @@
>>>  
>>>  u16 pm44xx_errata;
>>>  
>>> +extern unsigned long arch_timer_freq;
>>> +
>>>  /* SCU base address */
>>>  static void __iomem *scu_base;
>>>  
>>> @@ -66,6 +68,13 @@ static void __cpuinit omap4_secondary_init(unsigned int cpu)
>>>  							4, 0, 0, 0, 0, 0);
>>>  
>>>  	/*
>>> +	 * Configure the CNTFRQ register for the secondary cpu's which
>>> +	 * indicates the frequency of the cpu local timers.
>>> +	 */
>>> +	if (soc_is_omap54xx() || soc_is_dra7xx())
>> CNTFREQ programming was not supported on OMAP5 ES1.0 and that was one
>> of the reason this parameter came into picture. So you need to skip
>> the ES1.0 here. 
>  We do not intend to support ES1.0 right ? . ok then id.c needs to be
>  cleaned up to remove ES1.0 traces.
yes. With clean-up ES1.0, this can stay as is.

>>> +		omap_smc1(OMAP5_DRA7_MON_SET_CNTFRQ_INDEX, arch_timer_freq);
>>> +
>>> +	/*
>>>  	 * Synchronise with the boot thread.
>>>  	 */
>>>  	spin_lock(&boot_lock);
>>> diff --git a/arch/arm/mach-omap2/timer.c b/arch/arm/mach-omap2/timer.c
>>> index fd9238d..2c499f3 100644
>>> --- a/arch/arm/mach-omap2/timer.c
>>> +++ b/arch/arm/mach-omap2/timer.c
>>> @@ -55,6 +55,7 @@
>>>  #include "soc.h"
>>>  #include "common.h"
>>>  #include "powerdomain.h"
>>> +#include "omap-secure.h"
>>>  
>>>  #define REALTIME_COUNTER_BASE				0x48243200
>>>  #define INCREMENTER_NUMERATOR_OFFSET			0x10
>>> @@ -65,6 +66,7 @@
>>>  
>>>  static struct omap_dm_timer clkev;
>>>  static struct clock_event_device clockevent_gpt;
>>> +unsigned long arch_timer_freq;
>>>  
>>>  static irqreturn_t omap2_gp_timer_interrupt(int irq, void *dev_id)
>>>  {
>>> @@ -546,7 +548,11 @@ static void __init realtime_counter_init(void)
>>>  	reg |= den;
>>>  	__raw_writel(reg, base + INCREMENTER_DENUMERATOR_RELOAD_OFFSET);
>>>  
>>> +	arch_timer_freq = (rate / den) * num;
>>> +
>>>  	iounmap(base);
>>> +
>>> +	omap_smc1(OMAP5_DRA7_MON_SET_CNTFRQ_INDEX, arch_timer_freq);
>> Few problems with this approach..
>>
>> 1. CNTFREQ needs to be programmed on all CPUs. Even though arch-timer
>> code uses boot-cpu now thats not safe for guest which can start on
>> secondary CPUs. So please update this in secondary boot path as well.
> The patch is doing this for secondary cpu's as well in the
> call back  omap4_secondary_init.
>
NM just pointed out this off-list.

 
>> 2. When you power cycle CPU, you will loose the value of this register
>> so you need to reprogram them on every CPU power up. Both DRA and OMAP5
>> doesn't support that part yet but do remember to patch that when
>> such support gets added.
>  oh ok. did not realize this. Will add the change for this then.
Since PM isn't supported for now, the patch looks alright with ES1.0
support gets dropped. Just fix the change-log.

Feel free to add my ack on updated version.

Regards,
Santosh