[PATCH] irqchip/gic: Only allow the primary GIC to set the CPU map

[Jon Hunter]

On 30/07/15 16:05, Jon Hunter wrote:
> 
> On 30/07/15 15:33, Marc Zyngier wrote:
>> On 30/07/15 15:11, Jon Hunter wrote:
>>> The gic_init_bases() function initialises an array that stores the mapping
>>> between the GIC and CPUs. This array is a global array that is
>>> unconditionally initialised on every call to gic_init_bases(). Although,
>>> it is not common for there to be more than one GIC instance, there are
>>> some devices that do support nested GIC controllers and gic_init_bases()
>>> can be called more than once.
>>>
>>> A 2nd call to gic_init_bases() will clear the previous CPU mapping and
>>> will only setup the mapping again for the CPU calling gic_init_bases().
>>> Fix this by only allowing the CPU map to be configured for the primary GIC.
>>>
>>> For secondary GICs the CPU map is not relevant because these GICs do not
>>> directly route the interrupts to the main CPU(s) but to other GICs or
>>> devices.
>>>
>>> Signed-off-by: Jon Hunter <jonathanh at nvidia.com>
>>> ---
>>> This is a follow-up to the patch titled "irqchip: gic: Add a cpu map for
>>> each GIC instance" and discussed here [1]. Based upon the discussion I have
>>> re-worked and re-titled it approriately.
>>>
>>> [1] http://www.spinics.net/lists/kernel/msg2044421.html
>>>
>>>  drivers/irqchip/irq-gic.c | 43 +++++++++++++++++++++++++------------------
>>>  1 file changed, 25 insertions(+), 18 deletions(-)
>>>
>>> diff --git a/drivers/irqchip/irq-gic.c b/drivers/irqchip/irq-gic.c
>>> index a530d9a9b810..7566fe259d27 100644
>>> --- a/drivers/irqchip/irq-gic.c
>>> +++ b/drivers/irqchip/irq-gic.c
>>> @@ -416,19 +416,26 @@ static void gic_cpu_init(struct gic_chip_data *gic)
>>>  	int i;
>>>  
>>>  	/*
>>> -	 * Get what the GIC says our CPU mask is.
>>> +	 * Setting up the CPU map is only relevant for the primary GIC
>>> +	 * because any nested/secondary GICs do not directly interface
>>> +	 * with the CPU(s).
>>>  	 */
>>> -	BUG_ON(cpu >= NR_GIC_CPU_IF);
>>> -	cpu_mask = gic_get_cpumask(gic);
>>> -	gic_cpu_map[cpu] = cpu_mask;
>>> +	if (gic == &gic_data[0]) {
>>> +		/*
>>> +		 * Get what the GIC says our CPU mask is.
>>> +		 */
>>> +		BUG_ON(cpu >= NR_GIC_CPU_IF);
>>> +		cpu_mask = gic_get_cpumask(gic);
>>> +		gic_cpu_map[cpu] = cpu_mask;
>>>  
>>> -	/*
>>> -	 * Clear our mask from the other map entries in case they're
>>> -	 * still undefined.
>>> -	 */
>>> -	for (i = 0; i < NR_GIC_CPU_IF; i++)
>>> -		if (i != cpu)
>>> -			gic_cpu_map[i] &= ~cpu_mask;
>>> +		/*
>>> +		 * Clear our mask from the other map entries in case they're
>>> +		 * still undefined.
>>> +		 */
>>> +		for (i = 0; i < NR_GIC_CPU_IF; i++)
>>> +			if (i != cpu)
>>> +				gic_cpu_map[i] &= ~cpu_mask;
>>> +	}
>>>  
>>>  	gic_cpu_config(dist_base, NULL);
>>>  
>>> @@ -977,13 +984,6 @@ void __init gic_init_bases(unsigned int gic_nr, int irq_start,
>>>  	}
>>>  
>>>  	/*
>>> -	 * Initialize the CPU interface map to all CPUs.
>>> -	 * It will be refined as each CPU probes its ID.
>>> -	 */
>>> -	for (i = 0; i < NR_GIC_CPU_IF; i++)
>>> -		gic_cpu_map[i] = 0xff;
>>> -
>>> -	/*
>>>  	 * Find out how many interrupts are supported.
>>>  	 * The GIC only supports up to 1020 interrupt sources.
>>>  	 */
>>> @@ -1028,6 +1028,13 @@ void __init gic_init_bases(unsigned int gic_nr, int irq_start,
>>>  		return;
>>>  
>>>  	if (gic_nr == 0) {
>>> +		/*
>>> +		 * Initialize the CPU interface map to all CPUs.
>>> +		 * It will be refined as each CPU probes its ID.
>>> +		 * This is only necessary for the primary GIC.
>>> +		 */
>>> +		for (i = 0; i < NR_GIC_CPU_IF; i++)
>>> +			gic_cpu_map[i] = 0xff;
>>>  #ifdef CONFIG_SMP
>>>  		set_smp_cross_call(gic_raise_softirq);
>>>  		register_cpu_notifier(&gic_cpu_notifier);
>>>
>>
>> Looks good.
>>
>> I think there is a another bug caused by 322895062 ("irqchip: gic:
>> Preserve gic V2 bypass bits in cpu ctrl register"), where
>> gic_cpu_if_up() only acts on the primary GIC. In the secondary GIC case,
>> it will stay disabled.
>>
>> Mind fixing this while you're at it?
> 
> Ah yes, I see. Ok, I will resend this with the other fix as a series.

Hmmm, what about gic_cpu_if_down()? Looks like the only user is
vexpress-tc2. Ideally it should pass the gic_nr. I could make it pass 0
by default.

Jon