MQX Interrupt Shared Peripheral Routing

取消
显示结果 
显示  仅  | 搜索替代 
您的意思是: 

MQX Interrupt Shared Peripheral Routing

1,747 次查看
carlovalgocela
Contributor III

Reading the Vybrid AN4947.pdf,  and in 4.2 Dual-core communication and synchronization options,

• Interrupts. Routing options for both cores — none, one, or both cores can be notified for the event.

I wanted for  M4 MQX NVIC Interrupt 10 and 11 not to be shared on the A5 Linux.

I use this code to configure the routing.

/* Disable shared interrupt to A5 Core */

    _irouter_routing(26,0,0);

    _irouter_routing(27,0,0);

However it is not working. A5 Linux could still received and processed the interrupts.

Any ideas what I missed?

interrupt.png

I'm using MQX Version 4.1.1

Thanks,

Carlo Val

标签 (3)
0 项奖励
回复
7 回复数

1,392 次查看
karina_valencia
NXP Apps Support
NXP Apps Support

alejandrolozano​ are you able to help here?

0 项奖励
回复

1,390 次查看
alejandrolozan1
NXP Employee
NXP Employee

Hi,

I'd like to reproduce the behavior first. Carlo, which Linux BSP are you using?

Best Regards,

Alejandro

0 项奖励
回复

1,390 次查看
carlovalgocela
Contributor III

I'm using Linux version 3.0.15-ts-armv7l .

I saw yesterday in a file init_gic.c  routing all interrupts to Core 1.

Ill try it today and see if it will work

Here is my update would be:

volatile uint16_t *ptr26;

volatile uint16_t *ptr27;

ptr26 = &MSCM_IRSPRC26;

*ptr26 = MSCM_IRSPRC_CP1En_MASK; /* route interrupts to core 1*/

ptr27 = &MSCM_IRSPRC27;

*ptr27 = MSCM_IRSPRC_CP1En_MASK; /* route interrupts to core 1*/

0 项奖励
回复

1,390 次查看
alejandrolozan1
NXP Employee
NXP Employee

Hi Carlo,

Did you modification work?

Regards,

Alejandro

0 项奖励
回复

1,390 次查看
carlovalgocela
Contributor III

Hello alejandrolozano​,

No it didn't work :-(

0 项奖励
回复

1,390 次查看
alejandrolozan1
NXP Employee
NXP Employee

Hi,

On Linux side you could try to disable the interrupt routed to the A5 corresponding to the DMA.

void mvf_init_irq(void)

{

        unsigned int i;

        void __iomem *int_router_base =

                                MVF_IO_ADDRESS(MVF_MSCM_INT_ROUTER_BASE);

        struct irq_desc *desc;

        unsigned short route;

        /* start offset if private timer irq id, which is 29.

         * ID table:

         * Global timer, PPI -> ID27

         * A legacy nFIQ, PPI -> ID28

         * Private timer, PPI -> ID29

         * Watchdog timers, PPI -> ID30

         * A legacy nIRQ, PPI -> ID31

         */

        gic_init(0, 27, MVF_IO_ADDRESS(MVF_INTD_BASE_ADDR),

                MVF_IO_ADDRESS(MVF_SCUGIC_BASE_ADDR + 0x100));

        for (i = MXC_INT_START; i <= MXC_INT_END; i++) {

                desc = irq_to_desc(i);

                desc->irq_data.chip->irq_set_wake = mvf_gic_irq_set_wake;

        }

        mvf_register_gpios();

        int_router_base += 0x80;

        for (i = 0; i < 112; i++) {

                route = 1 | __raw_readw(int_router_base);

                __raw_writew(route, int_router_base);

                int_router_base += 2;

        }

}

This function is found at vim arch/arm/mach-mvf/irq.c

Let me debug the M4 (MQX side) to verify the possible problem.

Best Regards,

Alejandro

1,390 次查看
alejandrolozan1
NXP Employee
NXP Employee

Hi Carlo,

Did you try the previous suggestion?

Best Regards,

Alejandro

0 项奖励
回复