MSP432-RTOS: TI-RTOS scheduler locks/freezes after I2C slave interrupt

David Alessio

Part Number: MSP432-RTOS
Other Parts Discussed in Thread: SYSBIOS

Hello,

I'm running TI-RTOS (SYSBIOS) on an MSP432. I have a few threads configured to periodically poll. i.e.:

while (1) {
    do_something();     // toggle my LED
    Task_sleep(250);    // 1000 Hz tick, so wait 250ms
}

and another thread in a similar while (1) loop, but this one waiting on I2CSlave_Read(); Polling thread run fine until the I2C Slave thread wakes up, does its thing and comes around to I2CSlave_Read() again at which point all other threads cease to run/wake up.

I'm using GCC 9.2.1 (as per the many sample progs), compiled -O0. The thread are started "statically" via the .cfg file, but that shouldn't matter--the threads are created and start. Stack crashing is enabled and silent, I've increased stack size to 6144 for thread and 4096 for idle.

What's going on here? What's wrong with TI-RTOS?

Regards,

over 4 years ago

0 Allen Gao over 4 years ago

TI__Expert 7080 points

Hi David,

Does periodically poll work normally without I2C function? If so, you may check with the I2C threads configuration.

0 David Alessio over 4 years ago in reply to Allen Gao

Prodigy 160 points

Hi Allen,

Yes, of course the threads work fine. What do you suggest I "check with the I2C thread's configuration"? It's a thread, configured and started just as the other two threads.

I suspect there's a problem in the implementation of the I2CSlave_read(), I2CSlave_write(), and their underlying support functions. Or the way in which I2CSLAVE_MODE_BLOCKING is implemented.

Please advise.

0 Allen Gao over 4 years ago in reply to David Alessio

TI__Expert 7080 points

Hi David,

If you suspect underlying, you could try to note the I2CSlave_read() or I2CSlave_write() function. Whether it works well.

0 David Alessio over 4 years ago in reply to Allen Gao

Prodigy 160 points

Allen,

I'm not sure I understand your suggestion. Of course I2CSlave_read() and I2CSlave_write() work in that the thread receives and replies to an I2C master. If not, that would be the issue I report. The problem is that they seem to have an undesirable side effect in that returning from I2CSlave_read() and/or calling I2CSlave_write() freezes scheduling -- other threads fail to wake up from Task_sleep().

0 Allen Gao over 4 years ago in reply to David Alessio

TI__Expert 7080 points

I mean that could you try to mask these function, if you don't call I2C read or write in this threads, dose other threads fail?

0 David Alessio over 4 years ago in reply to Allen Gao

Prodigy 160 points

No, the other threads work correctly -- they return from Task_sleep() as and when expected.

0 David Alessio over 4 years ago in reply to David Alessio

Prodigy 160 points

Has anyone tested I2CSlave_read() and I2CSlave_write() in a multi-threaded environment? i.e. with other threads running at different priorities?

0 Allen Gao over 4 years ago in reply to David Alessio

TI__Expert 7080 points

Hi David，

Could you provide your code, I could review it.

0 David Alessio over 4 years ago in reply to Allen Gao

Prodigy 160 points

Hi, Allen,

I'll create a test case for the MSP-EXP432P401 Dev Kit. In the meantime please provide me your email address (I can not post proprietary code to a public forum).

Regards,

-david

0 Allen Gao over 4 years ago in reply to David Alessio

TI__Expert 7080 points

Hi David,

I will add you and send you private messages.

Arm-based microcontrollers

Arm-based microcontrollers forum

MSP432-RTOS: TI-RTOS scheduler locks/freezes after I2C slave interrupt