Issue with pin interrupt?

this.wiederkehr

I have an app which reads values from adxl362 and saves theme to the sd card. In order to know when to fetch data from the fifo in the adxl362 I am using pin interrupts. This interrupt triggers approximatly 3 times a second.

Additionally I have two slow timers and an other pininterrupt running.

The app fails approximatly all 60 second with the following guru mediation error:

Guru Meditation Error of type LoadProhibited occurred on core  0. Exception was unhandled.
Register dump:
PC      : 0x400eb244  PS      : 0x00060230  A0      : 0x800eb298  A1      : 0x3ffe1f50  
0x400eb244: mp_call_function_n_kw at pycom-micropython-sigfox/esp32/../py/runtime.c:562

A2      : 0x3ffe4560  A3      : 0x00000001  A4      : 0x00000000  A5      : 0x3ffe1f70  
A6      : 0x00050223  A7      : 0x00000000  A8      : 0x800eb244  A9      : 0x3ffe1f30  
A10     : 0x00000001  A11     : 0x00060223  A12     : 0x00060220  A13     : 0x3ffc5b20  
A14     : 0x00000003  A15     : 0x00060223  SAR     : 0x0000001a  EXCCAUSE: 0x0000001c  
EXCVADDR: 0x00000011  LBEG    : 0x400014fd  LEND    : 0x4000150d  LCOUNT  : 0xfffffffe  

Backtrace: 0x400eb244:0x3ffe1f50 0x400eb298:0x3ffe1f70 0x400dcca8:0x3ffe1fa0 0x400e1656:0x3ffe1fc0 0x400d8234:0x3ffe2050
0x400eb244: mp_call_function_n_kw at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400eb298: mp_call_function_1 at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400dcca8: pin_interrupt_queue_handler at pycom-micropython-sigfox/esp32/mods/machpin.c:300

0x400e1656: TASK_Interrupts at pycom-micropython-sigfox/esp32/util/mpirq.c:58

0x400d8234: freertos_entry at pycom-micropython-sigfox/esp32/mpthreadport.c:129

In the background the ftp server is running. By clicking update in an ftp app I can force the error to trigger much faster ...

Any ides? Am I doing something wrong or is this a bug?

this.wiederkehr

Ah yea, my callback functions are very easy instance methods which just do:

self.todo = 1

In the main loop i check for these flags and reset theme. at the end of the loop i call

machine.idle()

Interestingly, if I disable all the other callbacks except for the one related to the adxl362 I get:

Unhandled exception in callback handler
TypeError: 'slice' object is not callable

Even tough there is no slice object in the callback involved.

After that the program freezes. No further interrupts.

I'm not runnin out of memory: In the main loop I call gc.collect whenever there is less than 10000bytes available. At that point it goes up to around 50000 bytes.

this.wiederkehr

Sometimes I also get the following error:

Guru Meditation Error of type LoadProhibited occurred on core  0. Exception was unhandled.
Register dump:
PC      : 0x4019abec  PS      : 0x0006063f  A0      : 0x800f72a9  A1      : 0x3ffc5a90  
0x4019abec: mp_decode_uint at pycom-micropython-sigfox/esp32/../py/bc.c:50 (discriminator 1)

A2      : 0x3ffc5afc  A3      : 0x3f409d60  A4      : 0x3f40b6b8  A5      : 0x3f40b6b8  
A6      : 0x00000000  A7      : 0x40000000  A8      : 0x0023bccf  A9      : 0x00000000  
A10     : 0x0023bcd0  A11     : 0x00000019  A12     : 0x3f4078f8  A13     : 0x0000e79e  
A14     : 0x3f4078f8  A15     : 0x00000000  SAR     : 0x0000001a  EXCCAUSE: 0x0000001c  
EXCVADDR: 0x0023bccf  LBEG    : 0x400e3da4  LEND    : 0x400e3dd0  LCOUNT  : 0x00000eb3  
0x400e3da4: gc_realloc at pycom-micropython-sigfox/esp32/../py/gc.c:668

0x400e3dd0: gc_realloc at pycom-micropython-sigfox/esp32/../py/gc.c:685


Backtrace: 0x4019abec:0x3ffc5a90 0x400f72a9:0x3ffc5ab0 0x400eed03:0x3ffc5b50 0x400eb254:0x3ffc5b90 0x400eb2bc:0x3ffc5bb0 0x400f6e30:0x3ffc5bd0 0x400eed03:0x3ffc5c70 0x400eb254:0x3ffc5cb0 0x400eb281:0x3ffc5cd0 0x400d8c92:0x3ffc5cf0 0x400d8f61:0x3ffc5d90 0x400d7f8c:0x3ffc5db0
0x4019abec: mp_decode_uint at pycom-micropython-sigfox/esp32/../py/bc.c:50 (discriminator 1)

0x400f72a9: mp_execute_bytecode atpycom-micropython-sigfox/esp32/../py/vm.c:1309

0x400eed03: fun_bc_call at pycom-micropython-sigfox/esp32/../py/objfun.c:276

0x400eb254: mp_call_function_n_kw at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400eb2bc: mp_call_method_n_kw at pycom-micropython-sigfox/esp32/../py/runtime.c:591

0x400f6e30: mp_execute_bytecode at pycom-micropython-sigfox/esp32/../py/vm.c:980

0x400eed03: fun_bc_call at pycom-micropython-sigfox/esp32/../py/objfun.c:276

0x400eb254: mp_call_function_n_kw atpycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400eb281: mp_call_function_0 at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400d8c92: parse_compile_execute at pycom-micropython-sigfox/esp32/../lib/utils/pyexec.c:534

0x400d8f61: pyexec_file at pycom-micropython-sigfox/esp32/../lib/utils/pyexec.c:505

0x400d7f8c: TASK_Micropython at pycom-micropython-sigfox/esp32/mptask.c:243

or this one here:

Guru Meditation Error of type LoadStoreAlignment occurred on core  0. Exception was unhandled.
Register dump:
PC      : 0x400ec3d9  PS      : 0x0006003f  A0      : 0x800eb04d  A1      : 0x3ffc5b60  
0x400ec3d9: mp_map_lookup at pycom-micropython-sigfox/esp32/../py/map.c:152

A2      : 0x40370027  A3      : 0x000002ae  A4      : 0x00000000  A5      : 0x3ffe4040  
A6      : 0x40370027  A7      : 0x3ffc69bc  A8      : 0x800d82f4  A9      : 0x3ffc5b50  
A10     : 0x3ffc66bc  A11     : 0x00000019  A12     : 0x0006002f  A13     : 0x3ffe3da0  
A14     : 0x00000378  A15     : 0x00000000  SAR     : 0x00000002  EXCCAUSE: 0x00000009  
EXCVADDR: 0x40370027  LBEG    : 0x4000c2e0  LEND    : 0x4000c2f6  LCOUNT  : 0x00000000  

Backtrace: 0x400ec3d9:0x3ffc5b60 0x400eb04d:0x3ffc5b90 0x400eb0b4:0x3ffc5bb0 0x400f6356:0x3ffc5bd0 0x400eed03:0x3ffc5c70 0x400eb254:0x3ffc5cb0 0x400eb281:0x3ffc5cd0 0x400d8c92:0x3ffc5cf0 0x400d8f61:0x3ffc5d90 0x400d7f8c:0x3ffc5db0
0x400ec3d9: mp_map_lookup at pycom-micropython-sigfox/esp32/../py/map.c:152

0x400eb04d: mp_load_global at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400eb0b4: mp_load_name at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400f6356: mp_execute_bytecode at pycom-micropython-sigfox/esp32/../py/vm.c:278

0x400eed03: fun_bc_call at pycom-micropython-sigfox/esp32/../py/objfun.c:276

0x400eb254: mp_call_function_n_kw at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400eb281: mp_call_function_0 at pycom-micropython-sigfox/esp32/../py/runtime.c:562

0x400d8c92: parse_compile_execute at pycom-micropython-sigfox/esp32/../lib/utils/pyexec.c:534

0x400d8f61: pyexec_file at pycom-micropython-sigfox/esp32/../lib/utils/pyexec.c:505

0x400d7f8c: TASK_Micropython at pycom-micropython-sigfox/esp32/mptask.c:243

robert-hh

@this.wiederkehr Which version of the firmware are you using? It's always helpful to add the the response from:

import uos
uos.uname()

this.wiederkehr

@robert-hh sorry, forgot to mention.

it's a lopy and I rebuilt everything from the sources checkedout yesterday:
https://github.com/pycom/pycom-micropython-sigfox
https://github.com/pycom/pycom-esp-idf

So, it should be identical to:
(sysname='LoPy', nodename='LoPy', release='1.6.12.b1', version='v1.8.6-593-g8e4ed0fa on 2017-04-12', machine='LoPy with ESP32', lorawan='1.0.0')

But I don't have access to it to check right now.

danielm

It seems that you are having similar issues: https://forum.pycom.io/topic/935/unhandled-exception-in-interrupt-handler
It runs more stable with recent FW versions (tested with 1.6.12.b1) but it crashes anyway.

this.wiederkehr

@danielm: I also read your post. Indeed, it looks similar.

It looks like something is messed with the pointers to python objects... That would also explain the strange exception in the callback. Due to some reason it finds a slice although there should not be a slice.

may be due to the gc reallocating stuff or some error in the thread sync??? Honestly I have no clue ...

When I have time I'll try to run gc and interrupt exclusively and see whether the error persists: basically disabling gc in the main loop and only collect if interrupts are disabled...

this.wiederkehr

Might be totally unrelated but I see:

id(module) stays constant on multiple calls

id(module.Class) stays constant on multiple calls
id(module.Class.method) stays constant on multiple calls

inst = module.Class()

id(inst) stays constant on multiple calls

but

id(inst.method) changes continuously on each call (increasing)

Looks like inst.method is allocated each time we call it. Why is that? Is this the expected behavior?

interestingly you can easily run out of memory:

gc.disable()
while True:
    id(inst.method)

while on all other calls you will never run out of memory...

Just checked with cpython. Doing the above will result in two different ids but you can not exhaust memory like above.

This is all found on 1.6.12.b1

Just wonder if this causes problems

danielm

@daniel
Could you please provide a comment for the post above?

this.wiederkehr

Just to clarify: I checked with pure micropython unix port and it has the same behavior with id(inst.method).

Thinking about it in detail this behavior makes sense. The reason I think is, that its infeasible to hold a reference to each method of each instance in memory (there are possibly thousands of instances). And an instance method is pretty simple just the class method with self set.

In cpython they may have solved this with weakrefs but as long as I know micropython does not have weakrefs.

Anyway, the reason I did look into this is: I suspected that my interrupt handler (which is an instance method) is collected by the gc. The error will not be triggered immediately after the gc run, but only when that specific memory region where the instance.method was located before gets rewritten...
????

I'll try to hold a reference to the handler by assigning it to an instance variable (this should prevent it from being collected by the gc.

def __init__(self):
    self.callback = self.handler
    int_pin.callback(trigger=Pin.IRQ_RISING, handler=self.callback)
    

def handler(self):
    # do whatever has to be done =)

danielm

@this-wiederkehr
Did referencing of the handler help to prevent those issues?

this.wiederkehr

@danielm said in Issue with pin interrupt?:

Did referencing of the handler help to prevent those issues?

Actually I don't remember the outcome of this test. But it must have been unsatisfying.

Anyway, just found the following commit in github:

https://github.com/danicampora/pycom-micropython-sigfox/commit/d78bec2191358e3392f878392a19b1d9e3f002bc

Looks like this is exactly about the issue I ecountered. Going to compile this myself and test...

Explore Pybytes | Official Documentation | Report a Firmware Bug/Issue | GitHub

Issue with pin interrupt?

Pycom on Twitter