Incoming SMS on a SIM800C USB modem will crash

[ocalvo]

When using SMS callbacks with a SIM800C modem, the call to ReadDevice crash the process:

gammu/src/gammu.c:430

Fatal Python error: Segmentation fault

Current thread 0x00007fd295440700 (most recent call first):
  File "/mnt/s/dd/Repos/ha/core/homeassistant/components/sms/gateway.py", line 44 in sms_pull
  File "/mnt/s/dd/Repos/ha/core/venv/lib/python3.8/site-packages/python_gammu-3.2.3-py3.8-linux-x86_64.egg/gammu/worker.py", line 247 in run
  File "/usr/lib/python3.8/threading.py", line 932 in _bootstrap_inner
  File "/usr/lib/python3.8/threading.py", line 890 in _bootstrap

(gdb) backtrace
#0  0x00000000005f26bb in PyObject_Call ()
#1  0x00007fffe12601fb in CheckIncomingEvents (sm=0x7fffe061c030) at gammu/src/gammu.c:430
#2  0x00007fffe1262790 in StateMachine_ReadDevice (self=0x7fffe061c030, args=<optimized out>, kwds=<optimized out>)
    at gammu/src/gammu.c:893
#3  0x0000000000504939 in ?? ()
#4  0x000000000056acb6 in _PyEval_EvalFrameDefault ()
#5  0x000000000050b6fe in ?? ()
#6  0x000000000056fb87 in _PyEval_EvalFrameDefault ()
#7  0x00000000005f5956 in _PyFunction_Vectorcall ()
#8  0x000000000056acb6 in _PyEval_EvalFrameDefault ()
#9  0x00000000005f5956 in _PyFunction_Vectorcall ()
#10 0x000000000056acb6 in _PyEval_EvalFrameDefault ()
#11 0x00000000005f5956 in _PyFunction_Vectorcall ()
#12 0x000000000050b7f8 in ?? ()
#13 0x00000000005f2702 in PyObject_Call ()
#14 0x000000000065578c in ?? ()
#15 0x0000000000675d48 in ?? ()
#16 0x00007fffff586609 in start_thread (arg=<optimized out>) at pthread_create.c:477
#17 0x00007fffff6c2293 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
(gdb)

More info

(gdb) print *sm

$4 = {ob_base = {ob_refcnt = 5, ob_type = 0x7fffe127dca0 <StateMachineType>}, s = 0x7fffe0020010, DebugFile = 0x0,
  IncomingCallback = 0x0, SMSStatus = 0, MessageReference = 0, IncomingCallQueue = {0x0 <repeats 11 times>},
  IncomingSMSQueue = {0x0 <repeats 11 times>}, IncomingCBQueue = {0x0 <repeats 11 times>}, IncomingUSSDQueue = {
    0x0 <repeats 11 times>}, memory_entry_cache_type = 0, memory_entry_cache = 1, todo_entry_cache = 1,
  calendar_entry_cache = 1, mutex = 0x19534e0}

[nijel]

Having stack trace from the segfault would help. Most likely this is actually an issue in the Gammu library.

[ocalvo]

@nijel Where are the dumps? I have a system where I can repro this, but I dont know how to get the dump of stack trace.

[ocalvo]

I know how to get a dump, but I can only do it from my production system, but that system contains too many secrets. I will try with a VM over the weekend.

[nijel]

See https://wiki.archlinux.org/title/Debugging/Getting_traces#Getting_the_trace for guide to get trace from the core dump.

[ocalvo]

See https://wiki.archlinux.org/title/Debugging/Getting_traces#Getting_the_trace for guide to get trace from the core dump.

I am using WSL 1.0 and it does not support core dumps. However I updated the issue with the callback info.

Gamma.c is referencing the null pointer of the callback function.

[nijel]

Looking at https://github.com/home-assistant/core/pull/56552, the problem is probably in using Gammu functions from the callback, what is not safe. I'm not sure if this is somewhere documented, but it can definitely cause problems.