native segfaults when SIGIO occurs while booting (in valgrind)

[LudwigKnuepfer]

Testing:

Employ l2perf from https:/LudwigOrtmann/riot-tools like: echo 'tg 1000 1 1 0' | ./bin/l2perf.elf tap1. While this is running, start the default RIOT project (a few times) until (something like) this happens:

$ valgrind ./bin/default.elf tap0
==27262== Memcheck, a memory error detector
==27262== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al.
==27262== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
==27262== Command: ./bin/default.elf tap0
==27262== 
RIOT native interrupts/signals initialized.
RIOT native uart0 initialized.
LED_GREEN_OFF
LED_RED_ON
RIOT native board initialized.
RIOT native hardware initialization complete.

kernel_init(): This is RIOT!
Scheduler...[OK]
kernel_init(): jumping into first task...
==27262== Conditional jump or move depends on uninitialised value(s)
==27262==    at 0x804B0BE: ??? (tramp.S:40)
==27262==    by 0x804A523: native_irq_handler (irq_cpu.c:279)
==27262==    by 0x410109A: makecontext (in /usr/lib32/libc-2.18.so)
==27262== 
==27262== Use of uninitialised value of size 4
==27262==    at 0x80623A8: ???
==27262== 
==27262== 
==27262== Process terminating with default action of signal 11 (SIGSEGV): dumping core
==27262==  Access not within mapped region at address 0x0
==27262==    at 0x1: ???
==27262==    by 0x804A523: native_irq_handler (irq_cpu.c:279)
==27262==    by 0x410109A: makecontext (in /usr/lib32/libc-2.18.so)
==27262==  If you believe this happened as a result of a stack
==27262==  overflow in your program's main thread (unlikely but
==27262==  possible), you can try to increase the size of the
==27262==  main thread stack using the --main-stacksize= flag.
==27262==  The main thread stack size used in this run was 8388608.
==27262== 
==27262== HEAP SUMMARY:
==27262==     in use at exit: 0 bytes in 0 blocks
==27262==   total heap usage: 2 allocs, 2 frees, 104 bytes allocated
==27262== 
==27262== All heap blocks were freed -- no leaks are possible
==27262== 
==27262== For counts of detected and suppressed errors, rerun with: -v
==27262== Use --track-origins=yes to see where uninitialised values come from
==27262== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 0 from 0)
sh: segmentation fault (core dumped)  valgrind ./bin/default.elf tap0

I was not able to reproduce this either without valgrind or in gdb although I did not try very hard.

[LudwigKnuepfer]

apparently possible without valgrind as well: #498