Comment # 5 on bug 942297 from Takashi Iwai

Hmm, this looks rather like a deadlock of mutex in glibc's alloc/free
functions.  If I understand the backtrace correctly, the signal handler got
called during realloc() within a mutex lock, and then it tires to malloc the
memory again, which takes the mutex lock again.  (Also there was another free()
call in the middle, which also takes the mutex, too...)