MEMDBG=0 turns on f_malloc, disabling q_malloc. The token F_MALLOC
in the kamailio -V output shows f_malloc is in use.
q_malloc is more suitable for debugging, as it keep trace of
location in the source code that allocates/frees the memory chunks,
at the expense of a bit more memory usage due to tracking overhead.
q_malloc was the default for 3.1, that's why I asked because the
trace showed lines from f_malloc c code.
What is the size of shared memory you start kamailio? Is the default
32MB or you give a different -m parameter value? Was the serve very
loaded? How many such situations did you get so far?
Daniel, I don't remember to enable f_malloc... Actually I
do not know the difference between f_malloc and g_malloc to
choose compile kamailio with f_malloc instead of g_malloc.
My commands to compile kamailio were:
$ make PREFIX="/home2/local/kamailio" FLAVOUR=kamailio
include_modules="db_postgres tls" MEMDBG=0 cfg
$ make all
$ make install
If you need more informations, I
am available...
A minor release for 3.3 branch is going to be
soon, being there makes sure it will be reviewed
to see if it applies.
Cheers,
Daniel
On 7/27/12 3:00 PM, Bruno Bresciani
wrote:
Hi Users,
This is my ultimate try to ask for some
help to debug the two cores generated by
fm_realloc() function. If somebody can
give me a feedback about my questions or
doubts I will be very grateful, this user
list of kamailio always help me to solve
my problems.
send also the
version
'kamailio -V' to
match the proper
sources.
Cheers,
Daniel
On
7/17/12 7:11
PM, Bruno
Bresciani
wrote:
Hi all,
I observed two
at 0x08173bf0
in fm_realloc
(qm=0xb5f7d000,
p=0xb61f0bc0,
size=0) at
mem/f_malloc.c:536.
I don't know
exactly what
may be caused
this core
because the
debug of
kamailio is
disabled, but
looking at
backtrace
(gdb) and
source code I
could notice a
relationship
with TLS
connection
establishment
(handshake).
Other
important fact
is that core
was generated
by the kamalio
itself with
signal 6
(abort)
because it
can't find the
pointer to
memory
realloc.
I'd like to
know if
someone
already
observed this
issue and if
my analyse
above is
correct?