[Kamailio-Users] Core dump on writing dialog profile size into script var.

[Alex Balashov]

I also have another interesting problem with the aforementioned 
configuration (http://pastebin.com/f28051a5).

When I write the dialog profile size into an AVP, it works fine.

When I write it into a script var, i.e. replace $avp(S:dlg_sz) with 
$var(dlg_sz), it crashes:

Relaying INVITE from 210.23.22.23 to sip:2122222322 at 210.23.22.23:5060
[D] Added new dialog for 7709600101
[ONREPLY-ROUTE 1] Provisional reply 100 received.
[ONREPLY-ROUTE 1] 200 OK received for 7709600101
[ONREPLY-ROUTE 1] INVITE/200 is part of dialog for 7709600101
Segmentation fault (core dumped)

GDB reveals:

Program received signal SIGSEGV, Segmentation fault.
0xb7e7f2f8 in strncpy () from /lib/i686/nosegneg/libc.so.6
(gdb) where
#0  0xb7e7f2f8 in strncpy () from /lib/i686/nosegneg/libc.so.6
#1  0x080a4a07 in set_var_value (var=0x8188d98, value=0xbfae5f28, 
flags=<value o
     at script_var.c:122
#2  0xb7d6c4fe in w_get_profile_size (msg=0x81902b8, profile=0xb5ede180 
"\034??\
     value=0x818b6b8 "@?\030\b", result=0x818b6f8 "N") at dialog.c:668
#3  0x08054f15 in do_action (a=0x818b850, msg=0x81902b8) at action.c:850
#4  0x08053ed2 in run_action_list (a=0x818b5c8, msg=0x81902b8) at 
action.c:138
#5  0x08056365 in do_action (a=0x818bab8, msg=0x81902b8) at action.c:722
#6  0x08053ed2 in run_action_list (a=0x818b178, msg=0x81902b8) at 
action.c:138
#7  0x080572c2 in run_top_route (a=0x818b178, msg=0x81902b8) at action.c:118
#8  0xb7de3a64 in reply_received (p_msg=0x81902b8) at t_reply.c:1361
#9  0x08064793 in forward_reply (msg=0x81902b8) at forward.c:507
#10 0x08090d5b in receive_msg (
     buf=0x81600e0 "SIP/2.0 200 OK\r\nVia: SIP/2.0/UDP 
215.22.22.22;branch=z9hG4
08.52.173.18\r\nVia: SIP/2.0/UDP 
198.225.86.10:5060;branch=z9hG4bKragjo0207gm0dc
p:208.52."..., len=814, rcv_info=0xbfae65d4) at receive.c:203
#11 0x080cccfb in udp_rcv_loop () at udp_server.c:449
#12 0x0806b78d in main (argc=1, argv=0xbfae6764) at main.c:693

It would seem to me that there is some sort of buffer overflow issue 
that results in the garbage seen above.

Not sure that it makes a difference, but the glibc being linked against 
is a Xen-safe one that disables TLS functionality.  This is all running 
inside a Xen DomU.

-- 
Alex Balashov
Evariste Systems
Web    : http://www.evaristesys.com/
Tel    : (+1) (678) 954-0670
Direct : (+1) (678) 954-0671
Mobile : (+1) (706) 338-8599