amqp module causes syslog-ng (3.4 and 3.5) to segfault about syslog-ng

[bazsi]

It seems to have crashed because the username is unset.

It shouldnt crash in this case so its definitely a bug.

Bazsi

Steve Berryman [email protected] wrote:

destination config here:

amqp(
vhost("/")
host("rabbitmq-0.internal.example.com")
port(5672)
exchange("raw_logs")
exchange_type("fanout")
routing_key("")
body("")
persistent(yes)
value-pairs(
scope("selected-macros" "nv-pairs" "sdata")
)
);

Backtrace here:

Core was generated by `syslog-ng -dev -F --enable-core'.
Program terminated with signal 11, Segmentation fault.
#0 0x00007f1b1a65099f in ?? () from /lib/x86_64-linux-gnu/libc.so.6
(gdb) backtrace
#0 0x00007f1b1a65099f in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#1 0x00007f1b171b4bd2 in sasl_response (args=0x7f1b1b27da78,
method=AMQP_SASL_METHOD_PLAIN,
pool=0x7f1b10002040) at
../../../../../modules/afamqp/rabbitmq-c/librabbitmq/amqp_socket.c:149
#2 amqp_login_inner (vl=0x7f1b1b27da78,
sasl_method=AMQP_SASL_METHOD_PLAIN, heartbeat=0, frame_max=131072,
channel_max=0, state=0x7f1b10002000)
at
../../../../../modules/afamqp/rabbitmq-c/librabbitmq/amqp_socket.c:432
#3 amqp_login (state=0x7f1b10002000, vhost=0x152d9b0 "/",
channel_max=, frame_max=131072,
heartbeat=0, sasl_method=AMQP_SASL_METHOD_PLAIN)
at
../../../../../modules/afamqp/rabbitmq-c/librabbitmq/amqp_socket.c:517
#4 0x00007f1b171acb72 in afamqp_dd_connect (self=0x15235a0,
reconnect=)
at ../../modules/afamqp/afamqp.c:334
#5 0x00007f1b1ae18be8 in log_threaded_dest_driver_worker_thread_main
(arg=0x15235a0)
at ../../lib/logthrdestdrv.c:57
#6 0x00007f1b1ae1e22a in worker_thread_func (st=0x1514e20) at
../../lib/misc.c:586
#7 0x00007f1b1ab5c9b5 in ?? () from
/lib/x86_64-linux-gnu/libglib-2.0.so.0
#8 0x00007f1b1a8dde9a in start_thread () from
/lib/x86_64-linux-gnu/libpthread.so.0
#9 0x00007f1b1a60a3fd in clone () from /lib/x86_64-linux-gnu/libc.so.6
#10 0x0000000000000000 in ?? ()

---

Reply to this email directly or view it on GitHub:
#80

[bazsi]

@algernon can you have a look pls? thanks.

[steveberryman]

I can confirm that setting a username/password in the destination stops
the segfaulting :)

Balazs Scheidler wrote:

@algernon https://github.com/algernon can you have a look pls? thanks.

—
Reply to this email directly or view it on GitHub
#80 (comment).

[algernon]

I can't easily fix it to work without a username/password pair, but I can make those two mandatory, so I'll do that.

[algernon]

Merged to 3.5/master, closing here. #107 covers the fix for the master branch. Thanks for the report!