fix use-after-free: require new fsm deferred dealloc, check for term

API doc: require osmo_fsm_set_dealloc_ctx().

mgcp_client during delete: do not reparent the FSM when it is already
terminating.

I have recently discovered a vulnerability: if an endpoint FSM deallocates
during event handling of a successful MGCP response, this causes a
use-after-free; and once that is fixed, a state change on the already
terminated FSM causes a pointer corruption by using already cleaned data
structures. osmo_fsm_set_dealloc_ctx() fixes the use-after-free, and
osmo_fsm_set_term_stops_actions() fixes the pointer corruption.

Related: Ib7fce7b7d54dfb87af97544796680919e5929a50 (osmo-bsc),
         I08c03946605aa12e0a5ce8b3c773704ef5327a7a (osmo-msc)
Depends: Ief4dba9ea587c9b4aea69993e965fbb20fb80e78 (libosmocore),
         I0adc13a1a998e953b6c850efa2761350dd07e03a (libosmocore)
Change-Id: I7df2e9202b04e7ca7366bb0a8ec53cf3bb14faf3
This commit is contained in:
Neels Hofmeyr 2019-10-04 22:47:31 +02:00
parent e827831514
commit ca2aec0235
2 changed files with 6 additions and 0 deletions

View File

@ -255,6 +255,9 @@ struct osmo_mgcpc_ep *osmo_mgcpc_ep_fi_mgwep(struct osmo_fsm_inst *fi)
* MGCP messages to set up the endpoint will be sent on the given mgcp_client, as soon as the first
* osmo_mgcpc_ep_ci_request() is invoked.
*
* IMPORTANT: To avoid use-after-free problems, using this FSM requires use of deferred FSM deallocation using
* osmo_fsm_set_dealloc_ctx(), e.g. using osmo_select_main_ctx(OTC_SELECT) with osmo_select_main_ctx() as main loop.
*
* A typical sequence of events would be:
*
* ep = osmo_mgcpc_ep_alloc(..., mgcp_client_rtpbridge_wildcard(client));

View File

@ -708,6 +708,9 @@ void mgcp_conn_delete(struct osmo_fsm_inst *fi)
OSMO_ASSERT(mgcp_ctx);
if (fi->proc.terminating)
return;
/* Unlink FSM from parent */
osmo_fsm_inst_unlink_parent(fi, NULL);