Changeset 729 for pjproject/trunk/pjsip/src/pjsip/sip_ua_layer.c

Timestamp:

Sep 19, 2006 1:37:53 PM (18 years ago)

Author:

bennylp

Message:

Fixed race-condition/deadlock problems in the dialog/user agent layer
all the way up to PJSUA-API:

• standardized locking order: dialog then user agent, and dialog then PJSUA
• any threads that attempt to acquire mutexes in different order than above MUST employ retry mechanism (for an example, see acquire_call() in pjsua_call.c). This retry mechanism has also been used in the UA layer (sip_ua_layer.c) since it needs to lock user agent layer first before the dialog.
• introduced pjsip_dlg_try_inc_lock() and PJSUA_TRY_LOCK() to accomodate above.
• pjsua tested on Quad Xeon with 4 threads and 200 cps, so far so good.

File:

• pjproject/trunk/pjsip/src/pjsip/sip_ua_layer.c (modified) (9 diffs)

pjproject/trunk/pjsip/src/pjsip/sip_ua_layer.c

@@ -210 +210 @@
 
 
-/**
- * Lock the dialog's hash table. This function is normally called by
- * dialog code only.
- *
- * @return              PJ_SUCCESS on success or the appropriate error code.
- */
-PJ_DEF(pj_status_t) pjsip_ua_lock_dlg_table(void)
-{
-    return pj_mutex_lock(mod_ua.mutex);
-}
-
-
-/**
- * Unlock the dialog's hash table. This function is normally called by
- * dialog code only.
- *
- * @return              PJ_SUCCESS on success or the appropriate error code.
- */
-PJ_DEF(pj_status_t) pjsip_ua_unlock_dlg_table(void)
-{
-    return pj_mutex_unlock(mod_ua.mutex);
-}
-
-
 /*
  * Get the endpoint where this UA is currently registered.
@@ -500 +476 @@
     pj_str_t *from_tag;
     pjsip_dialog *dlg;
+    pj_status_t status;
 
     /* Optimized path: bail out early if request is not CANCEL and it doesn't
@@ -509 +486 @@
         return PJ_FALSE;
     }
+
+retry_on_deadlock:
 
     /* Lock user agent before looking up the dialog hash table. */
@@ -586 +565 @@
     rdata->endpt_info.mod_data[mod_ua.mod.id] = dlg;
 
-    /* Lock the dialog */
+    /* Try to lock the dialog */
     PJ_LOG(6,(dlg->obj_name, "UA layer acquiring dialog lock for request"));
-    pjsip_dlg_inc_lock(dlg);
+    status = pjsip_dlg_try_inc_lock(dlg);
+    if (status != PJ_SUCCESS) {
+        /* Failed to acquire dialog mutex immediately, this could be 
+         * because of deadlock. Release UA mutex, yield, and retry 
+         * the whole thing once again.
+         */
+        pj_mutex_unlock(mod_ua.mutex);
+        pj_thread_sleep(0);
+        goto retry_on_deadlock;
+    }
+
+    /* Done with processing in UA layer, release lock */
+    pj_mutex_unlock(mod_ua.mutex);
 
     /* Pass to dialog. */
@@ -596 +587 @@
     pjsip_dlg_dec_lock(dlg);
 
-    /* Done processing in the UA */
-    pj_mutex_unlock(mod_ua.mutex);
-
     /* Report as handled. */
     return PJ_TRUE;
@@ -610 +598 @@
     pjsip_transaction *tsx;
     struct dlg_set *dlg_set;
-    pjsip_dialog *dlg = NULL;
+    pjsip_dialog *dlg;
+    pj_status_t status;
 
     /*
@@ -621 +610 @@
      * the response is a forked response.
      */
+
+retry_on_deadlock:
+
+    dlg = NULL;
 
     /* Lock user agent dlg table before we're doing anything. */
@@ -783 +776 @@
     rdata->endpt_info.mod_data[mod_ua.mod.id] = dlg;
 
-    /* Acquire lock to the dialog. */
+    /* Attempt to acquire lock to the dialog. */
     PJ_LOG(6,(dlg->obj_name, "UA layer acquiring dialog lock for response"));
-    pjsip_dlg_inc_lock(dlg);
+    status = pjsip_dlg_try_inc_lock(dlg);
+    if (status != PJ_SUCCESS) {
+        /* Failed to acquire dialog mutex. This could indicate a deadlock
+         * situation, and for safety, try to avoid deadlock by releasing
+         * UA mutex, yield, and retry the whole processing once again.
+         */
+        pj_mutex_unlock(mod_ua.mutex);
+        pj_thread_sleep(0);
+        goto retry_on_deadlock;
+    }
+
+    /* We're done with processing in the UA layer, we can release the mutex */
+    pj_mutex_unlock(mod_ua.mutex);
 
     /* Pass the response to the dialog. */
@@ -792 +797 @@
     /* Unlock the dialog. This may destroy the dialog. */
     pjsip_dlg_dec_lock(dlg);
-
-    /* Unlock dialog hash table. */
-    pj_mutex_unlock(mod_ua.mutex);
 
     /* Done. */