sneaking up on that bug like a cat prowling in the jungle
[citadel.git] / webcit / context_loop.c
index 35e6ee7b34e5afcae4d83571c5c29f306171b4f5..3b1f30ae3969d7cb80d1a7fc654b7d5e9e6bf2d0 100644 (file)
@@ -39,6 +39,10 @@ extern HashList *HandlerHash;
 int num_threads_existing = 1;          /* Number of worker threads which exist. */
 int num_threads_executing = 1;         /* Number of worker threads currently executing. */
 
+extern void session_loop(void);
+void spawn_another_worker_thread(void);
+
+
 void DestroyHttpHeaderHandler(void *V)
 {
        OneHttpHeader *pHdr;
@@ -70,13 +74,14 @@ void do_housekeeping(void)
 
                /* Kill idle sessions */
                if ((time(NULL) - (sptr->lastreq)) > (time_t) WEBCIT_TIMEOUT) {
+                       syslog(3, "Timeout session %d\n", sptr->wc_session);
                        sptr->killthis = 1;
                }
 
                /* Remove sessions flagged for kill */
                if (sptr->killthis) {
 
-                       /** remove session from linked list */
+                       /* remove session from linked list */
                        if (sptr == SessionList) {
                                SessionList = SessionList->next;
                        }
@@ -96,11 +101,8 @@ void do_housekeeping(void)
         * Now free up and destroy the culled sessions.
         */
        while (sessions_to_kill != NULL) {
-               lprintf(3, "Destroying session %d\n", sessions_to_kill->wc_session);
-               pthread_mutex_lock(&sessions_to_kill->SessionMutex);
-               pthread_mutex_unlock(&sessions_to_kill->SessionMutex);
+               syslog(3, "Destroying session %d\n", sessions_to_kill->wc_session);
                sptr = sessions_to_kill->next;
-
                session_destroy_modules(&sessions_to_kill);
                sessions_to_kill = sptr;
        }
@@ -118,7 +120,7 @@ void check_thread_pool_size(void)
                (num_threads_executing >= num_threads_existing)
                && (num_threads_existing < MAX_WORKER_THREADS)
        ) {
-               lprintf(3, "%d of %d threads are executing.  Adding another worker thread.\n",
+               syslog(3, "%d of %d threads are executing.  Adding another worker thread.\n",
                        num_threads_executing,
                        num_threads_existing
                );
@@ -176,35 +178,46 @@ wcsession *FindSession(wcsession **wclist, ParsedHttpHdrs *Hdr, pthread_mutex_t
                                continue;
                        if ((!strcasecmp(ChrPtr(Hdr->c_username), ChrPtr(sptr->wc_username))) &&
                            (!strcasecmp(ChrPtr(Hdr->c_password), ChrPtr(sptr->wc_password))) ) {
+                               syslog(LOG_DEBUG, "-- matched a session with the same http-auth");
                                TheSession = sptr;
                        }
                        if (TheSession == NULL)
-                               lprintf(1, "found sessionkey [%ld], but credentials for [%s|%s] didn't match\n",
-                                       Hdr->HR.SessionKey,ChrPtr(Hdr->c_username), ChrPtr(sptr->wc_username));
+                               syslog(1, "found sessionkey [%d], but credentials for [%s|%s] didn't match",
+                                       Hdr->HR.SessionKey,
+                                       ChrPtr(Hdr->c_username),
+                                       ChrPtr(sptr->wc_username)
+                               );
                        break;
                case AUTH_COOKIE:
                        /* If cookie-session, look for a session with matching session ID */
                        if ( (Hdr->HR.desired_session != 0) && 
-                            (sptr->wc_session == Hdr->HR.desired_session)) {
+                            (sptr->wc_session == Hdr->HR.desired_session))
+                       {
+                               syslog(LOG_DEBUG, "-- matched a session with the same cookie");
                                TheSession = sptr;
                        }
                        break;                       
                case NO_AUTH:
+                       /* Any unbound session is a candidate */
+                       if ( (sptr->wc_session == 0) && (sptr->inuse == 0) ) {
+                               syslog(LOG_DEBUG, "-- reusing an unbound session");
+                               TheSession = sptr;
+                       }
                        break;
                }
        }
        CtdlLogResult(pthread_mutex_unlock(ListMutex));
-       if (TheSession == NULL)
-               lprintf(1, "didn't find sessionkey [%ld] for user [%s]\n",
-                       Hdr->HR.SessionKey,ChrPtr(Hdr->c_username));
+       if (TheSession == NULL) {
+               syslog(1, "didn't find sessionkey [%d] for user [%s]",
+                       Hdr->HR.SessionKey, ChrPtr(Hdr->c_username)
+               );
+       }
        return TheSession;
 }
 
 wcsession *CreateSession(int Lockable, int Static, wcsession **wclist, ParsedHttpHdrs *Hdr, pthread_mutex_t *ListMutex)
 {
        wcsession *TheSession;
-       if (!Static)
-               lprintf(3, "Creating a new session\n");
        TheSession = (wcsession *) malloc(sizeof(wcsession));
        memset(TheSession, 0, sizeof(wcsession));
        TheSession->Hdr = Hdr;
@@ -221,9 +234,11 @@ wcsession *CreateSession(int Lockable, int Static, wcsession **wclist, ParsedHtt
         */     
        if (Hdr->HR.desired_session == 0) {
                TheSession->wc_session = GenerateSessionID();
+               syslog(3, "Created new session %d", TheSession->wc_session);
        }
        else {
                TheSession->wc_session = Hdr->HR.desired_session;
+               syslog(3, "Re-created session %d", TheSession->wc_session);
        }
        Hdr->HR.Static = Static;
        session_new_modules(TheSession);
@@ -391,7 +406,7 @@ int ReadHTTPRequest (ParsedHttpHdrs *Hdr)
                        memset(pHdr, 0, sizeof(OneHttpHeader));
                        pHdr->Val = Line;
                        Put(Hdr->HTTPHeaders, HKEY("GET /"), pHdr, DestroyHttpHeaderHandler);
-                       lprintf(9, "%s\n", ChrPtr(Line));
+                       syslog(9, "%s\n", ChrPtr(Line));
                        isbogus = ReadHttpSubject(Hdr, Line, HeaderName);
                        if (isbogus) break;
                        continue;
@@ -479,6 +494,7 @@ void context_loop(ParsedHttpHdrs *Hdr)
        wcsession *TheSession;
        struct timeval tx_start;
        struct timeval tx_finish;
+       int session_may_be_reused = 1;
        
        gettimeofday(&tx_start, NULL);          /* start a stopwatch for performance timing */
 
@@ -488,20 +504,19 @@ void context_loop(ParsedHttpHdrs *Hdr)
        isbogus = ReadHTTPRequest(Hdr);
 
        Hdr->HR.dav_depth = 32767; /* TODO: find a general way to have non-0 defaults */
-       if (!isbogus)
+
+       if (!isbogus) {
                isbogus = AnalyseHeaders(Hdr);
+       }
 
-       if ((isbogus) ||
-           ((Hdr->HR.Handler != NULL) &&
-            ((Hdr->HR.Handler->Flags & BOGUS) != 0)))
-       {
+       if (    (isbogus)
+               || ((Hdr->HR.Handler != NULL)
+               && ((Hdr->HR.Handler->Flags & BOGUS) != 0))
+       {
                wcsession *Bogus;
-
                Bogus = CreateSession(0, 1, NULL, Hdr, NULL);
-
                do_404();
-
-               lprintf(9, "HTTP: 404 [%ld.%06ld] %s %s \n",
+               syslog(9, "HTTP: 404 [%ld.%06ld] %s %s \n",
                        ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) / 1000000,
                        ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) % 1000000,
                        ReqStrs[Hdr->HR.eReqType],
@@ -522,16 +537,12 @@ void context_loop(ParsedHttpHdrs *Hdr)
                /* How long did this transaction take? */
                gettimeofday(&tx_finish, NULL);
                
-#ifdef TECH_PREVIEW
-               if ((Hdr->HR.Handler != NULL) ||
-                   ((Hdr->HR.Handler->Flags & LOGCHATTY) == 0))
-#endif
-                       lprintf(9, "HTTP: 200 [%ld.%06ld] %s %s \n",
-                               ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) / 1000000,
-                               ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) % 1000000,
-                               ReqStrs[Hdr->HR.eReqType],
-                               ChrPtr(Hdr->this_page)
-                               );
+               syslog(9, "HTTP: 200 [%ld.%06ld] %s %s \n",
+                       ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) / 1000000,
+                       ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) % 1000000,
+                       ReqStrs[Hdr->HR.eReqType],
+                       ChrPtr(Hdr->this_page)
+               );
                session_detach_modules(Static);
                session_destroy_modules(&Static);
                return;
@@ -541,6 +552,10 @@ void context_loop(ParsedHttpHdrs *Hdr)
                CheckAuthBasic(Hdr);
        }
 
+       if (Hdr->HR.got_auth) {
+               session_may_be_reused = 0;
+       }
+
        /*
         * See if there's an existing session open with the desired ID or user/pass
         */
@@ -562,7 +577,7 @@ void context_loop(ParsedHttpHdrs *Hdr)
                }
                
                if (StrLength(Hdr->c_language) > 0) {
-                       lprintf(9, "Session cookie requests language '%s'\n", ChrPtr(Hdr->c_language));
+                       syslog(9, "Session cookie requests language '%s'\n", ChrPtr(Hdr->c_language));
                        set_selected_language(ChrPtr(Hdr->c_language));
                        go_selected_language();
                }
@@ -579,31 +594,38 @@ void context_loop(ParsedHttpHdrs *Hdr)
        CtdlLogResult(pthread_mutex_lock(&TheSession->SessionMutex));
        pthread_setspecific(MyConKey, (void *)TheSession);
        
+       TheSession->inuse = 1;                                  /* mark the session as bound */
        TheSession->lastreq = time(NULL);                       /* log */
        TheSession->Hdr = Hdr;
 
        session_attach_modules(TheSession);
        session_loop();                         /* do transaction */
 
-
        /* How long did this transaction take? */
        gettimeofday(&tx_finish, NULL);
-       
 
-#ifdef TECH_PREVIEW
-       if ((Hdr->HR.Handler != NULL) &&
-           ((Hdr->HR.Handler->Flags & LOGCHATTY) == 0))
-#endif
-               lprintf(9, "HTTP: 200 [%ld.%06ld] %s %s \n",
-                       ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) / 1000000,
-                       ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) % 1000000,
-                       ReqStrs[Hdr->HR.eReqType],
-                       ChrPtr(Hdr->this_page)
-                       );
+       syslog(9, "HTTP: 200 [%ld.%06ld] %s %s \n",
+               ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) / 1000000,
+               ((tx_finish.tv_sec*1000000 + tx_finish.tv_usec) - (tx_start.tv_sec*1000000 + tx_start.tv_usec)) % 1000000,
+               ReqStrs[Hdr->HR.eReqType],
+               ChrPtr(Hdr->this_page)
+       );
 
        session_detach_modules(TheSession);
 
+       /* If *this* very transaction did not explicitly specify a session cookie,
+        * and it did not log in, we want to flag the session as a candidate for
+        * re-use by the next unbound client that comes along.  This keeps our session
+        * table from getting bombarded with new sessions when, for example, a web
+        * spider crawls the site without using cookies.
+        */
+       if ((session_may_be_reused) && (!WC->logged_in)) {
+               WC->wc_session = 0;                     /* flag as available for re-use */
+               TheSession->selected_language = 0;      /* clear any non-default language setting */
+       }
+
        TheSession->Hdr = NULL;
+       TheSession->inuse = 0;                                  /* mark the session as unbound */
        CtdlLogResult(pthread_mutex_unlock(&TheSession->SessionMutex));
 }
 
@@ -632,20 +654,25 @@ void Header_HandleContentType(StrBuf *Line, ParsedHttpHdrs *hdr)
 
 void Header_HandleHost(StrBuf *Line, ParsedHttpHdrs *hdr)
 {
-       if (site_prefix == NULL) {
-               site_prefix = NewStrBuf();
-               StrBufAppendPrintf(site_prefix, "%s://", (is_https ? "https" : "http") );
-               StrBufAppendBuf(site_prefix, Line, 0);
+       if (hdr->HostHeader != NULL) {
+               FreeStrBuf(&hdr->HostHeader);
        }
+       hdr->HostHeader = NewStrBuf();
+       StrBufAppendPrintf(hdr->HostHeader, "%s://", (is_https ? "https" : "http") );
+       StrBufAppendBuf(hdr->HostHeader, Line, 0);
 }
 
 void Header_HandleXFFHost(StrBuf *Line, ParsedHttpHdrs *hdr)
 {
-       if ( (follow_xff) && (site_prefix == NULL)) {
-               site_prefix = NewStrBuf();
-               StrBufAppendPrintf(site_prefix, "http://");     /* this is naive; do something about it */
-               StrBufAppendBuf(site_prefix, Line, 0);
+       if (!follow_xff) return;
+
+       if (hdr->HostHeader != NULL) {
+               FreeStrBuf(&hdr->HostHeader);
        }
+
+       hdr->HostHeader = NewStrBuf();
+       StrBufAppendPrintf(hdr->HostHeader, "http://"); /* this is naive; do something about it */
+       StrBufAppendBuf(hdr->HostHeader, Line, 0);
 }
 
 
@@ -791,6 +818,7 @@ HttpDetachModule_CONTEXT
 (ParsedHttpHdrs *httpreq)
 {
        FlushStrBuf(httpreq->PlainArgs);
+       FlushStrBuf(httpreq->HostHeader);
        FlushStrBuf(httpreq->this_page);
        FlushStrBuf(httpreq->PlainArgs);
        DeleteHash(&httpreq->HTTPHeaders);
@@ -805,6 +833,7 @@ HttpDestroyModule_CONTEXT
        FreeStrBuf(&httpreq->PlainArgs);
        FreeStrBuf(&httpreq->this_page);
        FreeStrBuf(&httpreq->PlainArgs);
+       FreeStrBuf(&httpreq->HostHeader);
        DeleteHash(&httpreq->HTTPHeaders);
 
 }