Handle large raw strings from the server and properly track/rejoin long split/truncated messages

1 files changed, 51 insertions, 12 deletions

diff --git a/blabouncer.c b/blabouncer.c
index 2ac787e..751143d 100644
--- a/blabouncer.c
+++ b/blabouncer.c
@@ -10,9 +10,17 @@
 // - Keep track of changing user nicks/modes
 // - Should relay log do more than PRIVMSGs?
 // - Make certificate paths configurable
-// - Rename "ssl" to "arr_ssl" and remove the old send() based commented lines when ready
+// - Rename "ssl" to "arr_ssl"
+// - Implement TLS on real IRCd server side
+// - Make TLS optional
+// - Make cert/key filenames configurable
+// - Make listen port configurable
 // - Consider moving the three fd-related arrays into one struct
 // - Check authentication before even getting to the send functions to save unnecessary processing
+// - Configurable auto channels
+// - Comma separated channel list in JOINs/PARTs
+// - Long raw strings like a 321/322/323 channel LIST, and the initial server welcome, need to be read and split properly
+// - Only send some things to the requesting client (e.g. LIST replies)
 //
 // Example WHOIS reply:
 // BOUNCER-SERVER RECEIVED: :irc.tghost.co.uk 307 blabounce l_bratch :is identified for this nick
@@ -45,6 +53,9 @@
 #define SOURCE_CLIENT 1
 #define EXCEPT_NONE 0
 
+// It seems to be that *message length* is max 512 bytes, but a socket read from at least UnrealIRCd seems to be up to at least 2416 (+1 for null) bytes.
+// 1208 bytes with OpenSSL, 2416 bytes with plain text.
+#define MAXRCVSIZE 2417
 #define MAXDATASIZE 513 // max number of bytes we can get at once (RFC2812 says 512, plus one for null terminator)
 #define STDIN 0 // stdin is fd 0
 #define MAXCLIENTS 32 // maximum number of clients that can connect to the bouncer at a time
@@ -79,6 +90,7 @@ struct ircdstrings {
   char ircnick[MAXNICKLENGTH]; // TODO - Make sure this changes when nick changes
   char ircusername[MAXUSERNAMELEN];
   char ircrealname[MAXREALNAMELEN];
+  char currentmsg[MAXDATASIZE]; // Holding area for the current server-received IRC message being processed in case it needs building across multiple reads (i.e. a truncated/split message)
 };
 
 int debugmode = 0;
@@ -637,10 +649,10 @@ int processircmessage(int *serversockfd, int *clientsockfd, char *str, int sourc
 
           return 1;
         }
-
-        // Don't return if we got here because this means we didn't process something above
       }
 
+      // Don't return if we got here because this means we didn't process something above
+
       break;
     case SOURCE_CLIENT: // If message(s) were from a real IRC client
       // PASS received?  User is trying to log in, check their password.
@@ -915,6 +927,32 @@ int processrawstring(int *serversockfd, int *clientsockfd, char *str, int source
 
   free(strcopy);
 
+  // If there is a previous possibly truncated message still in the holding area, the prepend that to the first message of this new lot
+  // (Only if source was the server since we always strip \r\n from client messages when recving - TODO - Should we be doing that?
+  if (ircdstrings->currentmsg[0] && source == SOURCE_SERVER) {
+    // Make a copy since we can't have source and destination the same with snprintf
+    char *strtmp = strdup(messages[0]);
+    printf("processrawstring(): Previous truncated message detected, combining old '%s' with new '%s'...\n", ircdstrings->currentmsg, strtmp);
+    snprintf(messages[0], strlen(ircdstrings->currentmsg) + strlen(strtmp) + 1, "%s%s", ircdstrings->currentmsg, strtmp);
+    messages[0][strlen(ircdstrings->currentmsg) + strlen(strtmp)] = '\0'; // Make sure it's null terminated
+    printf("...into new string '%s' and clearing currentmsg holding area.\n", messages[0]);
+    ircdstrings->currentmsg[0] = '\0';
+  }
+
+  // If the final characters of the raw string weren't \r\n then assume the final token is a truncated message
+  // Copy to a holding area for continuation next time
+  // (Only if source was the server since we always strip \r\n from client messages when recving - TODO - Should we be doing that?
+  if ((str[strlen(str)-2] != 13 || str[strlen(str)-1] != 10) && source == SOURCE_SERVER) {
+    printf("processrawstring(): Truncated message detected, storing final token '%s' for later.\n", messages[messagecount - 1]);
+    strncpy(ircdstrings->currentmsg, messages[messagecount - 1], strlen(messages[messagecount - 1]));
+    ircdstrings->currentmsg[strlen(messages[messagecount - 1])] = '\0';
+    // Remove it from the message count so it's not processed time time
+    messagecount--;
+  } else {
+    // Otherwise, clear the holding area
+    ircdstrings->currentmsg[0] = '\0';
+  }
+
   // Go through each message, figure out what it is and if we're doing anything with it
   for (int i = 0; i < messagecount; i++) {
     // Copy to a temporary string so we still have the original in case it's not processed
@@ -934,20 +972,20 @@ int processrawstring(int *serversockfd, int *clientsockfd, char *str, int source
         case SOURCE_SERVER: // If message(s) were from the real IRC server
           // Relay/send to all clients ("except" = 0 because this should send to all clients)
           // TODO - Is this really going to send the original string if we have messed it with it in processrawstring() and friends!?
-          printf("bouncer-server: sending unprocessed server messasge \"%s\" to all clients, length %zd.\n", messages[i], strlen(messages[i]));
+          printf("bouncer-server: sending unprocessed server message \"%s\" to all clients, length %zd.\n", messages[i], strlen(messages[i]));
           sendtoallclients(clientsockfd, fdmax, arr_clients, messages[i], EXCEPT_NONE, arr_authed, ssl);
           break;
         case SOURCE_CLIENT: // If message(s) were from a real IRC client
           // Send to server
-          printf("bouncer-client: sending unprocessed client messasge \"%s\" to the server, length %zd.\n", messages[i], strlen(messages[i]));
+          printf("bouncer-client: sending unprocessed client message \"%s\" to the server, length %zd.\n", messages[i], strlen(messages[i]));
           sendtoserver(serversockfd, messages[i], strlen(messages[i]), sourcefd, arr_clients, arr_authed);
 
-          printf("bouncer-client: sending unprocessed client messasge \"%s\" to all other clients, length %zd.\n", messages[i], strlen(messages[i]));
+          printf("bouncer-client: sending unprocessed client message \"%s\" to all other clients, length %zd.\n", messages[i], strlen(messages[i]));
           // send the same thing to all *other* clients (all except for source fd)
           sendtoallclients(clientsockfd, fdmax, arr_clients, messages[i], sourcefd, arr_authed, ssl);
           break;
         default:
-          fprintf(stderr, "Unexpected raw IRC string source for unprocessed messasge \"%s\", length %zd.!\n", messages[i], strlen(messages[i]));
+          fprintf(stderr, "Unexpected raw IRC string source for unprocessed message \"%s\", length %zd.!\n", messages[i], strlen(messages[i]));
           exit(1);
       }
     }
@@ -960,8 +998,8 @@ int processrawstring(int *serversockfd, int *clientsockfd, char *str, int source
 
 // Where the big bouncing loop is
 void dochat(int *serversockfd, int *clientsockfd) {
-  char serverbuf[MAXDATASIZE]; // buffer for receiving data on server socket
-  char clientbuf[MAXDATASIZE]; // buffer for receiving data on client socket(s)
+  char serverbuf[MAXRCVSIZE]; // buffer for receiving data on server socket
+  char clientbuf[MAXRCVSIZE]; // buffer for receiving data on client socket(s)
   int servernumbytes; // Number of bytes received from remote server
   char outgoingmsg[MAXDATASIZE]; // String to send to server
   int outgoingmsgrc; // Return code from getstdin() for outgoing message
@@ -1007,6 +1045,7 @@ void dochat(int *serversockfd, int *clientsockfd) {
   ircdstrings.ircnick[0] = '\0';
   ircdstrings.ircusername[0] = '\0';
   ircdstrings.ircrealname[0] = '\0';
+  ircdstrings.currentmsg[0] = '\0';
 
   // Read required things from configuration file
   readnames(ircdstrings.ircnick, ircdstrings.ircusername, ircdstrings.ircrealname);
@@ -1067,7 +1106,7 @@ void dochat(int *serversockfd, int *clientsockfd) {
     if (FD_ISSET(*serversockfd, &rfds)) {
       printf("reading server socket!\n");
 
-      if ((servernumbytes = recv(*serversockfd, serverbuf, MAXDATASIZE - 1, 0)) == -1) {
+      if ((servernumbytes = recv(*serversockfd, serverbuf, MAXRCVSIZE - 1, 0)) == -1) {
         printf("receive error (-1), exiting...\n");
         perror("recv");
         exit(1);
@@ -1078,7 +1117,7 @@ void dochat(int *serversockfd, int *clientsockfd) {
       }
       serverbuf[servernumbytes] = '\0';
 
-      printf("BOUNCER-SERVER RECEIVED: '%s'\n", serverbuf);
+      printf("BOUNCER-SERVER RECEIVED: '%s', length '%d'.\n", serverbuf, servernumbytes);
 
       // Try to process received string (which should contain one or more server responses/commands)
       // TODO - What if there were two server respones/commands and only one didn't need relaying?
@@ -1200,7 +1239,7 @@ void dochat(int *serversockfd, int *clientsockfd) {
             // we got some data from a client
             // null terminate that baby
             clientbuf[clientnumbytes] = '\0'; // TODO make sure this can't overrun if some super long line (max bytes?) was received
-            // clear up any newlines
+            // clear up any newlines - TODO - Should we be doing this?  If not, we can stop only doing truncation checks for the server in processrawstring().
             while (clientbuf[strlen(clientbuf) - 1] == '\n' || clientbuf[strlen(clientbuf) - 1] == '\r') {
               clientbuf[strlen(clientbuf) - 1] = '\0';
             }