Clamp HEP capture payload to the buffer instead of dropping chunks (#321)

The capture builders broke out of the copy loop at the first iov
fragment that would exceed the buffer cap, before copying any of it,
so an oversize message lost everything from that fragment on. A UDP
datagram arrives as a single fragment, so the whole message was
dropped and the HEP frame carried an empty payload; a TCP read arrives
as several fragments, so some were kept and the tail lost.

Clamp each fragment's copy length to the remaining buffer room instead
of dropping it, in both the HEPv2 and HEPv3 builders, filling up to the
cap.

This also fixes an over-declared HEPv3 length: the sizing and copy
passes counted different totals against the cap, so `header.length`
could exceed the bytes actually sent. The copy pass now emits exactly
the counted `payload_len`, keeping declared and sent length in sync.

Author: morbit85

libsofia-sip-ua/tport/tport_logging.c

@@ -602,17 +602,19 @@ int tport_capt_msg_hepv2 (tport_t const *self, msg_t *msg, size_t n,
         buflen += sizeof(struct hep_timehdr);
    }                    
 
-   for (i = 0; i < iovused && n > 0; i++) {
+   for (i = 0; i < iovused && n > 0 && buflen < capt_bufsize; i++) {
        size_t len = iov[i].mv_len;
+       size_t room = capt_bufsize - buflen;
+
        if (len > n)
-            len = n;   
-       /* if the packet too big for us */
-       if((buflen + len) > capt_bufsize) 
-              break;
-
-      memcpy(*buffer + buflen , (void*)iov[i].mv_base, len);
-      buflen +=len;
-      n -= len;
+            len = n;
+       /* clamp to remaining capture buffer */
+       if (len > room)
+            len = room;
+
+       memcpy(*buffer + buflen, (void*)iov[i].mv_base, len);
+       buflen += len;
+       n -= len;
    }
 
    return buflen;
@@ -638,8 +640,7 @@ int tport_capt_msg_hepv3 (tport_t const *self, msg_t *msg, size_t n,
    su_time_t now;
    hep_chunk_ip4_t src_ip4 = {{0}}, dst_ip4 = {{0}};
    hep_chunk_t payload_chunk;
-   int orig_n = 0;
-      
+
 #if SU_HAVE_IN6
    hep_chunk_ip6_t src_ip6 = {{0}}, dst_ip6 = {{0}};
 #endif   
@@ -785,16 +786,19 @@ int tport_capt_msg_hepv3 (tport_t const *self, msg_t *msg, size_t n,
 
 
     /* Payload caclulation */
-    orig_n = n;
-    for (i = 0; i < iovused && n > 0; i++) {
-       	size_t len = iov[i].mv_len;
-	if (len > n) len = n;   
-	if((payload_len + len) > capt_bufsize) break;
-        payload_len +=len;
-    	n -= len;
+    for (i = 0; i < iovused && n > 0 && payload_len < capt_bufsize; i++) {
+        size_t len = iov[i].mv_len;
+        size_t room = capt_bufsize - payload_len;
+
+        if (len > n)
+            len = n;
+        /* clamp to remaining capture buffer instead of dropping the chunk */
+        if (len > room)
+            len = room;
+
+        payload_len += len;
+        n -= len;
     }
-    /* restore n */
-    n = orig_n;
 
     /* Payload */
     payload_chunk.vendor_id = htons(0x0000);
@@ -841,17 +845,17 @@ int tport_capt_msg_hepv3 (tport_t const *self, msg_t *msg, size_t n,
     memcpy((char*) *buffer+buflen, &payload_chunk,  sizeof(struct hep_chunk));
     buflen +=  sizeof(struct hep_chunk);
 
-   /* PAYLOAD */
-   for (i = 0; i < iovused && n > 0; i++) {
+   /* PAYLOAD: copy the payload_len bytes counted above (keeps buflen == tlen).
+    * payload_len is already folded into tlen/header.length, so consume it. */
+   for (i = 0; i < iovused && payload_len > 0; i++) {
        size_t len = iov[i].mv_len;
-       if (len > n) len = n;   
-       /* if the packet too big for us */
-       if((buflen + len) > capt_bufsize) 
-              break;
-
-      memcpy(*buffer + buflen , (void*)iov[i].mv_base, len);
-      buflen +=len;
-      n -= len;
+
+       if (len > payload_len)
+           len = payload_len;
+
+       memcpy(*buffer + buflen, (void*)iov[i].mv_base, len);
+       buflen += len;
+       payload_len -= len;
    }
 
    free(hg);