Unix Network Programming, Volume 1: The Sockets Networking API (3rd Edition)

22.3 Datagram Truncation

On BSD-derived systems, when a UDP datagram arrives that is larger than the application's buffer, recvmsg sets the MSG_TRUNC flag in the msg_flags member of the msghdr structure (Figure 14.7). All Berkeley-derived implementations that support the msghdr structure with the msg_flags member provide this notification.

This is an example of a flag that must be returned from the kernel to the process. We mentioned in Section 14.3 that one design problem with the recv and recvfrom functions is that their flags argument is an integer, which allows flags to be passed from the process to the kernel, but not vice versa.

Unfortunately, not all implementations handle a larger-than-expected UDP datagram in this fashion. There are three possible scenarios:

1. Discard the excess bytes and return the MSG_TRUNC flag to the application. This requires that the application call recvmsg to receive the flag.

2. Discard the excess bytes, but do not tell the application.

3. Keep the excess bytes and return them in subsequent read operations on the socket.

   The POSIX specification specifies the first type of behavior: discarding the excess bytes and setting the MSG_TRUNC flag. Early releases of SVR4 exhibited the third type of behavior.

Since there are such variations in how implementations handle datagrams that are larger than the application's receive buffer, one way to detect the problem is to always allocate an application buffer that is one byte greater than the largest datagram the application should ever receive. If a datagram is ever received whose length equals this buffer, consider it an error.