• Advertisement

Archived

This topic is now archived and is closed to further replies.

Another UDP question

This topic is 6539 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Well, I''ve looked in the windows socket documentation, the unix-socket faq and a bunch of tutorials, and can''t find the answer. What exactly does it mean when select() returns indicating a non-blocking UDP socket is available for writing? Does it mean that a single packet (smaller than the max packet size) can be writen to the socket safely? Or can any combination of packets be writen to the socket, as long as their total size does not exceed the max packet size? Or does it have another meaning? Should I check the return value to sendto(), even if select has indicated the socket is ready for a write? Thanks -genovov

Share this post


Link to post
Share on other sites
Advertisement
No one knows the answer? Does anyone have any ideas where I might be able to find the answer? Any good, advanced socket resources on the web?

Thanks again everyone,

- genovov

Share this post


Link to post
Share on other sites
I believe it means that the outbound buffer on that socket is empty.

That is to say once you''ve called a sento() on the socket, the function returns immediately as it is non-blocking. However, the data isn''t flushed immediately, so the select statement will signal when it has been flushed.

Share this post


Link to post
Share on other sites
Does that mean that it''s only possible to send one UDP packet between calls to select()?

Sorry if I''m being thick. I just want to figure out if I''ll need some kind of queue for outgoing packets...

Can anyone point me to some source that uses non-blocking udp sockets?

- genovov

Share this post


Link to post
Share on other sites
No, it doesn''t mean you can''t call more than one sendto() between select() calls. But it is possible to overflow the buffers if you don''t synchronize with the select calls. I believe if you do a buffer overflow on a non-blocking sendto() it will give some sort of error condition, either not the full length will be sent or some other explicit error condition.

Share this post


Link to post
Share on other sites
Guest Anonymous Poster
I didn''t see it specified as part of ther Berkley spec, or the Winsock 2 API docs. It''s probably one of those popular ''features'' of Sockets programming that is Implementation Specific.

To me, ''proper'' working order would mean that you should be able to send up to SO_MAX_MSG_SIZE bytes after a return of FD_WRITE from select(). If I get ambitious today, I''ll write some code to test it and post it here.

Share this post


Link to post
Share on other sites
I''m planning on porting my project when it''s finished, and was hoping this behaviour was defined in the spec. If you have a working version of Linux available to you, I would be very grateful if you could test how UDP behaves. I''ll do the same for winsock 1.1 over the weekend, and post the results here.

Does SO_MAX_MSG_SIZE include the size of the UDP header, or does it just refer to the data? If so, it may be that you can write SO_MAX_MSG_SIZE - ([number of calls to sendto()] * [UDP header size]) bytes after a call to select().

- genovov

Share this post


Link to post
Share on other sites
Ok, the spec says that select() will return with FD_WRITE set when send() or sendto() is guaranteed to succeed, so at least SO_MAX_MSG_SIZE bytes must be available in the buffer.

SO_MAX_MSG_SIZE refers to the maximum size of an individual datagram in the underlying transport layer. I believe that it doesn''t include headers. Therefore potentially multiple sendto()''s can be called with size up to SO_MAX_MSG_SIZE. SO_MAX_MSG_SIZE doesn''t refer to the buffer size for the socket.

Share this post


Link to post
Share on other sites

  • Advertisement