Hi,
Something that I'd spotted a while ago, and so far as I remember has not
yet been fixed, is the "extra socket creation" feature of accept(). This
too cannot help performance. I had an idea (which is on my "when 2.3.xx
comes out" list) to change the way accept() works at the socket layer by
implementing a protocol independent version of the tcp openreq scheme.
The basic idea was to have a queue on each socket of structures, one
per pending connection which can be managed with protocol independent
code.
I'd be interested to know who is considering the wake-one semantics since
it would seem an ideal thing to integrate with my accept() restructuring
idea,
Steve.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/