Messages in this thread | | | Date | Sun, 5 Apr 1998 19:40:16 +0100 | From | "Stephen C. Tweedie" <> | Subject | Re: I/O completion ports for Linux |
| |
Hi,
On Tue, 31 Mar 1998 19:43:38 -0500, "Theodore Y. Ts'o" <tytso@MIT.EDU> said:
> From: Robey Pointer <robey@lag.net>
> The following is a patch against Linux 2.1.84 to implement "I/O > completion ports" similar to VMS/NT.
> The only other choice is to spawn separate kernel threads for > each of the 30,000 connections, but given that most of these connections > are idle most of the time, this tends not to be a very good use of the > system's resources.
Bear in mind that there may be other ways we can do this. In particular, the posix RT signals mechanism in 2.1 seems an ideal way of dealing with this problem, since it allows us to queue a separate signal for each completion event and to pass the fd as an argument directly to the signal handler.
It would be relatively trivial to add this to the existing SIGIO mechanism --- all that would be required is to flag whether a SIGIO or a new signal, SIGRTIO with realtime status, is to be delivered, and to pass the fd in when queuing the signal.
--Stephen
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |