I think the socket-connect function in the SBCL backend is incorrectly calling bind when creating a client stream socket.
With the current code, when I create a lot of client stream sockets on a Linux system, the bind call in socket-connect eventually returns an error, address already in use. I don't think the bind call is needed for the client end of stream sockets.
The following patch appears to fix the problem for me.
bob
--- backend/sbcl.lisp.~1~ 2011-09-23 13:46:55.071765010 -0400 +++ backend/sbcl.lisp 2011-09-23 13:51:33.841811680 -0400 @@ -269,9 +269,8 @@ :protocol (case protocol (:stream :tcp) (:datagram :udp)))) - (local-host (host-to-vector-quad (or local-host *wildcard-host*))) - (local-port (or local-port *auto-port*)) - usocket ok) + (usocket nil) + (ok nil)) (unwind-protect (progn (ecase protocol @@ -284,8 +283,6 @@ ;; to pass compilation on ECL without it. (when (and nodelay-specified sockopt-tcp-nodelay-p) (setf (sb-bsd-sockets::sockopt-tcp-nodelay socket) nodelay)) - (when (or local-host local-port) - (sb-bsd-sockets:socket-bind socket local-host local-port)) (with-mapped-conditions (usocket) #+(and sbcl (not win32)) (labels ((connect ()
Robert Brown robert.brown@gmail.com writes:
With the current code, when I create a lot of client stream sockets on a Linux system, the bind call in socket-connect eventually returns an error, address already in use. I don't think the bind call is needed for the client end of stream sockets.
Bind is not needed unless we want a specific interface or a specific client port.
USOCKET code seems to take /some/ care of not calling bind() unless it's needed; however, *wildcard-host* and *auto-port* being non-NIL makes bind() happen.
(I guess that the underlying problem will go away if SO_REUSEADDR will be enabled on client socket, i.e. explicit bind() will succeed then).
On Fri, Sep 23, 2011 at 2:59 PM, Anton Kovalenko anton@sw4me.com wrote:> Bind is not needed unless we want a specific interface or a specific> client port. Seems like a better patch is the following, which calls bind if local-hostor local-port are specified in the call to socket-connect. --- third_party/lisp/usocket/backend/sbcl.lisp.~1~ 2011-08-12 16:23:26.718704693 -0400 +++ third_party/lisp/usocket/backend/sbcl.lisp 2011-09-23 15:11:24.481814325 -0400 @@ -269,9 +269,8 @@ :protocol (case protocol (:stream :tcp) (:datagram :udp)))) - (local-host (host-to-vector-quad (or local-host *wildcard-host*))) - (local-port (or local-port *auto-port*)) - usocket ok) + (usocket nil) + (ok nil)) (unwind-protect (progn (ecase protocol @@ -285,7 +284,9 @@ (when (and nodelay-specified sockopt-tcp-nodelay-p) (setf (sb-bsd-sockets::sockopt-tcp-nodelay socket) nodelay)) (when (or local-host local-port) - (sb-bsd-sockets:socket-bind socket local-host local-port)) + (sb-bsd-sockets:socket-bind socket + (host-to-vector-quad (or local-host *wildcard-host*)) + (or local-port *auto-port*))) (with-mapped-conditions (usocket) #+sbcl (labels ((connect () @@ -306,8 +307,7 @@ (:datagram (when (or local-host local-port) (sb-bsd-sockets:socket-bind socket - (host-to-vector-quad - (or local-host *wildcard-host*)) + (host-to-vector-quad (or local-host *wildcard-host*)) (or local-port *auto-port*))) (setf usocket (make-datagram-socket socket)) (when (and host port)
Hi, Robert
I'm sorry for replying your mail so late - a bit busy these days.
Thanks, your patch fixed a bug in SOCKET-CONNECT: in my original design, if calling SOCKET-CONNECT without LOCAL-HOST and LOCAL-PORT keyword argument, the calling of bind() won't happen, but ... it seems that I calculate LOCAL-HOST / LOCAL-PORT too early, which caused the following protection block actually won't work:
(when (or local-host local-port) ...)
Any way, I've committed your fixes as r674. And consider this is a important fix, I'll make a new release as soon as possible.
Besides, I agree Anton's opinion that the bind problem could be solved by enabling SO_REUSEADDR on client sockets, but I'm not sure if it's necessary to add another keyword argument to SOCKET-CONNECT ... when Robert's issue would be solved by current solutions.
Regards,
Chun Tian (binghe)
在 2011-9-24，03:20， Robert Brown 写道：
On Fri, Sep 23, 2011 at 2:59 PM, Anton Kovalenko anton@sw4me.com wrote:> Bind is not needed unless we want a specific interface or a specific> client port. Seems like a better patch is the following, which calls bind if local-hostor local-port are specified in the call to socket-connect. --- third_party/lisp/usocket/backend/sbcl.lisp.~1~ 2011-08-12 16:23:26.718704693 -0400 +++ third_party/lisp/usocket/backend/sbcl.lisp 2011-09-23 15:11:24.481814325 -0400 @@ -269,9 +269,8 @@ :protocol (case protocol (:stream :tcp) (:datagram :udp))))
(local-host (host-to-vector-quad (or local-host *wildcard-host*)))(local-port (or local-port *auto-port*))usocket ok)
(usocket nil)(ok nil))@@ -285,7 +284,9 @@ (when (and nodelay-specified sockopt-tcp-nodelay-p) (setf (sb-bsd-sockets::sockopt-tcp-nodelay socket) nodelay)) (when (or local-host local-port)
(sb-bsd-sockets:socket-bind socket local-host local-port))
(sb-bsd-sockets:socket-bind socket(host-to-vector-quad (orlocal-host *wildcard-host*))
(or local-port *auto-port*))) (with-mapped-conditions (usocket)@@ -306,8 +307,7 @@ (:datagram (when (or local-host local-port) (sb-bsd-sockets:socket-bind socket
(host-to-vector-quad(or local-host *wildcard-host*))
(host-to-vector-quad (orlocal-host *wildcard-host*)) (or local-port *auto-port*))) (setf usocket (make-datagram-socket socket)) (when (and host port)
usocket-devel mailing list usocket-devel@common-lisp.net http://lists.common-lisp.net/cgi-bin/mailman/listinfo/usocket-devel
-
Anton Kovalenko -
Chun Tian (binghe) -
Robert Brown