On Fri, 12 Nov 2010, Luigi Rizzo wrote:
Modified: head/sys/sys/socketvar.h
==============================================================================
--- head/sys/sys/socketvar.h Fri Nov 12 12:48:41 2010 (r215177)
+++ head/sys/sys/socketvar.h Fri Nov 12 13:02:26 2010 (r215178)
@@ -117,7 +117,14 @@ struct socket {
void *so_accept_filter_arg; /* saved filter args */
char *so_accept_filter_str; /* saved user args */
} *so_accf;
+ /*
+ * so_fibnum, so_user_cookie and friends can be used to attach
+ * some user-specified metadata to a socket, which then can be
+ * used by the kernel for various actions.
+ * so_user_cookie is used by ipfw/dummynet.
+ */
int so_fibnum; /* routing domain for this socket */
+ uint32_t so_user_cookie;
};
This seems like it should be an inpcb-layer field, set by an IP-layer socket
option, not a socket-layer field. Remember that, in some cases, we have
inpcb's floating around without sockets, if the socket has been closed but TCP
is still winding down (such as in TCPS_TIMEWAIT, where we've discarded not
only the socket, but also the tcpcb, and substituted a tcptw). In fact, the
code I pointed at in my previous e-mail will probably panic in that case,
since it improperly dereferences the inpcb's socket field without first
testing for NULL.
Robert
_______________________________________________
svn-src-head@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-head
To unsubscribe, send any mail to "svn-src-head-unsubscr...@freebsd.org"
