Date: Tue, 26 Jun 2001 15:25:51 -0400 From: FastPathNow@netscape.net To: luigi@info.iet.unipi.it Cc: FastPathNow@netscape.net, freebsd-net@freebsd.org Subject: Re: TCPS_HAVERCVDFIN not considering all possible conditions? Message-ID: <1DEC8D97.1B5FD06C.375A6AF3@netscape.net> References: <200106261837.UAA98582@info.iet.unipi.it>
next in thread | previous in thread | raw e-mail | index | archive | help
> > be possible to do interesting things things like state tracking
> > for individal connections, such that every time we move into the
> > next state instead of doing something like tp->t_state = TCPS_SYN_RCVD,
> > we can have something like tp->t_state |= TCPS_SYN_RCVD . The only
> > useful purpose that I can think of would be have a history of the
> > past states in the t_state variable.
>
> > Even without this ORing change, it would still make writing macros such as the above easier
>
> No way! that would be even more wrong.
>
> One state is one bit, and state changes are strict assignments.
> The bit-set only makes it more efficient to test for set membership
> ( "state \in {S1, S2, S3 ...}" becomes ( tp->t_state & (S1|S2|S3) != 0 )
>
> cheers
> luigi
When I mentioned the ORing change, I was implying that other portions of the code would also have to be in sync with this change. As in -- a simple switch statement like the follows will not be sufficient:
switch(tp->t_state)
{
case TCPS_SYN_RCVD:
}
As far as every possible succeeding state for a connection that is part of the FSM is denoted by a higher bit, the highest set bit in the t_state variable will always have the current state of the connection. The exception to this is .. when you move into TCPS_CLOSED state, clear out all bits.
So the states would look something like:
#define TCPS_CLOSED 0 /* closed */
#define TCPS_LISTEN 1 /* listening for connection */
#define TCPS_SYN_SENT 2 /* active, have sent syn */
#define TCPS_SYN_RECEIVED 4 /* have send and received syn */
/* states < TCPS_ESTABLISHED are those where connections not established */
#define TCPS_ESTABLISHED 8 /* established */
...
So switch based code fragments such as the above can be modified to
switch(HIGHEST_SET_BIT(tp->t_state))
{
case TCPS_SYN_RCVD:
}
Maybe I am missing something here, but theoretically wouldnt such a scheme work?. With this sort of a bit masking scheme the check that you make would still be valid because each state by itself is uniquely identified by its own bit position.
Regards
-AG
__________________________________________________________________
Get your own FREE, personal Netscape Webmail account today at http://webmail.netscape.com/
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1DEC8D97.1B5FD06C.375A6AF3>