[9fans] authentication protocol question

[9fans] authentication protocol question

[presotto]

The race is on the client side.  There can be a number
of simultaneous attaches through the same communications
channel.  Nothing synchronizes them so the authentications
can come out of order.  The important part is that the
id's aren't reused.

[9fans] authentication protocol question

[Mike Haertel]

I am puzzled by the following code in /sys/src/fs/port/auth.c.

	/*
	 *  the id must be in a valid range.  the range is specified by a
	 *  lower bount (idoffset) and a bit vector (idvec) where a
	 *  bit set to 1 means unusable
	 */
	lock(&cp->idlock);
	x = a.id - cp->idoffset;
	bit = 1<<x;
	if(x < 0 || x > 31 || (bit&cp->idvec)){
		unlock(&cp->idlock);
print("id out of range: idoff %ld idvec %lux id %ld\n", cp->idoffset, cp->idvec, a.id);
		return 0;
	}
	cp->idvec |= bit;

 From the auth(6) man page, I had the impression that the id
field was supposed to be a simple sequence number.  Is there a
race in some client in which the order of the sequence numbers
can get rearranged?  What is the race?