Path: utzoo!attcan!uunet!lll-winken!lll-tis!ames!ucsd!ucsdhub!hp-sdd!hplabs!hpda!hpcupt1!hpisoa1!vandys
From: vandys@hpisoa1.HP.COM (Andrew Valencia)
Newsgroups: comp.unix.wizards
Subject: Re: P/V using SysV semop(2)
Message-ID: <2200030@hpisoa1.HP.COM>
Date: 30 Aug 88 16:31:11 GMT
References: <1001@uwbull.uwbln.UUCP>
Organization: Hewlett Packard, Cupertino
Lines: 14

>Ah, but all processes sleeping awaiting a semaphore "unblock" sleep with the
>same priority (PSEMN)- the CPU consumption factor (p_cpu) is not taken into
>consideration when assigning priorities to sleeping processes.  Processes
>are removed from the run queue using the following rules:

    You're right; p->p_pri is used to choose the run queue (unless the process
is real-time, but that's another story).  Since it's set in sleep(), the
priority used to select a run queue is unchanging.  However, our VM system
(both paging and swapping sections) takes a bigger picture of the process's
behaviour.  If it isn't involved (for instance, on an empty system),
then you're quite likely to see something approaching FIFO.  Once it gets
involved, all bets are off, and the more starved processes are likely to
start catching up.

					Andy