Date: Mon, 19 Feb 2001 13:25:03 -0800 (PST) From: John Baldwin <jhb@FreeBSD.org> To: Matthew Jacob <mjacob@feral.com> Cc: alpha@FreeBSD.org, jake@FreeBSD.org Subject: RE: old fishwraps.. Message-ID: <XFMail.010219132503.jhb@FreeBSD.org> In-Reply-To: <Pine.LNX.4.21.0102191259410.19274-100000@zeppo.feral.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 19-Feb-01 Matthew Jacob wrote:
> (taking this to alpha where it belongs)
>
>>
>> On 19-Feb-01 Matthew Jacob wrote:
>> >
>> > Welcome back john- this is all *very* old news.....
>> >
>> > Right now we're more or less functional in the alpha-kernel front.. I
>> > still
>> > need to add the oncpu/lastcpu stuff to alpha- I've been waiting for a
>> > response
>> > from Jason Evans about making these integers (instead of chars) because
>> > that
>> > would make life on the alpha/genassym (or sparc, for that matter) a lot
>> > easier.
>>
>> Making p_stat an integer is something we might need to do as well as right
>> now
>> proc assumes that a 32-bit wordsize in its alighnment. And also, one thing
>> you
>> missed is that clear_resched()'ign in userret() as is currently the case is
>> wrong. It needs to be in mi_switch() instead, but mroe on that later.
>
> Oh- yes- sorry- I haven't checked my latest copy in because I've been
> cheesing
> it and jumping out of assembly to C-code to 'fix' this stuff- but it actually
> *is* being called from cpu_switch (see below for my current patches) but the
> hack currently is:
I mean sometihng like this:
Index: kern_synch.c
===================================================================
RCS file: /usr/cvs/src/sys/kern/kern_synch.c,v
retrieving revision 1.125
diff -u -r1.125 kern_synch.c
--- kern_synch.c 2001/02/09 06:09:50 1.125
+++ kern_synch.c 2001/02/12 21:55:06
@@ -969,6 +973,7 @@
PCPU_SET(switchtime, new_switchtime);
CTR4(KTR_PROC, "mi_switch: old proc %p (pid %d, %s), schedlock %p",
p, p->p_pid, p->p_comm, (void *) sched_lock.mtx_lock);
+ clear_resched();
cpu_switch();
CTR4(KTR_PROC, "mi_switch: new proc %p (pid %d, %s), schedlock %p",
p, p->p_pid, p->p_comm, (void *) sched_lock.mtx_lock);
(
Currently, p_oncpu and p_lastcpu are set in cpu_switch, but to be honest,
there's no reason to. It would be quite fine to do something like this:
curproc->p_lastcpu = curproc->p_oncpu;
curproc->p_oncpu = -1;
clear_resched();
cpu_switch();
curproc->p_oncpu = PCPU_GET(cpuid);
and just make it all MI and in mi_switch(). Some other stuff such as the
sched_lock dinking can also be done this way. In fact, there are several
things in cpu_switch() that don't really belong there, IMO. It should be for
MD stuff such as saving and restoring registers.
--
John Baldwin <jhb@FreeBSD.org> -- http://www.FreeBSD.org/~jhb/
PGP Key: http://www.baldwin.cx/~john/pgpkey.asc
"Power Users Use the Power to Serve!" - http://www.FreeBSD.org/
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.010219132503.jhb>