Skip site navigation (1)Skip section navigation (2)
Date:      22 Jan 2001 23:37:58 +0100
From:      Dag-Erling Smorgrav <des@ofug.org>
To:        "Bosko Milekic" <bmilekic@technokratis.com>
Cc:        <arch@FreeBSD.ORG>
Subject:   Re: Second zone allocator patch
Message-ID:  <xzpy9w3gqm1.fsf@flood.ping.uio.no>
In-Reply-To: "Bosko Milekic"'s message of "Mon, 22 Jan 2001 14:35:05 -0500"
References:  <xzp3decp50i.fsf@flood.ping.uio.no> <015201c084aa$6c356800$1f90c918@jehovah>

next in thread | previous in thread | raw e-mail | index | archive | help
"Bosko Milekic" <bmilekic@technokratis.com> writes:
> Dag-Erling Smorgrav wrote:
> >  - use atomic operations for updating the statistics counters.
>     Can you get away with grouping the changes to these counters while
> holding the zlist lock? If so, you can get rid of the necessity for the
> atomic() ops.

Well, right now they're updated inside _zget(), which holds the zone
lock, and the zlist lock shouldn't be held while holding the zone lock
(it'll cause a lock order reversal if vm_zone_sysctl() runs at the
same time). Are atomic operations too expensive? Since these are just
statistics, can we get away with leaving them as they are, at the risk
of getting incorrect stats once in a while?

> >  - make the rest of the VM system thread-safe so we can dispense with
>        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>     I can't believe you actually said this publically. :-) Now you're going
> to have to assume responsability for it. :-P

I know, I know. I'm planning to take a long hard look at the kmem
system, I'll pass it on to someone else if it proves to be beyond my
skills.

I've also been asked to implement some kind of zone destructor (which
appears to be needed to allow unloading the NFS kld) - it should be
simple enough. I'll do that first, right after I a) fix the bugs Bruce
pointed out and b) write a man page for the zone allocator.

> >    the Giant hackery in zinitna() and _zget().

The Giant hackery was not committed, btw - Jason objected. Since the
zone allocator still runs under Giant anyway, it shouldn't matter yet.

DES
-- 
Dag-Erling Smorgrav - des@ofug.org


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?xzpy9w3gqm1.fsf>