Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 21 Jan 2002 00:03:27 -0800
From:      Terry Lambert <tlambert2@mindspring.com>
To:        raman@hello.to
Cc:        stable@freebsd.org, current@freebsd.org
Subject:   Re: Extended paging bug in Athlon affect FreeBSD?
Message-ID:  <3C4BCB4F.60F1D4FC@mindspring.com>
References:  <20020121030018.7425.qmail@web14006.mail.yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Raman Ng wrote:
> I don't know what mailing list this mail should be
> posted to, sorry if it is posted to the wrong mailing
> list.
> 
> Recently I found Linux 2.4 kernel is affected by the
> bug of extended paging in AMD Athlon through the
> following link.  I don't know if FreeBSD is also
> affected.
> 
> http://linuxtoday.com/news_story.php3?ltsn=2002-01-21-001-20-NW-KN

I am well aware of this bug.

It does not affect FreeBSD, which only uses 4M pages for
the first 4M of the kernel itself.

I've worked on code that enables 4M pages on other memory
used in FreeBSD, that had this problem, but only if you
were really stupid in your allocation mechanism.

There's a workaround for this problem which is fairly
trivial to implement in software, and should probably be
done when 4M pages are enabled, if you are using an Athlon,
and are adding 4M pages.  The problem didn't bite my own
application, since I knew about it and used the TLBs in a
very specific way to get around the problem, but I'm told
that after I left the company, someone went in and used
them a different way, not knowing about the problem or the
workaround, and broke things.  This resulted in about a
14% performance loss.

If you are doing heavy networking on a machine with a lot
of memory, it's reasonable to move your mbufs into 4M pages,
which nets you a 4-14% improvement (depending on whether or
not you also change your allocator: 14% if you do).  If you
do that, knowing how to work around the problem nets you a
pretty big win in overall performance, for about 6-10 more
machine instructions at startup time.  8-).

In any case, this will not be a problem for FreeBSD, and is
only a problem for Linux because of the strange way they
initialize things.

-- Terry

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




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3C4BCB4F.60F1D4FC>