Date: Mon, 21 Sep 1998 14:27:32 -0500 From: Doug Ledford <dledford@dialnet.net> To: "Robert G. Brown" <rgb@phy.duke.edu> Cc: Heinz Mauelshagen <mauelsha@ez-darmstadt.telekom.de>, aic7xxx@FreeBSD.ORG, mge@u9ete.ez-darmstadt.telekom.de Subject: Re: IRQ deadlocks Message-ID: <3606A8A4.E4EC9BD8@dialnet.net> References: <Pine.LNX.3.96.980921114558.11003A-100000@ganesh.phy.duke.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Robert G. Brown wrote:
>
> On Mon, 21 Sep 1998, Heinz Mauelshagen wrote:
>
> > Does the vm bug show up in kernel > 2.1.120 on SMP too?
>
> I'm also curious; I know that there is a vm bug, but why does it strike
> one person and not another? I would have thought that if there was a
> really significant problem with the aic7xxx driver and virtual memory or
> memory mapping, the driver would fail for nearly everyone using the
> driver in an SMP system.
Because it isn't a driver issue at all. It's a race condition in the fork()
code that was worsened by the swapping of COW shared pages because now the
fork code has to call swap_duplicate() on any swapped out COW pages when it
forks off a new program, and then does a tlb flush afterwards, and somewhere
in there it is getting into a loop. When it didn't have to do the extra
work, it didn't happen as often. Also, the faster your disks subsystem
gets, the worse the problem is. Presumably this is because that with the
faster disks, something that should be happening after the fork completes,
is happening in the middle of the fork and causing problems. Anyway, that's
what I've gathered from some testing and from what Stephen Tweedie has had
to say about the problem.
--
Doug Ledford <dledford@dialnet.net>
Opinions expressed are my own, but
they should be everybody's.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-aic7xxx" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3606A8A4.E4EC9BD8>