Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 03 Mar 2005 07:49:33 -0700
From:      Scott Long <scottl@samsco.org>
To:        Ryan Sommers <ryans@gamersimpact.com>
Cc:        David Schultz <das@freebsd.org>
Subject:   Re: Removing kernel thread stack swapping
Message-ID:  <422723FD.6030103@samsco.org>
In-Reply-To: <42271B6A.4070802@gamersimpact.com>
References:  <20050303074242.GA14699@VARK.MIT.EDU> <42271B6A.4070802@gamersimpact.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Ryan Sommers wrote:
> David Schultz wrote:
> 
>> Any objections to the idea of removing the feature of swapping out
>> kernel stacks?  Unlike disabling UAREA swapping, this has the
>> small downside that it wastes 16K (give or take a power of 2) of
>> wired memory per kernel thread that we would otherwise have
>> swapped out.  However, this disadvantage is probably negligible by
>> today's standards, and there are several advantages:
> 
> 
> I like the idea of fixing a lot of possible panics. However, I don't 
> know if we should nix it completely. Wasting this little memory won't 
> hurt anyone on a contemporary computer. However, our embedded systems 
> folks don't look at memory in the same light, and 16K here or there can 
> begin to really add up in a memory tight architecture. Of course it 
> could be argued that embedded systems probably don't have many threads, 
> many threads that can be swapped, or even swap space in the first place.
> 
> I guess it's a judgment call that one of our embedded systems engineers 
> could better answer.
> 

Please stop thinking about the process table in terms of what you see on
your desktop or laptop.  FreeBSD is a whole lot more than that, and we
have to keep it in perspective.  16k or 32k might not seem like much,
but it really does make a difference when you look at the process
workload on a real server and is one of the reasons that FreeBSD has
always performed well.  Yes, it's hard to get it right, but it's not as
hard as David Xu and David Shultz are trying to project it, and it's a
worthwhile feature.  I formally object to removing this feature, and I
heavily encourage the current problem (sigwait) to be fixed and the
uninformed FUD to stop.

Scott



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?422723FD.6030103>