Date: Sat, 2 Aug 2003 10:56:33 -0400 (EDT) From: Robert Watson <rwatson@freebsd.org> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: Andy Farkas <andyf@speednet.com.au> Subject: Re: headsup: swap_pager.c Message-ID: <Pine.NEB.3.96L.1030802104607.98114F-100000@fledge.watson.org> In-Reply-To: <8035.1059809404@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 2 Aug 2003, Poul-Henning Kamp wrote: > In message <20030802100150.H39348-100000@hewey.af.speednet.com.au>, Andy Farkas > writes: > > >I wasn't going to say anything (you can delete now) but the more I think > >about it, the more I think "if it aint broke, dont fix it". > > It is broken, it contains a bogo-vnode and it wastes RAM (it may be > cheap but you shouldn't use 4 times the necessary RAM). For me, the biggest benefit of this change is the removal of the bogo-vnode; I'd just like to make sure we don't see an observable performance hit in interesting cases. To clarify the notion of bogo-vnodes, for those less familiar with it, there are a few vnodes that are randomly pulled out of hats for some arbitrary things. They break some of the normal working assumptions of vnodes -- for example, almost every vnode in the system has a non-NULL v_mount pointer, and is associated with a mountpoint. The vnodes I know of that don't meet this property are the swap striping vnode, and the vnodes "pulled out of a hat" to refer to devices when you don't yet have access to the device file system. This occurs in bdevvp(), which converts a dev_t into a dummy vnode pointing at the device. Under normal circumstances, non-dead vnodes always have a file system they are associated with. Robert N M Watson FreeBSD Core Team, TrustedBSD Projects robert@fledge.watson.org Network Associates Laboratories
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.96L.1030802104607.98114F-100000>