On Mon Dec 03 2007 at 09:53:26 +1100, Sarton O'Brien wrote:
> Hi Andrew,
> 
> Andrew Doran wrote:
> >This is a bug in LFS. It spends so long traversing mnt_vnodelist and 
> >syncing
> >all the vnodes that the other CPU gives up trying to acquire kernel_lock.
> >
> >Workaround: don't use LFS, it will always give you a headache in the end.
> >  
> 
> Should this be considered a rule until further LFS improvements are made 
> or is this unlikely to change?
> 
> I'm starting to wonder, with all the LFS tweaking required for 
> performance and then the underlying issues with regards to the above and 
> trigerable panics using lfs -> smb/cifs,  is it worth using LFS at all? 
> I'm not sure if I'm seeing the benefits.

I use it for /obj all the time, no tweaking, works great.

-- 
Antti Kantee <pooka@iki.fi>                     Of course he runs NetBSD
http://www.iki.fi/pooka/                          http://www.NetBSD.org/
    "la qualité la plus indispensable du cuisinier est l'exactitude"