Subject: Question on some kernel options, NetBSD 3.0
To: None <netbsd-users@netbsd.org>
From: Mark Cullen <mark.r.cullen@gmail.com>
List: netbsd-users
Date: 06/10/2006 21:10:33
I've successfully recompiled a kernel and removed un-needed options etc, 
but I see a few options with "experimental", and was just wondering if 
they really are still experimental or not (from GENERIC on 3.0).

1) options UFS_DIRHASH       #UFS Large Directory Hashing - Experimental
2) pseudo-device   nsmb      # experimental - SMB requester
3) #options        NEW_BUFQ_STRATEGY

Are UFS_DIRHASH and nsmb really still experimental? Are there known bugs 
causing panics (or worse) that I should know about, before I start using 
them? :-)


As for NEW_BUFQ_STATEGY, the comments sound like it would be very 
beneficial:

"Enable experimental buffer queue strategy for better responsiveness 
under high disk I/O load. Likely stable but not yet the default."

I am noticing some pretty severe slow-down's when I, for instance, untar 
pkgsrc. It was really quite bad actually, while it was doing it I did a 
`ps ax`, which took around 30 seconds (possibly more) to return and 
print to the console! If it makes any difference, I am using UFS2 and 
softdep.

I get the feeling that the GENERIC config supplied may be a little bit 
dated, as I seem to remember some mention of another, different strategy 
option in NetBSD 3.0? Does this mean that NEW_BUFQ_STATEGY should 
probably be avoided, much like the soon-to-be-removed vinum? I've read 
that people were experiencing lock-up's under certain situations with 
this option, but I think that was some time ago now. It might have got 
fixed since then?

Thanks in advance,
Mark