Subject: Re: "swap lock" problem with NetBSD VM
To: None <laine@MorningStar.Com>
From: John S. Dyson <email@example.com>
Date: 09/17/1996 21:17:57
> It would sure be nice if this was in the released code so that the
> performance of stock NetBSD under heavy load was not so embarrassing
That is *exactly* the reason why I sent out the fix. I did NOT
want any of the *BSD's to look bad for such a silly reason...
When I saw the code, I felt that a one line pseudo-fix wouldn't be
too much of a conflict of interest :-).
> Thanks John! You'll be disappointed to know that you have singlehandedly
> prevented me from further investigating FreeBSD :-)
There are other reasons why FreeBSD could be better for you,
there has been alot of work on it to improve the performance in many
ways. Not only does FreeBSD have many such bug fixes, but it also
has significant structural improvments. I can't remember if FreeBSD
ever had this bug -- but it was probably one of the first that we
fixed (about 3yrs ago or so), if it did exist.
Frankly the original Lite/(1,2) pageout daemon still needs alot of
work, even with that bug fix. But if the worst of your problems are
gone, then at least you can continue along. (Note that I have a
purist attitude of solving the problem correctly, but also we can
help with making the problem less severe until the problem is
"correctly" solved.) At the risk of sounding nasty, and I really
don't mean it, the pageout performance is probably now as good as
Linux was on 1.2.13 or FreeBSD (V1.1,V2.0), FreeBSD V1.1.5 was a
"sweet spot"... (Both FreeBSD and Linux have come a very long way
since those days.)
> By the way, this same fix works on OpenBSD (as would be expected), so it
> should probably be put in there too.
Yep, I informed the OpenBSD people privately in parallel with the NetBSD
discussions, since I wasn't sure if any of them were on the appropriate
mailing lists. Since they were going to include the patch anyway, I
thought that it would be useful to them to explain why it was needed.