Re: More 2.2.17pre9 VM issues

[Rik van Riel]

On Thu, 29 Jun 2000, Andrea Arcangeli wrote:

> >Another thing which I would like to receive comments from you and other
> >people is my patch to fix the IO completion wait problem at: 
> >
> >http://bazar.conectiva.com.br/~marcelo/patches/v2.2/2.2.17pre9/wait-dirty.patch
> 
> I still don't love this 64 (aka random ;) thing:
> 
> +                               wait = ((gfp_mask & __GFP_IO)
> +                                       && (!(nr_dirty % 64)));

Oh, you are right here, but there are two important
things to keep in mind:

1) it's not like the algorithm depends on the value being
   64, you may as well use 16 or 128, it doesn't matter
   (this is the magic number test ... if the algorithm
   DEPENDS on the magic number it's most likely wrong, but
   if the value of the magic number doesn't matter thatm uch
   it may just be right)

2) this fix, no matter the downsides, has made 2.2 stable
   under heavy VM load.

Of course we'd all like to see some change that keeps the
stability and makes performance higher, but until we've
figured out how to do that we should leave this thing in
the kernel.

regards,

Rik
--
The Internet is not a network of computers. It is a network
of people. That is its real strength.

Wanna talk about the kernel?  irc.openprojects.net / #kernelnewbies
http://www.conectiva.com/		http://www.surriel.com/