On Mon, 5 Jan 2004 18:42:57 -0500 Thor Lancelot Simon <tls@rek.tjls.com> wrote:

> On Tue, Jan 06, 2004 at 09:27:54AM +1000, George Michaelson wrote:
> > 
> > I know this kind of non-specific whinge is (worse than) useless, but the new
> > mem alloc/kernel thing seems to suck. laptop user, current as of 2-3 days 
> > ago. 256Mb memory.
> 
> I don't know which change you think you're referring to, but I suspect that
> you mean PK's recent changes to the buffer cache code.

Yes. I think thats what I'm referring to.

> 
> I just looked at that code again and I'm a bit curious about some aspects
> of it -- such as the use of, essentially, different LRUs for buffers
> of different sizes, which seems a bit bogus -- but I can't see how it
> could possibly cause additional demand for *pageable* memory, which is
> what would be necessary to cause an increase in swap utilization.
> 
> I think some other recent change must be causing your problem.

Its bizarre. the apparent symptoms are: slowness re-painting X11 (eg screen
refresh from a virtual desktop re-position for mozilla, which is widget rich and
requires lots of discrete X callbacks to repaint) and a lot more disk noise than
usual, coupled to long stretches where interactive X response is slow, no window
repaints, no kbd/mouse input works, huge amounts of disk noise, and then a
splurge of activity. top/xosview don't show much I can detail. the reaper and
ioflush tasks in kernel are (to my eyes) unusually active.

In the past, NetBSD users with the luxury of more than one host have said X and
compile/disk don't sit well together, but for a single-instance user this mix is
quite normal.

Oh well. somebody with more clue than me will see this, and debug it better!

cheers
	-George

> 
> Thor


-- 
George Michaelson       |  APNIC
Email: ggm@apnic.net    |  PO Box 2131 Milton QLD 4064
Phone: +61 7 3858 3150  |  Australia
  Fax: +61 7 3858 3199  |  http://www.apnic.net