On 27-Apr-98 Jason Thorpe wrote:

> Um, I wasn't trying to be insulting... Just trying to point out that your
> evidence doesn't really point to a memory leak in the kernel...
>
> I have been reading your messages in their entirety.

Well, I didn't think you were trying.  It was trying, though, being late....

But the "evidence" (if you could call it that) really does -- on the surface at
least -- point to _something_ in the kernel.  The reason is (as I strove to
make clear) that while hundreds of kilobytes at a time are piling up on the
active list, only a couple of pages of that can be found (according to ps)
adding to the vm space of ANY program, including the X server.  Eventually, in
multi-megabyte herds they migrate to the inactive list, then finally to swap,
where they accumulate until you restart the session or it grinds to a thrashing
halt.  Through all of that, NO program has accounted for them, including the X
server.

I used to assume the X server was causing this, or mallocs, or window managers,
or whatever.  More recently I discovered that is not the case, that the memory
cannot be explained as accumulating in ANY user-level program.

That is the point many of you seem to be having trouble getting.  This is
perhaps understandable....I'm trying to explain something I've been looking at
for a long time, and I'm throwing it at you in a form that says it all to me
while probably not articulating so well for you.

Again, Chuck just suggested a way to use ddb with X, so I may learn more about
this soon....


--Jim Howard  <jiho@mail.c-zone.net>


----------------------------------
E-Mail: jiho@mail.c-zone.net
Date: 27-Apr-98
Time: 12:29:31

This message was sent by XFMail
----------------------------------