Subject: Re: uvm_pageidlezero()
To: matthew green <mrg@eterna.com.au>
From: Chuck Silvers <chuq@chuq.com>
List: tech-kern
Date: 06/21/2005 06:39:10
On Tue, Jun 21, 2005 at 10:02:44AM +1000, matthew green wrote:
>
> > IIRC, some ports don't call uvm_pageidlezero as it's considered harmful.
>
> Why? Cache poisoning?
>
>
> it also interfered with UBC when it arrived. enable it and run some
> tests and see if it works better for you :-)
I've had it on in my test environment for a couple years now, I think.
I meant to go check if it had any discernable negative impact anymore,
but, uh, I forgot. the problem before was that we were mapping the
page to be zeroed with an uncached mapping, which slowed down DMA a lot.
now we use a cached mapping, so the problem (if there is one) will be
excessive cache flushing instead. it'd be nice if someone looked into this.
-Chuck