2009-03-20 08:16:54

by Raymond Jennings

[permalink] [raw]
Subject: question about pt locking

Just curious, but does a process effectively have a read-lock on its
page tables when it's in userland?

-- a curious newbie


2009-03-20 08:53:17

by Peter Zijlstra

[permalink] [raw]
Subject: Re: question about pt locking

On Fri, 2009-03-20 at 01:16 -0700, Shentino wrote:
> Just curious, but does a process effectively have a read-lock on its
> page tables when it's in userland?

No, page reclaim can fiddle with a running process' page tables
concurrently.

2009-03-20 09:24:35

by Raymond Jennings

[permalink] [raw]
Subject: Re: question about pt locking

On Fri, Mar 20, 2009 at 09:52:10AM +0100, Peter Zijlstra wrote:
> On Fri, 2009-03-20 at 01:16 -0700, Shentino wrote:
> > Just curious, but does a process effectively have a read-lock on its
> > page tables when it's in userland?
>
> No, page reclaim can fiddle with a running process' page tables
> concurrently.

Yeah just now remembered why the kernel has tlb inval/flush routines

...duh...

Thanks for the answers though, I learned something today.