Re: UVM and the NULL page

To: tech-kern%NetBSD.org@localhost
Subject: Re: UVM and the NULL page
From: Matt Thomas <matt%3am-software.com@localhost>
Date: Tue, 27 Dec 2016 18:50:51 -0800

Older ARM CPUs had the exeception page (page 0) that had to be mapped into the user's address space.  Fortunately that was solved in more recent ARMs with HIGHVECTOR and now you actually set the base address for the exception vectors.

PowerPC has completely split user/kernel space so that address 0 while in the kernel is a different mapping that user access would use.  In fact, page 0 on OEA was where all the exception vectors were.

One a different note, on many ports I've made access to a 0xdeadbeef address always generate a fault.  This was done by reserving +/- 32KB on either side preventing it from being returned by UVM.

References:
- Re: UVM and the NULL page
  - From: Joerg Sonnenberger
- Re: UVM and the NULL page
  - From: Paul Goyette
- Re: UVM and the NULL page
  - From: Wolfgang Solfrank
- Re: UVM and the NULL page
  - From: Maxime Villard
- Re: UVM and the NULL page
  - From: Thor Lancelot Simon
- Re: UVM and the NULL page
  - From: Eduardo Horvath
- Re: UVM and the NULL page
  - From: David Laight
- Re: UVM and the NULL page
  - From: Mouse
- Re: UVM and the NULL page
  - From: David Holland
- Re: UVM and the NULL page
  - From: Mouse
- Re: UVM and the NULL page
  - From: David Holland

Prev by Date: Re: UVM and the NULL page
Next by Date: In kernel mixing - audio cloning
Previous by Thread: Re: UVM and the NULL page
Next by Thread: re: UVM and the NULL page
Indexes:

Home | Main Index | Thread Index | Old Index