Re: [patches 2/2] struct switchframe/pcb improvements

[christos%astron.com@localhost (Christos Zoulas)]

In article <52F92D38.8070303%bbn.com@localhost>, Richard Hansen  
<rhansen%bbn.com@localhost> wrote:
>-=-=-=-=-=-
>
>Hi all,
>
>I discovered a couple of issues in one of the patches I sent.  I've
>attached a new version of the patch, and commented on the issues below:

You really cannot (or should not) dereference kernel/user addresses
(even when it happens to work), and use the db_read_foo() and
pointer + offsetof() style so that things always work. If it is not
possible to avoid the ifdef for the base pointer (perhaps doing the
same thing we did for i386 to amd64), then it is best if the ifdef
is centralized in one place and hidden behind a macro. But dsl@ should
really take a look at all of this.

christos