Re: RFC: Slab-o-rama (after Rene's suggestion)

On 4/23/07, Axel Dörfler <[email protected]> wrote:
> Oh, I left out a few things here:
> you'll have to solve many circular dependencies (especially during
> bootstrapping code); ie. creating an area needs memory for some
> structures, too, and you'll also need memory for the slab structures as
> well.
> If you want to tackle this, have a look at the FreeBSD implementation
> (sys/vm/uma_core.c) - I think they solved those issues in a nice way.

  Yeah, we need to have a couple static structures during bootup to
init the caches that will serve the VM and heap allocators. This is
something that will have to be done eventually when the allocator gets
deeply merged into the system. Right now i think we could have a more
iterative approach, incrementing the number of users without changing
the core bits. I still haven't looked at some of the kernel's VM
details yet, and i hope Travis will help there, as eventually we'll
have to be able to block to obtain memory (thus the flags argument to
the Backend, as to consider the possibility of sleeping).

  Hugo

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/