[Gllug] Re: Anyone having broadband (probably PlusNet) problems today?

Liam Smit liam.smit at gmail.com
Mon Apr 4 17:18:54 UTC 2005


I believe that AMD has made an effort at an IOMMU with their AMD64
line of processors, I believe Intel didn't quite get that done for
their equivelant processor.

cheers
Liam


On Mon, 4 Apr 2005 12:41:56 +0100 (BST), Christian Smith
<csmith at micromuse.com> wrote:
> On Wed, 30 Mar 2005, Nix wrote:
> 
> >On Tue, 29 Mar 2005, Russell Howe whispered secretively:
> >> On Tue, Mar 29, 2005 at 10:35:13AM +0100, Nix wrote:
> >>> Actaully, in general, if the hardware is DMA-capable, the firmware
> could
> >>> do anything in RAM it liked, including (if the firmware were malicious
> >>> or buggy) introducing security holes.
> >>>
> >>> Disk controllers tend to be DMA-capable. :)
> >>
> >> True, and I suppose this would bypass the memory protections afforded by
> >> the CPU and/or MMU as I guess the DMA transfer would bypass this?
> >
> >Yeah, DMA is direct-to-physical-memory: you lose.
> >
> >(This is why scatter-gather support is so important: contiguous virtual
> >memory blocks tend to be discontiguous in physical memory...)
> 
> 
> It's also why machines such as SUNs have a IOMMU, which provides
> protection and linear access to sparse physical memory addresses from IO
> cards. The kernel can then only allow IO cards to access the proper memory
> required for operation.
> 
> Christian
> 
> -- 
>     /"\
>     \ /    ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
>      X                           - AGAINST MS ATTACHMENTS
>     / \
> -- 
> Gllug mailing list  -  Gllug at gllug.org.uk
> http://lists.gllug.org.uk/mailman/listinfo/gllug
>
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list