Re: On PATH_MAX

[Jonathan S. Shapiro]

On Tue, 2005-11-08 at 17:34 +0100, Bas Wijnen wrote:
> On Tue, Nov 08, 2005 at 08:51:09AM -0700, Christopher Nelson wrote:
> > > Not exactly.  If a server wants to support arbitrary long 
> > > paths, it's not going to map the whole thing into its address 
> > > space.  It'll accept a container capability and map parts of 
> > > it in, unmapping other parts of it.
> > 
> > What mechanism allows it to map *parts* of a single transfer in?
> 
> There is no transfer of data, there is only transfer of the container
> capability.  This capability gives access to a set of pages, which can be
> mapped in or out of the address space when the process likes.

Operationally this is correct, but the mechanism described is
unfortunate if your description is intended literally.

If I hold a capability to a memory object, what I have is the authority
to map a *window* on each page within that object, subject to the
permission restrictions of the memory object.

The difference between mapping such a window and mapping the page
directly is that the window becomes invalid if the memory object is
destroyed.

The distinction is subtle, but important.


shap