[ros-kernel] Re: CVS commit: ReactOS

[Phillip S. <ps...@cf...>]

This shouldn't be able to happen because the page fault path should 
reference/lock the address space/page tables as it is updating them, and 
that would prevent it from being unmapped.  It might be a good idea for the 
unmap code to be able to cancel pending paging IO though, and then 
unmap.   This way paging IO to a slow device, such as a slow network 
server, couldn't prevent a process from being terminated quickly.

At 10:20 PM 5/8/2002 +0000, you wrote:
>On Wed, May 08, 2002 at 03:29:18PM -0400, Phillip Susi wrote:
> > It should not be possible to delete a section while a page is being 
> faulted
> > in from it, because when the section is mapped, the section object should
> > be referenced, thus preventing it from being freed.  The fault doesn't 
> need
> > to reference the section, because it is referenced as soon as it is mapped.
> >
>Yes but one thread in a process could be faulting on a section while the
>other unmaps it. Of course this would be a very stupid application and
>it still might cause an exception but it shouldn't cause the kernel
>itself to crash or exhibit unboundedly undefined behaviour.