[coyotos-dev] Scratch Pages, Fault on First Reference

[Jonathan S. Shapiro]

On Sun, 2006-01-15 at 20:29 -0800, Charles Landau wrote:

> Normal state: The process/thread is assuming that all scratch pages 
> are initialized, and wants to be faulted if that is not the case. 
> Such a fault changes the process state to:
> 
> Initializing state: The process is aware that some scratch pages are 
> uninitialized. The process will arrange to initialize the pages (and 
> should not be faulted while doing so), synchronizing with other 
> threads as necessary. As each page is initialized, the process 
> changes its state (there might be a race here). When the process 
> thinks it is done, it changes its state back to normal.

This design relies on the assumption that all processes are
collaborating. Perhaps Marcus/Neal can comment on whether this fits
their use case?

shap