Re: constant data rate streaming

[James Simmons]

On Tue, 5 Oct 1999, Andreas Beck wrote:

> > the DRI method of coorperative locking from userland. I think this method
> > sucks. 
> 
> Depends. If you have a good card, this is the way to go. If it cannot be
> crashed/locked/burned by a bad program, all is well.
> 
> If it can, we have to protect it.

I agree. I think the best way is to create a struct mem_locks that has a
linked list of locks to test and a field to discribe what card is using
mmio thats mmapped to user space. Place this in the task_struct. In the
schedular test to see if mem_lock is their then see if its lock by another
process. If it is then skip picking this process.

> 
> CU, ANdy
> 
> -- 
> = Andreas Beck                    |  Email :  <[EMAIL PROTECTED]> =
>