> > Well I understand about 50% of what you just said. I can see how it
> > could be moved to the end of the task structure to make things easier
> > for re-sizing, but I don't undertand the 8K hack you mention. Will
> > this affect our patch that allocates a full 8Kb I/O bitmap?
>
> 2.1.10x keeps a pool of 8K blocks handy for task forks - thats because
> it may be unable to get two adjacent pages of memory at any given instant -
The TSS needs not to be _physically_ in adjacent pages. If all TSS were
(virtually) in the valloc area with address holes for the not allocated 8K
IO map, then to map some more pages into later should be easy.
i.e.: you would need 16K address space for each TSS but you need not to
allocate pages for all of them.
This technique would also have the advantage that you won't need to keep
the above mentioned pool for the TSSs.
Hans
<lermen@fgan.de>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html