Re: [PATCH] Double-level file descriptor allocation

[Skip Ford <skip.ford@xxxxxxxxxxx>]

>     This patch is a lot cleaner then what DES did, but doesn't quite
>     solve the problem either.
> 
>     Would either you or Skip like to adapt *just* the fd_used() and 
>     fd_unused() pieces, without any of the bitmap stuff?  i.e. don't
>     implement fd_lomap or fd_himap or anything like that, just create
>     the fd_used() and fd_unused() abstraction and the structural
>     field manipulation they do on the original structural fields.  The
>     allocation, reallocation, and freeing code also has to be abstracted
>     into API functions.

I'd be willing to give it a shot.  I've gotten familiar with the code so I
understand the API you want.  If nobody else jumps in to do it, I'll try
to do it.  Sounds much cleaner.

>     might as well use a radix tree with simple first-free hinting

I might need some pointers on that part though.

-- 
Skip