Re: "raw" I/O....

Alan Cox (alan@lxorguk.ukuu.org.uk)
Fri, 18 Oct 1996 21:59:31 +0100 (BST)


> This isn't an argument for raw I/O: it's just an observation that
> the model of secondarymem<>primarymem<>userapp is not necessarily
> the one to make the most efficient - perhaps this why more kernel
> process/daemons have crept into the kernel since the 1.2.X days?

We've acquired some NFS readahead daemons and an asynchronous pager. I don't
think either of those could do well in userspace ;)

The NFS daemon now does use mmap so that cuts down the performance problems.
Until I can persuade Linus that I want to allocate virtual address space and
fill the pages in for it on interrupts however there will be problems with
large network buffers. Thats something a pile of ATM folk are playing with
ideas for resolving