On Sat, 12 Dec 1998 14:09:48 -0800 (PST), Linus Torvalds
<torvalds@transmeta.com> said:
> Quite frankly, nobody has EVER given me a reason that makes any kind of
> sense at all for supporting raw devices in any other way than we already
> do. Nobody sane uses a disk without a filesystem, and the insane people
> that do I feel we can and should ignore. Insanity has a way of dying off
> over time, when Darvin starts to look into it.
Don't read too much into the implementation's restriction to raw
devices: the core rw_raw_io as implemented already takes, and uses, an
optional bmap() argument, and the _only_ thing missing for a full
ext2fs O_DIRECT right now is cache coherency.
The aim here is O_DIRECT, for filesystem access and not just for raw
devices. The block device interface just happens to be sufficiently
simple to be worth testing first.
> I've also become almost completely uninterested in zero-copy schemes. They
> make for great benchmarks, but have little relevance to anything else. You
> have two cases:
> - sendfile() like copying things around. We want to do zero-copy here,
> but it's easy because everything is already in kernel space.
> - real work. If zero-copy makes a real difference, you'd better buy a new
> computer, because your memory subsystem is too slow. Either you're IO
> bound or you aren't, and if you're IO bound and your memory is slower
> than your IO then it's not something the OS should do all that much
> about. And if you aren't IO-bound, then the whole discussion is moot.
I really don't think it's an issue of memory bandwidth: it's about CPU
utilisation, and to some extent about cache trashing. Right now, on
this P200, reading from /dev/sda1 (a paltry 5MB/sec over ncr scsi)
takes 18% system-time cpu utilisation, and reading from /dev/rsda1
takes 4%. That's a heck of a lot of cpu time which my database
application could be using for something more useful. Large
data-shuffling applications _are_ often cpu bound, or people wouldn't
be running them on 4-way SMP boxes.
--Stephen
-
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.tux.org/lkml/