To: davem@dm.cobaltmicro.com, sct@redhat.com, alan@lxorguk.ukuu.org.uk, linux-kernel@vger.rutgers.edu, number6@the-village.bc.nu, torvalds@transmeta.com, Subject: Re: PATCH: Raw device IO for 2.1.131 From: lm@bitmover.com (Larry McVoy) Date: Sat, 12 Dec 1998 10:44:38 -0800 > > Finally, I'd put all this lock down pages stuff into a mm/*.c file. > > I want to use it for networking if it goes in (zero copy TCP, NFS > > directly into page cache, etc.). > > Well for sendfile() it seems ideal for zero copying. I'd second it being in > mm/*.c - I want it for video capture I came into this late but I want this too. I need to bring you up to date on some stuff I'm trying to make happen with the disk/network people. I'm working with one of the big disk drive companies. They want to put networking on the disk drives. In order for this not to be a joke, there needs to be a way to essentially do DMA over the network - standard networking won't be fast enough (in other words, if a 386 can set up a DMA to move data off the disk at 20MB/sec, then the same 386 had better be able to do the same thing with a network disk - that isn't true today). There is a way of accomplishing the goal - SGI does it with something called scheduled transfer (you can sort of read about it at http://bitmover.com/talks/pp/slide01.html ) - they can do 791MBytes/sec over their GSN network using this technique. The technique requires that you can lock down pages and set things up such that the card does all the work - the OS sets up the "descriptors" and then stands back and lets the "network DMA" happen. I've agreed to work with SGI and the disk & networking people to try and make a ST-lite specification which would work for everyone. Linux is the primary target for this, so as soon as there is an implementation, you'll all get it. So, Linus, I know you haven't been thrilled about "raw I/O" and that you have some very valid reasons for not wanting it in the kernel. Does the patch for raw device I/O come anywhere near being good enough that you are OK with it? If not, can you suggest what it would need to look like? - 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/