From mboxrd@z Thu Jan 1 00:00:00 1970 From: lm@mcvoy.com (Larry McVoy) Date: Mon, 20 Nov 2017 21:20:23 -0800 Subject: [TUHS] UNIX on S/370 In-Reply-To: References: <20171121025646.380EB18C0B0@mercury.lcs.mit.edu> <05e701d36277$0db803b0$29280b10$@ronnatalie.com> <20171121035112.GC9093@mcvoy.com> Message-ID: <20171121052023.GE9093@mcvoy.com> On Mon, Nov 20, 2017 at 10:14:41PM -0700, Warner Losh wrote: > I assume you aren't talking about things like mmap where you can't really > bcopy it... Correct. > At Fusion I/O we had hooks into our PCIe flash card driver that would do > DMA directly into user buffers (since we wanted IOPS and any extra copies > got in the way of that). Yeah, cool, sorta like the SGI stuff, but that was stuff that you had to add and SGI had to add. The only place I can see dma right into user buffers would be if the user blocked in the read, it was to a raw device, *and* the OS took care to make sure that those page[s] were locked. I can see the 1st 2 but I've not run into the last one except for specialized stuff like O_DIRECT.