IRC logs for #openrisc Monday, 2016-10-24

--- Log opened Mon Oct 24 00:00:11 2016
shorneolofk: I have access to some not all.  Would it be possible to grant me to all?  I am trying to do some doc cleanups i.e. finish todo's on and add some readmes to repo like openrisc/doc etc04:50
shorneHello, anyone have any problem with me adding a moving banner to the wiki?  i.e.05:54
ZipCPU|LaptopYeah ... I tend to ad-block anything that moves.  Such things are just too annoying.06:48
-!- Netsplit *.net <-> *.split quits: stekern, SMDwrk, olofk, rhythmx08:45
-!- Netsplit over, joins: SMDwrk08:52
SilentHello :)16:22
SilentI have just discovered OpenRisc, and I have to say that it looks very interesting16:23
kc5tjaGreetings.  Just FYI, due to time-zone issues, you might not get a response to your comments here for many hours, so please be patient.  :)  I happened to be at keyboard to see your text.16:40
SilentIt's ok, I can live with that16:41
SilentAlso sometimes I am on very late or very early so that is ok16:41
kc5tjaJust wanted to chime in, as I've seen folks ask for help, then leave 15 minutes later expecting a real-time response.  :)16:41
SilentThanks for the heads-up16:42
SilentI am currnetly watching lectures to get familiar with OpenRISC16:42
SilentIs desktop or workstation usage also a target, or only space exploration? ;)16:42
kc5tjaI'm actually more familiar with RISC-V than with OpenRISC, but from what little I know about it, I don't think anyone ever had a real "target" application for it.16:47
kc5tjaIt was just a project that the creators had hoped to spread to different areas beyond embedded.16:48
SilentWhat brings me here is the quest to secure hardware16:50
SilentIt may seem like it is a battle lost already, but if we can use FPGAs this may change16:51
SilentMaybe as thecnology advances FPGAs and similar technologies become powerful enough to support a full desktop CPU implementation16:52
SilentOr at least other applications, such as security gateways or IoT devices16:54
ZipCPUSilent: If you are interested in secure hardware, you're starting to knock on the right doors.17:38
ZipCPUAs with kc5tja, I'm not an OpenRISC guy myself.  I'm more of a ZipCPU guy, but I tend to hang around this bunch.17:40
SilentSo are you saying I am in the right place?17:41
ZipCPUGoing back to your question of desktop or workstation, I think the specification was generic enough to support a workstation, but that the entire spec has never been implemented.17:41
ZipCPUSilent: Well, I was going to say you are in the right place, but to do that might mean that there were no other applicable places.17:41
ZipCPUThis is certainly an applicable place to be, and those here will have all the tools you need to do your research.17:42
SilentThat it very good to hear17:42
ZipCPUYou might also wish to hang around the #riscv forum.  There's been some security papers discussed there, from researchers using the RISC-V platform.17:42
ZipCPUOne interesting one was a researcher who encrypted not only the memory data bus, but also the memory address bus in such a way that the memory never needed the key, if you know what I mean.17:43
ZipCPUI would also encourage you to find a copy of the wishbone spec.  I've fallen in love with the B4 version, but I may be the only one using B4 rather than B3.17:44
ZipCPUThe neat thing about that is, when you take a CPU plus a bus, you can do ... just about whatever security research is on your mind.17:44
SilentI will certainly look into that17:45
SilentOpen source hardware is a very interesting concept, I am very interested so far17:46
ZipCPUI have personally thought about doing something with security but ... haven't decided what would be worthwhile to try (yet).17:46
ZipCPUOr ... who would be worthwhile to impress with it.17:46
ZipCPUOr ... how to handle the key exchange(s).17:46
ZipCPUI mean, I built an SD card controller some time ago, and just recently an ethernet controller.  It shouldn't be too difficult to couple both of these controllers with some nice strong hw encryption, but ... then what?17:48
SilentPublish, be merry17:49
* ZipCPU laughs heartily17:49
SilentSeems to me like the community of people interested in such things is very small17:50
ZipCPUI remember going to a course, years ago, into how to build disk drives.  The speaker said, basically, if you can't make something work then publish it.17:50
SilentBut it's a start17:50
ZipCPUAnother thought I had was to create a truly multi-level secure O/S.17:52
ZipCPURemember how all that research died back in the 80's?17:52
kc5tjaZipCPU: I'm using B4, but I'm not using all of its features.  I'm staying compatible with B3.  (re: who's using what version of Wishbone)17:53
ZipCPUkc5tja: I'm using B4, but only in the pipeline mode.  I'm loving the performance, but I was surprised to discover at ORCONF that no one else is.17:53
ZipCPUSeems like there's a fear of patent issues with it.17:54
ZipCPUCool!  Julius just posted to the ORCONF attendees that all of the presentations have been posted online.17:55
kc5tja"multi-level secure O/S"?  Not sure what you mean here.  Most OS/systems programming R&D I'm familiar with that era aims to reduce as much as possible both the number of supported levels and trusted footprint of the kernel.17:56
kc5tja(e.g., small kernel, no more than a user and supervisor mode, possibly with kernel in completely separate address space as any other user-mode program)17:57
ZipCPUSo "multi-level secure O/S" goes back to a DOD effort to create a system that could handle secure information at multiple classification levels.17:57
ZipCPUUsers at higher classification levels were not allowed to "leak" information to lower levels and so forth.17:57
ZipCPUEven today, the issue of crossing classification domains (network at one level to a second network at another) is a ... very difficult problem.17:58
ZipCPUFew computers/systems are able to do such.17:58
kc5tjaI would think an object-capability security system would be ideal here.17:58
kc5tjaIt seems general enough to address this need.17:59
ZipCPUIt may well be.  A *lot* has been learned since those initial efforts back in the 80's failed.17:59
ZipCPUFor example, no one likes the Bell-Lapadula model of being able to write up and read down, but nothing more.18:00
ZipCPUIt failed pretty hard in practice.18:00
ZipCPUGoogle the "Trusted Computer System Evaluation Criteria" for a fascinating read.18:02
ZipCPUSteven Lipner wrote a very fascinating history of what happened with it, titled "The Birth and Death of the Orange Book."18:07
-!- Netsplit *.net <-> *.split quits: imphil, nurelin, heroux, LoneTech19:43
-!- Netsplit over, joins: LoneTech19:44
-!- Netsplit over, joins: imphil, nurelin, heroux19:44
--- Log closed Tue Oct 25 00:00:12 2016

Generated by 2.15.2 by Marius Gedminas - find it at!