5 ms·
It's not a laptop with no secrets or fully open-sourced. There's the hardware components internal operation, the FPGA tools, and so on that need to be open. Inc
by nickpsecurity 11y ago
It's not a laptop with no secrets or fully open-sourced. There's the hardware components internal operation, the FPGA tools, and so on that need to be open. Incidentally, the things with the biggest risk, too. Freescale operates DOD-certified fabs in U.S. on top of it with major sales to U.S. govt. Whatever subversion risk Intel poses we must assume Freescale does as well.
It's actually a laptop for FPGA hackers with incremental steps in the direction of an open-source laptop. That's great. That they made FPGA development a bit easier plus made a more open laptop is extra great. So, let's give them props they deserve while acknowledging it's still untrustworthy hardware and there's plenty of ways to go.
- JoachimS 11y agoYou don't need to use the FPGA to use the Novena, it is a peripheral you could even remove (albeit it would be hard since its a BGA). The tool chain for the FPGA is proprietary and a black box. There are some reverse engineering efforts, but hard to do any meaningful. By and large, open HW is in a fairly sorry state. Most users seem to think zero cost is good enough.
- nickpsecurity 11y agoI know. I'm just assuming FPGA is trusted in TCB, esp with DMA, plus implying many people will want to use it. Your assessment of open HW is accurate. Far as FPGA's, there's progress on several fronts. Some for you to check out that your people might even consider using given the continual payoff of a FPGA w/out high unit costs. Open-source FPGA architecture at 45nm http://www.eecs.berkeley.edu/Pubs/TechRpts/2014/EECS-2014-43.pdf http://www.eecs.berkeley.edu/Pubs/TechRpts/2014/EECS-2014-43... Open-source bitstream generation for Xilinx w/out EULA violation or R.E. (!?) http://www.isi.edu/~nsteiner/publications/soni-2013-bitstream-fccm13.pdf http://www.isi.edu/~nsteiner/publications/soni-2013-bitstrea... Open tools for Lattice http://www.clifford.at/icestorm/ http://www.clifford.at/icestorm/ Open-source HW synthesis flow http://opencircuitdesign.com/qflow/ http://opencircuitdesign.com/qflow/ Note: Cliff is on a winning streak, eh? I recently did a write-up showing what it could or would take at both ASIC and production levels here: https://news.ycombinator.com/item?id=10468534 https://news.ycombinator.com/item?id=10468534 https://news.ycombinator.com/item?id=10468624 https://news.ycombinator.com/item?id=10468624 So, paths are clear and plenty of potential. Just no uptake. Will take government, corporate, or private sponsors working with academic (low NRE) and professional (experienced) HW designers that open-source stuff as they go. Proprietary, dual-licensed OSS is the only way to go for HW that I know of.
- tonyarkles 11y agoI haven't looked at the actual board design, but I'm making a few educated guesses based on people mentioning Freescale. If this is using an i.MX6, one great way to interface the FPGA with the CPU is through the EIM (external interface memory). It allows you a quite fast interface, without giving the FPGA direct access to RAM. It might not be fast enough to connect a 100MSPS ADC, but for many things it's quite great and easy to write both the Verilog/VHDL and the kernel-side driver for.
- nickpsecurity 11y agoThanks for the tip!
- JoachimS 11y agoThe FPGA on the Novena is connected to the EIM. The EIM is a bit of a pain to work with. Pavel (part of Cryptech core team) has done a good top level with much cleaned up constraints compared to what Bunnie provides. If anybody wants to build FPGA designs for the Novena I would recommend to start with our top level design and build setup, just clone and go. We get good performance out of the FPGA as configurable coprocessor.
- JoachimS 11y agoThe FPGA on the Novena does not have access to main men via dma or anything else besides the connector. It sits in its own EIM memory space as a slave. (It has its own local external memory though.) Thanks for the links. I knew about the efforts to generate for Lattice but haven't kept up. The big issue is really good P&R with working optimizer for anything bigger than toy designs and chips. It seems Cliff has done some real progress yes. And it is interesting that it is basically easier to go open with ASIC than FPGA. The problem is getting your chip manufactured. What would you say would be the best path for Cryptech to try and go open with the flow? Alternatively, to mitigate black boxes in the FPGA of commercial chips? We currently fill the Spartan6lx45 on the Novena to max. We target the biggest Artix7 you can design for with zero cost versions of Xilinx tools for our Alpha board. I've toyed with the idea of mapping a synthetic FPGA into a commercial FPGA and then implement the real use case core in the synthetic FPGA. This would allow greater control on how the design is implemented. But then you need a good optimizer. And the cost to implement a gate will require even more transistors than the FPGA already do. Mapping the synthetic FPGA to the underlying tech can be done pretty efficiently, but configuration support is costly.
- nickpsecurity 11y agoCurious, what's your level of HW development experience for FPGA's JoachimS? I'm keeping my eye out for people who might put high-level synthesis tools I spot through their paces. There's two that I think haven't got much review but have potential.
- sliverstorm 11y agoWhat do you want to build with these high level synthesis tools?
- nickpsecurity 11y agoThat's like asking me what do I want to build with this C programming language and compiler thing. You really don't have time to read it all. :) Main objective is to find a tool flexible enough to handle about arbitrary algorithms and dramatically lower the bar for people/companies wanting to do FPGA acceleration. The less hardware design they need to know the better. That takes HLS. Worst case, the design happens faster and cheaper using fewer hardware developers possibly with less experience.
- JoachimS 11y agoFairly big. I am part of the Cryptech core team and has designed most of the crypto cores and the TRNG. I have worked on several FPGA designs including VoIP gateways, transcoders, radar systems, networking etc. focus on use case arch more than low level interface issues. Done some ASICs to.
- nickpsecurity 11y agoSweet. That means you have the experience for this. I'm ignoring the ones that seem like OSS/academic toys or commercial that are focused on too narrow a model. The two candidates I stumbled upon recently are Synflow's Cx language and Synthagate HLS. Synflow [1] is interesting because it's a step up from Verilog/VHDL w/ open-source tooling. Their strategy sounds smarter and easier than many where it seems to try to augment the human brain rather than replace it. That it outputs synthesizable Verilog/VHDL is nice. Downside is that one of the founders wrote an article [2] expressing that HLS is dead and they transitioned to an internet-of-things company instead. Upside is they probably keep supporting this, anything OSS can be picked up by another team, and he was on HN very happy to answer any questions. I just lack hands-on experience to properly put Cflow through paces and ask him the right questions, you know. What got me interested, other than Cflow's design, was that they were offering I.P. blocks for around $500. I never see I.P. for three digits. So, I'm hoping it's because they're onto something good. :) The other to check out is Synthagate. Syntagate's web site [3] details their methodology using Abstract State Machines with two different types for control and data. That's interesting and it's probably expensive. More interesting was Baranov's book [4] that details his exact methodology for doing HLS. Correct me if I'm wrong, but full details of a HLS process good enough to synthesize microcontrollers rarely leaves trade secret stage. The real value here, other than possibly good HLS tool, is a specific method of doing HLS described in enough detail to clone with OSS tools. Just need someone with HW design expertise, esp FPGA's, to evaluate what's in the book to tell me if it could be used for algorithms, protocol engines, etc. Does Baranov's method look like it can work for real-world use cases and does his tool do the job for a set of I.P. representative of what people will really use? The former you can just read but the latter might be available with free trial or something. I'm thinking it is his secret sauce because he (a) sells it and (b) seems to have pulled it off his web site. Goofball forgot about Wayback Machine. Enjoy. :) [1] http://cx-lang.org/ http://cx-lang.org/ [2] http://www10.edacafe.com/blogs/disruptedhard/2015/04/13/numbers-dont-lie-there-is-virtually-no-interest-in-high-level-synthesis/ http://www10.edacafe.com/blogs/disruptedhard/2015/04/13/numb... [3] http://synthezza.com/what-makes-synthagate-different1/ http://synthezza.com/what-makes-synthagate-different1/ [4] https://web.archive.org/web/20150314182353/http://synthezza.com/download/AboutUs/Book2011.zip https://web.archive.org/web/20150314182353/http://synthezza....
- SomeCallMeTim 11y agoIt kills me that it has no working GPU. :( I found the wording about XFce to be weird: They claimed to not use Ubuntu because they needed XFce, but ... http://xubuntu.org/ http://xubuntu.org/
- dikaiosune 11y agoXubuntu's Xfce is weird, and heavily changed from upstream, last time I tried using it. Broken themes, incompatible panel applets, etc.
- JoachimS 11y agoA GPU is actually one of the things you could use the FPGA for fairly easy. The question is if you could gain any serious performance benefits though. The CPU cores runs at 1.2 GHz, for reliable transfer you want to run the EIM interface at 66 MHz. You could potentially run the GPU at something like 100 MHz in the Spartan6. That is a 10x difference compared to the CPU. So your GPU must provide more than 10x compared to SW to be useful. Not taking parallel processing into account though.
- nickpsecurity 11y agoI don't think that's a good idea. Doing it on the FPGA will suck so bad that people will immediately remember why we use custom designs for GPU's. I'm sure most people want to be able to do graphics acceleration and do specialized tasks on FPGA's at same time. Need the GPU for those use-cases. I could see a compromise where they used a CPU w/ onboard GPU or multimedia acceleration, though. Not sure if it will meet their cost or NDA-free datasheet requirements though.
- xobs 11y agoThe GPU is being worked on. The Mesa driver is being worked on as we speak by a number of people including rmk and jnettlet, with austriancoder's experimental repo located at https://github.com/austriancoder/mesa-1 https://github.com/austriancoder/mesa-1. If you want to get technical, we already have fully open-source 2D acceleration, and we routinely use the 2D GPU to e.g. do colorspace conversion and 2D scaling on video. I went with Debian because we knew some Debian people. At one point I considered moving to Ubuntu, because they package "Firefox" and "Thunderbird" (instead of rebranding them) and they have Chromium as well. There were lots of teething problems, though, ranging from their Xorg modesetting driver refusing to use generic KMS on a non-PCI system, to strange oddities when no initrd was present, to the desktop straight-up crashing when logging in (turns out Unity assumes GL is present and has no fallback.) Ubuntu seems to make many assumptions about running on standard x86 hardware, and was rough around the edges. We could have polished it, but it seemed easier to stick with Debian.
- pjc50 11y agoThis whole thread should be cited on https://en.wikipedia.org/wiki/Perfect_is_the_enemy_of_good https://en.wikipedia.org/wiki/Perfect_is_the_enemy_of_good Fully open hardware is never going to be quite as achievable as open source software because the manufacturing and distribution costs cannot be driven to zero in the same way. Furthermore, ASICs can't be rebuilt by the end user in the same way as software. You're always going to be dependent on a device coming out of someone else's factory. Openness remains a minority interest, and therefore the NRE cost amortized in every unit is going to be higher. Signs of people being willing to pay for open hardware (the original designers who did the development, not the cloners who didn't) are scarce. And when someone does offer something like this to the community, what do they get? Negativity. It's not even an unachievable amount of money that would be required to do a fully open CPU+GPU by paid professionals. $10m at the bottom end. Much smaller than, say, the Star Citizen kickstarter. What's unachievable is the community unity of purpose behind it.
- sneak 11y agoIt's not negativity; it's a practical analysis of the fact that the project failed to achieve its singular purpose. He even acknowledges the effort and that it's a good incremental step; but a verifiable computer it is not.
- pjc50 11y agoVerifiable is not the same as the stated goal of "as open as possible at a reasonable price using obtainable components". The latter is somewhat flexible, the former is extremely hard to achieve even for milspec gear.
- nickpsecurity 11y agoIt's extremely easy to achieve. It just costs money and takes effort.
- pjc50 11y ago"Going to the moon is easy to achieve, it just costs money and takes effort." I don't think it is easy to achieve verifiability to be above nitpicking of the kind we're seeing here. Assume we start with open RTL in some HDL. You then have to run it through layout and get someone with a Formality license to check it's equivalent (trusting that particular piece of software). You then take your GDSII to a fab. If you want to check that they've not inserted something, you have to resort to sampling production and sending it to the specialists with microscopes to check it - but then you have to trust their reports. And none of this is provable to doubting strangers on the internet. It's always possible for someone to assert that the whole group of people responsible is a clever fake by the CIA. You can't check it on your desk the same way you might be able to check software with reproducible builds (not that more than a few people do this anyway).
- skaevola 11y agoFreescale's fabs in the U.S. are all very old (like early 1990's old). AFAIK they only manufacture analog and MEMS type stuff there. I think TSMC manufactures their MCU, MPU's, and most other digital electronics. This is the 'fab-lite' model.
- nickpsecurity 11y agoThanks for the info. I'll try to double-check mine on that. :)