Hardware for OpenWrt
wrtnode.comThis is an interesting project, but it's a shame they're using a Mediatek chipset instead of an Allwinner one. Mediatek is one of the biggest GPL violators out there, while Allwinner chips are among the most open hardware out there (perhaps the most open). See:
http://linux-sunxi.org/Main_Page
The only non open hardware on Allwinner chips is the ARM-licensed Mali GPU, but it's been reversed engineered to the point of producing a working, third-party, open source driver (the lima driver).
Both are very inexpensive Chinese chip manufacturers that use ARM-licensed MCUs, so it's not like the Allwinner is a vastly different product.
I personally prefer licensing projects with a permissive license that permits closed commercial development, such as the MIT or BSD license, over the GPL. However, when a company like Mediatek chooses to use GPL'ed software and then fails to follow the license terms, I think that displays a contempt for the entire open source community that I cannot support.
Does Allwinner actually manufacture router-oriented chips? Ralink (now owned by MediaTek) has been making networking parts for over 10 years now.
> The only non open hardware on Allwinner chips is the ARM-licensed Mali GPU
I seriously doubt that any part of Allwinners chips can be classified as open hardware. Just to clarify, this is wikipedias view on what open (source) hardware means:
> The term usually means that information about the hardware is easily discerned. Hardware design (i.e. mechanical drawings, schematics, bills of material, PCB layout data, HDL source code and integrated circuit layout data), in addition to the software that drives the hardware, are all released with the FOSS approach.
> Both are very inexpensive Chinese chip manufacturers that use ARM-licensed MCUs
The chip in this device has a MIPS core. Besides, MCU != IP-core != SoC.
I should have differentiated between "open source hardware" and open hardware. By using the later term, I meant to describe hardware that uses open source software to drive the hardware, and that can be easily modified.
And I screwed up with my description of this project -- you're right that it's a MIPS not ARM core. Sorry about that.
A fun comparison: 20 years ago you could spend a large number of dollars (anyone remember how much?) on a dreamy Silicon Graphics Indigo2 Unix workstation. From http://en.wikipedia.org/wiki/SGI_Indigo%C2%B2_and_Challenge_... its specs included:
* MIPS CPUs ranging from the the R4000 to the R10000, with clock rates of up to 250MHz. (This project uses a related MIPS CPU that clocks at 600MHz.)
* Up to 512Mb of RAM, equal to that in this project. The Wikipedia articles explains that the Indigo2 hardware could theoretically support 1Gb of RAM, but the thermal output of the DRAM available at the time was too great for the enclosure.
* 100Mbit/s wired Ethernet network interface (this project provides 802.11n wireless networking, which Wikipedia claims will reach 600Mbit/s).
I find it staggering that the hardware that was once powering a high-end workstation is now being put to good use as a low power router. It's good to see that after all this time you can still run a version of Unix on the same hardware though.
> * 100Mbit/s wired Ethernet network interface (this project provides 802.11n wireless networking, which Wikipedia claims will reach 600Mbit/s).
This project has 2x 100Mbps wired Ethernets in addition to 300Mbps WiFi.
edit:
> * Up to 512Mb of RAM, equal to that in this project. The Wikipedia articles explains that the Indigo2 hardware could theoretically support 1Gb of RAM, but the thermal output of the DRAM available at the time was too great for the enclosure.
As user Wicher below notes, there seems to be bit/byte confusion here. So Indigo has still significantly more memory than this project.
SGI's secret sauce was their 64-bitness and their 3D hardware. Their CPU/networking wasn't really anything special even at the time compared to Sun/DEC.
Not true in my experience. My R4400 and R10000 hardware always ran circles around SPARC workstations of the era and I wasn't a graphics developer.
all the SGIs i got on enterprise/university auctions were used for networking tasks... i did got a couple full color cards and couple 3d cards (indy had 8 bit cards, 24 bit cards, and 8bit Z cards -- which were good for 3d. there was no 24bit 3d)
Didn't they sell a stripped down Indy with no graphics capability under the Webforce line?
There was a stripped down, headless, Indigo2-based server. I'm not sure how much you could really strip down an Indy, though. It was already kind of their entry-level thing.
i think that still had the 8bit card. but i may be wrong. also, i never had U$20k to pay for it when it launched, so i never saw their marketing :)
I have an old SGI catalog, an Indigo2 Impact used to be about USD 30 000 to 40 000 in 1996, depending upon the configuration (Solid, High or Max Impact).
Luckily hardware doesn't have feelings. I use mine as a footrest.
It is a pretty color, though.
Hello people, I am Noel from WRTnode.com. Very glad that you notice us. Actually, we have opened public Beta version sale in China for RMB$148(as US$24), and closed today as two month later. We have donated 20 boards for the BattleMesh V7 http://battlemesh.org/BattleMeshV7#Donated_OpenHardware_-_WR.... And the Schematic is: http://cn.wrtnode.com/wp-content/uploads/2014/05/WRTnode_sch..., and we will give all the BOMlist and PCBlayout soon. Maybe it will be sold publicly at the ends of July. All OpenWrt firmware source comes from dev.openwrt.org. The board is officially supported by OpenWrt.org as one of the three devices on MT7620n. the only thing bothered is that the Wi-Fi driver of MT7620n is close-sourced by the MTK, so we have to provide the .ko binary driver in https://github.com/WRTnode/openwrt-packages/tree/master/rali... , We are trying to communicating with MTK to open the driver's source.
It seems unusual to pick a Mediatek SoC, if the goal is open source compatibility. It would have made much more sense to go with an Atheros SoC and Atheros modems, which can be used with no proprietary binaries whatsoever.
I was thinking the same thing. Atheros has been very good with Open Source friendly hardware documentation and licensing.
The downside is that most of the Atheros chips have a single <1Ghz MIPS core, when most of the rest of the router SoC market is switching to multi core 1GHz+ ARM devices, or is focused on other parts of the device space.
isn't there binaries burned onto the chip?
Check this out too: http://8devices.com/carambola-2
The hardware design isn't open-source, but all the software is, and it's available now for 19euro per module.
It's built on Atheros' AR9331 chipset which means you get to use the awesome open-source ath9k driver.
Everytime i see those things, i dream that i will wake up the next day and there will be no signs of ISP or cable companies. let alone telcos. only mesh network and IPv6.
Frankly I'm getting tired of these empty posts. Where is the source?
Can we stop posting "news" like this unless it is actually significant, ie contains the thing that they claim is open?
Its almost to the point where these posts should be flagged IMO.
I think you have a point. We took "open source" out of the title as misleading, since the project apparently isn't there yet.
Even if they release the original design files, it won't be open source hardware. Requiring binary blobs clearly violates the Necessary Hardware Criterion. See the OSH Definition 1.0:
Err, Necessary SOFTWARE Criterion.
Is OpenWRT safe to use? The latest release, at least according to https://openwrt.org, was in April 2013. I'm not intimately familiar with what software packages it uses, but it seems like it could, at the very least, be vulnerable to Heartbleed.
OpenWRT has a pretty decent developer base and in general has always been very stable for me (I've been using many WRT54GLs and Foneras over the years). That said it seems off that they wouldn't have fixed heartbleed by now, it would be significant for the web administration interface - especially for people who make it accessible externally via the Internet.
If anybody wishes to contribute to OpenWRT, this talk from 30c3 should give you a good overview on the "current" state of the project (hint: apart from system programmers, web developers are also very much welcome/needed):
OpenWRT's OpenSSL package has been updated to 1.0.1g, so it's not vulnerable to Heartbleed. The default install doesn't use OpenSSL so there is no need to update the install images.
Additionally, many packages can use alternate, embedded focused SSL libraries (PolarSSL, etc.) that will also ship with OpenWRT.
Their latest stable/official releases are indeed lagging. One of the things openwrt does (for good and bad) is to spend a lot of time testing and stabilizing before a release.
Many people therefore prefer to run the nightly builds to get a (much) more up to date build. It may sound scary, but it's not really.
Anyway: openwrt is definitely not dead, definitely safe and I consider it about the only routing firmware I fully trust.
I'm missing something: where is the source code (design files?) for this "open source hardware"?
They claim they are going to open source it. This is (disappointingly) common practice among "open source" hardware projects. See, for example, most "open source" kickstarters.
Sometimes it's the chipset vendors holding things up, but I agree - it's getting frustrating.
I'm holding out hope for http://www.securifi.com/almondplus
Why would you describe your project as open source if you didn't have the rights to release it as such?
Nothing on that page or their Kickstarter says that Almond+ is meant to be OSH.
Idea: OSHW wall of shame. Quick poll: which existing website would be most likely to volunteer to curate such a page?
edit: OSH --> OSHW
edit: probably http://www.oshwa.org (seems to be down right now)
It seems very hard to find an high-performance (i.e., not TI's CC3000 module), open Wifi module for a project like this. I assume that's why they choose a chip that requires a binary blob Wifi driver. Has anyone researched the open alternatives? Novena for example didn't integrate Wifi but left it to a PCIe module. Are there open PCIe modules?
edit: I should say, TI's module isn't open either. As far as I know the firmware, which includes a TCP/IP stack, isn't open.
Atheros AR7010/AR9271 has open drivers and firmware, which is the closest you'll get to "fully open" WiFi hardware today. Performance is decent, although I don't think PCIe is supported by the open firmware (only USB).
I'm sorry to bust the fun on this, but I would never rebuild this device (or anything other with radio capability)... simply because of the costs of certification which is mandatory for any RF transmitter.
I don't think that's the point.
The problem with a lot of home automation and internet-of-things stuff is proprietary lock in, either from new young companies that could be dead in 18 months or established monsters looking to slurp your data. Having an open source foundation is the only viable long term approach to something that needs to live as long as your home infrastructure. If this lot cease being able to make the boards then at least, in principle, you can. (Or more likely someone else can pick up the ball without any license cost). If you've built around a Nest and that stops being made then you're screwed.
I have to admit I don't think either in extreme is going to work. You need the combination of standardisation with non-cloud dependency but also commercial viability. i.e. an Android of Internet-of-Things without Googly influence is needed, but without a rich backer it's hard to see where it will come from.
Isn't standard commercial hardware tat you can install opernwrt enough?
Not at all. All successful end user platforms contain provisions for the easy installation of commercially sold software, including at the point of purchase. Tying yourself to OpenWRT prevents this.
This is the real reason the GPL has a sort of built in success limit for end user deployments, and why so much of the code in Android is Apache or BSD licensed.
How do you mean?
What are the costs exactly? Wifi operates in the 2.4Ghz band which is license free worldwide.
(As I understand in some countries you need to get a license to sell wifi devices - some Arduino's have issues with this - but not to use or build them).
You still need to meet the FCC specs for not causing interference (in the US at least), which can be costly. And you can be sure that a bare board with no shielding around the RF sections is going to be pretty noisy.
I think the OP's talking about those who want to implement the boards on their own, where you do have to worry about all those certification costs.
Can you clarify what you mean here? Are you saying that certification would be required to build this device in the U.S. (not sell or market it)?
If so, this is incorrect. There is no FCC certification required for low-powered, non-licensed devices like this one if you don't sell or market them. You can run up to 5 such devices without any certication. See page 3 on:
http://transition.fcc.gov/Bureaus/Engineering_Technology/Doc...
I'm living in Europe, here you are liable for any RF device you use - and even in the license free bands, if you operate a self built jammer (or accidentally build a jammer instead of a working device) you're liable.
Less open than many competitors. But more choice is always nice. Hopefully they get the module FCC/CE certified.
The latest post there that I was able to find was from April 2014 - is there any news? Am I missing something?
There are newer posts on the Chinese-language version of the site. Click 中文 on the top right. :P
This looks very useful, but price, RAM and flash sizes are not readily apparent. Am I missing something?
Can't find price but http://wrtnode.com/tech-specs.html says 512M RAM and 128M flash
It says 'Mb'. Chip sizes are frequently expressed in bits, not bytes, so this could mean 64MiB of ram and 32MiB of flash (I'm assuming mebibits (2^20 bits) instead of megabits (10^6 bits), for which 'Mb' is the SI symbol).
512Mbit ram (so 64MB)
witness RAM part number here: http://wrtnode.com/wp-content/uploads/2014/01/DSC_0038.jpg http://wrtnode.com/wp-content/uploads/2014/01/DSC_0046.jpg http://wrtnode.com/wp-content/uploads/2013/10/2s.jpg
decode here: http://www.elixir-memory.com/Download/2007%201H.pdf
Very likely the 128M flash is 128Mb, so 16MB
The price is $24.80
https://twitter.com/WRTnode/status/483611574796447744
In the <title> it's about $20 so I'm guessing $25 includes a VAT tax or something: <title>PREORDER WRTnode for $19.99 now | WRTnode</title>
Did they choose a wireless chipset that requires "binary blobs"?
Yes, it appears they did.
Would this be a good router?
Probably not. It doesn't seem to have any Ethernet ports for a start. I think it's aimed more at being a wifi-enabled device that can do somewhat more complicated things locally.
The board has an ethernet switch built in, with all the pins for it broken out. You're going to have to supply your own magnetics and jacks for it though. The chip in it [1] is very much built around routing lots and lots of packets. I kinda wish they'd have released a board based on its bigger sibling with PCIe, though.
Doesn't appear so. It has one 802.11 interface - so it could receive from upstream, or send downstream, but not both.