It's not openwrt even though they pretend it is in their marketing. It's based on openwrt and might be "compatible" to some level with other openwrt packages.
When asked for full source code they seem transparent about it:
The issues regarding GPL compliance or lack thereof are worth noting, however. I made a point of asking for native OpenWrt firmware for the products I have from them, only to discover after the fact that due to closed source firmware blobs, it will likely never be available in that format, which was somewhat disappointing.
Given the fairly low/competitive price point of their hardware, I think it’s worth taking the time to make sure that the device suits your needs in that regard, if it’s important to you.
Are you referring to open source BL2/BL31 for GL-iNet products, and/or for OpenWrt One? I’m not sure it’s possible with either, as haven’t looked into the One in detail yet, as I wasn’t aware it had launched until TFA was posted, though I was aware of it since January or so.
I’d like to run fully open source network stack if possible myself, though I’m not sure if that possible without moving the goalposts and virtualizing something or doing it in software, and even then I’d have to figure out some kind of boot attestation ideally, thought I'm not sure how that's going to pan out. Isn't Intel SGX/AMD SEV/ARM CCA required for that?
Some links I thought we interesting on that topic, as it's adjacent to the discussion:
> A comparison study of intel SGX and AMD memory encryption technology
This isn't entirely accurate. It absolutely is running a full OpenWRT instance. In addition to that, they have produced their own UI/shell, which is the default that you'll land on, but it's not difficult to get into LuCI.
That said, I'm not stating that it's only running OpenWRT, or that the OpenWRT instance it is running is unmodified, or trustworthy.
That said, I have struggled to get gigabit wireguard VPN throughput on other devices that support OpenWRT.
I love FOSS, I love self-hosting, I love DIY-friendly tinkerer-friendly, and I love high levels of user control, I just wish the ecosystem that prioritized these things had a stronger emphasis on high-end hardware that offers high performance.
It's not really an issue, since OpenWrt has full support for this device. IIRC the support in OpenWrt 23.05 is pretty good, I have been running 24.10 on a Flint 2 that we use as an AP and I've had zero issues.
I'd argue that even though by default the Flint 2 has a nicer interface for beginners, vanilla OpenWrt is much better. E.g. their old OpenWrt 21.02 build with the proprietary Mediatek SDK does not support baby jumbo frames, which are used by a bunch of providers that still use PPPoE (to get better performance).
Be careful of GL.iNet's products - some of them say they run OpenWRT, but they don't run stock OpenWRT and instead offer a version of OpenWRT supplied by the CPU manufacturer, with binary drivers and no source code.
Yeah, it's probably best to pick something with upstream OpenWRT support and flash it when you get the device. I think that's what grandparent was saying, given that they link to the ToH.
Also, 2 x 2.5Gbe, so if you have a > 1Gbit && <= 2.5Gbit connection, you can do line speed without having to resort to more complicated router on a stick configurations.
Besides that it has quite a bit better CPU (quad core rather than dual core), so if you do anything that cannot be hardware-offloaded (e.g. Cake), the performance will be better.
The OpenWrt has better hackability though. USB-C serial is very handy if you manage to mess up your configuration in a way you can't access the device (though LuCI have this features where it can revert changes if a change makes the router non-responsive to the user).
I currently use the Flint 2 as an AP, so I cannot easily test it. I tested it at some point, but it was far less than the 2.5Gbe line speed. I vaguely remember something like 700-800Mbit, but I might misremember, plus disabling hardware acceleration also disabled PPPoE offloading, so it might do better without PPPoE.
The other question is whether it is worth it. As far as I understand, the Filogic in the Flint 2 has hardware support for fq_codel. When doing buffer bloat tests with the waveform test, the score would always be A, whereas on the same connection a Fritz!Box 5590 Fiber would show pretty bad buffer bloat (grade D on the waveform test).
This is somewhat true but their fork of Openwrt is horribly out of date, very hacked, and has a baby interface that doesn't fully provide what openwrt can do. I found myself having to manually edit files on the routers shell. I mean my use-cases aren't exactly normal since I'm testing very arcane networking stuff. But I found myself disappointed.
I want something that has like actual good packages already installed for common internet standards, that are configured by experts (so they work), that supports IPv6 perfectly, and is user-friendly so I can use it for what I need without having to work on router firmware. It's like... maybe I should write my own firmware at this point. Cause everything is actually just shit.
and has a baby interface that doesn't fully provide what openwrt can do. I found myself having to manually edit files on the routers shell. I mean my use-cases aren't exactly normal since I'm testing very arcane networking stuff. But I found myself disappointed.
???
Devices like Flint 2 have LuCI preinstalled. It's even linked in GL.iNets interface (IIRC on the Advanced page).
With shipping and tax the One cost me £94.
The Flint 2 has been on sale for £115.51 twice in the last couple of months, 23% more, so not a huge price difference.
Huge +1 to this. Managed switching with hardware offload is the only thing standing between me and a fully open source* homelab.
I'm vaguely aware of some enterprise options. I want something quieter and reasonably power efficient. Price is no object within reason. I would pay low 4 figures for something equivalent to a Mikrotik CRS326, but with upstream kernel drivers and an X86 CPU.
* I don't run libreboot or anything, I'm sure there's a fair bit of closed blobs in my lab, but almost all my devices can boot any standard Linux boot image.
My desires are driven by ideology rather than any specific requirement. My needs are met by my Mikrotik CRS328, but I loathe all devices that I pay for but do not own (if I owned it, I could run software of my own choosing on it).
I'm not sure where you got 48 from. I need 4 or 5 VLANs. I need a little bit of PoE. I need a little bit of SFP+. I want it all to fit in 1U of rack space.
The VLANs are to keep my network segmented: wan, management, internet of shit, trusted, testing.
Funny things many of those commercial switches available now run customized version of OpenWRT with different UI and custimization that make it hard to tell it's oprnwrt
Mikrotik, Ubiquiti, TrendNet, Mokerlink and all the other Chinese brands usually have something in this space (4-48 port managed L2 or L3 switches). A lot of the chipsets are mentioned in reviews on sites like ServeTheHome.
The sub-enterprise, prosumer, SOHO-type market is pretty strong, but OpenWRT isn't a part of it -- not sure if that was a requirement.
QNAP has the QSW-M2116P-2T2S-US. Has 16 ports of 2.5G, 2x10G (copper), and 2x10G SFP+. The 10G ports support 10G/5G/2.5G unlike the older 10G chipsets.
True, and I admit I didn't think that was a literal requirement, more so just the availability of "some" mGig or 10G ports. What would the use case be for 24px2.5G switches, you think? They seem rare.
The "home network with a NAS" is typically served with a small 4-8p 10G, while the 2.5/5G switches (particularly high density ones) are more of a bridge for orgs that have a lot of AP density and to get more out of that physical cable plant as they upgrade to WiFi 6e/7 devices on those existing wires. Gaming motherboards ship with up to 2.5/5, but again, how many switchports are needed for the SOHO market, especially when users can chain a few smaller switches together?
The price point for mostly/full mGig switches is almost an order of magnitude over what the prosumer/SOHO space prices at. UI offers a 48p but with only 16x2.5 ports at $1300. Cisco Meraki has 48p models with full mGig but that's $7-10k based on Google, and now you're out of prosumer/SOHO and squarely in enterprise platforms.
Ultimately, I do think such switches will pop up, but it'll be a little in the future as those enterprise dollars drive the cost of chipsets and transceivers down. Just not there yet to have that density at SOHO prices.
Most likely an IO limitation of the MediaTek MT7981B SoC used—it looks like it only has one lane of PCIe Gen 2, a USB 3.0 port, and one built-in 1 Gbps Ethernet controller (among a few other common low-speed interfaces)[1]. Typically these chips have bandwidth constraints, and it seems this one is meant for small routers, not the typical multi-2.5G or 10G setups some homelabbers would prefer.
The difference isn't that much (around 10% IIRC), but gigabit internet service needs a 2.5 GbE connection to your router if you want the full speed that it's capable of.
For my gigabit fiber, it's actually 1.3 Gbps (they overprovision a bit), so having 2.5G to my network is nice, in that I don't miss out on the extra 400 Mbps of bandwidth :)
For 99% of the time, though, it makes no difference. Anything over about 100 Mbps is adequate for a lot of what I do.
That's just how it works — Gigabit Ethernet doesn't mean gigabit data transfer rates. This is why all fiber providers I'm aware of provide ONTs with 2.5 GbE for their gigabit internet service.
If you're paying for gigabit internet and getting 925 Mb/sec through a gigabit Ethernet router, you're doing great. But you need to use a 2.5 GbE-capable router/switch to get advertised speeds.
Total deal breaker for me. I don't understand this design decision, this router is clearly targeted to advanced users so I would expect most of those would have multiple devices they'd want wired?
Not a downvoter, but it’s probably because it doesn’t really matter. If you want high speed connections for multiple devices, use a separate switch. Devices that are purely routers often have only a few ports, and you can use VLANs to multiplex subnets.
External ISP connections rarely exceed 1Gb in many parts of the world (for home connections especially), so there’s not much point in a faster link speed.
And the chip itself only supports 1x 2.5Gb and 1x 1Gb, so the choice was made for them. Maybe they could have limited both ports to 1Gb to avoid a perception that one port is hobbled.
This one doesn't have WiFi, just 5 Ethernet ports. It can be flashed openwrt with a very reasonable amount of tweaking. It's actually quite powerful and has 256M of RAM and 128M of flash memory. I have one, it's very cool.
Seems like they’re starting to do this as mesh is becoming common. In my apartment, my ISP has a permanently installed router without wifi. Then they just send you a wifi router or mesh devices depending on your requirements.
I never buy routers w wifi. Running PC Engines APU4 with Ubiquiti AP is more rock solid/stable than anything else I ever ran which had wifi+wired in a single router.
As an aside: Why can't DSL modems be a single USB dongle?
Those of us with DSL connections must suffer either an extremely limited selection of DSL modem/routers that can run Linux/OpenWRT, or have to suffer running a Linux/OpenWRT router behind a DSL modem (that often has proprietary and out of date firmware).
I'm just about to cancel my DSL, but when you run the modem in bridge mode, and run PPPoE on your actual NAT gateway if needed (which is sadly often the case), the modem firmware doesn't matter very much.
I put together some stuff so I could transfer PPPoE sessions to a backup system and then I could reboot the NAT boxes for upgrates with minimal downtime. Sometimes, it even worked ;)
There are fiber PON/ONU/UT/$JARGON in shape of an SFP module, though most customers don't appreciate such offering and therefore it'll be an upsell.
As for why not USB specifically, probably because such a device is inherently much faster and responsive in upload to the Internet than downloads, and therefore it makes less sense.
When I last used DSL, I was using AT&T uverse (which is/was VDSL with multicast video layered in).
I configured the provided gateway/router-widget to provide a "DMZ Plus" mode for my router (a custom box running Tomato or OpenWRT or something), and I called to get ports 25 and 80 unblocked. And then, plus-or-minus some completely-surmountable difficulty with making dynamic DNS behave properly it all worked fine.
For years.
I never connected anything other than my router to the ISP-provided device.
There's probably some corner cases where this configuration falls flat, but I never ran into them.
What might be some practical advantages of what you suggest?
A device runs on electricity and performs a function. It consumes power at a rate of x.
You're telling me that a device that performs the same function will consume power at a rate of precisely x/2 simply by virtue of being plugged into USB?
> As an aside: Why can't DSL modems be a single USB dongle?
They definitely existed in the UK for a time and were often supplied by ISPs. IIRC they were only supported in Windows XP, and drivers were never provided for Vista.
I tried for years to do similar when I lived in Australia. Though with a PCI/PCI-E card
ZyXEL if I remember correctly did make an ADSL2+ at the time PCI-E card. Literally just a DSL modem wired to a Realtek 8139 NIC. You could slap it in a Linux (or BSD, or Windows) PC and just use PPPoE to connect to the internet
Naturally it was impossible to order the damn thing and I never got to realize my dream of an "all-in-one" DSL Linux router.
After moving to DOCSIS (Cable) internet I ran into the same confusing problem. "Thankfully" with Fiber everything is just ethernet (more or less) now. But it was an infuriating time in the 2010's
+1 - I have a "beryl" one I think and it worked well in a few situations (backup router at home, and replicating home WiFi SSID when travelling to places with random hotel wifi so I didn't need to reprogram all the kid's wifi-connected white noise machines/night-lights/cameras/monitors/etc - yes I realise this is absurd but we find it all useful for a stress-free life with tiny kids)
Only complaint is the USB LTE dongle was super unintuitive to setup/use, and I felt like some of the translations were just flat out wrong so it was a bit of trial by error and factory resetting things when inadvertently bricking/locking myself out etc.
In addition to the wifi, I recall the preloader at the start of the boot chain is also a binary blob, which handles some of the chip init and memory calibration for the DDR4.
Assuming there's an active uplink that can reach a timeserver, then yes.
The batter-powered RTC allows the router to serve more accurate time and maintain working HTTPS in the event it gets rebooted as a troubleshooting step for when the internet goes down, possibly for an extended period.
> Industry “conventional wisdom” often argues that FCC requirements somehow conflict with the software right to repair. SFC has long argued that's pure FUD. We at SFC and OpenWrt have now proved copyleft compliance, the software right to repair, and FCC requirements are all attainable in one product!
This is not FUD, regulatory compliance requires that devices which make unlicensed use of radio spectrum (such as Wi-Fi) must contain non-user modifiable software for their radios in order to comply with regulations, and the MT7976C chip used in this device is no exception. While most of the mt76 driver is open source, it does contain a binary blob containing proprietary firmware which must be uploaded to the chip in order for it to function.
The claim that all of the software in the OpenWRT One is open source and copyleft is therefore false, it cannot be any other way, FCC regulatory compliance and being fully open source are mutually exclusive by definition.
> regulatory compliance requires that devices which make unlicensed use of radio spectrum (such as Wi-Fi) must contain non-user modifiable software for their radios in order to comply with regulations
Couldn't one restrict the chip to only accept signed firmware and then publish the source with reproduceable build instructions? This would prevent user modification while still being open source.
I think it's unfortunate given the audience I imagine will make up most of its purchases. For example, the NBN in Australia just announced earlier this year it's first 2 Gbps residential plans (previously 1 Gbps being the maximum) planned for availability some time next year[1].
Where I live (Japan), 10 Gbit availability is starting to spread, I just got upgraded a few weeks ago, the monthly cost is the same. Ended up building a router out of a used ThinkCentre Tiny.
Worse still, Wi-Fi *7* devices are already out. Qualcomm's QCN9274's running on the "ath12k" chipset [1] have been a tempting DIY option. Though that module requires an external 3 amps at 5 Volt (15 watts!) to be soldered to it to operate
Decent but sort of annoying to set up;
https://openwrt.org/docs/guide-user/network/wifi/wifiextende... appears to have combined what used to be 2 pages with different instructions, but it still seems to have both sets of slightly different steps. IMHO there really should be a single button in LuCI to make the box just an AP without DHCP/DNS/firewall.
That's my preference as well. Keep the router simple, don't put a transmitter near the sensitive bits to cause interference with the CPU, ethernet, or USB bits. I want my AP to be dumb and do nothing besides connect ethernet to wifi clients. That way when you need more coverage you add APs, newer WIFI upgrade only the APs, etc.
- Switches frequently offer PoE by default whereas routers (especially consumer or PC-grade stuff) would need a PoE adapter
- Devices connected to the switch can communicate directly without the router, saving a hop
- Modern switches may only need to read the first few bytes of the frame before cutting the packet over to the appropriate port, whereas routers tend to need to read the whole packet and apply rules/policy, incurring higher processing costs
An ordinary home LAN may not find any noticeable benefits, but an enterprise setup might benefit.
> Am I the only one who doesn’t even want their router to have wifi?
I'd prefer to separate the two as well, but I'd rather not admin them separately. It'd be convenient to manage them all in one (fully local, non-cloud) place.
But yeah, my ideal setup would be a 10Gbps Ethernet router the upstream Internet connection, firewall, NAT, and LAN DHCP, and then one or more separate Wifi 7 routers whose only job is to bridge transparently to Ethernet but not handle DHCP or NAT.
I used to think that way, but now my router is a Ryzen 5700G running NixOS so there is no reason to not hang some wifi radios off of it. Additional APs are low power amd64 motherboards (which also run Kodi to drive "TVs"), because I got tired of the ARM dumpster fire. Although I just ordered a pile of RPis to make security cameras, so maybe I'm back looking for a little punishment.
Not required, but you can solder a Multi Link Operation wire between multiple cards (within a single system), to enable them to work in concert together (basically as a 4x4 or whatnot). In case you still really miss soldering extra shit onto your cards!!
Do you mean open wrt? If yes, the devs have serious problems whenever they need to deal with broadcom because of all the proprietary blobs. So the number of supported chipsets is low.
Seems almost ideal if you don’t use it as an ethernet switch. A third radio for wireless mesh trunking is the only thing I see missing from my OpenWRT setup.
So was the recent Linksys WRT series. The product description on Amazon (which Linksys presumably provided) explicitly advertises the WRT3200ACM as "[o]pen source ready with OpenWrt and DD WRT for complete flexibility and customization of networking functions, or to optimize your router for specific use cases". Likewise for the WRT1900AC: "For advanced users who want to customize their firmware, the WRT1900AC stays true to its lineage by offering an Open WRT compatible firmware with Open Source ready capabilities that allow advanced users to expand the capabilities of the Router.".
The headline's (and article's) claim of the OpenWRT One being the "first router designed specifically for OpenWRT" is misleading, at best.
> The headline's (and article's) claim of the OpenWRT One being the "first router designed specifically for OpenWRT" is misleading, at best.
This is the first “official” OpenWrt first-party developed and supported router. It also is intended to be the initial “blessed by OpenWrt devs” hardware configuration that will hopefully lead to further devices in the future. It’s also intended to help drive awareness and interest in the platform, with a known-good hardware and software platform, help build brand loyalty, and raise money for development of OpenWrt for the One and all other compatible devices. It’s something the community and the project maintainers have wanted to do for many years now.
You can read more about the discussion leading up to this on the OpenWrt mailing list and forum.
That's great, but they should actually go with that claim of being the first router OpenWRT designed themselves, instead of incorrectly claiming that their router is the first designed specifically for OpenWRT.
I think there's some confusion, but I'll admit that you're not confused about what you want, but you might be about what correctness and/or a reasonable remedy looks like.
TFA is a post, a press release really, by Software Freedom Conservancy, of which the OpenWrt project is a member. That said, the post doesn't have a byline, so we can't really point the finger at anyone. I doubt that the OpenWrt project would even characterize the OpenWrt One in the same manner as does the title of the post in TFA.
That said, the title is largely correct to my reading, as other devices that precede the One, like the Turris Omnia, for instance, aren't designed specifically for OpenWrt, but rather a derivative, Turris OS.
To my knowledge, the OpenWrt One is the first router designed specifically for mainline/stock OpenWrt, which is what it ships with, and any comparable router meeting your definition must also meet these marks.
That rules out every Linksys or GL-iNet router ever sold, because they don't ship with stock OpenWrt, regardless of whether or not they run a derivative out of the box, or whether they may be reflashed to run stock OpenWrt, whereas the OpenWrt One disticntly does.
I will admit I could be mistaken on this point or any other made herein, and will happily admit my mistake if so, but as it sits, the burden of proof is on the accuser, which would be you in this case.
> To my knowledge, the OpenWrt One is the first router designed specifically for mainline/stock OpenWrt
The issue is that there never really was (and arguably still ain't) such a thing as "mainline/stock OpenWrt". From its very conception, OpenWrt has been developed with the expectation that it'd be tailored to specific devices, much like (for example) LineageOS and other custom Android versions are tailored to specific devices. Every device has different flash layouts, different onboard devices (meaning different required drivers and different initial configurations to ensure that the user can access a fresh install at 192.168.1.1 via Ethernet), even different CPU architectures. The closest thing to "mainline/stock OpenWrt" would maybe be builds for generic non-router platforms like QEMU or x86 PCs, but even these are just additional variations on the same "shared base + device-specific customization" development and distribution model. Even the OpenWrt One's builds of OpenWrt are no exception to this; they're still tailored to the specific hardware and quirks of the underlying BananaPi-based hardware, and not some generic "mainline/stock OpenWrt" image.
In light of this:
> That rules out every Linksys or GL-iNet router ever sold, because they don't ship with stock OpenWrt
They don't need to ship with stock OpenWrt to have been designed for OpenWrt, because "stock OpenWrt" ain't really a thing (per above), and because being designed for OpenWrt is a matter of whether or not the manufacturer endorses/supports the user going to openwrt.org and downloading/flashing some actively-maintained image for that particular router - and in the Linksys case at least, that was and is true (and trivially so; you just upload the OpenWrt "factory" image as if it was any other Linksys firmware update - which is unsurprising, since the stock Linksys firmware was itself (allegedly) a customized OpenWrt).
And in light of that:
> as it sits, the burden of proof is on the accuser, which would be you in this case.
That burden of proof has been satisfied, by quoting Linksys' own marketing materials (plus some firsthand corroboration, having ran OpenWrt on the WRT1900AC and WRT3200ACM).
I'll admit I'm being a little pedantic here, but nowhere near as much as one would need to be to insist that the OpenWrt One is literally the "first router designed specifically for OpenWrt" despite there clearly having been numerous preceding routers designed specifically for OpenWrt.
> So if OpenWrt One isn't first, who is?
No idea. I only know that the Linksys WRT line was designed specifically for OpenWrt multiple years before the OpenWrt One existed. Others probably existed before then. That's going by my already-pretty-strict definition of "vendor explicitly advertises OpenWrt compatibility and ships with firmware derived from OpenWrt"; loosening that to one or the other would probably extend that timeline by quite a bit.
Technically you could argue that the ol' reliable WRT54G was the first, in a backwards sort of way: OpenWrt (and DD-WRT, and Tomato, and probably others) descend from the open-source parts of the WRT54G's original Linux-based firmware, after all, and the router was clearly designed to work with that firmware. But that's certainly stretching the meaning of "designed for" quite a bit :)
The difference is the first party development and support. The OpenWrt One is the first device to have it. The One is the first device to be developed to have native support for OpenWrt and in fact run it with no changes. That’s also part of the difference. The other devices are unsupported when running “third party” firmware, which would be OpenWrt is that case.
OpenWrt is saying “this is the first router made for us, by us,” and I agree with them. Nothing you have said above discounts that or changes the truth of that statement.
In my experience WRT3200ACM with OpenWrt is garbage. Extremely unstable on OpenWrt, changing various settings to non-default values simply crashed the router. In addition, the physical ethernet ports have incorrect dimensions causing the cables to be stuck in the ports. Pulling them out damages the metal frame around the ports.
My experience was the opposite (or at least: not worse than the stock firmware, or than most routers running OpenWRT), but there are a lot of confounding variables that'd affect that difference in experience - namely, a silent hardware revision that broke things for awhile.
The sad thing is, the WRT3200ACM has more or less an unmaintained wifi driver, with 802.11w (and thus WPA3) possibly broken within the radio firmware itself. I believe there are other issues regarding regulatory settings being hardcoded in there too.
They were good enough to actually put out some tests and specs! I included the performance numbers below; heise.de also includes physical dimensions and other stats that I didn't include below.
https://openwrt.org/toh/gl.inet/gl-mt6000
reply