If you do that I'd be interested in seeing a custom-build done.  I started on 
this myself last year with a Netgear R6050 but due to my inexperience with 
setting up an OpenWRT toolchain and lack of time, and in particular the low 
price of other, supported, devices, I never got a "Round Tuit"   I did post my 
findings here:

https://forum.openwrt.org/t/openwrt-support-for-a-netgear-r6050/179491

And it looks to be a very easy port job, just a few mods of the R6020 build.  
If there's any interest I'd be happy to solder on a serial port to the thing 
and show up with it and a laptop and tested with nmrpflash ready to go.

Unfortunately, while I had asked some specific questions on what to expect from 
setting up a build system for openwrt on that forum - none of the dev's 
responded to any of them.  Sigh.

Since then I've gotten a lot more experience installing OpenWRT builds.

In my experience openWRT compliant devices either fall into 1 of 2 camps:

You got the ones like the Cisco Meraki MR52, or many of the TP-Link devices, or 
the 4Mb flash devices like the WNR2000, which require special, long procedures 
to install - usually a rs232 serial port, etc. or with the tp-links, pulling 
the factory firmware down to a Linux box, slicing off the uboot header and 
sticking it on the front of the openwrt firmware.  Or finding older versions of 
openwrt.

This can take a lot of time and reading to figure out the secret sauce for the 
specific device.

Many of the custom builds I've seen people do in the openwrt forum for these 
edge cases involve cutting out extras to get newer firmware releases to fit in 
4MB fash devices.  I've used a number of these builds people have produced and 
found them to be all of high quality.  Needless to say, I used them on devices 
that are behind firewalls, and are just being used as client bridges or access 
points since many of them are of older releases.

Then you got everything else in which installing openwrt is a snap - 5 minutes 
and you are done.  Because I use OpenWRT on a LOT of devices (I have probably 
50 of them I manage now) I have found it just easier to buy supported devices 
that are in the "installation is a snap" camp.

But it still would be interesting to see a few border case installs, like my 
R6050


Now, in the SOHO router vein, here are 3 other SOHO router presentation PLUG 
ideas I've been considering over the years:


Packet Per Second speed

There's a huge dearth of information out there on just how fast these devices 
work.  With PC-based routers, it's easy.  Even a crappy Intel Core i5 
generation 4 CPU desktop - the kind of stuff companies today are forklift 
chucking into the garbage - can run a full ethernet-to-ethernet Internet 
routing for a 1Gbt Internet connection.  And, at least in the United States, 
very few people still have this fast an Internet connection.  Most are 
asymmectrical.

So what tends to happen is that people who need real power, real speed - they 
go to used 1U rack mounted servers.  For example something like this:

https://www.ebay.com/itm/235527553078

For $80 you get a Xeon with 2 gigE ports in it ready to rock and roll.  You can 
install pfSense or OPNsense on that and have a router ready to handle a 1000 
users on a business Internet connection without even knowing hardly anything 
about routers.

But, the ARM chipsets in these SOHO devices are surprisingly powerful in 
comparison.  However, setting up a test bench with 2 servers on each side of a 
router that can spew high rates of packets of various sizes at each other 
through the router isn't that simple.  Not a lot of people do it and test these 
devices, nor do they even do this and test "commercial" routers like various 
Cisco offerings.

It would be really cool to setup a presentation with a collection of devices 
some with OpenWRT some with manufacturer's firmware and ran them through the 
gauntlet until they fell down on their faces.

I know most people buy these for their wifi chips and wifi just plain old isn't 
very fast, but they work as ethernet-to-ethernet routers also.


Debricking

When I first started out playing with alternative router firmware if I made a 
mistake and bricked a router that I couldn't unbrick with the usual tftp 
methods, I'd then put it in a box to get to "one of these days"

Over time that box got larger and finally, attempting to avoid serial port 
nonsense, I tried some of the "pin short" methods of debricking.  This worked 
OK for a few models that supported it until one day I fried a router doing that 
and had to throw it out.

So then I finally bit the bullet, bought the proper 3v+ level serial adapter 
and started in with the soldering iron.  And I found that actually soldering on 
a serial port was not that difficult.  There's some tricks to the trade of 
course - but one weekend I did an unbrickfest and worked through my cardboard 
box of devices until I ran out of them.

Now I'm slowly building that box up again and one of these days I'll do 
another.  It's actually quite easy to do these once you get the hang of it and 
a procedure worked out.

I even built, years ago, a jtag adapter out of a parallel port and used that 
with the handmaid software on a wrt54g, but never did it for debricking, only 
just to see how to do it.  Doing it taught me that if the router is so scotched 
that you have to jtag it - your time is better spent tossing it in the garbage 
and buying a replacement.


Big 3 overview:

The last idea I had concerned a huge giant hole in OpenWRT:  it's support for 
Broadcom-based chips.  IT STINKS.  And in some cases it's downright evil.  
There are devices where you install OpenWRT and it boots, you enable the 2.4Ghz 
radio, everything works fine, you enable the 5Ghz radio and you brick the 
device since it goes into a bootloop.  And there is no warning on the hardware 
page for that device that this will happen.

Love them or hate them, Broadcom-chap based wifi routers probably have close to 
50% of the market.  So there are tons and tons of older yet very good router 
hardware gear out there that is Broadcom.

For example I picked up a nice Netgear R6700v3 AC1750 yesterday for all of 
$7.50 from Goodwill.  This has a 1Ghz CPU, hardware address translation that 
will allow the device to get to 95% throughput of an ethernet-to-ethernet 
routing of a 1Gbit Internet connection, and can source or terminate a wireguard 
or OpenVPN vpn.

The catch is it's Broadcom based.  Happily, there's 2 alternative firmwares out 
there  FreshTomato and dd-wrt.

FreshTomato uses the original K2.6 linux kernel.  Dd-wrt uses the newer K3.0 
kernel.  And these firmwares are all completely patched for the latest 
vulnerabilities just like OpenWRT.

While it is true that using OpenWRT allows you to run newer kernels, is it 
really worthwhile to be constantly chasing the latest kernel that the embedded 
Linux project releases?  These devices are toasters and who cares what the 
kernel version is as long as the firmware of the device isn't vulnerable.

Possibly a presentation on "the big 3" alternative router firmwares might be of 
interest?  Something along the lines of:

FreshTomato:

Limited set of features but extremely stable since the developer uses the 
manufacturers binary blob hardware drivers that are released with their OSS 
source.  Developer did not sign an NDA with Broadcom.  Compilation process for 
firmware is easy.

dd-wrt:

Much more full featured including support for the Broadcom hardware NAT that 
allows for high speed packet routing however requires more lengthy testing 
periods for firmware used in production since the developer is using newer 
kernels with binary blobs released by Broadcom rather than vetted by the router 
manufacturer.  Developer did sign an NDA with Broadcom.  (note that dd-wrt also 
supports open chipsets like Atheros/Qualcomm with much newer kernels however 
this is a separate dev branch)  Compilation process for firmware is complicated 
requiring oddball toolchains and so on but instructions are available on how to 
set it up.  The very best wiki and forum support of all the three (my opinion 
obviously)

OpenWRT

Probably the most full featured but support is spotty on the wiki and forums 
particularly for newbies and particularly for the oddball hardware/corner cases 
(Like I cited above)  Easier to load binaries since the flash isn't in 
read-only mode like the other 2.  Compilation is probably in between the other 
2 in level of difficulty.  Definitely easier to produce custom-builds of 
OpenWRT with embedded programs that aren't in the default OpenWRT since by 
cutting off all the proprietary wifi chipsets out there,


Possibly a presentation with a demo of 3 of these devices with the 3 major 
alternative firmwares and their web interfaces, along with some common 
scenarios (static IP's assigned via DHCP, special DHCP options used for phones, 
VLAN routing, multiple secondary IP assignments to the same interface, 
subnetting different interfaces, etc.)

Anyway, just some random thoughts.

Ted

-----Original Message-----
From: PLUG <[email protected]> On Behalf Of Russell Senior
Sent: Friday, October 11, 2024 3:17 PM
To: [email protected]
Subject: [PLUG] Next meeting topic?

I think I mentioned this at the last meeting, but I don't have a speaker or 
prospect for November yet.

So,

   a) Does anyone have a topic they want to do a presentation on or a plausible 
lead on a speaker?

If not, then:

   b) I am thinking about doing a kind of OpenWrt clinic at the November 
meeting. That is, I/we bring some devices, lay them out on a table, show what 
installing OpenWrt looks like, show what building OpenWrt looks like, maybe 
people can bring devices they want help with. That sort of thing.

Thoughts?


--
Russell Senior
PLUG Volunteer
[email protected]

Reply via email to