A Popular European Cable Modem Is Now Open-Source

[kaliszad]

microcode you weren't expecting miracles, were you? The blobs are exactly the point, why the OpenWRT and mainline kernel support is pretty much non-existent. If you don't have the blob, you can't run your device. If you can build something, which kind of works, you can start your reverse engineering from there. Nobody starts from scratch nowadays, also nobody takes a huge pile of binary data and tries to reverse engineer drivers and firmware from it, at least from what I know.

tigerroast you're welcome

[darkbasic]

You can check out the Turris Omnia by CZ.NIC https://www.indiegogo.com/projects/t...-router#/story that router should fulfil most of your needs. I was wondering, why that open source SW&HW router didn't get more attention on Phoronix

Turris Omnia doesn't have and open source modem and that's why I didn't buy it.

[cthart]

I hope this will actually be usable and not be nixed by cable providers who won't talk to a remote modem that's not one of "theirs". All the cable modems I've had have been remotely configured and controlled by the cable provider...

[kaliszad]

darkbasic yes, that is one thing that concerns me too. See this discussion on Turris Omnia discussion/ QA website https://discourse.labs.nic.cz/t/devi...urris-omnia/45 if one could at least bridge correctly and easily with modern modem/ router devices from the provider, Turris Omnia would be just the next device after the now dummy modem bridge (basically just a smarter adapter to translate for instance DOCSIS over coaxial to ethernet over TP, what ever, this is not my expertise currently.

cthart yes, that might be an issue. Cable providers can be very cocky about people messing with their network in their opinion. People just need to push, so that the provider will just publish the needed config on their website. Basically this should be something like "send us the MAC address of the device, configure DHCP client on your device". You just wouldn't do the "public" Wi-Fi feature and maybe no telephony but in the age of cellular phones, skype, WebRTC etc. who cares?
There is something seriously wrong nowadays with providers having a black box in your home through which you route most or all traffic and phone calls (usually). This approach is so 20th century. Today's networks should be robust enough, that a group of devices cannot bring the network down and just not be able to connect if configured badly (like if somebody sets static IP address). This approach should spread to cellular networks as well but that is another problem entirely.

The thing is with Turris Omnia, I actually have a symmetric 1 Gigabit WAN connection and it is accessible as a normal RJ-45 port from the wall. Therefore I can use the power it provides.

[microcode]

microcode you weren't expecting miracles, were you? The blobs are exactly the point, why the OpenWRT and mainline kernel support is pretty much non-existent. If you don't have the blob, you can't run your device. If you can build something, which kind of works, you can start your reverse engineering from there. Nobody starts from scratch nowadays, also nobody takes a huge pile of binary data and tries to reverse engineer drivers and firmware from it, at least from what I know.

tigerroast you're welcome

Didn't mean to imply a moral or technical failure. Just didn't want anyone to get the impression that all the source was there rather than just the GPL bits.

It's good work, and the kernel should be easy to rebase, just need to sync the changes on top of the v2.6.30 tag. once you have that, you can inch it up to 2.6.32, which has LTS patchsets.

At a first glance, a lot of the changes are really ugly though.

[kaliszad]

microcode we might also question concrete persons code. See this Github search of the repository in question: https://github.com/search?l=cpp&q=ru...utf8=%E2%9C%93 just don't be to harsh on him, he might be the person, who actually contributed the source code to Github. I haven't seen any concrete signature in the emails from Technicolor. Also their reading skills are not that great, the addressed me "Dear M. Kalisz" even though in an answer, I have written explicitly "Mr. Adam Kalisz" as my signature.

I know, you haven't meant your answer badly. I haven't answered in a harsh way, have I? I haven't seen any company to make a 180° turn and really open source everything correctly though, so your comment was just pointing out something, which a knowledgeable person, a reader of this site, should imply anyway.
Your comments on the porting seem to be informed. Why wouldn't you try to port it to something like 3.18 or 4.1 which are also longterm support kernels, especially if a lot of the code would have to be refactored anyway? I don't have any expertise with kernel development, porting or reverse engineering so a bit more background or a link to a good article/ book about it would be nice ;-)

Btw. Turris Omnia will be running the 4.1 kernel initially, at least that's what I heard.

[microcode]

microcode we might also question concrete persons code. See this Github search of the repository in question: https://github.com/search?l=cpp&q=ru...utf8=%E2%9C%93 just don't be to harsh on him, he might be the person, who actually contributed the source code to Github. I haven't seen any concrete signature in the emails from Technicolor. Also their reading skills are not that great, the addressed me "Dear M. Kalisz" even though in an answer, I have written explicitly "Mr. Adam Kalisz" as my signature.

I know, you haven't meant your answer badly. I haven't answered in a harsh way, have I? I haven't seen any company to make a 180° turn and really open source everything correctly though, so your comment was just pointing out something, which a knowledgeable person, a reader of this site, should imply anyway.
Your comments on the porting seem to be informed. Why wouldn't you try to port it to something like 3.18 or 4.1 which are also longterm support kernels, especially if a lot of the code would have to be refactored anyway? I don't have any expertise with kernel development, porting or reverse engineering so a bit more background or a link to a good article/ book about it would be nice ;-)

Btw. Turris Omnia will be running the 4.1 kernel initially, at least that's what I heard.

I don't mean to particularly pick on the folks at TC; anyone who writes software professionally has written some that they wish they'd never show to others.

More specifically, I was noting that since it's currently based on the 2.6.30 tag, there's a clear path to 2.6.31 and 2.6.32.
If there's some specific major feature improvement you'd be looking for (maybe iptables improvements, eBPF, or new filesystems work) I would suggest refactoring the patches before attempting to forward-port them that far.

I'm not in EU, so I only took a look at this out of curiosity. Even if I would do this, I likely wouldn't have the hardware or any personal reason to do the work.