So, if we ever manage to get proper open-source control of the cellular radios of random smartphones, would it be possible to use those cheap devices to create a GSM mesh network?

Not legally, since all of the frequencies in question require licensing and most are allocated to mobile phone providers.

But yes, this is already possible, in fact, Osmocom has a fully open baseband running on certain phone chipsets.

Note that the Osmocom baseband project isn't aimed at end-user devices, only at researchers. For example it runs part of the stack on the laptop the phone is attached to.

I guess this is GSM, but I heard 5G operates on some license free bands.

The CBRS band (48, 3.6Ghz) could be used for this. Support for this band only exists in the very newest smartphones.

Due to the frequency division/time division duplexing nature of 3GPP protocols, a mesh network is very challenging and non-trival. The 5G standardization has some study items on multi-hop networks, but that is a ways out and I’m not sure how and if it will even manifest in the standard.

it's funny how we have two hundred dozen squared projects claiming their goal is mesh networking to fight the police state du jour, while on the other hand we have a bunch of badly known full scale GSM AGPL platforms. :/

I used this once to configure a small GSM network using a LimeSDR.

Has anyone started a project like this for LTE or 5G?

There are a couple projects of differing scope:

- srsLTE (https://www.srslte.com/)

- OpenAirInterface (https://www.openairinterface.org/)

I've successfully used srsENB as a base station, largely to generate "model" LTE waveforms so that I could then decode them and verify my understanding. srsLTE is quite well written code-wise and is actively improving/growing. Seems to have a supportive single company and community behind it.

OpenAirInterface is more geared towards 5G research and is a lot more ad-hoc in terms of documentation and capabilities.

Please note: OpenAirInterface is not open source software. Large parts of it (particularly everything related to the RAN) are under a license that does not comply with either DFSG, nor OSI Open Source Definition, nor the FSF Free Software Definiton.

A couple more links to such projects:

https://open5gs.org/ https://github.com/nextepc/nextepc

indeed, if you want a 100% open source 4G network, you can use srsLTE for the eNodeB, and open5gs for the EPC. No proprietary bits needed, and no non-free licenses needed either.