Jump to content

Orange pi 4G-IOT

Ondřej Glogar

By Ondřej Glogar
in Off-topic

 Share

Recommended Posts

chwe

chwe

Posted
Posted

My personal opinion: I don't think so, the SoC has (to my knowledge) no mainline support and it doesn't look that someone pushes it to mainline. So you have to deal ~15GB android BSP SDK to figure out what works and what needs to be adjusted so that you can boot up Linux. Probably most of the 'nice features' (e.g. GSM, LTE, Camera) need some additional attention to bring it up properly..  Therefore you need a developer which will do all this work but chances are rather low if we talk about a SoC which can be only found in one SBC. To my knowledge @zador.blood.stained tried this once for the 'little brother', the 2G-IoT, but stopped too cause it was to much crap to deal with. It's a bunch of work for only a limited use-case. 

IMO cheapest way to get 4G connectivity is to buy a USB dongle (with known good linux drivers), together with a 'cheap' well supported SBC. 

TonyMac32

TonyMac32

Posted
Posted

From where I'm sitting I can't see paying that much for it when I can get a 4G USB dongle for a well-supported chip

chwe

chwe

Posted
Posted
6 minutes ago, TonyMac32 said:

From where I'm sitting I can't see paying that much for it when I can get a 4G USB dongle for a well-supported chip

 

well, a lot of people didn't saw it coming that trump would be your president ~1.5 years ago.. :D It has some nice features (e.g. mipi csi if I've it right in mind, probably a nice pmic with some powersave functions etc). But since I got fooled by rockchips csi which is for sure more maintained, I wouldn't mess with this one...

lrou

lrou

Posted
Posted

> From where I'm sitting I can't see paying that much for it when I can get a 4G USB dongle for a well-supported chip

Could you please share what's that cheap 4g-gps dongle, comparing to 4g-iot? Or just 4g dongle?

TonyMac32

TonyMac32

Posted
Posted
16 hours ago, lrou said:

Could you please share what's that cheap 4g-gps dongle, comparing to 4g-iot? Or just 4g dongle?

 

What I mean to say is, I'd prefer to have a widely supported SBC and add 4G rather than gamble on something at that price.  Were it $20, sure thing, I'd buy one for an experiment.  But at $45, well...

lrou

lrou

Posted
Posted
On 5/19/2018 at 9:06 PM, chwe said:

My personal opinion: I don't think so, the SoC has (to my knowledge) no mainline support and it doesn't look that someone pushes it to mainline. So you have to deal ~15GB android BSP SDK to figure out what works and what needs to be adjusted so that you can boot up Linux. Probably most of the 'nice features' (e.g. GSM, LTE, Camera) need some additional attention to bring it up properly..  Therefore you need a developer which will do all this work but chances are rather low if we talk about a SoC which can be only found in one SBC. To my knowledge @zador.blood.stained tried this once for the 'little brother', the 2G-IoT, but stopped too cause it was to much crap to deal with. It's a bunch of work for only a limited use-case. 

IMO cheapest way to get 4G connectivity is to buy a USB dongle (with known good linux drivers), together with a 'cheap' well supported SBC. 

On orangepi site there is some message about Linux ETA - it's in about mid-July. So, let's see, it's no long time to wait

Dan25

Dan25

Posted
Posted

Problem is, sourcing decent LTE dongles is a challenge, and generally costs more than $50 per dongle. Plus, the dongles that are availble nearly all run an insecure version of Linux onboard, which is great when you want to create a malicious USB LTE dongle, but not so great when your just looking to have a backup internet connection. If this board just had mainline support for the USB ports and LTE modem, it'd be awesome to throw OpenWRT onboard and some USB to ethernet adapters, as it would make for a decent all in one LTE modem.

chwe

chwe

Posted
Posted
52 minutes ago, Dan25 said:

Plus, the dongles that are availble nearly all run an insecure version of Linux onboard,

hmm..  do you really think the kernel from this bsp saw ever a patch? :rolleyes: I didn't even spot one commit about mt6737, it's an elcheapo android phone SoC.. Mt starts to mainline some of their SoCs (e.g. 7623) but I think they have minor interests in mainlining this one.. And even if they start, it would probably need a lot of time until you get decent support. So to summarize... you have only one board using this SoC, you have a not maintained BSP kenel, you probably deal with a binary drivers for the whole LTE stuff (I didn't look into the 14GB android SDK). Do you still think it's worth the efforts to clean up everything for just one board? This is just the efforts needed to bring it up.. You still have to maintain the the board & kernel...  There is no linux-sunxi around MT mobilphone SoCs doing all this stuff for free.. There's no bootlin mainlining the VPU and Xunlong is not famous for maintaining their boards.. (they have cheap nice hardware, but 'doing software' doesn't seem to be their focus..)

 

 

Dan25

Dan25

Posted
Posted
2 hours ago, chwe said:

So to summarize... you have only one board using this SoC, you have a not maintained BSP kenel, you probably deal with a binary drivers for the whole LTE stuff (I didn't look into the 14GB android SDK). Do you still think it's worth the efforts to clean up everything for just one board? This is just the efforts needed to bring it up.. You still have to maintain the the board & kernel...

I think it might be worth it to have this board mainlined, the price point is significantly cheaper than any other SBC with LTE, of which there aren't terribly many. A 2nd revision with ethernet ports would be notably more useful for an integrated router application, which is my ideal use case.

 

Here in the US, 2G GSM has been shut down by AT&T, with only T-Mobile still offering it (till 2020, then who knows), and 3G UMTS/HSPA+ is already shut down in many areas on T-Mobile to make way for LTE. Is it really a good idea to deploy things like the OrangePi 2G-IOT when the only network that supports their cellular chipset has only committed to 18 months more support? Even if you get till 2021 before 2G networks shutdown, an inexpensive SBC with LTE could probably run for many years more than older 2G and 3G boards.

chwe

chwe

Posted
Posted
18 hours ago, Dan25 said:

I think it might be worth it to have this board mainlined,

for users... for sure.. for developers, this might look a bit different..  Mainlining a SoC isn't done in one or two days.. Just look how old those H3 boards are and they're still not fully mainlined (good enough for a bunch of use cases but people still complain that *random feature* doesn't work.. :D )...

18 hours ago, Dan25 said:

Here in the US, 2G GSM has been shut down by AT&T, with only T-Mobile still offering it (till 2020, then who knows), and 3G UMTS/HSPA+ is already shut down in many areas on T-Mobile to make way for LTE.

I think 2G will live a way longer than 3G... A lot of 'IoT' before people called it IoT stuff is based on 2G... Here in switzerland for example those little orange 'emergency call boxes' you find on highways:

notrufsaeule.jpg

have to my knowledge such 2G modems inside and a bunch of 'sensor nodes' for whatever use-case... As long as bigger players happily throw as much money as needed to T-Mobile and there's no other opportunity than t-mobile, they will happily take your money for providing 2G... :D 

 

18 hours ago, Dan25 said:

an inexpensive SBC with LTE could probably run for many years more than older 2G and 3G boards.

correct, but the mt6737 is from early 2016,  new cellphone soc are more or less thrown out every year and no cellphone SoC maker decided to my knowledge to bring such a chip upstream (you can be mostly happy when your android phone gets more than one update after you bought it... :D). If somebody seriously works on mainline for this SoC, we should see first commits on the mailing lists (the board is long enough available) but I didn't spot one related to it.. So chances are hight that mainline kernel developers aren't interested in this chip and therefore IMO armbian shouldn't try to support it. If somebody is patient enough, download the SDK, strip out everything you need and throw away the crap, reorganize it properly on different repos, try to build, try to fix stuff and write a 'board bring up' proposal what's working, what's not.. I tried this the first time for the BPi R2, a more or less well supported SoC on mainline, with some public available information, which is known that it boots up linux and it still took me days to bring it to a point where it boot an image built with armbians buildscript (the kernel is IMO well maintained, others work on it and solved a lot of the issues I had prior to me)... and it's still far away from usable.. 

Dan25

Dan25

Posted
Posted

Android 8.1 for this device is unstable from my testing, though the prebuilt image is using the older 3.18 kernel. Perhaps a build with kernel 4.4 will be a bit more stable! Also, the LTE bands listed on this board are inaccurate, and are aimed squarely at the Chinese market. Support for common/required bands in the Americas and Europe is mostly missing.

TonyMac32

TonyMac32

Posted
Posted

@Ondřej Glogar I am going to say that for myself no, I won't be supporting this board.  The lack of engagement by the other devs tells me they have no plans either.  Without additional devs I think we're probably at a critical mass for number of platforms and "unique" boards.

Igor

Igor

Posted
Posted
8 hours ago, TonyMac32 said:

I think we're probably at a critical mass for number of platforms and "unique" boards.


Most likely above :) It is already hard to provide support for existing hardware or do any serious development.

 

14 hours ago, Ondřej Glogar said:

Still nothing..


Majority of current engagement is sponsored by our private resources which means we probably can't expand this project further since it can become more fragile and could collapse. We actually need to step down a bit and from my personal perspective, I need to get few weeks off to recharge. 

1
chwe

chwe

Posted
Posted

Just for those interested in this board, based on their android 6 sources (http://www.orangepi.org/downloadresources/orangepi4G-IOT/2018-03-28/1a019a7d51c01f91998a1c96eba7f5cb.html):

 

  • bootloader lk, seems there are no u-boot sources
  • preloader no idea seems that it needs to be compiled too 
  • some atf garbage I didn't even have a closer look into it
  • 3.18.19 kernel (at least it looks like DT is somehow there, but EOL since more than a year)
  • modem - just another pill of crap I didn't looked at it.. 

I don't think their android 8.1 sources will look much better but I might have a look into it too. But to summarize, this would be a board with a kernel which is EOL before support even starts.. There is only an LK bootloader no u-boot and not much documentation. This might be a reason why xunlong never delivered a Linux for this board. 

chwe

chwe

Posted
Posted

Shortly update of the oreo sources:

unexpected, there is a 4.4 kernel: 

  • VERSION = 4
PATCHLEVEL = 4
SUBLEVEL = 95
  • bootloader seems to be still more or less the same (preloader and lk, no u-boot)

At least there would be a kernel which isn't EOL yet.. Quality of the 4.4 sources are unknown. and LK might be the reason why there's no 'official' linux provided by xunglong. It's not u-boot and projects where people booted something like debian with LK aren't well that often provided with some how-to's. 96boards have done it in the past. 

hjc

hjc

Posted
Posted
4 hours ago, chwe said:

96boards have done it in the past. 

That can be viewed as Android with rootfs replaced with GNU/Linux, thus no real Linux experiences (like updating the kernel with "apt upgrade" instead of connecting to a PC to run a set of fastboot commands)

Though it is possible to load u-boot from LK, like those working on dragonboard 410c/820c. In this way, u-boot acts as (pretends to be) the kernel.

TonyMac32

TonyMac32

Posted
Posted

So I'm guessing there is no documentation on the SoC either?  Perhaps those Mediatek guys who were given the... task of supporting Bpi could be helpful?

chwe

chwe

Posted
Posted
2 hours ago, hjc said:

That can be viewed as Android with rootfs replaced with GNU/Linux, thus no real Linux experiences (like updating the kernel with "apt upgrade" instead of connecting to a PC to run a set of fastboot commands)

maybe disable LKs trusted zone (from what I saw, xunglong did this anyway) and then doing some hacky stuff to replace the kernel otherwise on the flight.. LK knowledge on my side is 'non-existing' based on some 'tutorials/ forumstuff' for the dragonboard.. It also seems that xunlong still uses the 3.18 kernel for their android oreo (there's a 4.4 but not sure if they use it due to 3.18 sources are also there). I guess proper linux support will be painful for this board.. Even some android/linux mixup will be a pain... I guess there aren't many people experienced in LK here and you've to start with 20gb android 'sources' without any documentation. Such an android BSP looks like a complete mess to me... No idea how people can deal with it.. 

2 hours ago, TonyMac32 said:

So I'm guessing there is no documentation on the SoC either?  Perhaps those Mediatek guys who were given the... task of supporting Bpi could be helpful?

probably an other business unit? If I've it right in mind @Ryder.Lee is related to mediatek and can comment on this? 

VikramSN

VikramSN

Posted
Posted

i am facing problem to flash android 6.1 or android 8.0 img file. 

i follow orange pi user manual instruction but at last i download SMART PHONE FLASH TOOL and trying to flash .img file but when i click download button it stuck and nothing happen.

i also install driver for win7 64 bit.

and also connect Laptop and Orange pi using USB Data Cable.

Please help me.

 

Orange Pi version     -    4G IoT

SD Card               -    Class 10, A1, 16GB

tool                  -    Smart Phone Flash tool as in user manual instruction  

TonyMac32

TonyMac32

Posted
Posted

I can't help, since I've never installed Android on a single board computer (usually I'm deletiglng it), and this is a Linux project, after all, and this is an unsupported board.

Guest
This topic is now closed to further replies.
 Share
Go to topic listing
×
×
  • Create New...

Important Information

Terms of Use - Privacy Policy - Guidelines

  I accept