OESF Portables Forum
Model Specific Forums => Gemini PDA => Gemini PDA - Linux => Topic started by: Nimrod on February 16, 2020, 04:16:31 am
-
Hi there!
I have had a look onto different pages of mainlining efforts, but all of them don't contain a lot of information on what is working, what is not working and what needs to be working to be able to run a mainline kernel.
Does anyone have an overview of the linux mainlining efforts?
Best regards,
Nimrod
-
The last thing I did was:
https://lists.infradead.org/pipermail/linux...ber/024041.html (https://lists.infradead.org/pipermail/linux-mediatek/2019-October/024041.html)
No response so the project is somewhat wedged : (
-
I had a look at your email to the list (which I'm watching now to see if relevant patches hit it...) and also the discussion going on there in September 2019 [1]
As I understand it, one can already boot mainline 5.3 or 5.4 to a busybox shell over UART via USB but the device starts overheating. Your latest effort of mainlining can be found in the gemian linux repo [2] and enables thermal management of the CPUs. You posted that on the list but for some reason there was no feedback there. Did I get this right?
Does the implementation of thermal management mean that one can now safely boot a Gemini with mainline without overheating?
As far as I get it, the linux-mediatek list is operating on rather high volume... is there any chance your (understandably rather long) email just drowned in the load? It would be great to see more support going upstream!
Either way, thanks a lot for both your effort and the useful information!
[1] http://lists.infradead.org/pipermail/linux...ber/023478.html (http://lists.infradead.org/pipermail/linux-mediatek/2019-September/023478.html)
[2] https://github.com/gemian/linux/ (https://github.com/gemian/linux/)
-
The overheating thing was what I thought at the time, but having done the CPU temperature work I've not noticed it going up, I'm not certain what was going on.
The device did get low on power when plugged in via the USB-UART cable and when rebooted to a working kernel it refused to mount read write. It could have been because of their not being enough juice, or some other part than the CPU doing the overheating. There are other temperature sensors in other parts, not yet added.
The physical external temperature of the device was the highest I've ever felt it at that time, giving it a few minutes off and recharging would get it working again. With the later work I was careful to always plug it back into a proper power source to keep it topped up between tests so have never had the same flat-bat/overheating/read only problems.
Matthias maintains an official overview: https://mtk.bcnfs.org/doku.php?id=linux_mainline_effort (https://mtk.bcnfs.org/doku.php?id=linux_mainline_effort) though as per my list mail I consider it to be in need of an update.
My mainline branch is re-based upon the latest each time I go back to the project. If anyone else wants to help out I'll happily do another re-base given a nudge. Its probably easier for me as I know the changes.
In terms of the list not noticing my email, that is possible though Matthias did ping me in early December to see if I'd made any progress, so I pointed him back to my October email, which also lead to no response. Having someone else being interested in this work, and preferably replying to the list with thoughts would help bring it back to the top of mailboxes etc. If there is anyone willing to jump through the upstream hoops the cpu-temp stuff could be cut out (ignore the test code for now) and pushed upstream.
In terms of next steps, I'm keen to verify the IO/I2C 'magic numbers' that appear to me to be failing and no one will answer my questions of where and how these things are supposed to be configured. We obviously have a working kernel to compare against but the mediatek mainline community appear to want things done differently but are unwilling/unable to answer my questions.
It would also be good to start on the MT6771 too so that we can bring them up in parallel.
Though for me the project is basically shelved pending the finding of collaborators/answers.
-
Altough I am interested in booting mainline Linux on the Gemini, I have no prior experience in Linux kernel development and only rudimentary understanding of how drivers interact with the hardware. So, I don't think I can make reasonable contributions in that regard...
Also, looking at the MediaTek proprietary codebase in the Gemian 3.18 kernel makes me feel like there is a lot to port to even get close to a useful barely working system on mainline... especially concering the display driver, I guess, which leaves debugging only possible via USB UART cable.
Concering the MT6771 chipset of the Cosmo, I don't plan onto getting one, so I wouldn't be able to test things there.
-
I wish Planet were trying to ensure that devices get into the hands of people who would be interested in working in this area. It shouldn't have to be a one man effort! (I'm sorry to say I also have no experience of working in this area.)