Independent Developers Tackling Snapdragon 630/660 SoC Support For The Upstream Linux Kernel

[Haxk20]

isn't it ironic how the big silicon companies do a sh1t for linux support and it's up to volunteers to clean the mess up, reverse engineer missing bits and bring it upstream?

It is indeed sad. We dont even access to documentation that linaro gets. So its seriously reverse enginnering

[konradybcio]

isn't it ironic how the big silicon companies do a sh1t for linux support and it's up to volunteers to clean the mess up, reverse engineer missing bits and bring it upstream?

This really is an issue. With enough upstream support, you could buy a few-years-old "broken" phone and turn it into a low-power-consumption server equipped with its own UPS (well, the battery), a more than capable GPU, video encoder/decoder hardware, Wi-Fi, BT, a cellular modem(!) and all that jazz. Sadly, most of the SoCs are unsupported. But hopefully that'll change.

[Deetwenty]

This really is an issue. With enough upstream support, you could buy a few-years-old "broken" phone and turn it into a low-power-consumption server equipped with its own UPS (well, the battery), a more than capable GPU, video encoder/decoder hardware, Wi-Fi, BT, a cellular modem(!) and all that jazz. Sadly, most of the SoCs are unsupported. But hopefully that'll change.

Not only that my experience from having a Jolla with sailfish for some times (before I cracked the screen) tells me that native Linux apps are a lot more powerful than many Android apps, so long as I kept to native apps and didn't browse to much it was a reasonable capable daily driver. So I think a lot of those older phones with something like plasma mobile (or the equivalent GTK project) will still be pretty capable as daily drivers.

[S.Pam]

This whole issue is because of two things

1. The vendor didn't open source the drivers or didn't provide updates
2. Linux doesn't load drivers/mods made for earlier kernels. This is a big issue in Linux... If they had a stable interface for drivers we could have had much better support for older stuffy on newer kernels.

[konradybcio]

This whole issue is because of two things

1. The vendor didn't open source the drivers or didn't provide updates
2. Linux doesn't load drivers/mods made for earlier kernels. This is a big issue in Linux... If they had a stable interface for drivers we could have had much better support for older stuffy on newer kernels.

Well, both of your points are wrong.


1. They *do* provide source, but it's of so bad quality that it's not going to be accepted upstream. Though this is changing with Google's GKI concept (as in the drivers are likely going to be distributed as proprietary kernel modules)..

2. Linux _does_ have a stable-ish ABI (yeah, it could be better though..), but downstream modifications break and contaminate it by adding things on top of it (just diff any Qualcomm kernel against corresponding torvalds version...) making it essentially impossible to plug-and-play older drivers. Also, until this year most downstream kernels were based on ancient Linux releases (most SoC vendors jumped to the next LTS version every flagship SoC release, Qualcomm hit 4.19 this year)

[S.Pam]

Well, both of your points are wrong.


1. They *do* provide source, but it's of so bad quality that it's not going to be accepted upstream. Though this is changing with Google's GKI concept (as in the drivers are likely going to be distributed as proprietary kernel modules)..

2. Linux _does_ have a stable-ish ABI (yeah, it could be better though..), but downstream modifications break and contaminate it by adding things on top of it (just diff any Qualcomm kernel against corresponding torvalds version...) making it essentially impossible to plug-and-play older drivers. Also, until this year most downstream kernels were based on ancient Linux releases (most SoC vendors jumped to the next LTS version every flagship SoC release, Qualcomm hit 4.19 this year)

Right, that's why we all can run our old phones with new OS on them?

Stable interface is what we have in Windows Driver Model, where a driver will work for a very very long time. This is severely lacking. Have you tried to load a kernel module from an older kernel? How often does that work?

[ed31337]

Stable interface is what we have in Windows Driver Model, where a driver will work for a very very long time.

Uh huh... Tell that to my CanoScan LiDE 30 flat bed scanner. Hasn't worked in Windows since Windows XP. But works great, right out of the box, on any recent Ubuntu Linux. Took a little downloading to get it to work on my Raspberry Pi 4, but still, it really works well. Meanwhile, Windows doesn't work at all.

[skeevy420]

Right, that's why we all can run our old phones with new OS on them?

Stable interface is what we have in Windows Driver Model, where a driver will work for a very very long time. This is severely lacking. Have you tried to load a kernel module from an older kernel? How often does that work?

Have you actually ever diffed the sources for Motorola or LG kernels with upstream Linux of the same version? They've never diffed cleanly. I'm using Motorola and LG in that example because I've done that with the kernel sources released for the Motorola Bravo, Defy, and Atrix HD as well as the LG G3 and V20. I used to be big into roms and porting them to my phones.

I haven't bothered with my Motorola G8 Power since I have to void the warranty to unlock the bootloader. GRRRRR.

I wonder how much these bootloader tied to warranty policies stifle innovation?

[pal666]

Linux doesn't load drivers/mods made for earlier kernels. This is a big issue in Linux...

it's a big issue with stupid windows slaves like you. with linux it is one of its advantages

If they had a stable interface for drivers we could have had much

worse linux. just use any other inferior kernel with stable interfaces and keep your hands off linux

[topolinik]

On GSMArena, when searching for just Snapdragon 630+660 SoCs you get 62 phones (includind mine, great!), this will be great for XDA people!