google-blueline: init #361
Conversation
samueldr
force-pushed
the
device/google-blueline/init
branch
2 times, most recently
from
0bc15fc
to
972165d
Compare
|
I want all the Mainly they're generic helpful changes for a vendor/mainline split device. I have thought about it... and I think the best way forward is to add a "provenance" option that allows a device to define multiple "provenances" for the sources. I'm open for other suggestions for wording. It could also have some values for a totally-vendor-based workflow. E.g. there's the official vendor sources, which have some compatibility issues, and an alternative community effort that has better compatibility. |
|
|
||
| # FIXME: apply patchsets directly on top of mainline | ||
| src = fetchFromGitHub { | ||
| # Mirror of http://git.linaro.org/people/vinod.koul/kernel.git/commit/?h=pixel/dsc_rfc&id=dcf128e83ddd25e461a5080a349de5a0552eed32 |
There was a problem hiding this comment.
Why mirror? To use fetchFromGitHub rather than a lengthy clone.
And anyway the goal is to just use the mainline kernel and add patches... So this will not stay... Though maybe the patch will come from that mirror as a span of commits.
| ./0001-HACK-Add-back-TEXT_OFFSET-in-the-built-image.patch | ||
| ]; | ||
|
|
||
| # FIXME: generic mainline build; append per-device... |
There was a problem hiding this comment.
I probably want to look at asus-dumo and pine64-pinephone (and pinetab by extension) at the same time. To figure out a good pattern for a generic mainline builder?
samueldr
force-pushed
the
device/google-blueline/init
branch
6 times, most recently
from
b600134
to
71d6a65
Compare
|
(edited the main body with the current status...) |
|
Finally circling back to this. Getting errors building with mainline selected: |
|
Oops, didn't comment on this port:
In theory you just need to normalize. IIRC normalizing automatically handles setting the right values for that new correctness harness. |
This is after I even copied the |
|
so this script reaches into the device dir for the config, but goes back up and out through the entire default/release mechanisms just to get to the config via nix? |
|
I hacked something together for the normalize script to support variants: colemickens@5405095 is where I'm at now. |
This scheme has to be implemented by any device implementation that wants to use the scheme. `grep` for `provenance` under the devices folder to get an idea as to how this can be used.
Co-authored-by: Cole Mickens <cole.mickens@gmail.com>
It's the whole content of the vendor partition.
This is used to recover from bad flashes from the mainline driver.
samueldr
force-pushed
the
device/google-blueline/init
branch
from
71d6a65
to
91c1bfb
Compare
|
@colemickens updated the normalization script to:
It ended up sharing less than I'd hoped with your initial try.
Forgot I had temporarily committed changes not to include... Undoing! Next step: update to latest release! |
This comment has been minimized.
This comment has been minimized.
samueldr
force-pushed
the
device/google-blueline/init
branch
from
91c1bfb
to
477ee59
Compare
This comment has been minimized.
This comment has been minimized.
Seems like fbcon doesn't work with the updated kernel/patchset
|
Updated to 5.15-rc4... but lost fbcon console. Having fbdev emulation enabled completely broke the display. Though it's not like everything's all peachy. It might be that the moment the DRM display is closed, or once it's tried to be opened back after having been closed, the kernel unceremoniously crashes with a null pointer dereference. |
|
Dead-end; see #511. |
Early WIP, basically first hour of work, not finished cleaning-up the history yet, but I don't want to flatten the config commits. To show how I ended up making it work.
Current status is:
It is likely that the device needs to have had an upgrade to Android 11 before this works. Just in case, download the factory image and reset to a clean Android 11 from Google.
This will not relock the bootloader.
Tested against Nixpkgs
6933d068c5d2fcff398e802f7c4e271bbdab6705Will obsolete #191. Also ran: #44.
cc @colemickens @craigem
TODO
Current state
Not sure there's really anything to do, the vendor port is about as good as other devices of the same vintage... and really mainline is a further improvement type of thing.
And I'm happy with the provenance option to allow multiple kernel sources to be defined.
Mainline
rq2atorq3abootloaders and partitions on the device.Doing the userspace stuff listed here is likely the only thing left for WLAN, and maybe even other things like cellular radio and GPS to work.