mobile-nixos icon indicating copy to clipboard operation
mobile-nixos copied to clipboard

make modem work on motorola potter

Open telent opened this issue 2 years ago • 3 comments

This is not mergeable or even useful yet, this is me developing in public

telent avatar Jan 13 '24 21:01 telent

something like this also needs doing (populate /tmp/tqftpserv with handset-specific state)

https://gitlab.com/postmarketOS/pmaports/-/blob/master/device/testing/firmware-motorola-potter/moto-modem-rfs-setup.openrc

telent avatar Jan 15 '24 14:01 telent

this is proceeding steadily in the direction of "worse not better" ...

  1. GPS (or to use the more accurate term, "GNSS") on this SoC is provided by the modem coprocessor (this makes sense inasmuch as it's a radio receiver).

  2. According to https://wiki.postmarketos.org/wiki/Qualcomm_Snapdragon_450/625/626/632_(MSM8953)#Modem , "Motorola devices need IPA support for the modem to work". IPA on this SoC isn't supported by CONFIG_QCOM_IPA because that only understands SDM845, instead there's a WIP kernel branch of 6.x that provides CONFIG_QCOM_IPA_LEGACY

  3. However, upgrading to the 6.5.2/ipa branch has not only made wifi stop working, but also the screen. In 5.16 we saw a small centered graphic with scrolling text under it, then the screen cleared (at around the time control is transferred to stage-2) and further log output filled the whole screen. With this kernel, once the screen clears, the phone resets back to the bootloader

  4. with boot.stage-1.gui.enable = false then the screen clears but the device does not reboot, and is available via adb shell. Nothing on the screen.

4a) fwiw recovery.img crashes in the same way and boot.stage-1.gui.enable makes no difference there, so the problem is probably not specific to stage-2

  1. postmarketos aren't using this kernel branch either, fwiw. https://app.element.io/#/room/#mainline:postmarketos.org/$rCNsBmFKG_KO4GZslX0WSPfUj-ELx5SYymfYSit28OE

  2. I flashed the postmarketos prebuilt binaries for my device, and, haha, that has the same screen-doesn't-turn-on problem

telent avatar Jan 15 '24 16:01 telent

After a few more runs, the screen blanking seems to happen at varying points - sometimes we get all the way to login.

It appears to be contemporaneous with the line "ibb: disabled" being logged, which refers to the labibb regulator which is one of the regulators that powers the panel.

telent avatar Jan 17 '24 17:01 telent