Skip to content

[Device support request] OnePlus 8 Pro (instantnoodlep, SM8250 / Snapdragon 865) #808

Open

Description

@inumineq

Device

Model OnePlus 8 Pro, EU variant IN2023 (siblings: IN2020 CN, IN2021 IN, IN2025 NA)
Codename instantnoodlep
SoC Qualcomm SM8250 (Snapdragon 865), Adreno 650
RAM / storage 8/12 GB LPDDR5, 128/256 GB UFS 3.0
Display 6.78" 1440x3168, 120 Hz AMOLED (Samsung AMS678UW01 on my EU unit; other panel variants exist)
Bootloader Unlockable out of the box, fastboot oem unlock, no token or waiting period; unlock wipes data

I own this phone (EU, IN2023) and can have it out of action for a while.

Why a port looks unusually plausible

SM8250 precedent in this project. The POCO F3 (alioth, SM8250 / SD870) was proposed as the first non-Pixel port in #25, paused by its author and closed on 2026-11-11, so the SoC is not new ground for this tree.

Most of the mainline kernel work already exists.

  • Upstream Linux carries the SM8250 OnePlus family: the bindings, a common devicetree, panel and Synaptics TCM touchscreen drivers, merged with the initial OnePlus 8T support series (https://lwn.net/Articles/979482/).
  • The community SM8250 mainline fork used by postmarketOS/Nura (https://gitlab.postmarketos.org/soc/qualcomm-sm8250/linux) carries sm8250-oneplus-instantnoodlep.dts and the fixes the phone needs, for example the Wi-Fi/BT DTS regression introduced around v6.11.
  • Nura (postmarketOS under its new name since 2026-09-27) maintains a device package for oneplus-instantnoodlep, and an unofficial Nura image (https://github.com/naorunaoru/oneplus-instantnoodlep-nura-unofficial, releases October and November 2026, kernel 7.2) demonstrates on an IN2023 EU unit: display at 1440x3168 and 120 Hz, GPU, speakers and microphones, Wi-Fi, Bluetooth, modem/SIM/mobile data/SMS, GPS, NFC, sensors, and all five cameras through libcamera.
  • Firmware has a precedent too: Nura packages firmware-oneplus-instantnoodlep (ath11k and modem firmware), and per CONTRIBUTING section 7 the per-driver firmware could instead come from the owner's own factory image.

A/B slots look compatible with the safety model. The phone has A/B boot slots (boot_a/b, dtbo_a/b, vbmeta_a/b, ...) with Android's dynamic partitions inside a shared super; userdata is a physical, non-slotted partition. A slot-B install (the phone's boot images on slot B, the Arch root on userdata, super untouched) looks compatible with keeping stock on slot A. To be checked in stage 0: the exact GPT/LUN layout, whether the slot flags live in GPT attributes or devinfo, and what the storage guard's allowlist would need here.

Stage-0 facts (read-only, to follow in this issue)

Following the process from #220, I will collect from my unit and post here:

  • Unlock status: "OEM unlocking" in developer options, the fastboot oem unlock result, and fastboot getvar all including both slots' bootloader version.
  • The partition list: ls -l /dev/block/by-name/ (both slots, plus the LUN layout).
  • A stock dmesg and the device tree, from a temporary-root session (fastboot boot of a patched image, nothing flashed).
  • How stock is restored: full-OTA payload images via fastbootd as the normal path, and the Qualcomm EDL (MSM Download Tool) service packages that circulate for IN2023 as the last resort, with the usual goal of never needing it.

Caveats I already know

  • My EU unit has the Samsung AMS678UW01 panel the Nura port was tested with; OnePlus shipped other panel variants, which may need separate panel/DT work.
  • Voice calls (VoLTE) on mainline are not yet verified working on this device; SMS and mobile data are.
  • I am new to this project's tooling but comfortable with Linux, fastboot and kernel work, and I will follow the bring-up order in CONTRIBUTING and the [V]/[I] evidence discipline of the plans.

If these facts look right, would you add the phone to the Wanted list in DEVICES.md? Happy to be the hardware-and-testing half if someone more experienced wants to lead the port.

Activity

  1. inumineq commented on Nov 16, 2026

    @inumineq
    Author

    Status from the instantnoodlep port workspace (OnePlus 8 Pro, IN2023 EU, stock IN2023_13.1.0.591(EX01)), plus the questions we need answered before any slot-B write.

    Where we are

    • Stage 0 is done. The bootloader is unlocked. Stock runs from slot A, and slot B holds the same .591. A read-only SHA-256 comparison (taken with temporary root) shows all 19 slotted bootloader partitions (abl, xbl, xbl_config, tz, hyp, aop, devcfg, keymaster, modem, ...) identical across _a/_b. Radio/calibration partitions are backed up off-repo.

    • M1a, RAM-only first light, passes with two kernels. Both runs used fastboot boot with the shared initramfs/init and the storage guard and no allowlist:

      • K1: the community SM8250 7.2 tree.
      • K2: linux-omarchy-mobile omamobile (317d87dc2db0) plus our four DTS commits (7.3.0-rc6).

      Results: USB NCM and SSH come up, all 8 CPUs, the watchdog is fed, and every UFS partition is ro=1 with /dev/sda removed. Stock comes back on slot A with unchanged slot flags.

    • Two ABL findings other ports may hit:

      1. OnePlus's ABL picks the base DTB by oplus,dtsi_no and then applies its dtbo entry. It refuses a RAM boot it cannot overlay, and does so silently: Booting OKAY, then it stays in fastboot.
        • What boots: our DTB plus oplus,dtsi_no/oplus,pcb_range copied from stock, plus a hand-written __symbols__ that points every label the stock overlay needs at disabled nodes under a root node with no compatible.
        • The overlay applies, and no real node is touched (zero diff outside the sink).
      2. A second fastboot boot in the same fastboot session, after a refused one, failed AVB. ABL then ran HandleActiveSlotUnbootable: it marked slot A unbootable, set B active and cold-reset (generic Qualcomm ABL VerifiedBoot.c does the same). We recovered with set_active a. Our rule now: fastboot reboot bootloader between RAM boots.
    • AVB: stock vbmeta has Flags 0. boot (location 4) and dtbo (location 3) are chain partitions with their own footers.

    Numbers for a GPT-attribute entry (from read-only stage-0 dumps)

    LUN sde (UFS LUN 4): 4096-byte sectors, last LBA 655359, 96 entries x 128 bytes
    headers/entry CRCs valid, primary and backup arrays identical
    boot_a entry 10, boot_b entry 34; bit 54 (successful) = byte 310 of the entry sector, mask 0x40
    SLOT_B_MARK="sde boot_b 1 3 655356 655359"
    

    userdata is sda23 (LUN 0); boot_b/dtbo_b are on sde, so the descriptor would need UFS_RW_LUNS=sde. Generic ABL SetActiveSlot(..., TRUE) clears the successful bit on the slot it activates, so a mark is still needed after set_active b.

    Questions

    1. Process. Same path as enchilada? We supply the numbers and a gated plan, you write the instantnoodlep §3.1 entry, and the device lines (UFS_RW_LUNS, SLOT_B_MARK) follow in a PR. Does instantnoodlep reuse tools/mark-slot-b (tools/mark-slot-b and push over UFS_RW_LUNS, inert until the §3.1 entry (part 1 of #616) #651) unchanged?
    2. Rule shape. Is "LUN + boot_b + P_HDR, P_ENT, B_ENT, B_HDR" (the SLOT_B_MARK form above) what the entry should state?
    3. --set-active=b. On this ABL it also swaps the _a/_b type GUIDs and moves the UFS boot LUN. Should the plan record a before/after GPT of every LUN and bBootLunEn around the first --set-active=b, and should the entry name that write too?
    4. AVB on boot_b. Given the chain descriptors, do you accept upstream's test-key add_hash_footer on boot_b (partition size 100663296, rollback 0, os_version 12, security_patch 2024-06-05)? For dtbo_b: a footed image of our own, or leave stock dtbo_b and rely on the overlay sink above?
    5. Way back. Slot A holds stock, and the owner has the EU MSM package plus a native Windows install for EDL. Is that enough, or do you want written EDL acceptance as on OnePlus 6 (enchilada, SDM845) port: RAM-only first light, contributors welcome #432?
    6. No-USB panic vs retry count. An unmarked B spends a try per boot. Should a boot with no laptop be a gate before the mark, as you asked for enchilada?
    7. Interim. Until the entry exists we plan to run M2 tethered: Arch root on userdata, booted each time by fastboot boot, with UFS_RW = userdata only and no slot-B or GPT write. Any objection?
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions