diff --git a/device/dipper/NOTES.md b/device/dipper/NOTES.md new file mode 100644 index 000000000..3a0ec67dd --- /dev/null +++ b/device/dipper/NOTES.md @@ -0,0 +1,309 @@ + +# dipper (Xiaomi Mi 8) — bring-up notes + +Facts only, each read off the phone. Captures remain outside the repository +because they contain device identifiers and Wi-Fi keys; this file records only +the facts needed for review. + +## RAM sessions (newest first) + +| Date | `boot.img` SHA-256 | Source, as recorded | `/dev/sda` and UFS partitions | Checks and gaps | +| --- | --- | --- | --- | --- | +| 2026-11-10 | `540b5d31ea73a31ade9a8d8cb5152cf701dc3c2ad8256734790b0ef2db6dbdd0` | Combined #503/#505; kernel/DTS #220 `055a4019998f22cc0e2a8afef7650274a6ce0657` | Node absent, disk `ro=0`; all 85 partitions `ro=1`; `sdb`–`sdf` `ro=1` | Watchdog/feeder, simpledrm, USB/SSH, reserved TZ range, pinned host key and `reboot-to bootloader` passed; physical keys not captured. | +| 2026-11-10 | `a2d435a7430e7e4c145c61cdb9866d16929731859d91c7105e6bfcaf9e3e9dd7` | Rebased #503 NFS image with #220 `055a4019998f22cc0e2a8afef7650274a6ce0657` | `/dev/sda` present, set `ro=1`; six disks and all 85 partitions read back `ro=1` | `uname`, NFS root, multi-user systemd, watchdog and SSH passed; generic Arch test root only; link-loss recovery and Omarchy desktop untested. | +| 2026-11-10 | `0ee4f11d1973085c434e741a2e9dd73af5a9cd157296c25801ed5a92a6b9e9d9` | #505 `994944d7ecbacd69444221e6951153cb329a374e`; kernel/DTS #220 `055a4019998f22cc0e2a8afef7650274a6ce0657` | Node absent, disk `ro=0`; all 85 partitions `ro=1`; `sdb`–`sdf` `ro=1` | Watchdog/feeder, framebuffer, USB/SSH, reserved TZ range and `reboot-to` passed; physical keys not captured. | +| 2026-11-10 | `e2cdb765de21a5869a1db314c19f1fddb668219a51d3c9ce81a1a083536d711e` | Older rescue image; exact source commit unknown; its init differs from the shared init | Node present, disk `ro=0`; all 85 partition flags recorded `ro=1`, including `sda21` | Kernel `7.3.0-rc6-omamobile-push1`, initramfs root; no `logdump` or whole-disk removal in init, so not the October diagnostic init. SSH host key verified but key was not authorized. No UFS write. | +| 2026-11-09 | `e2cdb765de21a5869a1db314c19f1fddb668219a51d3c9ce81a1a083536d711e` | Image was reported booted; exact build/init source unknown | `/dev/sda` reported present with `ro=0`; partition details for this session unknown | The retained 08:53 capture is a dirty-kernel writable-root observation; whether it came from this exact image is unknown. | +| 2026-11-09 | `0f0ec9d62ebe57bef9ae8a4804bb23728f624daab824a0bf7dee18f63c2a3d59` | `linux-omarchy-mobile` panel test tree with EA8074 from #218; exact combined revision unknown | `/dev/sda` absent; all 85 partitions `ro=1`; `sdb`–`sdf` present, `ro=1` | Panel/DRM, backlight 0→511, DPMS off/on passed; `vdda` used a dummy regulator, GPU rendering and touch untested. | +| 2026-11-09 | Same boot as panel test, `0f0ec9d6…` | Same as panel test | Same as panel test | Full backup of 84 non-`userdata`, non-`super` GPT partitions was read from this guarded boot and every source hash/size verified; separate userdata archive was file-level and non-atomic. | +| 2026-11-09 | Unknown; exact image hash not retained | Omarchy Mobile RAM kernel; exact build unknown | Disk and partition flags not recorded | Power-only 10-second restart passed from Android and RAM kernel; no UFS write was attempted. | +| 2026-11-09 | `2244d74aea472b0c595b2056ad8618f614e6a9161b5b264c7f898939e626f318` | Initial NFS-root image reused the earlier #505 kernel and DTB | `/dev/sda` present and set `ro=1`; all six disks and 85 partitions logged `ro=1` | NFS root and SSH worked; systemd reported `RuntimeWatchdogUSec=20s`. Direct `uname`/sysfs reads were not captured. | +| 2026-11-09 | `f2f5cb2e3da8fe992d366dba36e059c0d9106cf14a7d1c8872c61ee144a4474b` | Earlier #505 image after `qcom-pon` was built in; exact kernel/DTS heads unknown | `/dev/sda` absent, disk `ro=0`; all 85 partitions and other disks `ro=1` | PMIC keys registered and Volume Down produced `KEY_VOLUMEDOWN`; watchdog/feeder, framebuffer and USB/SSH worked. | +| 2026-11-09 | `93484a91103c294f6c73799ae6cb7161019129b9f89ab331222ae04f10ecf3e1` | First #505 image before `qcom-pon` was built in; exact kernel/DTS heads unknown | `/dev/sda` absent, disk `ro=0`; all 85 partitions and other disks `ro=1` | Console, framebuffer, USB/SSH and watchdog/feeder worked; no physical key event was captured. | + +The original 10-29 count of 92 `/dev/block/by-name/` entries was not saved as a +name list. The retained backup and kernel records cover 84 non-userdata GPT +partitions plus `userdata`; the other seven names cannot be identified from the +retained artifacts and remain unknown. + +## 2026-11-10 — final PR #505 RAM image + +Booted image SHA-256 `0ee4f11d1973085c434e741a2e9dd73af5a9cd157296c25801ed5a92a6b9e9d9` with `fastboot boot`; it was built from PR #505 head `994944d7ecbacd69444221e6951153cb329a374e` and kernel/DTS PR #220 head `055a4019998f22cc0e2a8afef7650274a6ce0657` (touch PR #219 is merged). `uname -r` was `7.3.0-rc6-omamobile-push1`; `/` was the initramfs (`rootfs`). The guard reported the UFS devices read-only. Sysfs showed six disks: `sda` had `ro=0` but its whole-disk `/dev/sda` node was absent; `sdb`–`sdf` each had `ro=1` and present nodes. All 85 partitions had `ro=1`. This matches PR #505's intended missing-node case for `sda`; no UFS write was attempted. + +`qcom_wdt` loaded and `/usr/bin/watchdog -t 5 -T 30 /dev/watchdog0` was running. simpledrm registered `fb0`; USB gadget and authenticated SSH worked. `/proc/iomem` marked `0x85fc0000–0x8a8fffff` reserved, covering the former TZ range. The final image's physical button events were not captured, so the earlier Volume Down/Up observations are not evidence for this kernel. The image's `reboot-to bootloader` helper successfully returned the phone to fastboot; fastboot reported product `dipper` and unlocked `yes`. The only boot operation was `fastboot boot`; no UFS partition was written. + + +## 2026-11-10 — rebased #503 and combined #505 RAM runs + +The rebased PR #503 init was RAM-booted with PR #505's Dipper builder and +kernel/DTS PR #220 at `055a4019998f22cc0e2a8afef7650274a6ce0657`. NFS image +SHA-256: `a2d435a7430e7e4c145c61cdb9866d16929731859d91c7105e6bfcaf9e3e9dd7`; +`uname -r` was `7.3.0-rc6-omamobile-push1`, and `/` was NFS. The initramfs +used BusyBox `mount`; the export was a generic Arch Linux ARM test root, not an +Omarchy rootfs. Direct sysfs reads returned `ro=1` for all six UFS disks and +all 85 partitions. The guard logged each read-back, including `/dev/sda` set +read-only for NFS. Systemd reached the multi-user target with +`RuntimeWatchdogUSec=1min`; SSH remained available. `reboot-to bootloader` +returned the phone to fastboot. No UFS write was attempted. The NFS link was +not deliberately dropped, so hard-mount link-loss recovery remains untested. + +The combined #503/#505 non-NFS RAM image was also booted. SHA-256: +`540b5d31ea73a31ade9a8d8cb5152cf701dc3c2ad8256734790b0ef2db6dbdd0`. +`uname -r` was `7.3.0-rc6-omamobile-push1`; `/` was the initramfs (`rootfs`). +Direct sysfs reads returned `ro=1` for all 85 partitions and disks `sdb`–`sdf`; +`sda` remained `ro=0`, while the guard removed `/dev/sda`. The `qcom_wdt` +module and feeder ran, simpledrm registered `fb0`, USB/SSH worked, and the +former TZ range was reserved in `/proc/iomem`. The rescue SSH host key matched +the build output, and `reboot-to` returned the phone to fastboot. No UFS write +was attempted. Physical input events were not captured on this exact image. + +## 2026-11-09/10 — recovery and historical RAM observations + +The current `recovery` partition matches the `recovery.img` in the +checksum-verified 15-partition backup of the installed system captured on +2026-10-29 at 19:01; +both have SHA-256 +`7691a33e8b65d1ffcbf67b16fd4dc5c4c421420230f230db312512c73abc2221`. Its +Android boot header contains a gzip-compressed kernel that identifies as +`Linux version 4.9.337-perf-g0b3e38ef6c40`, not the 7.3 diagnostic kernel. This +confirms the current partition holds the installed ROM's recovery image. The order of the +snapshot and the two diagnostic writes remains unknown. `logdump` is not in the +15-partition backup. A separate live, file-level userdata archive was +verified off-device; it is not a raw partition image or an atomic snapshot, +and the filesystem remained mounted during the copy. + +The recovery image hash is recorded from the backup, but the backup's `boot.img` +was not present among the accessible artifacts for this review. Its kernel +string and whether it matches the recovery build remain unknown. + +Before entering fastboot, the handset answered over USB SSH; no restart was +needed to restore SSH. Holding Power alone for 10 seconds restarted the phone +back into Android both from the installed system and under the Omarchy Mobile +RAM kernel. Volume Down + Power then entered fastboot. Read-only fastboot +variables reported product `dipper` and an unlocked bootloader. + +A separate NFS-root attempt used a RAM-only image with SHA-256 +`2244d74aea472b0c595b2056ad8618f614e6a9161b5b264c7f898939e626f318`, reusing +the earlier #505 kernel and DTB and repacking its initramfs with +`OMAMOBILE_NFSROOT`. The persistent NFS-root journal recorded NFS mode and +read-only confirmations for all 85 unique UFS partitions and all six UFS +disks; the guard reads each `ro` value back as `1` before logging success. SSH +confirmed `/` was NFS, PID 1 was systemd, and the system was running with +`RuntimeWatchdogUSec=20s`. The exported root was a generic Arch Linux ARM base, +not an Omarchy rootfs. No UFS write was attempted. Direct `uname -r` and +sysfs `ro` output were not captured, and this pre-#484 image does not verify +the post-#484 whole-disk-node absence check. + +An earlier read-only SSH observation, reported on 2026-11-09 at 08:53 UTC, +was on the dirty kernel +`7.3.0-rc6-omamobile-push1-g2a318a96ae9d-dirty`. It reported `/dev/sda21` +mounted as the root filesystem read-write, with `/sys/block/sda/ro=0` and +`/sys/block/sda/sda21/ro=0`; the other partition `ro` values were `1`. Thus +`/dev/sda` existed and the whole main LUN was writable. The surviving record +does not say how that image was started, the device uptime at the check, or +whether its diagnostic init's `logdump` mirror was running. Do not boot that +old image again. Systemd reported failure to set the `qcom_wdt` timeout to one +minute, but the exact journal line was not retained and the feeder was not +checked. This old-kernel observation is not evidence for PR #505's image. + +The image with SHA-256 +`e2cdb765de21a5869a1db314c19f1fddb668219a51d3c9ce81a1a083536d711e` was +previously described here as PR #505's RAM image. On 2026-11-10, I booted that +exact file again with `fastboot boot` and read it over the USB serial shell. +It reported `7.3.0-rc6-omamobile-push1`, but `/dev/sda` existed with +`/sys/block/sda/ro=0`; `/sys/block/sda/sda21/ro=1`. Its initramfs root was +`rootfs`; no UFS write was attempted. Offline extraction found init SHA-256 +`ab6db476f5cd26cb84a8346e18da5b2137530091be504c47a23b0a21934277c5`, different +from the checked-in shared init. It contains neither the October `logdump` +mirror nor whole-disk-node removal, so it is not the October diagnostic init +and does not verify PR #505's guard. Do not treat it as #505's image or boot it +again. SSH host-key verification succeeded, but the configured key was not +authorized. The separate rescue image still provides authenticated SSH +and `reboot-to bootloader` returned the phone to fastboot without writing UFS. + +`pm8941_pwrkey` and `pm8941_resin` registered. On the earlier final image, a +physical Volume Down press produced `KEY_VOLUMEDOWN` (code 114, value 1), and +Volume Up produced `KEY_VOLUMEUP` (code 115, value 1). The current kernel +image's button events have not yet been rechecked. The owner reported seeing +the console during the first boot. + +An earlier test record attributed removal of `/dev/sda` to the PR #505 image, +but the recorded image hash is the `e2cdb765…` artifact checked above, where +`/dev/sda` was present and writable. The earlier whole-disk-removal result +cannot be tied to this artifact. The final #505 image and its successful +disk-node check are recorded below. The `e2cdb765…` boot did report all 85 UFS +partition flags read-only; no UFS writes were made. +`/proc/iomem` reports `0x86200000–0x88efffff` as System RAM and +`0x88f00000–0x890fffff` as reserved, so the TZ range is not reserved in this +RAM boot's map. A separate EA8074 panel test used a RAM-only overlay on +2026-11-09 with kernel `7.3.0-rc6-omamobile-push1` and boot image SHA-256 +`0f0ec9d62ebe57bef9ae8a4804bb23728f624daab824a0bf7dee18f63c2a3d59`. The panel +driver bound to `ae94000.dsi.0`, DRM reported `connected` at `1080x2248`, and +the owner confirmed the console was visible on the lit panel. All 85 partition +flags were `ro=1`; `/dev/sda` was absent, while `/dev/sdb` through `/dev/sdf` +remained present and had `ro=1` on this pre-#484 image. Backlight value 0 was +accepted and restored to 511 (maximum 1023). Framebuffer blanking changed +DRM `dpms` from `On` to `Off` and back to `On`. The FTS5 touch input still needs +a test on the current kernel. The panel log included a missing DSI `vdda` +supply (dummy regulator) and missing `a630_sqe.fw`; the display initialized, +but GPU rendering was not tested. The RAM image had no Arch root, and the probe +did not write UFS. The installed-system backup checksums were verified before +boot. +After this session, a read-only backup of all 84 non-`userdata`, non-`super` GPT +partitions was captured from this guarded RAM boot. Each partition's source +SHA-256 and byte count were checked against the decompressed local copy, and +the compressed archive checksums were rechecked. The separate live userdata +archive remains file-level, non-atomic, and is not a raw partition backup. +The backup is stored outside the repository; it must not be published. After +the last Android boot and before any future flash, the on-device partition +checksums must be taken again and compared with the off-device backup because +Android can rewrite modem partitions. A reviewed button-by-button restore +procedure and supported single-slot writer are still required before a flash. + +## 2026-10-29 — installed-system facts, read-only over SSH with the owner's Magisk root + +- **SoC** SDM845. Live DT root: `model = "Xiaomi Technologies, Inc. Dipper new MP v2.1"`, + `compatible = "qcom,sdm845-mtp", "qcom,sdm845", "qcom,mtp"`, + `qcom,msm-id = <0x141 0x20001>`, `qcom,board-id = <0x37 0x00>`. + 6 GB RAM. `androidboot.cert=M1803E1A` (Mi 8 Global), `hwversion=3.13.0`. +- **The installed system** is a custom Android 14 (SDK 34) ROM with Magisk 30.7, not MIUI; + its build fingerprint is a spoofed MIUI 8.1 string, and `ro.boot.verifiedbootstate=green` + with `flash.locked=1` contradict the owner's unlocked bootloader. **No boot-state claim + may be made from Android on this phone; read it in fastboot.** Running kernel 4.9.337. +- **Not A/B for the OS.** 92 entries under `/dev/block/by-name/`: a single `boot` (64 MiB), + `recovery` (64 MiB), `dtbo` (8 MiB), `system`, `vendor`, `cache` and `userdata` + (113 GiB). Only the bootloader chain is slotted: `xbl`, `xbl_config`, `abl`, `tz`, `hyp`, + `aop`, `cmnlib`, `cmnlib64`, `devcfg`, `keymaster`, `qupfw`, `storsec`. + **There is therefore no known-good OS slot to fall back to**, which the shared Pixel + safety model assumes. A full backup of the installed system is a precondition, not a nicety. +- **Boot image** uses Android header v1, page size 4096, `second_size = 0`, and the kernel + blob ends with **two concatenated FDTs of ~480 KB each** (`d00dfeed`, version 17): the + base DTB is appended to the kernel and ABL selects one by msm-id/board-id. +- **Display** Samsung **EA8074**, 1080x2248, command mode, 60 Hz, notch; physical + 68 x 142 mm. Confirmed twice: `msm_drm.dsi_display0=dsi_ss_ea8074_notch_fhd_cmd_display` + on the kernel command line, and the live DT node `qcom,mdss_dsi_ss_notch_fhd_ea8074_cmd` + (`qcom,mdss-dsi-panel-width = <0x438>`, `height = <0x8c8>`, `framerate = <0x3c>`). +- **Touch is STMicroelectronics, not FocalTech.** Live DT, on `i2c@a98000` (mainline + `i2c14`): `fts@49 { compatible = "st,fts"; fts,irq-gpio = <&tlmm 31 ...>; + fts,reset-gpio = <&tlmm 32 ...>; fts,default-fw-name = "st_fts_e1.ftb"; }`, with + `fts,tp-vendor = <0x48>`. The running kernel's input device is `N: Name="fts"`. + An `atmel,mxt-ts@4a` node exists as the alternate vendor and is not the active one. + **Open:** mainline's ST driver is `st,stmfts` (`drivers/input/touchscreen/stmfts.c`), + a different driver from Xiaomi's downstream `st_fts`; the current FTS5 driver + has not been tested with Dipper's DTS node. No sibling board helps — + polaris uses `syna,rmi4-i2c@20`, beryllium a FocalTech/Novatek part. +- **Audio** `sdm845-tavil-snd-card` (WCD9340). **Fingerprint** Goodix, capacitive + (`fingerprint_goodix`, `uinput-goodix`); no mainline driver, out of scope. + **Buttons** `qpnp_pon` plus `gpio-keys`. +- **UFS** host at `1d84000.ufshc`. The installed Android `system` is dm-verity over the + `system` partition. + +## 2026-10-29 — first light, and what the hardware said + +**The phone reached a shell on an unpublished kernel tree.** `fastboot boot` of +the image used kernel release +`7.3.0-rc6-omamobile-push1-g2a318a96ae9d-dirty` with its matching DTB: UFS up +(all six LUNs, ext4 mounts), USB NCM gadget, dropbear, gpio-keys, 8 CPUs and +5.7 GB RAM. This records the historical first-light tree; kernel PRs [#218](https://github.com/derekross/linux-omarchy-mobile/pull/218), +[#219](https://github.com/derekross/linux-omarchy-mobile/pull/219) and +[#220](https://github.com/derekross/linux-omarchy-mobile/pull/220) are later +successor changes and are not the code tested in that boot; FTS5 PR #219 is now +merged into linux-omarchy-mobile. The diagnostic image's guard allowed `userdata` and `logdump` +writes. A temporary init widened the allowlist to include `logdump` and used +`dd` to copy the kernel log there every two seconds during the 2026-10-29 +diagnostic boot. The 2026-11-09 read-only check was on this dirty October +kernel; it found `/dev/sda21` as the read-write root and therefore `/dev/sda` +present and writable. The saved record does not establish whether the mirror +process was still running during that later check, nor the device uptime then. +Two diagnostic-image writes to the `recovery` partition from rooted Android were also +made on 2026-10-29. The existing 15-partition backup did not include `logdump`; +an unverified `logdump.img` file is not evidence of a checked phone backup. +The current recovery SHA matches the verified 2026-10-29 19:01 backup +described above; the order of that backup and the two writes is unknown. The +image previously attributed to PR #505 was `e2cdb765…`; its `/dev/sda` node was +present and writable, so it does not verify #505's guard. The final #505 image +and its successful check are recorded above. The mount options and whether +`persist` was mounted during first light are not recorded here. + +The recorded 15-partition backup of the installed system before the radio runs +was checksum-verified. + +The 2026-10-30 comparison found `boot`, `recovery`, `dtbo`, `vbmeta`, +`fsc`, `devinfo`, `misc`, `logo`, `splash`, `persistbak`, `modemst2` +and `fsg` byte-identical; only `persist`, `modemst1` and `frp` changed. +All five NV partitions were read-only before and after the recorded runs. The +cause of the three differences is not established. Other recorded activity: +NFS-root Omarchy sessions on 2026-10-30, 2026-10-31 and 2026-11-02; modem +starts under `rmtfs -r`, followed by an XPU violation crash and remoteproc +recovery loops. The sequence of boots between the two backups (date, installed +Android or RAM image, and actions) has not been reconstructed, so the later +backup is not a baseline and the changes are not attributed to Android or +Omarchy. The installed ROM's current recovery image matches the 10-29 19:01 +backup and its kernel identifies as `4.9.337-perf-g0b3e38ef6c40`; the backup's order +relative to the two diagnostic writes remains unknown. The 10-29 first-light +boot is recorded as `fastboot boot` from RAM, not a recovery boot. The separate +08:53 old-kernel observation's launch method remains unknown; the surviving +record does not establish its source. + +### What had to be fixed before any of that +Six defects in the recovered [sdm845-mainline MR !71](https://gitlab.com/sdm845-mainline/linux/-/merge_requests/71) +work, three of them fatal: + +1. **`qcom,board-id = <69 0>` (0x45) is polaris's.** dipper reports `0x37`. ABL matches on + it and silently refuses the whole boot image otherwise — no kernel output, + no panic, just a reboot. Two blind diagnostic-image attempts were written to + the `recovery` partition from rooted Android before this mismatch was + corrected. +2. `drm_panel_init()` no longer exists; the panel probe needed the refcounted + `devm_drm_panel_alloc()`. +3. `panel-samsung-ea8074.ko` had no `MODULE_LICENSE()`, so modpost refused to link. +4. The touch node named `focaltech,ea8074` at `0x38`; the chip is ST at `0x49`. +5. Its Kconfig entry was copied from an SPI panel (`depends on SPI`, `select DRM_MIPI_DBI`) + for what is a `mipi_dsi_driver`. +6. `ts_int_default` carried `bias-pull-down` from beryllium; dipper's vendor tree says + `ts_active { pins = "gpio31", "gpio32"; drive-strength = <16>; bias-pull-up; }`. + +### And two kernel-config lessons, both found on the hardware +Devices probe once, early. A driver that only appears later as a module never gets a +second chance — the device is left unbound and is not even on the deferred list. Proven +twice: the Qualcomm PHYs (so no USB and no storage: `dwc3: failed to initialize core`, +`sync_state() pending due to 88e8000.phy`), then i2c (`modprobe i2c-qcom-geni` bound to +nothing, and a hand-written `bind` failed). That is why every driver needed for first +light was built in (`=y`). + +### Touch: first failed probe (historical) + +The original FTS4-only driver bound and timed out. These observations describe that +kernel revision; they are not results from the current FTS5 driver: + +- `i2c-14` and `i2c-5` come up; the touch device **exists as `14-0049`**, so the bus and + address in the DTS are right. +- Both supplies are enabled and in use: `ldo28` at 2.856 V, `vreg_tp_vddio` at 1.8 V, + one user each. +- `stmfts` binds and fails with **`-110`** from `wait_for_completion_timeout` — it sends + its first command and never sees the reply interrupt. +- The interrupt pin bias was corrected to pull-up per the vendor tree. **No change.** +- `i2cdetect` output is incoherent: in `-r` mode every even address reads as present and + every register as `0x00`; in quick-write mode `0x49` is not detected at all, while + `stmfts`'s own `i2c_smbus_write_byte` to `0x49` returns success. +- On i2c5 the driver logs **`geni_i2c_gpi_xfer:I2C timeout` / `GPI transfer failed: -110`**, + so the GPI DMA path is not reliable here. +- GPI DMA cannot simply be turned off: this kernel's `geni_i2c` requires the channels and + defers forever without them (`geni_i2c: Failed to get tx DMA ch`). There is no FIFO + fallback to fall back to. + +At the time, it was unknown whether GENI/GPI or the FTS5 protocol caused the timeout. + +### FTS5: historical tap and published driver series + +The live controller is FTS5. A separate historical RAM boot on 2026-10-30, using +the unpublished kernel release above, registered `stmfts` and a physical tap +produced `ABS_MT_POSITION_X`. That observation is not a test of the current +kernel PRs. The public successor series and its binding are in +[kernel PR #219](https://github.com/derekross/linux-omarchy-mobile/pull/219). +The upstream v8 cover letter states it was tested on Xiaomi Mi 8: +[FTS5 v8 cover letter](https://lists.openwall.net/linux-kernel/2026/09/25/2211), +[FTS5 driver patch](https://lists.openwall.net/linux-kernel/2026/09/25/2213), +[controller-ready fix](https://lists.openwall.net/linux-kernel/2026/09/25/2214), +[binding update](https://lists.openwall.net/linux-kernel/2026/09/25/2212). +Firmware loading, low-power touch and wake-on-touch/AP-SLPI handoff are not implemented.