q 4a7224c41f B9: built-in display (ST7735S SPI) in Linux and persistent B9C cache

- B9A: read-only audit of stock LK/DT: the panel is an ST7735S 128x128 on SPI (BLSP1 QUP4,
  16 MHz, mode 3, D/C GPIO116, RESET GPIO118), backlight = PM8916 MPP4 current sink 40 mA.
- B9B: upstream panel-mipi-dbi with the stock init sequence as firmware; MPP4 sink via pinctrl
  + gpio-backlight; first light with correct colours/orientation/offsets.
- B9C: fbcon (6x8, 21x16) on tty1 + getty, UART console kept, aurora-display service;
  cache B9C b9ce13c9 written and verified, cold boots with LTE + Wi-Fi + display.
- B9L (RAM only): lk2nd second stage with an msm8916 SPI panel port shows a picture before Linux.
- Sanitizer: whitelist for b8/ and b9/.

2026-10-03 01:45:02 +03:00

2.4 KiB

Raw Blame History

B9C cache write — preflight (2026-10-03)

Candidate

cache-extlinux-b9c.img b9ce13c9… (134217728 B, ext2, label aurora-b9c). Built on 480s by mkfs-cache-b9c.sh = exact B8F recipe; only the kernel/fdt/initrd files, the append line, label and UUIDs differ. The rebuild is bit-identical and e2fsck -fn is clean. Contains the AP PSK (B8F initramfs) — never publish. Component diff: cache-diff-b8f-vs-b9c.txt.

Evidence before the write

RAM pretest aurora-b9c.img fa855e1c uses the same kernel/DTB/initramfs/cmdline as the candidate. Run: fastboot, ram1, class A (RTC 6 s).
- fbcon 21x16 (6x8) took over and the panel pipe came up by itself; aurora-display did no unblank, backlight on → DISPLAY READY at 8.0 s.
- getty on tty1 (pid 624); hostname aurora; /dev/console = ttyMSM0 (console active tty0 ttyMSM0); UART shell on ttyMSM0.
- START OK 69.2 s, AP ENABLED 73.5 s, MPSS/WCNSS running, display errors 0, eMMC w=0.
- Operator: text visible and readable, orientation correct, no clipping; backlight OFF→ON via sysfs confirmed.
- The login prompt scrolls away under later kernel messages (ignore_loglevel) — accepted (option A).
lk2nd emulator (emu/):
- S6 regression on the live B8F cache == the B8F emu run.
- S7 candidate: "Trying to boot 'b9c'", kernel 3afc6c77 (24686600 B, image_size 0x1820000, ends at DDR+0x1820000), DTB 7ca9cc79 @DDR+0x20A9000, ramdisk ac518e27 @DDR+0x22A9000 — no overlap.
- Fault variants F1/F3/F10 → android boot (emmc1 fallback).

Write / rollback

B9C-write.sh = B8F-write.sh with only the image, expected hashes, rollback references and log dir changed. Gates, in order:

loader;
candidate sha + size;
rollback image B8F 857c9c30 and the B8F readback backup;
9008 present; identity/geometry;
GPT == A3; boot == emmc1; aboot == lk1st-autoboot;
live cache == B8F 857c9c30 — the pre-write readback is saved as logs/b9/b9c/W/pre-cache.bin (backup).

Then: write cache only → readback cmp → verify every other partition → B9C_WRITE_VERIFIED. Rollback: B9C-rollback.sh → B8F cache (via HW EDL). Entry: HW EDL (battery on, D+→GND, plug USB, release).

After the write

3 normal class-A cold boots (no RESET, no fastboot, no UART input): power off ≥ 90 s → normal power-on → b9c-boot.sh K cold. In one of them, check backlight ON → OFF → ON visually via aurora-display bl (sysfs).

2.4 KiB Raw Blame History