##Not a bootlog! Some extra notes! Board - Main board silkscreen: "AK3919EV300 SC2232H V401", dated 2019.09.29; sensor on the same PCB. - SoC: Anyka AK3918EV300, chip id , chip marking "AK3918EN064 V300-J". Kernel machine string Cloud39EV3_AK3918E80PIN_MNBD; the platform self-check prints "AK3919Ev300(V310)" (why the silkscreen says AK3919 perhaps?); ipctool reports ak3918ev300. - Clocks: CPU 400 / MEM 400 / ASIC 120 MHz. - Fixed dome: the AK Motor Driver loads (vendor platform supports PTZ) but no motors fitted. - Memory layout: stock Linux loads at 0x81e08000 (runs in the top ~34 MiB bank, media pool below). OpenIPC's EV300 layout uses kernel 0x80008000 + media 0x81400000, and the stock 1.5 MiB KERNEL slot (0x80000-0x200000) won't hold the larger OpenIPC uImage. Sensor - SmartSens SC2232H, 1920x1080, MIPI 2-lane. These are read straight from the vendor sensor driver in the dump — sensor_sh2232.ko (note the filename: it is the SC2232H driver; every symbol is sc2232h_*, so grepping for "sc2232h.ko" would miss it): I2C address : 0x30 (7-bit) / 0x60 8-bit write [immediate in sc2232h_sensor_read_register] MCLK : 24 MHz [sc2232h_get_mclk returns 0x18] Resolution : 1920 x 1080 [sc2232h_get_resolution -> 0x780 x 0x438] Chip-id probe: regs 0x3107 / 0x3108 / 0x3109 expected 0x22 / 0x38 / 0x20 [sc2232h_probe_id] - No live sensor ID was possible: ipctool and the stock aksensor driver read 0xffff for every sensor even with the full vendor app (anyka_ipc) running — with the app's power/clock sequence not up in a captured shell the SC2232H is unpowered. Model confirmed from the PCB silkscreen, the ISP tuning file, and the driver above — not from a live register read. - Stock SC2232H ISP tuning is in this backup at /usr/local/isp_sc2232h_mipi_2lane.conf (70924 B, binary, header "4.002_MIPI", in USR.SQUASH) — OpenIPC has no SC2232H tuning yet, so this is the raw material for one. The generic firmware also ships drivers/tunings for sc2145, sc2232, sc2235, sc4236, f35, f37 and gc2053. Other hardware - Flash: SPI-NOR GD25Q64 (JEDEC id 0x00c84017), 8 MB. - RAM: 64 MB DDR2 (ipctool media pool 30 M). - Wi-Fi: Realtek RTL8188FTV on USB (8188fu.ko + otg-hs); daughterboard marked "8188 FTV V112". - Ethernet: AK3918 GMAC (MAC base 0x20300000, per /proc/iomem "ak_ethernet.0") + Realtek RTL8201F PHY (id 0x001cc816), RMII — present and working (eth0 pulls a DHCP lease). The PE3020-W twin is Wi-Fi-only. - Audio: on-chip AD/DA codec (akpcm) + mic; speaker amp enabled via IOC_SET_SPEAKER ioctl. - aw9523b I2C GPIO/LED expander — confirmed live at I2C-0 address 0x58 (/sys/bus/i2c/devices/0-0058; controller "ak39-i2c" @ 0x20150000). - GPIO assignments — from the stock kernel's board GPIO table. This SoC has no Linux gpiolib (confirmed live: /sys/class/gpio and /sys/kernel/debug/gpio are both absent; it uses a custom Anyka GPIO framework), so there is no runtime GPIO map to read — the kernel table is the source. The table was read out of the dump: KERNEL partition -> zImage -> xz -> vmlinux, then the gpio-* label strings resolved to their numbers. It is a multi-board kernel carrying 4 variants selected by board-detect at boot (machine string is the single build Cloud39EV3_AK3918E80PIN_MNBD); values identical across all variants are certain, the rest are variant-dependent and our unit's variant is not determinable from the dump alone: IR LED GPIO 1 (gpio-ir) [same in every variant that defines it] Wi-Fi power GPIO 5 (gpio-wifi_power) [same in all 4 variants] TF/SD power GPIO 81 (gpio-tf_power) [same in all that define it; defined though no SD slot fitted] IR-cut filter GPIO 41 & 42 (gpio-ircut_a / gpio-ircut_b, dual-coil) — the pair is certain in every variant; which coil is a vs b flips between variants, so not asserted key LED GPIO 23 (gpio-keyled) in 3 of the 4 variants (GPIO 2 in the other) (Note: /sys/class/leds is empty with the camera app disabled, so no LED-class device is exposed live; the LED GPIOs above are from the kernel table, not a live sysfs read.) Partitions (SPI-NOR, 64K erase) - boot 0x000000-0x060000 (384K) - ENV 0x060000-0x070000 (64K) - MAC 0x070000-0x080000 (64K) blank -> random MAC assigned at boot - KERNEL 0x080000-0x200000 (1536K) Linux-3.4.35 uImage - ROOT 0x200000-0x300000 (1024K) squashfs, mounted / - USR.SQUASH 0x300000-0x780000 (4608K) squashfs, mounted /usr - USR.JFFS2 0x780000-0x800000 (512K) jffs2 rw, mounted /etc/jffs2 Console - UART: TX/RX/GND pads near the corner (see the annotated rear photo), 115200 8N1. - U-Boot 2013.10.0-AK_V3.0.06 (2019), console-unlocked (no password). Linux 3.4.35, uClibc 0.9.33.2. - The kernel serial console goes dead early: a board-init routine de-muxes the UART pin-mux (SYSCTRL reg 0x08000074, U-Boot value 0x0200040c). To get it back, flash a ROOT patch that writes 0x08000074 back to 0x0200040c in rcS and disables the camera app (it re-grabs that pin for MIPI). U-Boot's own console is fine. OpenIPC - This appears to be a distinct variant of the newly-adopted HeimVision PE3020-W (builder #238, same SoC marking AK3918EN064 V300-J): this unit is SC2232H + Ethernet, where the PE3020-W is JX-F37 + Wi-Fi-only. SC2232H has no OpenIPC tuning yet (the stock one is in this backup, path above), and EV300 still has no open video encoder/streamer (the encoder is an Allegro DVT core). - Of openipc-ai's builder #234 evidence list: flash dump (this backup) and the SC2232H tuning (in the backup) are provided, the GPIO map is above. - I can provide an OpenIPC-U-Boot boot log if a dome profile is wanted.