Boot slots & recovery (since 2026-11-01)

The Kindle no longer runs Amazon’s OS. Stock U-Boot is unchanged (boot0, env untouched); only raw kernel slots and partitions changed. All slots share one layout: DTB at +0 (≤ 0xC8 sectors), uImage at +0xC8, ≤ 0x7000 sectors total.

Slot	Base sector	Holds	How it boots
normal	0x208 (DTB) / 0x2D0 (kernel)	mainline, normal variant	U-Boot autoboot (bootcmd → testboot)
diags	0x7208 / 0x72D0	test kernels	tools/stage.py --boot-only (or by hand, below)
rescue	0xF1000 (= start of p2)	mainline, rescue variant	by hand from U-Boot

normal variant: CONFIG_CMDLINE_EXTEND appends root=/dev/mmcblk0p4 rootwait rw + g_ether MACs to U-Boot’s console=ttymxc0,115200 root=/dev/mmcblk0p1 rootwait rw uart_at_4M → last root= wins → Alpine on p4. rescue variant: no built-in cmdline, U-Boot’s bootargs decide.

Partitions

| Part | Now | Notes | |—|—|—| | p1 | stock Amazon rootfs (untouched, ro) | rescue userland: busybox, wget, g_mass_storage/g_ether modules (3.10 only) | | p2 | raw rescue slot (old diags ext3 overwritten) | | | p3 | stock /var/local (untouched) | | | p4 | ext4 “alpine”: persistent Alpine 3.24 root | was FAT userstore | | boot0/1, sectors < 0x208, 0xE208+ | never touched | U-Boot, MBR, unknown data |

Day-to-day

• Power on → Alpine; tools/ssh.sh (root@192.168.7.2, key ~/.ssh/kindle8_ed25519, strict host key).
• PC side: NetworkManager kindle8-usb, shared mode (NAT+DNS) on 192.168.7.1 → Kindle has internet, apk add works.
• Kernel change: tools/test-kernel.sh (build → diags slot over SSH → reboot → boot diags once w/ watchdog → SSH check), then tools/test-kernel.sh --promote to make it the autoboot kernel.

Recovery (serial, from the uboot > prompt — hit a key during the 3 s countdown)

Test kernel hung → watchdog resets after 60 s → reset / power-cycle autoboots normal again.

Normal slot broken, boot the rescue kernel into Alpine:

mmc dev 1; mmc read 0x83000000 0xF1000 0xC8; mmc read 0x80800000 0xF10C8 0x7000
setenv bootargs console=ttymxc0,115200 root=/dev/mmcblk0p4 rootwait rw g_ether.dev_addr=02:4b:38:00:00:02 g_ether.host_addr=02:4b:38:00:00:01
bootm 0x80800000 - 0x83000000

Alpine broken too → same, but root=/dev/mmcblk0p1 rootwait ro init=/bin/sh (stock userland under mainline), then mount -t proc p /proc; ifconfig usb0 192.168.7.2 up and fetch/dd images with busybox wget.

Full stock restore (everything is in dumps/mmcblk0.img): write back sectors 0x208–0x7207 (normal slot), 0x7208–0xE207 (diags), p2 and p4 from the image.