Reorganize source tree: kernel/, v3/, v4/ split and board infrastructure
Source tree reorganization: - Move StarForth v3 engine to v3/ (src/, include/, Makefile) - Move kernel to kernel/ (src/, include/, linker/, Makefile) - Create v4/ skeleton for F18-ISA golden model (DECOMPOSITION.md, JUSTIFICATION.md) - Move FABRIC-0..4.md to docs/fabric/ - Move ONTOLOGY.md and ROADMAP.md to docs/ Board infrastructure: - Add boards/ser5/, boards/raspi/, boards/milkv/, boards/zynq7020/ - Each board has board.mk (ISA, CPU flags, boot recipe) and README.md - Root Makefile becomes thin dispatcher: boot_image, all, clean, docs take TARGET - make boot_image TARGET=SER5|RASPI|MILKV builds one GPT/MBR image per board - ZYNQ7020 target exists but stops with clear error (ARMv7 port not built yet) - scripts/mkdiskimage.sh builds disk images for all boards Docs pipeline: - docs/book/ with LaTeX master (main.tex) and Makefile - pandoc converts Markdown to LaTeX at build time - Two Lua filters: table-widths.lua (wide tables wrap), code-breaks.lua (inline code breaks) - make docs builds single PDF (754 pages, 0 missing characters) - make docs TARGET=<board> adds board appendix - build/docs/<book|board>/meta.tex stamps git commit into PDF Bug fixes: - 42 include paths that only worked by accident now use correct relative paths - clang-18 hardcode replaced with configurable CC variable (fixed aarch64 build) - Pi 5: kernel_2712.img linked at 0x80000, .bss zeroed, memory reserved - Doxyfile, .clang-tidy, README.md, Kconfig paths updated Verified: - Hosted v3 build passes 1012 tests, 0 failures - SER5 image boots in QEMU (OVMF), POST passes, K exact (65536 = Q48_ONE) - Milk-V image boots in QEMU (OpenSBI + U-Boot + bootefi), POST passes - make clean TARGET=<board> removes only that board and its ISA objects - make all builds all boards, hosted v3, and docs in one run Co-authored-by: Junie <junie@jetbrains.com>
This commit is contained in:
rajames
@@ -0,0 +1,20 @@
|
||||
# Raspberry Pi 5 native boot media
|
||||
|
||||
`make boot_image TARGET=RASPI` builds `build/boards/raspi/lithos-raspi.img`: an MBR disk
|
||||
with one FAT32 (LBA) partition. Write it to an SD card with `dd`. The partition holds:
|
||||
|
||||
- `config.txt`: checked in here (FABRIC-3.md §IV.3 item 6). It pins `kernel_address=0x80000`
|
||||
because Pi 5 firmware loads `kernel_2712.img` at 0x200000 by default.
|
||||
- `kernel_2712.img`: the aarch64 kernel objects relinked at 0x80000 by
|
||||
`kernel/linker/starkernel-native-rpi5.ld`, entered at `rpi5_native_start`
|
||||
(`kernel/src/arch/aarch64/native_rpi5_entry.S`, which zeroes `.bss` first), and flattened
|
||||
with `objcopy -O binary`.
|
||||
- `bcm2712-rpi-5-b.dtb`: the Raspberry Pi firmware's own stock DTB. It is not built by this
|
||||
project; it is fetched once from the firmware release pinned in `board.mk`
|
||||
(`RPI_FIRMWARE_TAG`) into `build/cache/`. To use a local copy, pass `RPI5_DTB=/path/to/dtb`.
|
||||
- `cmdline.txt`: only present when `KERNEL_ARGS="..."` is given. The firmware copies it into
|
||||
`/chosen/bootargs`, which `rpi5_native_boot()` parses.
|
||||
|
||||
Not yet verified on hardware. The native path still assumes everything `kernel_main()`
|
||||
needs after a UEFI handoff. The firmware enters at EL2 with the MMU off, which is not what
|
||||
UEFI hands over, and that difference has not been exercised.
|
||||
@@ -0,0 +1,18 @@
|
||||
# boards/raspi/board.mk -- Raspberry Pi 5 (BCM2712, Cortex-A76, aarch64).
|
||||
#
|
||||
# Native firmware boot, no UEFI (FABRIC-3.md §IV): the Pi 5 bootloader
|
||||
# lives in EEPROM and reads config.txt, kernel_2712.img and the board DTB
|
||||
# from the first FAT partition of the SD card. kernel_2712.img is the
|
||||
# aarch64 kernel relinked at 0x80000 (kernel/linker/starkernel-native-rpi5.ld)
|
||||
# and flattened to a raw binary.
|
||||
#
|
||||
# The DTB is the Raspberry Pi firmware's own stock file, not built here. It
|
||||
# is fetched once from the pinned firmware release below into build/cache/
|
||||
# unless RPI5_DTB points at a local copy.
|
||||
BOARD_DESC := Raspberry Pi 5 (BCM2712, Cortex-A76, native firmware boot)
|
||||
BOARD_ARCH := aarch64
|
||||
BOARD_BOOT := rpi-native
|
||||
BOARD_IMAGE := lithos-raspi.img
|
||||
RPI_FIRMWARE_TAG ?= 1.20260915
|
||||
RPI5_DTB_URL ?= https://raw.githubusercontent.com/raspberrypi/firmware/$(RPI_FIRMWARE_TAG)/boot/bcm2712-rpi-5-b.dtb
|
||||
RPI5_DTB ?= build/cache/raspi-$(RPI_FIRMWARE_TAG)/bcm2712-rpi-5-b.dtb
|
||||
@@ -0,0 +1,69 @@
|
||||
# LithosAnanke native (non-UEFI) boot config for Raspberry Pi 5 (BCM2712).
|
||||
# FABRIC-3.md §IV.3 item 6 -- see that section's own doc comment for the
|
||||
# research trail behind each line (confirmed against raspberrypi.com's
|
||||
# current config.txt reference and a real, working Pi 5 bare-metal
|
||||
# project's own checked-in config.txt, not assumed from general Pi
|
||||
# knowledge).
|
||||
|
||||
# Load our own kernel image directly -- no UEFI, no Linux. kernel_2712.img
|
||||
# is the real filename Pi 5/500/CM5 firmware looks for first (its own
|
||||
# 16K-page-size build); naming our own image this exactly, rather than
|
||||
# falling back to kernel8.img, removes any ambiguity about which file the
|
||||
# firmware picks up.
|
||||
kernel=kernel_2712.img
|
||||
|
||||
# Pin the load address to the one the image is linked at
|
||||
# (kernel/linker/starkernel-native-rpi5.ld). Pi 5 firmware's own default
|
||||
# for kernel_2712.img is 0x200000, not the classic 0x80000 -- leaving it
|
||||
# implicit would load a 0x80000-linked flat image 1.5 MiB away from where
|
||||
# every absolute address in it points.
|
||||
kernel_address=0x80000
|
||||
|
||||
# Disable the firmware's Linux-compatible-image sanity check. Without
|
||||
# this, official docs describe it as checking for "a compatible Device
|
||||
# Tree file before attempting to boot" and warn that "older non-compatible
|
||||
# kernels would be loaded and then hang" -- documented specifically as the
|
||||
# bare-metal-development escape hatch. One real working Pi 5 bare-metal
|
||||
# project's own config.txt does NOT set this and still boots, so it may
|
||||
# not be strictly required for every non-Linux image; kept here anyway
|
||||
# since the official doc names this exact use case and there is no
|
||||
# hardware in hand yet to verify which behavior is actually gated.
|
||||
os_check=0
|
||||
|
||||
# Pin the exact devicetree explicitly rather than relying on the
|
||||
# firmware's own board-revision auto-selection (which would also resolve
|
||||
# to this file on real Pi 5 hardware, per general config.txt convention --
|
||||
# explicit is certain, not just probably-correct).
|
||||
device_tree=bcm2712-rpi-5-b.dtb
|
||||
|
||||
# Enable the dedicated debug UART (Pi 5's own JST connector, not the GPIO
|
||||
# 14/15 header pins). Not because serial capture hardware is set up for
|
||||
# this bring-up (that decision is HDMI-only, FABRIC-3.md §IV) -- this is
|
||||
# a firmware-side switch that costs nothing when nothing is listening, and
|
||||
# it is the only diagnostic channel that survives a framebuffer failure.
|
||||
# rpi5_uart_base() (rpi5_dtb.c) already exists and is currently uncalled;
|
||||
# this is the hook a UART-based fallback console would use if the mailbox
|
||||
# framebuffer driver's two documented unverified details turn out wrong on
|
||||
# real hardware.
|
||||
enable_uart=1
|
||||
|
||||
# Deliberately NOT set, researched but not included this pass:
|
||||
# dtoverlay=uart0 -- GPIO 14/15 header UART; needs a USB-serial
|
||||
# adapter this bring-up doesn't have (HDMI-only
|
||||
# decision). enable_uart=1 above covers the JST
|
||||
# debug UART instead, which needs no extra cable.
|
||||
# dtparam=pciex1 -- one real Pi 5 bare-metal project's own config.txt
|
||||
# comments this as "Required for RP1 access", which
|
||||
# would be surprising (RP1 is always-on in stock Pi
|
||||
# 5 boots) and is community-reported, not confirmed
|
||||
# against official docs. Not needed for this boot
|
||||
# (no RP1 access yet -- FABRIC-3.md §IV.3 item 5 is
|
||||
# still re-scoped, not implemented). Revisit if that
|
||||
# item's own real Broadcom-indirect PCIe work needs
|
||||
# it.
|
||||
# arm_64bit=1 -- researched and dropped: official docs state
|
||||
# "Models that only support a 64-bit kernel ignore
|
||||
# this flag" -- inert on Pi 5, not wrong to include,
|
||||
# just meaningless. The original item 6 wording
|
||||
# named this; corrected here rather than carried
|
||||
# forward.
|
||||
Reference in New Issue
Block a user