Repository navigation
hi3516cv300, hi3516ev100: the split NOR image, as on the gk7205v200 family - #2645
Merged
Merged
Conversation
…amily hi3516cv300_lite, hi3516cv300_ultimate and hi3516ev100_lite build firmware.bin with general/scripts/post-image-split.sh -- the kernel's uImage (zImage + appended hi3516cv300-demb.dtb), then the squashfs at the next 64 KiB -- and publish it as the -split package. The kernel configs turn on the splitters and the root from the split (MTD_SPLIT_FIRMWARE, MTD_SPLIT_UIMAGE_FW, MTD_SPLIT_SQUASHFS_ROOT, MTD_ROOTFS_ROOT_DEV); the kernel (OpenIPC/linux hisilicon-hi3516cv300, the split NOR layout commit) finds the layout from its device tree, whatever mtdparts and root= the bootloader passes. hi3516cv300_neo keeps its own image. Like the Goke boards, these keep shipping the plain fixed-slot package (BR2_OPENIPC_NOR_FIXED_PACKAGE): cameras in the field upgrade from fixed slots. Its uImage is now held a 64 KiB block under the 2 MiB slot (1983K), which is how the kernel, U-Boot and sysupgrade tell the layouts apart; the field kernels are about 1.6 MiB. The kernel-read limit stays at the default 0x200000. The published U-Boot 2010.06 reads 3 MiB from 0x50000 on either chip size, but the fixed-slot 8 MiB layout has a 2 MiB kernel slot, so 2 MiB is what holds on every camera. Sizes (local builds with that kernel): - hi3516cv300_lite: uImage 1659K (1983K slot, 2048K read), squashfs 4956K (5120K slot), firmware.bin 6656K of 7616K (8 MiB less 320K and a 256K overlay); - hi3516cv300_ultimate: uImage 1659K, squashfs 6748K (8192K), firmware.bin 8448K of 15808K; - hi3516ev100_lite: uImage 1577K, squashfs 4948K, firmware.bin 6592K of 7616K. Tested in qemu-hisilicon (hi3516cv300; hi3516ev100, a machine added for this) from flash: the split firmware.bin at 0x50000 under the published U-Boot 2010.06 (its own bootcmd and mtdparts, 8 and 16 MiB) and under OpenIPC U-Boot 2022.10 with CONFIG_OPENIPC_NOR_SPLIT -- "hi_sfc: split NOR layout, device tree partitions", root on the split rootfs, /init's overlay found on rootfs_data on the next boot and a file written there kept across a reset; the plain package's uImage and rootfs under both, in fixed slots as before. Verified on a hi3516cv300 board (imx291, W25Q128 16 MiB, 128 MiB DRAM), after a full flash backup: - the split-aware kernel booted from RAM over the board's fixed-slot flash (its U-Boot 2010.06 env, 3072k kernel / 10240k rootfs): the command line's five partitions, nothing split, root on 31:3, login prompt; - the hi3516cv300_lite split firmware.bin written at 0x50000 under the same U-Boot 2010.06 and its saved 16 MiB mtdparts: "hi_sfc: split NOR layout, device tree partitions", kernel 0x50000-0x1f0000, rootfs from 0x1f0000, rootfs_data from 0x6d0000, "root=/dev/mtdblock3 ignored", root on 31:4, login prompt; /init's overlay marker (a JFFS2 inode and dirent ".openipc-overlay") on rootfs_data, kept across a power cycle; - OpenIPC U-Boot 2022.10 with CONFIG_OPENIPC_NOR_SPLIT written at 0, env default and saved: "Found Kernel at 0x00050000", "Found RootFS at 0x001f0000", command line "mem=32M console=ttyAMA0,115200 panic=20 rootfstype=squashfs init=/init", the same split, root on 31:4.
widgetii
requested review from
cronyx,
flyrouter and
viktorxda
as code owners
October 20, 2026 19:14
PR Summary by QodoEnable split NOR images for hi3516cv300 and hi3516ev100
AI Description
Diagram
High-Level Assessment
Files changed (5)
|
Code Review by Qodo🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)
Great, no issues found!Qodo reviewed your code and found no material issues that require reviewTip of the day💡 Did you know, you can open the rule or relationship behind a finding in one click |
This was referenced Oct 20, 2026
widgetii
added a commit
to OpenIPC/defib
that referenced
this pull request
Oct 21, 2026
hi3516cv300 and hi3516ev100 now ship the split NOR package (OpenIPC/firmware#2645, on OpenIPC/linux#127), and their OpenIPC/u-boot NOR build scans for it (OpenIPC/u-boot#42). Both join `NOR_SPLIT_SOCS`: one `firmware.bin.<soc>` at 0x50000, the rest of the chip erased. - Their U-Boot comes from OpenIPC/u-boot as `u-boot-<soc>-nor.bin` (plus `-nand.bin` for cv300), so they move from `AVAILABLE_FIRMWARE` to `PER_FLASH_TYPE_UBOOT`. The `u-boot-*-universal.bin` still on the release is the U-Boot 2010.06 build, which knows no split layout. - hi3516ev100 publishes no NAND build: the new `NOR_ONLY_UBOOT` makes asking for one find none, not a URL that 404s. - Tests that used hi3516cv300 as their example of a universal image or of the legacy NAND layout now use hi3516cv200. CLAUDE.md lists the two SoCs. **Verified end to end on a hi3516cv300 board** (imx291, W25Q128, 128 MiB), after a full flash backup: ``` defib install -c hi3516cv300 --firmware openipc.hi3516cv300-nor-lite-split.tgz \ --uboot u-boot-hi3516cv300-nor.bin -p /dev/ttyUSB0 --power-cycle --poe-port ether3 ... ``` U-Boot at 0 (flash CRC 9507112D), firmware.bin at 0x50000 with 0x50000+0xFB0000 erased (flash CRC BD92A6E5), default env with the factory ethaddr restored. The next boot showed `Found Kernel at 0x00050000`, `Found RootFS at 0x001f0000`, `Verifying Checksum ... OK`, `hi_sfc: split NOR layout, device tree partitions`, root on 31:4, then the login prompt. The install used local artifacts. The `u-boot-hi3516{cv300,ev100}-nor.bin` on the release predate OpenIPC/u-boot#42 and the split packages come with the next nightly, so the download path works once both are published. `pytest tests/ --ignore=tests/fuzz`: 1082 passed. ruff and mypy are clean.
widgetii
added a commit
to OpenIPC/website
that referenced
this pull request
Oct 22, 2026
…yout (#491) * Catalogue: hi3516ev300 on the split NOR layout, edition by edition hi3516ev300's lite and ultimate builds now ship only the split NOR package (OpenIPC/firmware split/hi3516ev300): its kernel fills the 2 MiB slot, so no fixed-slot package can ship beside a split one. neo keeps its own fixed-slot build. nor_split, without nor_split_only: SplitOnlyEdition then takes the split package for every edition that has one and no plain package, and keeps neo on fixed slots; nor_split_only would have dropped neo from the installer. That needs the stale plain openipc.hi3516ev300-nor- {lite,ultimate}.tgz gone from the release once the split packages publish; until then the site keeps offering them as fixed slots, as today. Tests: the fixture release index is moved to that state (the two -split packages in, the plain lite/ultimate out; neo kept), and the golden regenerated: hi3516ev300 still offers nor lite, ultimate, neo and nand ultimate, lite and ultimate as split images. The two firmware tests and the wizard test that used hi3516ev300 as a fixed-slot NOR SoC use hi3516ev200 (same bootloaders and NAND layout, fixed slots kept). go test ./... passes. * Catalogue: hi3516cv300, hi3516ev100, hi3516ev200, hi3518ev300, hi3516dv200 on the split NOR layout They ship -split packages since OpenIPC/firmware#2645, #2655 and #2656 (hi3518ev300 split-only, like hi3516ev300); nor_split lets each edition use its split package where no plain one is published, and keeps the fixed-slot one where it is. go test ./... (uncached) passes. * Catalogue: hi3516cv300 and hi3516ev100 install their -nor U-Boot From review: their bootloader was still the -universal U-Boot 2010.06, which does not scan the NOR layout, so the wizard never gave them the split install (scansLayout) and fell back to fixed slots. They take u-boot-<soc>-nor.bin (OpenIPC/u-boot 2022.10, scanning; cv300 also -nand.bin), as defib does since OpenIPC/defib#158. The test index lists those images, the boards reference names them, the golden is regenerated for the two SoCs, and the test that used hi3516cv300 as its -universal example uses hi3516cv200. go test ./... (uncached) passes.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
hi3516cv300 lite/ultimate and hi3516ev100 lite build the split NOR image as the gk7205v200 family does:
general/scripts/post-image-split.sh,BR2_OPENIPC_NOR_FIXED_PACKAGE=y(the plain fixed-slot package unchanged, plus-split), and the kernel's split options (MTD_SPLIT_FIRMWARE,MTD_SPLIT_UIMAGE_FW,MTD_SPLIT_SQUASHFS_ROOT,MTD_ROOTFS_ROOT_DEV). neo is unchanged.Depends on, both merged: OpenIPC/linux#127 (
hisilicon-hi3516cv300: mtdsplit, DT partitions, the hisi-sfc split hook) and OpenIPC/u-boot#42 (CONFIG_OPENIPC_NOR_SPLITfor cv300/ev100). Kernel read limit left at the 2 MiB default: 8 MiB chips have a 2 MiB kernel slot.Sizes (local builds): cv300_lite uImage 1659K, squashfs 4956K of 5120K, firmware.bin 6656K of 7616K; cv300_ultimate uImage 1659K, squashfs 6748K, firmware.bin 8448K of 15808K; ev100_lite uImage 1577K, firmware.bin 6592K.
Verified on a hi3516cv300 board (imx291, W25Q128 16 MiB, 128 MiB), after a full flash backup:
hi_sfc: split NOR layout, device tree partitions, kernel 0x50000-0x1f0000, rootfs 0x1f0000, rootfs_data 0x6d0000,root=/dev/mtdblock3 ignored, root on 31:4, login prompt; the overlay marker kept across a power cycle;Verifying Checksum ... OK, the same split;run fwupdover TFTP rewrote it and erased the rest of the chip, and it booted again;Also tested in qemu-hisilicon on hi3516cv300 and hi3516ev100 (8 and 16 MiB slots, both bootloaders, the fixed-slot package unchanged).