Repository navigation
hi3516ev300: the split NOR image only, like hi3516cv6xx - #2656
Merged
Merged
Conversation
hi3516ev300's kernel is 2009K with the split options: padded to 64 KiB it ends exactly at the 2 MiB slot, so a fixed-slot package cannot ship beside a split one (the kernel would take the fixed image as split while U-Boot reads fixed slots). #2655 therefore left it out; it now ships the split package alone, as hi3516cv6xx does. Its lite images sat 16 KiB short of the 5 MiB rootfs slot, so it is the board that needs it most. - lite, ultimate, dev and glibc use general/scripts/post-image-split.sh without BR2_OPENIPC_NOR_FIXED_PACKAGE: openipc.hi3516ev300-nor-<variant> -split.tgz, no plain NOR package; ultimate still packages its NAND image. dev and glibc share hi3516ev300.generic.config, so they go with it rather than keep fixed-slot images a split-capable kernel would misread. neo keeps its own config and fixed-slot package. - hi3516ev300.generic.config: MTD_SPLIT_FIRMWARE ("firmware"), MTD_SPLIT_UIMAGE_FW, MTD_SPLIT_SQUASHFS_ROOT, MTD_ROOTFS_ROOT_DEV (OpenIPC/linux#131, #134). Built locally (kernel at hisilicon-hi3516ev200 e7008a95c): - hi3516ev300_lite: uImage 2008K/2048K read, firmware.bin 7168K/7616K; - hi3516ev300_ultimate: firmware.bin 10240K/15808K, rootfs.ubi 18304K/24576K, plus the NAND package. Tested in qemu-hisilicon (hi3516ev300) with the published u-boot-hi3516ev300-nor.bin (2022.10-gbaa15068), split firmware.bin at 0x50000, 8 MB lite and 16 MB ultimate, booted twice each: - U-Boot reads it as fixed slots (the squashfs is exactly 2 MiB after the uImage): "Found RootFS at 0x00250000", root=/dev/mtdblock3 and fixed mtdparts; the kernel: "hi_sfc: split NOR layout, device tree partitions", "root=/dev/mtdblock3 ignored", root on 31:4, login; rootfs_data at 0x750000 (lite) / 0xa50000 (ultimate) formatted once and kept (/init's .openipc-overlay marker there after the second boot). - .github/scripts/test_sysupgrade.sh passes (split-only boards take the -split package by default). After the first nightly with these packages, the stale plain openipc.hi3516ev300-nor-{lite,ultimate}.tgz on latest should be deleted: older cameras whose sysupgrade predates the split URL fallback would otherwise keep installing them. Not run on hardware.
widgetii
requested review from
cronyx,
flyrouter and
viktorxda
as code owners
October 21, 2026 18:15
PR Summary by QodoShip split-only NOR images for hi3516ev300
AI Description
Diagram
High-Level Assessment
Files changed (5)
|
Code Review by Qodo
1. Real-camera boot evidence is missing
|
![qodo-free-for-open-source-projects[bot]](https://avatars.githubusercontent.com/in/1068000?s=60&v=4){kind=small}
- hi3518ev300 lite and ultimate drop BR2_OPENIPC_NOR_FIXED_PACKAGE: the fixed-slot lite package no longer fits its 5120K rootfs slot (4 KiB over after today's merges), which fails the whole nightly. They ship the -split package only, like hi3516ev300 and hi3516cv6xx. - sysupgrade's default_upgrade_url applied the -split fallback only for packages from OpenIPC/firmware. dev and glibc come from OpenIPC/builder, so on a split-only board they asked for a plain archive that is no longer built (from review). nor_split_url now applies wherever the archive lives; it switches only when a -split archive exists. - test_sysupgrade.sh: a hi3516ev300 dev camera gets builder/.../openipc.hi3516ev300-nor-dev-split.tgz when one is published, the plain one otherwise. All checks pass.
This was referenced Oct 21, 2026
Merged
widgetii
added a commit
to OpenIPC/defib
that referenced
this pull request
Oct 22, 2026
hi3516ev300's firmware now ships only the split NOR package, like hi3516cv6xx (OpenIPC/firmware#2656): its uImage (2008K with the split options) fills the 2 MiB slot, so a fixed-slot package cannot ship beside a split one. hi3518ev300 went split-only in the same PR; it is already in `NOR_SPLIT_SOCS` (#160). - hi3516ev300 joins `NOR_SPLIT_SOCS`, so `install` writes `firmware.bin.hi3516ev300` at 0x50000 and erases the rest of the chip; its U-Boot is the published `u-boot-hi3516ev300-nor.bin` (split mode since OpenIPC/u-boot#48). - The refusal test that used hi3516ev300 as a chip without the split layout now uses hi3518ev200. CLAUDE.md lists it. The split package appears with the next nightly. Not run on hardware; the firmware side was checked in qemu-hisilicon (8 MB lite, 16 MB ultimate: root on the split rootfs, overlay kept across a reboot). `pytest tests/ --ignore=tests/fuzz`: 1084 passed. ruff and mypy are clean.
This was referenced Oct 23, 2026
widgetii
added a commit
to OpenIPC/website
that referenced
this pull request
Oct 23, 2026
…ex (#505) * Builds: a plain NOR package a newer split one replaced leaves the index The index takes, for each file name, the newest retained build that published it, so a package the firmware stops publishing stays on offer for the whole retention period. hi3516ev300 and hi3518ev300 lite and ultimate ship only the split package since OpenIPC/firmware#2656. Their last plain packages (nightlies of 2026-10-19 to 21) are still in the index, so the wizard gives them fixed slots, and SplitOnlyEdition, which looks for "split published, plain not", never sees them split-only. The stale assets are gone from the latest release; the index did not notice, because it does not read the release. LoadIndex now leaves out a plain NOR package whose split package came from a newer build than the plain one's newest. A board that publishes both does so in one build and keeps both (the gk7205v200 and gk7205v500 families). An edition with no split package (hi3516ev300 neo) and a board whose build failed, publishing neither, keep what they had. Test: TestLoadIndexDropsPlainSupersededBySplit stores two firmware builds, the 21st with the plain ev300 lite/ultimate, neo, gk7205v200 lite and cv300 lite, and the 22nd with the ev300 split pair, neo and gk7205v200 plain plus split. The ev300 plain pair is gone; split, neo, gk7205v200's both, and cv300's from the 21st stay. It fails on the old LoadIndex. service/run.sh test (all packages, Postgres) passes. * Builds: a split package replaces a plain one by the query's build order From review: two builds with one built_at rank by id in LoadIndex's query; the split check now does the same, comparing (built_at, id). Test: TestLoadIndexSplitSupersedesOnTiedTimestamp, two builds at one time, the later id carrying the split package. service/run.sh test 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.
hi3516ev300 joins the split NOR layout, split-only. Its lite rootfs is 16 KiB short of the 5 MiB slot and its kernel 41 KiB short of 2 MiB; with the split options its uImage is 2008K, over the 1983K a fixed-slot package beside a split one may use (#2655), so it ships
-splitalone, as hi3516cv6xx does.hi3516ev300_lite,_ultimate,_devand_glibcusegeneral/scripts/post-image-split.shwithoutBR2_OPENIPC_NOR_FIXED_PACKAGE. dev and glibc share the kernel config, so leaving them on fixed slots would give them images their split-capable kernel misreads. neo has its own config and stays fixed-slot.hi3516ev300.generic.config:MTD_SPLIT_FIRMWARE(named "firmware"),MTD_SPLIT_UIMAGE_FW,MTD_SPLIT_SQUASHFS_ROOT,MTD_ROOTFS_ROOT_DEV(hi3516ev200 family: the split NOR layout, from the device tree linux#131). Its U-Boot has split mode since nor: split layout on the hi3516ev200 family u-boot#48 (published).-split.tgzonly, and ultimate still produces its NAND package (hi3516ev200, hi3518ev300, hi3516dv200: the split NOR image, as on gk7205v200 #2655's fix).Sizes: lite uImage 2008K (2048K read), firmware.bin 7168K of 7616K; ultimate firmware.bin 10240K of 15808K, rootfs.ubi 18304K of 24576K.
Tested in qemu-hisilicon (hi3516ev300, published
u-boot-hi3516ev300-nor.bin2022.10-gbaa15068, 8 MB lite and 16 MB ultimate, two boots each): U-Boot reads the image as fixed slots (Found RootFS at 0x00250000: the squashfs lands exactly 2 MiB after the uImage) and passesroot=/dev/mtdblock3with fixed mtdparts; the kernel printshi_sfc: split NOR layout, device tree partitionsandroot=/dev/mtdblock3 ignored, root on 31:4, login prompt; rootfs_data at 0x750000 / 0xa50000 formatted once and kept across the reboot (overlay marker present).test_sysupgrade.shandci-matrix --self-testpass.After merge: the plain
openipc.hi3516ev300-nor-{lite,ultimate}.tgzonlateststay until deleted; once deleted, CLIsysupgrade(which updates itself first) takes-split.tgz, while web-UI upgrades with a sysupgrade older than the split-URL fallback find no package and write nothing. Not run on hardware.