Efuse logical-map override: end-user loadable patch file (RAM-only)
維護者通常 1 天內回覆
還沒有人認領這個 Issue。
評估
- 難度
- 5/5
- 預估耗時
- 一週以上
- 新手友好度
- 35/100
- Issue 類型
- 功能
- 描述清晰度
- 描述清楚
- 活躍度
- 冷清
- 技術堆疊
- c, json
- 領域
- embedded-iot
研究方向
Start with the four materialization sites named in the issue: src/jaguar1/EepromManager.cpp, src/jaguar2/HalJaguar2.cpp, and both variant paths in src/jaguar3/HalJaguar3.cpp; also inspect examples/common/env_config.cpp and the existing replay_wseq handling. Begin by establishing machine-readable logical-map dumps and headless tests, then cover sparse patch parsing, each generation’s apply point, and the 8814A late re-read. Done means tests cover ID behavior and reapplication, and diagnostics make overrides visible; the issue also calls for hardware validation.
由索引模型根據 Issue 內容生成。
描述
Motivation
The efuse logical map is the root of almost every per-adapter parameter we consume: MAC, xtal cap, thermal baseline, RFE type, PA/LNA types, per-channel TX-power PG base. Today it is read-only and invisible except for the stderr DEVOURER_EFUSE_DUMP text (Jaguar2/3 only). A RAM-only override lets us:
- test "what would this adapter do with rfe_type=N / xtal=M" without touching hardware state,
- resurrect adapters with corrupted efuse (force the autoload-OK ID, supply a sane map),
- act as the validation harness for physical efuse programming (see follow-up issue): dump → edit → run with override → validate on-air → only then burn.
Shape
A sparse patch file applied after logical-map materialization, before field parsing.
DeviceConfig::Debug::efuse_override(path,std::optional<std::string>), env-mapped asDEVOURER_EFUSE_OVERRIDEinexamples/common/env_config.cpp— same pattern asreplay_wseq, which is the existing precedent for a library-loaded file path.- Grammar:
0xCA=0x04single bytes,0x22:8=xxxxxxxxxxxxxxxxruns. A full-map binary replace can be a second, explicitly named mode later — it must not be the default (a full dump from adapter X replayed on adapter Y silently clobbers Y's calibration). - The patch parser and the apply helper are generation-neutral pure functions → live in
src/next toPhyTableLoader, unit-testable headlessly (ctest).
Implementation details / hazards
- Four materialization sites, not one. Jaguar1 decodes into
efuse_eeprom_data(src/jaguar1/EepromManager.cpp:1382), Jaguar2 into_efuse_map(src/jaguar2/HalJaguar2.cpp:305), Jaguar3 has two variant paths — 8822C standard walk and 8822E packed 2-byte headers into a one-shot cache (src/jaguar3/HalJaguar3.cpp:621-719). One sharedapply_efuse_override(map, len)called at each site. - 8814A re-reads efuse after FW download (
LateInitFor8814A,src/jaguar1/EepromManager.cpp:89). The override must re-apply on that second read or the 8814 silently loses the patch. Needs a dedicated test. - 8822E PA bias bypasses the logical map — it reads physical addresses 0x5c5–0x5c8 during the power-cut bracket (
src/jaguar3/HalJaguar3.cpp:580). A logical override cannot affect it. v1: document the hole; aphys:patch plane doubles the surface and is not worth it yet. - Autoload/ID semantics cut both ways. Jaguar1 gates on ID
0x8129at offset 0x00 (Hal_EfuseParseIDCode8812A). A wrong override flips the device into autoload-fail defaults; forcing it right is how you resurrect a corrupt-efuse adapter. Both directions need tests. - Dump tooling is a prerequisite. You cannot author an override without a dump in the same grammar. Extend
DEVOURER_EFUSE_DUMPto emit a machine-readable logical map on all three generations (JSONLefuse.mapevent or file) — today it is stderr text and Jaguar2/3 only; Jaguar1 only has the physicalReadPhysicalEfuseMaphealth probe. - Footgun visibility. One-shot
Wdiagnostic diffing every overridden byte + anefuse.overrideJSONL event, so test scripts can assert the patch took effect and nobody benchmarks on doctored calibration without noticing.
Suggested milestones
- Machine-readable logical-map dump, all generations (
efuse.mapevent). - Sparse-patch parser +
apply_efuse_overrideinsrc/, headless selftests. - Wire into the four materialization sites (incl. the 8814A late re-read) +
efuse.overrideevent. - Hardware validation: override rfe_type / xtal on a bench adapter, confirm the parsed-field logs and on-air behavior follow.
- 主要語言
- C
- 星號
- 56
- 分支
- 29
- 平均合併
- 3 小時 43 分鐘
- 30 天內合併 PR
- 75
環境準備
這個專案沒有提供開發容器、Dockerfile 或貢獻指南,環境需要你自己搭建:先看它的 README,通用步驟見我們的新手貢獻指南。
從這裡開始
- 先讀完整個 Issue,再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫,在一個分支上完成修改。
- 送出 Pull Request,並在描述裡引用這個 Issue 編號。
OpenIPC/devourer 的其他 Issue
-
難度 2/5 1-3 小時 新手友好度 82/100
維護者通常 1 天內回覆
-
Follow-ups left open by the 8822B fast bandwidth toggle fix (#584)可能已有人在做 @josephnef 今天認領。 未關閉enhancement
難度 4/5 3-5 天 新手友好度 42/100
維護者通常 1 天內回覆
-
難度 5/5 一週以上 新手友好度 12/100
OpenIPC/devourer#579 · 1 則留言 ·
維護者通常 1 天內回覆
-
enhancement
難度 5/5 一週以上 新手友好度 12/100
維護者通常 1 天內回覆
-
enhancement
難度 5/5 一週以上 新手友好度 8/100
維護者通常 1 天內回覆
相似的 Issue
-
難度 2/5 1-3 小時 新手友好度 72/100
維護者通常 1 天內回覆
-
Needs-Triage
難度 2/5 1-3 小時 新手友好度 66/100
維護者通常 2 天內回覆
-
issue: bug report
難度 2/5 1-3 小時 新手友好度 82/100
維護者通常 2 天內回覆
-
cross-platform fix-now
難度 2/5 1-3 小時 新手友好度 76/100
維護者通常 1 天內回覆
-
feature:index kind:crash status:confirmed triaged
難度 2/5 1-3 小時 新手友好度 77/100
維護者通常 1 天內回覆