docs(zhihu): refresh part0b numbers; rewrite part4 to avoid overlap#4
Conversation
Two updates to the unpublished 知乎 articles, prepping for the next daily publication wave. part0b-ai-workflow.md — fix stale numbers: - 33 → 34 test suites - ~282 → ~457 assertions - CLAUDE.md size: note the recent compression to ~280 lines part4-war-stories.md — full rewrite to "四大坑 · 收尾篇": The original "四大坑" article had 坑一 (SIMD), 坑二 (NS bit), 坑三 (跨核 buffer) all later superseded by deep-dive articles (part6 covers NS bit in depth; part7 covers SIMD + SMC-barrier). Publishing the original side-by-side with part6/part7 would feel ~75% repeated to readers. The rewrite drops 坑一/二/三 (with explicit cross-links to part6/part7) and develops 坑四 — SPMD per-CPU handshake, secondary CPU CPU_ON hanging — into a full standalone story (~220 lines). New framing: this is the closer of the "四大坑" arc, the one bug whose root cause is in TF-A's SPMD state machine (not in any ARM spec) and you only find it by reading TF-A source for `spmd_cpu_on_finish_handler`. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 53bc5c2e07
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| ### 第一步:注册 secondary entry point | ||
|
|
||
| 第一步，在 SPMC 初始化阶段注册 secondary CPU 的入口地址： | ||
| primary CPU 初始化完成后,调用 `FFA_SECONDARY_EP_REGISTER`(0x84000087)把 secondary entry 的物理地址告诉 SPMD: |
There was a problem hiding this comment.
The implementation registers a 64-bit secondary entry address and defines FFA_SECONDARY_EP_REGISTER as 0xC4000087 in src/ffa/mod.rs:73; documenting it here as 0x84000087 names the 32-bit SMC ID. If a reader follows the article literally or substitutes this numeric value for the constant, SPMD receives the wrong FF-A call for a 64-bit EP registration and the secondary entry point is not registered.
Useful? React with 👍 / 👎.
1 participant
Why
Prepping the unpublished 知乎 articles for the upcoming daily publication wave:
Changes
part0b-ai-workflow.md — small fixes
make runoutput)part4-war-stories.md — full rewrite as "四大坑 · 收尾篇"
spmd_cpu_on_finish_handler).How to read
docs/zhihu/part4-war-stories.mdagainst the v0 history to see the new framingdocs/zhihu/part0b-ai-workflow.mddiff is 3 lines🤖 Generated with Claude Code