feat(i2c): Add c_board I2C2 transport support#47
Closed
gqsdjhh wants to merge 1 commit into
Closed
Conversation
总体概览该PR为机器人控制系统添加了完整的I2C（Inter-Integrated Circuit）通信支持，涵盖跨越核心协议层、固件硬件驱动和主机代理的集成实现。包括新增数据结构、序列化/反序列化流程、STM32硬件初始化和固件驱动等内容。 变更清单
代码审查难度评估🎯 4 (复杂) | ⏱️ ~60 分钟 说明：本 PR 涵盖多个分层次的变更（协议、硬件驱动、固件、主机集成），涉及复杂的异步I/O流程、中断处理、DMA管理、错误恢复逻辑和跨边界的回调机制。序列化/反序列化逻辑、固件驱动的请求队列和超时管理、以及主机端的模板/继承链都需要仔细审查以确保正确性。 可能相关的PR
诗
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (3)
core/include/librmcs/protocol/i2c.hpp (1)
8-10: 建议加一条防御性 static_assert
kI2cMaxDataLength通过static_cast<std::uint16_t>((1U << kI2cDataLengthBits) - 1U)计算，当前kI2cDataLengthBits = 9时结果为 511，正确。但若将来有人将位宽改为 > 16，强制转换会静默截断导致协议常量错误。建议增加一个编译期断言以防止此类回归：♻️ 建议的小幅加固
inline constexpr std::size_t kI2cDataLengthBits = 9; +static_assert(kI2cDataLengthBits < 16, + "kI2cDataLengthBits must fit in uint16_t without truncation"); inline constexpr std::uint16_t kI2cMaxDataLength = static_cast<std::uint16_t>((1U << kI2cDataLengthBits) - 1U);🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@core/include/librmcs/protocol/i2c.hpp` around lines 8 - 10, Add a compile-time guard to prevent silent truncation of kI2cMaxDataLength if kI2cDataLengthBits is increased: insert a static_assert that kI2cDataLengthBits is <= std::numeric_limits<std::uint16_t>::digits (and optionally > 0) before the kI2cMaxDataLength definition so the expression static_cast<std::uint16_t>((1U << kI2cDataLengthBits) - 1U) cannot overflow/truncate; reference the constants kI2cDataLengthBits and kI2cMaxDataLength when adding the assertion.firmware/c_board/app/src/i2c/i2c.hpp (2)
175-183:PendingUplink对错误条目保留了整块kMaxDataLengthpayload,内存浪费可观
pending_uplinks_为kPendingUplinkQueueSize = 32条固定大小的PendingUplink,每条内含alignas(uint32_t) std::array<std::byte, kMaxDataLength> payload{}。当kMaxDataLength等于协议允许的最大帧长度(看起来是 512 字节),即使一条队列中全部只是错误上报(不携带 payload),也要占用约 16 KB RAM。对 F4 级 MCU 这是相当大的一块。可选改进(按影响由小到大):
- 将
kReadResult与kError拆成两条独立队列,错误队列只存元数据;- 或把读结果的 payload 单独放进一个小池(如 4~8 个
kMaxDataLength槽),条目用 chunk 索引引用,类似现在的payload_chunks_设计;- 或降低
kPendingUplinkQueueSize(32 条对错误/读结果积压而言偏多)。此为可选优化,不影响功能正确性。
Also applies to: 682-682
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@firmware/c_board/app/src/i2c/i2c.hpp` around lines 175 - 183, PendingUplink currently embeds a full kMaxDataLength payload causing large RAM use for error-only entries; split uplink types into a small metadata-only ErrorUplink and a ReadUplink that references payload storage: create a new lightweight struct (e.g. ErrorUplink) containing type/slave_address/data_length/has_register/reg_address/is_read without the large payload, change pending_uplinks_ usage to maintain two queues (one for ErrorUplink, one for ReadUplink or keep pending_uplinks_ only for reads), and implement a small fixed-size payload pool (reuse existing payload_chunks_ design or add a payload_chunks_ array and have ReadUplink store a chunk index and actual length) so actual kMaxDataLength buffers are only allocated in the pool; update enqueue/dequeue logic around functions that push/pop PendingUplink to use the new structs and pool.
511-531: 在InterruptLockGuard作用域内执行可达kMaxDataLength字节的 memcpy第 525 行在关中断状态下把整段 payload 拷进 pending 槽。当
kMaxDataLength取最大值时,这段拷贝会把整个系统的中断响应拖后数个微秒。建议把 payload 拷贝挪出锁外:先在锁内分配/保留一个条目槽(占位并推进 head/count),记录目标槽指针,锁外再 memcpy payload,再在锁内将条目置为"已就绪"供
try_flush_uplink消费。或者使用单生产者-单消费者的 ring,head 由 ISR 推进、tail 由主循环推进,拷贝本身不需关中断。🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@firmware/c_board/app/src/i2c/i2c.hpp` around lines 511 - 531, enqueue_read_result_uplink holds InterruptLockGuard while doing a potentially large memcpy of up to kMaxDataLength bytes into pending_uplinks_, which extends interrupt-disabled time; fix by reserving a slot inside the lock (use pending_uplink_head_, pending_uplink_count_ and advance_pending_uplink_index to increment head/count and obtain the target entry index or a pointer), release the lock, perform std::memcpy to that target entry's payload outside the InterruptLockGuard, then re-enter the lock and mark the entry fields (type, slave_address, data_length, has_register, reg_address or a ready flag) so try_flush_uplink can consume it; alternatively implement single-producer-single-consumer ring semantics where ISR advances head and main loop advances tail so the copy does not need interrupts disabled.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@core/src/protocol/deserializer.hpp`:
- Around line 44-52: Add no-op overrides for the four missing I2C callbacks in
the Vendor class so it fulfills the DeserializeCallback interface: implement
i2c_write_deserialized_callback(FieldId id, const data::I2cDataView& data)
override, i2c_read_config_deserialized_callback(FieldId id, const
data::I2cReadConfigView& data) override,
i2c_read_result_deserialized_callback(FieldId id, const data::I2cDataView& data)
override, and i2c_error_deserialized_callback(FieldId id, const
data::I2cErrorView& data) override in
firmware/rmcs_board/app/src/usb/vendor.hpp; each should have an empty body
(no-op), similar to existing rmcs_board GPIO/accelerometer callbacks, to satisfy
the pure-virtual methods in DeserializeCallback and unblock compilation.
In `@firmware/c_board/app/src/i2c/i2c.hpp`:
- Around line 412-419: recover_active_request_locked currently performs heavy
HAL operations (abort_dma_if_enabled, HAL_I2C_DeInit/HAL_I2C_Init) while holding
InterruptLockGuard, which blocks other ISRs; change it to only perform quick
locked updates: set recovery_pending_ = true, set internal state to Idle,
clear_stale_irq_sources_locked(), and call finish_active_request_locked() — do
not call abort_dma_if_enabled or HAL_I2C_DeInit/Init inside the lock. Move the
actual DMA abort and HAL reinit into the non-ISR path: have update() perform
abort_dma_if_enabled and HAL_I2C_DeInit/HAL_I2C_Init when recovery_pending_ is
true (mirroring error_callback's deferred recovery pattern at lines ~124-128).
Also modify callers (recover_active_request, recover_timed_out_transfer,
consume_deferred_recovery and the try_start_next_request failure branch) to set
recovery_pending_ instead of performing immediate HAL reinit so all heavy HAL
work occurs unlocked in update().
---
Nitpick comments:
In `@core/include/librmcs/protocol/i2c.hpp`:
- Around line 8-10: Add a compile-time guard to prevent silent truncation of
kI2cMaxDataLength if kI2cDataLengthBits is increased: insert a static_assert
that kI2cDataLengthBits is <= std::numeric_limits<std::uint16_t>::digits (and
optionally > 0) before the kI2cMaxDataLength definition so the expression
static_cast<std::uint16_t>((1U << kI2cDataLengthBits) - 1U) cannot
overflow/truncate; reference the constants kI2cDataLengthBits and
kI2cMaxDataLength when adding the assertion.
In `@firmware/c_board/app/src/i2c/i2c.hpp`:
- Around line 175-183: PendingUplink currently embeds a full kMaxDataLength
payload causing large RAM use for error-only entries; split uplink types into a
small metadata-only ErrorUplink and a ReadUplink that references payload
storage: create a new lightweight struct (e.g. ErrorUplink) containing
type/slave_address/data_length/has_register/reg_address/is_read without the
large payload, change pending_uplinks_ usage to maintain two queues (one for
ErrorUplink, one for ReadUplink or keep pending_uplinks_ only for reads), and
implement a small fixed-size payload pool (reuse existing payload_chunks_ design
or add a payload_chunks_ array and have ReadUplink store a chunk index and
actual length) so actual kMaxDataLength buffers are only allocated in the pool;
update enqueue/dequeue logic around functions that push/pop PendingUplink to use
the new structs and pool.
- Around line 511-531: enqueue_read_result_uplink holds InterruptLockGuard while
doing a potentially large memcpy of up to kMaxDataLength bytes into
pending_uplinks_, which extends interrupt-disabled time; fix by reserving a slot
inside the lock (use pending_uplink_head_, pending_uplink_count_ and
advance_pending_uplink_index to increment head/count and obtain the target entry
index or a pointer), release the lock, perform std::memcpy to that target
entry's payload outside the InterruptLockGuard, then re-enter the lock and mark
the entry fields (type, slave_address, data_length, has_register, reg_address or
a ready flag) so try_flush_uplink can consume it; alternatively implement
single-producer-single-consumer ring semantics where ISR advances head and main
loop advances tail so the copy does not need interrupts disabled.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 72a1c41d-62bf-49f3-a7d9-58760f13cde5
📒 Files selected for processing (25)
core/include/librmcs/data/datas.hppcore/include/librmcs/protocol/i2c.hppcore/src/protocol/deserializer.cppcore/src/protocol/deserializer.hppcore/src/protocol/protocol.hppcore/src/protocol/serializer.hppfirmware/c_board/app/src/app.cppfirmware/c_board/app/src/i2c/i2c.cppfirmware/c_board/app/src/i2c/i2c.hppfirmware/c_board/app/src/usb/interrupt_safe_buffer.hppfirmware/c_board/app/src/usb/vendor.hppfirmware/c_board/bsp/cubemx/Core/Inc/i2c.hfirmware/c_board/bsp/cubemx/Core/Inc/stm32f4xx_hal_conf.hfirmware/c_board/bsp/cubemx/Core/Inc/stm32f4xx_it.hfirmware/c_board/bsp/cubemx/Core/Src/dma.cfirmware/c_board/bsp/cubemx/Core/Src/gpio.cfirmware/c_board/bsp/cubemx/Core/Src/i2c.cfirmware/c_board/bsp/cubemx/Core/Src/main.cfirmware/c_board/bsp/cubemx/Core/Src/stm32f4xx_it.cfirmware/c_board/bsp/cubemx/cmake/stm32cubemx/CMakeLists.txtfirmware/c_board/bsp/cubemx/rmcs_slave.iochost/include/librmcs/agent/c_board.hpphost/include/librmcs/agent/detail/i2c0_common.hpphost/include/librmcs/protocol/handler.hpphost/src/protocol/handler.cpp
Comment on lines
+44
to
+52
| virtual void i2c_write_deserialized_callback(FieldId id, const data::I2cDataView& data) = 0; | ||
|
|
||
| virtual void | ||
| i2c_read_config_deserialized_callback(FieldId id, const data::I2cReadConfigView& data) = 0; | ||
|
|
||
| virtual void | ||
| i2c_read_result_deserialized_callback(FieldId id, const data::I2cDataView& data) = 0; | ||
|
|
||
| virtual void i2c_error_deserialized_callback(FieldId id, const data::I2cErrorView& data) = 0; |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# 检查两个 Vendor 类是否 override 了新加的 4 个 i2c_*_deserialized_callback 方法。
fd -a 'vendor.hpp' firmware | while read -r f; do
echo "=== $f ==="
rg -n 'i2c_(write|read_config|read_result|error)_deserialized_callback' "$f" || echo " (未实现)"
doneRepository: Alliance-Algorithm/librmcs
Length of output: 410
🏁 Script executed:
#!/bin/bash
# 检查 rmcs_board vendor.hpp 的内容和继承关系
echo "=== rmcs_board vendor.hpp 内容 ==="
cat -n firmware/rmcs_board/app/src/usb/vendor.hpp | head -100
echo -e "\n=== rmcs_board Vendor 类定义 ==="
rg -A 5 'class Vendor' firmware/rmcs_board/app/src/usb/vendor.hpp
echo -e "\n=== c_board 四个 callback 实现的内容（示例） ==="
rg -A 8 'i2c_write_deserialized_callback' firmware/c_board/app/src/usb/vendor.hpp | head -20Repository: Alliance-Algorithm/librmcs
Length of output: 4810
🏁 Script executed:
#!/bin/bash
# 检查是否存在其他继承 DeserializeCallback 的类
echo "=== 所有继承 DeserializeCallback 的类 ==="
rg 'class \w+.*:\s*.*DeserializeCallback' firmware --type hpp --type cppRepository: Alliance-Algorithm/librmcs
Length of output: 150
🏁 Script executed:
#!/bin/bash
# 读取 rmcs_board vendor.hpp 的全部内容，确认是否有 i2c 回调实现
wc -l firmware/rmcs_board/app/src/usb/vendor.hpp
echo "=== 完整文件内容 ==="
cat -n firmware/rmcs_board/app/src/usb/vendor.hpp
echo -e "\n=== 检查是否含有 i2c_write_deserialized_callback ==="
grep -n 'i2c_write_deserialized_callback' firmware/rmcs_board/app/src/usb/vendor.hpp || echo "未找到"Repository: Alliance-Algorithm/librmcs
Length of output: 6532
🏁 Script executed:
#!/bin/bash
# 正确的方式：用 --type-list 查看支持的类型，或直接用 rg 的默认设置
echo "=== 所有继承 DeserializeCallback 的类 ==="
rg 'class \w+.*:\s*.*DeserializeCallback' firmwareRepository: Alliance-Algorithm/librmcs
Length of output: 105
🏁 Script executed:
#!/bin/bash
# 查看 DeserializeCallback 中新增的 i2c 方法签名
echo "=== DeserializeCallback 中的 i2c 方法 ==="
rg -A 2 'i2c_(write|read_config|read_result|error)_deserialized_callback' core/src/protocol/deserializer.hpp
# 检查 rmcs_board 是否有 I2C 支持
echo -e "\n=== rmcs_board 是否有 I2C 支持 ==="
find firmware/rmcs_board -name '*i2c*' -type f 2>/dev/null || echo "(未找到 I2C 文件)"Repository: Alliance-Algorithm/librmcs
Length of output: 604
🏁 Script executed:
#!/bin/bash
# 检查 c_board vendor.hpp 中 i2c 回调的完整实现
echo "=== c_board i2c_write_deserialized_callback 完整实现 ==="
rg -A 6 'i2c_write_deserialized_callback' firmware/c_board/app/src/usb/vendor.hpp
echo -e "\n=== c_board i2c_read_result_deserialized_callback 完整实现 ==="
rg -A 6 'i2c_read_result_deserialized_callback' firmware/c_board/app/src/usb/vendor.hpp
echo -e "\n=== c_board i2c_error_deserialized_callback 完整实现 ==="
rg -A 6 'i2c_error_deserialized_callback' firmware/c_board/app/src/usb/vendor.hppRepository: Alliance-Algorithm/librmcs
Length of output: 1048
firmware/rmcs_board/app/src/usb/vendor.hpp 中缺少四个 I2C 回调实现，导致构建失败。
Vendor 类继承自 DeserializeCallback（私有继承），必须实现所有纯虚方法。firmware/c_board/app/src/usb/vendor.hpp 中的实现已齐全（downlink 写操作路由到 i2c0，读结果和错误帧作为仅上行数据类型被忽略）。但 firmware/rmcs_board/app/src/usb/vendor.hpp 缺失下述四个 override：
i2c_write_deserialized_callbacki2c_read_config_deserialized_callbacki2c_read_result_deserialized_callbacki2c_error_deserialized_callback
rmcs_board 没有 I2C 硬件，这些方法应为 no-op（可参考 rmcs_board 现有的 GPIO 或加速度计回调实现），以便解除编译阻塞。
🧰 Tools
🪛 GitHub Actions: Lint
[error] 44-52: Pure virtual functions in DeserializeCallback: i2c_write_deserialized_callback, i2c_read_config_deserialized_callback, i2c_read_result_deserialized_callback, i2c_error_deserialized_callback (unimplemented by Vendor).
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@core/src/protocol/deserializer.hpp` around lines 44 - 52, Add no-op overrides
for the four missing I2C callbacks in the Vendor class so it fulfills the
DeserializeCallback interface: implement i2c_write_deserialized_callback(FieldId
id, const data::I2cDataView& data) override,
i2c_read_config_deserialized_callback(FieldId id, const data::I2cReadConfigView&
data) override, i2c_read_result_deserialized_callback(FieldId id, const
data::I2cDataView& data) override, and i2c_error_deserialized_callback(FieldId
id, const data::I2cErrorView& data) override in
firmware/rmcs_board/app/src/usb/vendor.hpp; each should have an empty body
(no-op), similar to existing rmcs_board GPIO/accelerometer callbacks, to satisfy
the pure-virtual methods in DeserializeCallback and unblock compilation.
Comment on lines
+412
to
+419
| void recover_active_request_locked() { | ||
| abort_dma_if_enabled(hal_i2c_handle_->hdmarx); | ||
| abort_dma_if_enabled(hal_i2c_handle_->hdmatx); | ||
| core::utility::assert_always(HAL_I2C_DeInit(hal_i2c_handle_) == HAL_OK); | ||
| core::utility::assert_always(HAL_I2C_Init(hal_i2c_handle_) == HAL_OK); | ||
| clear_stale_irq_sources_locked(); | ||
| finish_active_request_locked(); | ||
| } |
There was a problem hiding this comment.
recover_active_request_locked() 在关中断期间执行 HAL_I2C_DeInit/Init,会放大中断延迟
该函数通过 recover_active_request()、recover_timed_out_transfer()、consume_deferred_recovery() 被调用时,都处于 InterruptLockGuard 作用域内。HAL_I2C_DeInit + HAL_I2C_Init 涉及多次寄存器写、可能等待 BUSY 位清零,关中断期间执行会阻塞所有其它 ISR(CAN/UART/USB/定时器)。
考虑把真正的 DMA abort + HAL 重初始化移到锁外:锁内只快速"标记为需要恢复 + 状态置 Idle",锁外再做 HAL 重建。error_callback() 已经采用了这种延迟恢复的思路(行 124-128),可以把 try_start_next_request 启动失败分支和超时恢复也统一改走 recovery_pending_ 路径,让所有重型 HAL 操作都在 update() 主循环中(且锁外)执行。
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@firmware/c_board/app/src/i2c/i2c.hpp` around lines 412 - 419,
recover_active_request_locked currently performs heavy HAL operations
(abort_dma_if_enabled, HAL_I2C_DeInit/HAL_I2C_Init) while holding
InterruptLockGuard, which blocks other ISRs; change it to only perform quick
locked updates: set recovery_pending_ = true, set internal state to Idle,
clear_stale_irq_sources_locked(), and call finish_active_request_locked() — do
not call abort_dma_if_enabled or HAL_I2C_DeInit/Init inside the lock. Move the
actual DMA abort and HAL reinit into the non-ISR path: have update() perform
abort_dma_if_enabled and HAL_I2C_DeInit/HAL_I2C_Init when recovery_pending_ is
true (mirroring error_callback's deferred recovery pattern at lines ~124-128).
Also modify callers (recover_active_request, recover_timed_out_transfer,
consume_deferred_recovery and the try_start_next_request failure branch) to set
recovery_pending_ instead of performing immediate HAL reinit so all heavy HAL
work occurs unlocked in update().
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
1 participant
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.
Summary
This PR adds end-to-end I2C transport support for
c_board.It introduces the I2C wire format in the shared protocol layer, exposes host-side I2C APIs for
CBoard, and implements the firmware-side transport onc_boardby mapping the protocol-facing logicalI2C0channel to STM32
I2C2with DMA-backed transfers.What Changed
1. Shared protocol layer
core/include/librmcs/protocol/i2c.hppI2cDataViewI2cReadConfigViewI2cErrorViewDataCallbackto report structured I2C errors instead of only a slave addressI2cHeaderprotocol layout with:2. Host-side support
host::protocol::Handler::PacketBuilderwith I2C write/read/read-result helpersagent/detail/i2c0_common.hppagent::CBoard3.
c_boardfirmware transportfirmware/c_board/app/src/i2c/module4. USB downlink/uplink integration
5. CubeMX / BSP integration
I2C2peripheral filesI2C2rmcs_slave.ioc) accordinglyDesign Notes
I2C0in shared/host APIs.c_board, that logical channel is implemented by STM32I2C2.Scope
This PR is focused on
c_boardI2C transport support and the shared protocol/host pieces required for it.It does not introduce a new public multi-I2C board model yet; it adds the first concrete
c_boardI2C transport path on top of the shared protocol.I2C2传输支持
本PR为c_board添加了端到端的I2C传输支持，跨越共享协议、主机端、固件和USB/CubeMX集成四个领域。
协议层
新增I2C协议定义 (
core/include/librmcs/protocol/i2c.hpp)数据接口扩展 (
core/include/librmcs/data/datas.hpp)DataId::kI2c0 = 16标识符I2cDataView（从地址、数据载荷、可选寄存器）、I2cReadConfigView（从地址、读长度、可选寄存器）、I2cErrorView（从地址、数据长度、可选寄存器、读写标志）DataCallback接口新增两个虚函数：i2c_receive_callback()和i2c_error_callback()协议编码 (
core/src/protocol/protocol.hpp,serializer.hpp,deserializer.cpp)I2cHeader位域结构体，包含载荷类型选择器、寄存器标志、错误标志、7比特从地址和9比特数据长度kWrite)、读请求(kReadRequest)、读结果(kReadResult)、错误(kError)write_i2c_write()、write_i2c_read_config()、write_i2c_read_result()、write_i2c_error()（两个重载）process_i2c_field()处理I2C帧解析与验证，包括从地址范围检查、长度验证、标志组合验证等主机端
代理接口 (
host/include/librmcs/agent/c_board.hpp)CBoard现继承自detail::SingleI2c0DataCallback替代直接继承DataCallbackPacketBuilder继承自detail::I2c0PacketBuilderMixin<PacketBuilder>I2C0工具库 (
host/include/librmcs/agent/detail/i2c0_common.hpp)SingleI2c0DataCallback实现，针对DataId::kI2c0路由I2C数据和错误回调I2c0PacketBuilderMixin<Derived>提供i2c0_write()和i2c0_read()方法，含参数验证（从地址≤0x7F，数据长度验证）协议处理器 (
host/include/librmcs/protocol/handler.hpp,src/protocol/handler.cpp)PacketBuilder新增三个公共方法：write_i2c()、write_i2c_read_config()、write_i2c_read_result()process_result_strict辅助函数改进错误处理（kBadAlloc现视为失败）c_board固件
I2C驱动 (
firmware/c_board/app/src/i2c/i2c.hpp和i2c.cpp)I2c类实现DMA驱动的排队写/读传输，单例i2c0绑定到STM32 I2C2handle_downlink_write()、handle_downlink_read_config()）tx_complete_callback()、rx_complete_callback()）、错误处理和恢复HAL_I2C_MasterTxCpltCallback等）应用集成 (
firmware/c_board/app/src/app.cpp)USB处理 (
firmware/c_board/app/src/usb/vendor.hpp,interrupt_safe_buffer.hpp)i2c::i2c0try_unlock_and_clear()更名为try_unlock()，移除清除行为以防丢弃队列中的上链消息硬件和BSP配置
CubeMX配置 (
firmware/c_board/bsp/cubemx/)HAL_I2C_MODULE_ENABLED）MX_I2C2_Init()）、MSP初始化/反初始化GPIO和DMA
设计说明：协议层I2C通道标识为I2C0，在c_board硬件上映射到STM32 I2C2外设。