probe-rs on LPC-Link2 CMSIS-DAP v1.0.7: series of upstream quirks to patch through #930

Tracks the ongoing "Path A" work identified during #921 investigation: patching upstream probe-rs until it can drive a full SWD deploy against the LPC-Link2 CMSIS-DAP v1.0.7 firmware that ships on the NXP LPC845-BRK.

Context

probe-rs is dramatically better than pyOCD for our use case:

• Doesn't hang: probe-rs list returns instantly against v1.0.7. pyocd list hangs indefinitely inside hidapi_backend.py::read() (see fbuild deploy: own nxplpc/LPC8xx flash end-to-end (FastLED is removing its direct pyocd path) #921 for full pyOCD failure trace).
• Doesn't wedge the CDC sibling: verified — [System.IO.Ports.SerialPort]::Open('COM10') succeeds immediately after probe-rs touched the HID interface. pyOCD leaves COM10 in Device Manager error 31 requiring a physical USB replug.
• Actively maintained Rust project, statically-linked binary via cargo install probe-rs-tools.

BUT: the LPC-Link2 v1.0.7 firmware (2013 vintage) is so old that it fails to implement large chunks of the CMSIS-DAP protocol probe-rs relies on. Making it work end-to-end requires a series of upstream patches, each turning a hard "timeout error" into a "log fallback, continue with spec-default."

Repro environment

• Windows 10, LPC845-BRK on COM10
• Debugger: LPC-Link2 running factory CMSIS-DAP v1.0.7 (USB\VID_1FC9&PID_0132\0B03400A)
• probe-rs 0.31.0 (installed via cargo install probe-rs-tools --locked)

Verified by: probe-rs list → returns instantly with LPC11U3x CMSIS-DAP v1.0.7 -- 1fc9:0132-0:0B03400A (CMSIS-DAP).

Blockers (running list)

Each item below fires a specific Timeout in USB communication from probe-rs against the v1.0.7 firmware. The fix pattern is uniform: log-and-fall-back to the CMSIS-DAP v1 HID spec default rather than error out.

1. find_packet_size (DAP_Info sub-command 0xFF) — DRAFT PATCH READY

Symptom: 16 attempts to DAP_Info(PacketSize=0xFF) all time out at 50 ms; probe-rs returns CmsisDapError::NoPacketSize.

Fix: on all-attempts-timeout, fall back to the spec-mandated 64-byte HID report size (CMSIS-DAP 2.1.0 §Interface Configuration → HID Interface — every conforming probe MUST accept 64-byte reports). Log a warn! so the fallback is visible.

Status: applied locally at probe-rs/src/probe/cmsisdap/commands/mod.rs::find_packet_size and verified — probe-rs now moves past the packet-size query. COM10 still opens cleanly after. Ready to upstream to probe-rs as a small PR (targeting probe-rs #973 which tracks the same class of bug against a different v1.x firmware).

2. DAP_Info(Vendor) sub-command 0x01 — NEXT

Symptom: after fix #1, probe-rs fails at Error handling CMSIS-DAP command Info. Timeout in USB communication. at the vendor-string query.

Fix: same pattern — treat a timeout as "returns empty string" (which is what a compliant probe with no vendor field would return anyway).

Status: untested but likely identical fix path.

3. DAP_Info(Product) sub-command 0x02 — LIKELY

Same pattern as #2.

4. DAP_Info(SerialNumber) sub-command 0x03 — LIKELY

Same pattern as #2.

5. DAP_Info(Capabilities) sub-command 0xF0 — CRITICAL

Symptom (predicted): timeout on capability query.

Fix: probe-rs picks its SWD/JTAG strategy from this response. A safe fallback is "SWD only, no JTAG, no SWO trace" — the minimum viable capability set for LPC845 flashing.

Status: NOT YET REACHED. This will require careful decision — the default has downstream effects on which transfer mode probe-rs uses.

6+. Post-init SWD transfer quirks — DISCOVERY LOOP

Once init passes, actual SWD read/write transfers may hit their own timing / packet-format quirks specific to this old firmware. Each will surface incrementally as we run probe-rs info --chip LPC845M301JBD48 end-to-end.

Packaging (after all blockers close)

Once probe-rs works end-to-end against v1.0.7:

• Cross-compile the patched build for 6 targets: x86_64-pc-windows-msvc, aarch64-pc-windows-msvc, x86_64-apple-darwin, aarch64-apple-darwin, x86_64-unknown-linux-gnu, aarch64-unknown-linux-gnu.
• Cosmopolitan-libc "actually portable executable" is NOT viable — cosmocc is a C-only toolchain and probe-rs' HID/USB stack depends on platform-native APIs (Windows SetupAPI, Linux libusb, macOS IOKit) that don't exist in Cosmopolitan's runtime. Real-world universal binary would require running Rust against APE, which no production Rust target supports as of 2026.
• Host binaries alongside the existing LPC-Link2 debugger reflash bundle at FastLED/framework-arduino-lpc8xx tools/lpc-link2-debugger/ or a sibling tools/probe-rs/ directory.
• Wire fbuild's LpcDeployer to prefer probe-rs over lpc21isp when a CMSIS-DAP probe is enumerated on the target port — probe-rs would then flash via SWD without any ISP button press even on the factory firmware.

Alternative — Path B (dramatically less work)

Reflashing the on-board LPC-Link2 to CMSIS-DAP V2 (via the DFU jumper + dfu-util --download flow already staged in FastLED/framework-arduino-lpc8xx#37) makes ALL of the above obsolete: once V2 is running, probe-rs works out of the box AND lpc21isp -control auto-enters ISP. That path requires one physical JP1 short from the user + ~10 s of tool time; this Path A avoids that hardware step at the cost of maintaining a probe-rs fork indefinitely.

This issue is the checkbox tracker for whichever path we ship first. Path B lives in #921's remaining requirements list.

Deliverables

• Blocker Esp32-s3 PSRAM use #1: find_packet_size fallback — draft patch ready, needs upstream PR to probe-rs/probe-rs
• Blocker cdc on boot warning #2: DAP_Info(Vendor) timeout fallback
• Blocker Fix ESP32-S3 PSRAM board configuration in board JSON database #3: DAP_Info(Product) timeout fallback
• Blocker Add CDC on boot warning to ESP32 builds #4: DAP_Info(SerialNumber) timeout fallback
• Blocker Add VSCode extension scaffold (vscode-fbuild/) #5: DAP_Info(Capabilities) timeout fallback + minimal-capabilities default
• Blocker 6+: whatever surfaces in actual SWD transfers
• Cross-compile patched probe-rs for 6 targets
• Host binaries under a FastLED-org repo
• Wire fbuild's LpcDeployer to prefer probe-rs when a CMSIS-DAP probe is enumerated
• Regression test suite: exercises the patched build against the v1.0.7 firmware
• End-to-end test: fbuild deploy tests/platform/lpc845brk -e lpc845brk --port COM10 completes without a button press against the factory firmware

Related

• fbuild deploy: own nxplpc/LPC8xx flash end-to-end (FastLED is removing its direct pyocd path) #921 — LPC845-BRK deploy end-to-end (the parent tracking issue)
• LPC845-BRK: deploy must always unwedge the shared VCOM (pyocd/SWD flash leaves CDC in Windows error 31) #565 — VCOM unwedge acceptance (this path also satisfies it)
• Could not determine a suitable packet size for hs-probe and nrf52840 probe-rs/probe-rs#973 — upstream tracking issue for the same class of packet-size negotiation bug against a different v1.x firmware
• feat(tools): add LPC-Link2 debugger reflash bundle (fbuild#921) framework-arduino-lpc8xx#37 — the (Path B) LPC-Link2 debugger reflash bundle already merged and available