Race condition/memory corruption on CH32V307

[PoroCYon]

Operating System

Linux

Board

CH32V307 (using both the SCDZ devboard and a custom board)

Firmware

Provided at https://github.com/PoroCYon/tinyusb-bug-repro --- both a CH32V307 version, and an RP2040 version which does functions correctly.

What happened ?

On the CH32V307, using DFU, the following effects may be observed:

• The data buffer received by the DFU functions is 64 bytes off (can be in either direction)
  • This feels like a race condition thing, as it is very repeatable with CFG_TUSB_DEBUG=2 or ASAN enabled, while without it happens much more randomly
• Random bytes in the data buffer are corrupted (watch the CRC)
• Sometimes it hangs (due to a crash due to the memory corruption, I think)
  • With payloads larger than 512 bytes, crashes seem to almost always happen.

How to reproduce ?

Compile and load the provided firmware onto the chip. Then run dfu-util -D payload.py. Watch how the data in the hexdump changes, the CRC changes (for hard to spot single-byte errors), and how sometimes, hangs/crashes occur.

Debug Log as txt file (LOG/CFG_TUSB_DEBUG=2)

This gist contains three log files:

• log-vkart-rp2040.log: RP2040, working correctly
• log-vkart-ch32v307.log: CH32V307, showing erratic behavior, with CFG_TUSB_DEBUG=0
• log-vkart-ch32v307.2.log: CH32V307, with CFG_TUSB_DEBUG=2, showing wrong but repeatable output

Screenshots

No response

I have checked existing issues, dicussion and documentation

• I confirm I have checked existing issues, dicussion and documentation.

[PoroCYon]

(cc @gregdavill since he is the author of the driver/backend/port?)

[HiFiPhile]

There is a PR regarding ch32 issues, you can give a try: #2392

[PoroCYon]

#2392 indeed seems to fix the issues, thanks! (Should this issue be closed as a duplicate then?)

[HiFiPhile]

Glad it works, it can be closed once PR merged.