Hello Marc et al., I have discovered a rare but annoying crash in mcp251xfd. I don't know how to replicate this directly, and the occurrence rate is very low (on the order of once per 1e7 frames or so). I cannot be too specific on our hardware, but we are running ARMv8, and the CAN chip is the MCP2518FD. The kernel version is 6.1.21. The crash always happens immediately following a CRC read error. I've attached a sample kernel log of the crash (hopefully correctly - I've not interacted with mailing lists for a while). So far, the call trace is always the same. To be clear, this is not after every CRC read error; those are several orders of magnitude more frequent than this crash. I've done a tiny bit of digging myself, but I'm not familiar with the territory. I guess something in the SPI dev struct is being set to NULL, while an SPI transaction is ongoing - but it surely must be related to the CRC read error and the timestamp read. Thanks for reading :)
Attachment:
oops.log
Description: Binary data
