search

oxidecomputer / hubris

A lightweight, memory-protected, message-passing kernel for deeply embedded systems.
Mozilla Public License 2.0
2.92k stars 165 forks source link

stm32h7 qspi: reading out of range sectors appears to hang driver #848

Open cbiffle opened 1 year ago

cbiffle commented 1 year ago

If the QSPI driver / auxflash task think they have a much larger chip than they actually have, they may issue out-of-range requests, which appears to stall the driver.

The bringup team patched the size and moved on, so I don't have a lot of diagnostic information to include here, but the driver winds up hanging in this loop waiting for interrupts here: https://github.com/oxidecomputer/hubris/blob/2df5f323963423b322d1a42a02f1cc0cc433ca52/drv/stm32h7-qspi/src/lib.rs#L286

That commit should repro the issue on a stock B-rev Sidecar with an erased auxflash, if anyone wants to dig into it.

cbiffle commented 1 year ago

FWIW, if someone decides to dig into this, my suspicion is that the read loop needs to be checking the TEF bit in SR instead of only watching for the "happy condition" bits. So I think currently in the event of a transfer error we've got an unconditional infinite loop.