GitLab

Summary

The current software implementation of the recovery might fail because of the semaphore being occupied by previously running applications that failed, for which the semaphore is not released.

In fact, using the old legacy recovery procedure with manual mpoke of the registers to load the oh firmware worked fine. In the development case, we had to cancel the file /dev/shm/sem.memhub.gem in the stuck amc.

What is the expected correct behavior?

the recovery should work.