Chapter 5.4 - Interruptible TSM with no re-entrancy - TSM enforcement

[eckhard-delfs-qualcomm]

Text states "The TSM-driver supports a TEERESUME ECALL which enables the TSM to enforce that the resumption of the interrupted TSM security routine is initiated by the OS/VMM on the same hart."

The TSM is scheduled to run by the VMM, so how can the TSM "enforce" that it is running on a specific hart, considering that it cannot "schedule itself"?
More in general, forcing the resume to happen on the same hart is too restrictive.
In practice the processing of the interrupt in the VM might cause the calling thread to migrate to a different core/hart, which is where it will initiate the resume from. (this would be typically the case when the interrupt was a timer for the VM OS scheduler and the VM is running on multiple cores/harts).