Clarification regarding validity of concurrent host-device access of mapped memory

[physics-enthusiast]

The documentation for vkMapMemory states that

vkMapMemory does not check whether the device memory is currently in use before returning the host-accessible pointer. The application must guarantee that any previously submitted command that writes to this range has completed before the host reads from or writes to that range, and that any previously submitted command that reads from that range has completed before the host writes to that region

Does the "previously" in "previously submitted command" mean "before the call to vkMapMemory" or "before any read/write"?
i.e. Is concurrent host-device access to a mapped memory range considered to be undefined behaviour, or merely "externally synchronized" (valid but with no guarantees on visibility/availability)?

[spencer-lunarg]

The application must guarantee that any previously submitted command that writes to this range has completed before the host reads from or writes to that range

Is saying any calls to vkQueueSubmit trying to write to the memory must be complete before doing a CPU reads or write to the host memory

that any previously submitted command that reads from that range has completed before the host writes to that region

Is saying any calls to vkQueueSubmit trying to read to the memory must be complete before doing a CPU writes to the host memory