Answer - This is due to the reason that a context switch implementation is done by the kernel. During this process the state information is copied between the processor and the PCB (process control block) or the TCB (thread control block). Since the kernel does not know anything about user-level threads, technically it is not possible for it to be a user level thread context switch. The user level scheduler can do some limited state copying on the behalf of a thread prior to the control being handed to that thread. But this copying of state information is smaller compared to that of a kernel-level process. Also the process does not involve going into the kernel mode with the help of a system call.