5. Interrupt and exception
   - System calls
     - Why do we need the system call?
	- Allow the kernel to expose certain key pieces of functionality to user programs
     - What kinds of work that the system call can do?
	- Access the file system (open(), write(), read())
	- Destroy processes (kill())
	- Allocate more memory
     - How does the system call issue a signal to the kernel space?
	- Trap
	- System call table
	  - e.g. fwrite("hello", 5, 1, f) -> test.c
	       write() -> glibc
	       sys_write() -> kernel space
	  - A system call table contains and the system call number and types
	    e.g. 0 sys_read, 1 sys_write ...
     - What are the orders of a system call ?
	- Program puts syscall parameters in registers
	- Program executes a trap
	- Processor state (PC, PSW (program status word)) pushed on stack
	  - PSW is used to describe the condition of a processor at each instant
	  - e.g. use to switch from the user to the kernel space
	- CPU switches to kernel mode
	- Trap handler uses parameters to jump to desired handler
	- When a system call completes its work, reserve operation
	  - place return code in register
	  - return from exception
	- See system call slide 10
     - System call number
	- used to distinguish between system calls
	- See syscall.h
     - A prototype of a typical system call
	- int system_call (resource_descriptor, parameters)
	  - resource_descriptor: file, device ..., means the current process if not specified
	  - return value indicates the completion status of the system call
	  - sometimes return value means the number of bytes written to file
	  - How do pass parameters of the system call to the kernel space ?
     - Passing parameters in system calls
	- Typical method
	  - pass by registers (Linux)
	    - % eax (sys call number), %ebx, %ecx, %esi, %edi, %ebp
	    - e.g. mov x, %eax
		   INT 64
	  - pass via user mode stack (xv6)
	    - use user stack and trapframe, see slide 18
	  - pass via a designated memory region
     - Trap instruction
	- jumps into the kernel and raises the privilege level to kernel mode
	- use return-from-trap instruction to return the user space when finished
	- How does the trap know which code to run inside the OS?
	  - Using the trap table
	  - When does the OS kernel create the trap table?
	    - at boot time
	- When will we use trap?
	  - invalid memory access
	  - float point exception ...
     - wait()
	- delay the parent's process until the child finishes executing
	- How to implement wait() in the OS kernel?
	  - see slide 32
     - exit()
	- exit all other process except "init", the first process
	- see slide 34
   - Interrupt
     - An event that alters the sequence of instructions executed by a processor
     - Raised by hardware or programs to get OS attention
     - Hardware interrupt: 
	- a device (programmable interrupt controller (PIC)) asserts a pin in the CPU
	- When will we use hardware interrupt ?
	  - the I/O device signals to the CPU that it wants to be serviced
	  - .e.g tell CPU that a keypad has been pressed
	- 2 pins on the CPU
	  - INT (interrupt) - maskable interrupt
	  - NMI - non maskable interrupt for very critical signal
     - Software interrupt: INT x, an executed instruction causes an interrupt
     - Synchronous interrupt
	- The control unit issues interrupts only after terminating the execution of an instruction
     - Asynchronous interrupt
	- Generated by hardware devices at arbitrary time 
     - What happens when there is an interript?
	- Basic program state saved
	  - save SS, ESP, EFLAGS, CS, EIP ..
	  - CPU suspends current task
	- Jump to interrupt handler
	- Interrupt handler (top half) -- software
	  - Respond to interrupt
	  - storing more program state
	  - schedule bottom half
	  - IRET (interrupt return)
	- Return from interrupt -- CPU
	  - Restore flags and registers saved eariler
	  - Restore running task
	- Interrupt handler (bottom half) -- software
     - Interrupt vectors
	- Each interrupt/exception provides a number
	- The index of an interrupt descriptor table (IDT)
	- IDT provides the entry point into an interrupt/exception handler
	- 0 to 255 vectors possible
	  - 0 - 31: exception and nonmaskable interrupts
	  - 32- 47: maskable interrupts caused by IRQs
	  - 48 - 255: identify software interrupt
	  - e.g. Linux uses 128 (0x80) vector to implement system calls
     - interrupt descriptor table (IDT)
	- store in memory
	- initialized by OS at boot time
     - small interrupt handlers
	- Top and bottom half technique (Linux)
	- Top half
	  - Do minimum work and return from interrupt handler
	- Bottom half
	  - deferred processing
	  - implemented in Linux as softirqs, tasklets or workqueues
   - Exceptions
     - Due to illegal operations 
     - Generated when the CPU detects an anomalous condition while executing an instruction
     - Exception sources
	- program-error exceptions
	  - divide by zero
	- Software generation exceptions
	  - INT 3: a break point exception
	  - INT 0: overflow instruction
	- Machine-check exceptions
	  - Hardware error such as system bus error