290 likes | 305 Views
Input and Output. Programmed IO Interrupts and Exception Handling 0 – Programmed IO 1 – Memory mapped vs isolated 2 - The Exception Mechanism 3 - The Role of the Operating System 4 - The MIPS RISC Exception Mechanism 5 - A Sample Exception Handler 6 - Operating System Issues.
E N D
Input and Output Programmed IO Interrupts and Exception Handling 0 – Programmed IO 1 – Memory mapped vs isolated 2 - The Exception Mechanism 3 - The Role of the Operating System 4 - The MIPS RISC Exception Mechanism 5 - A Sample Exception Handler 6 - Operating System Issues
MAL Input Example • WaitLoop: lw $14, KeyboardStatus • andi $14, 0x1 • beqz $14, WaitLoop • lw $2, KeyboardData • Note: KeyboardStatus is 0xffff0000 • KeyboadData is 0xffff0004 • Bit 0 of KeyboardStatus is the ready bit 1 means ready, 0 means not ready. • Hardware changes the ready bit.
MAL Output Example • WaitLoop: lw $14, DisplayStatus • andi $14, 0x1 • beqz $14, WaitLoop • sw $2, DisplayData • Note DisplayStatus is 0xffff0008 • DisplayData is 0xffff000c • Bit 0 of DisplayStatus is the ready bit 1 means ready, 0 means not ready. • Hardware changes the ready bit.
Programmed I/O • The previous examples are known as programmed I/O. This means that I/O is initiated by the CPU which checks status bits periodically. • This wastes CPU time while it waits for the device to become ready. (spinwaiting) • CPU can miss characters that are typed when the program is doing something else. • Two other types of I/O control we will look at are interrupt I/O and DMA (Direct Memory Access).
Kbd Echo • Reads the keyboard input register, then echos the input to the display
Interrupts CPU executes Interrupt Handler 2 Interrupt Handler Some code Jr $27 to return to Main 3 Program Main PC When IRQ arrives, Interrupt Handler called. 1 Add a,b,c 4 Next instruction is executed after return from Handler Sub a,b,c
OS - Kernel • Responsible for all processes on the machine • Is a protected resource • User programs must make requests to Kernel. • This includes interrupt and exception handling.
MIPS RISC Exception Mechanism 0x80000080 • Two modes of operation: User or Kernel. • Upper half of memory reserved for OS. • 0x8000 0000 .. 0xFFFF FFFF • Exception Handler starts at 0x8000 0080 OS 0x80000000 Stack Segment 0x7fffffff Data Segment Test Segment 0x400000 Reserved
MIPS RISC Exception Mechanism • Only special procedures are allowed access to kernel • User’s are not allowed to modify code in the kernel. • Nor, access Interrupt handler. • System with lots of power. • Kernel memory and User Memory are distinct. • Security rights prevent users from accessing kernel memory.
Asynchronous Exceptions (Interrupts) • Generated by h/w devices. • Interrupts, can appear at any time • State of system is not known in advance. • We cannot count on any general purpose register being free. • 2 registers $k0, $k1 ($26, $27) are kernel reserved to cleanly transition to and from an interrupt. • This prevents clobbering information stored in user program registers.
Synchronous Exceptions (Traps) • Traps are exceptions generated by software. • Thus are Predictable • Handed by the OS. • Device drivers • Buffer flush • Service routines • VM Page Fault • Fatal coding errors • Divide by Zero • Overflow • Memory violation. User program tries to access kernel memory, or another user programs memory.
Processors • Main Processor (CPU) • - 32 general registers $0 .. $31 • - LO & HI for multiply & divide • Co-processor 1(FPU) • Physically there are 32 32-bit registers • - 16 floating point registers $f0, $f2, .. $f30 • Co-processor 0 - BadVAddr, Cause, Status, Exception Program Counter (EPC) registers
Co-processor 0 • $8 - Bad Virtual Address • $12 - Status Register • $13 - Cause Register • $14 - Exception Program Counter • mfc0 Rdest, C0src move from co-processor 0 • mtc0 Rsrc, C0dest move to co-processor 0
31 - 16 15 14 13 12 11 10 9 8 7 6 - 2 1 0 . . . IP(7) IP(6) IP(5) IP(4) IP(3) IP(2) IP(1) IP(0) 0 ExcCode 0 0 Cause Register
Cause RegisterInterrupt Pending • IP - Interrupt Pending bits • Six interrupts in bits 10 .. 15 • IP(2) .. IP(7) • Two simulated interrupts in bits 9 .. 8, s/w • IP(0) .. IP(1) • By examining the bits software can determine which device is requesting service. • 1 = interrupt has occurred but not yet serviced
Status RegisterKernel/User Mode • Bit subscript meaning • 1 c current • 3 p previous • 5 o old • KU - kernel or user mode. bits 1,3,5 • 0 = kernel, 1 = user. • Kernel must return it to the state it was in before the exception occurred. • c p o bits act as a 3 layer h/w stack.
Status Register Interrupt Enable • Bit subscript meaning • 0 c current • 2 p previous • 4 o old • Interrupt Enable. bits 0,2,4 • 1 means interrupts are allowed • c p o are a simple 3 layer h/w stack to store the state of the Interrupt enable
Status Register Interrupt Mask • IM - Interrupt Mask bits - Six interrupts in bits • 10 .. 15 OR IM(2) .. IM(7) • 2 simulated interrupts in bits • 9 .. 8 OR IM(0) .. IM(1) • By setting the bits software can specify which devices are enabled/disabled. • 1 = interrupts allowed, 0 = interrupts disabled
EPC Register ($14) • Holds return address for exception handler • Exceptions can come from anywhere • user program (syscall) • kernel (trap) • external (device) • When an exception is raised: • Must save the return address in EPC • Normally return address is done with JAL but it is not used when an exception occurs. • An exception can occur between JAL and the saving of $ra. Making the use of $ra problematic.
EPC Register ($14). • Example • Interrupt is received • PC is saved in EPC • PC is modified to point to the handler. • When done PC is reloaded with EPC.
Interrupt Process • Interrupt Occurs • Return address is automatically stored in EPC • Disable Interrupts (done by OS) • Save status of machine (create an AR) • Save current state of interrupts and KU mode • May enable interrupts at this point • Handler address is calculated by indexing the Jump table using ExcCode • Execute Handler • Return from exception. • Enable interrupts • Restore registers • $k0 is loaded with EPC • jr $k0
Returning from the Handler • - Register Values are restored. • - The value in the EPC, $14 of C0 is loaded into $k0 ($26) • - rfe instruction Restore From Exception restores the user mode and interrupt enable bits by "shifting" the 6 ls bits in the status register • - jump to the location in register $k0
Re-entrant and Non Re-entrant Handlers • Interrupts maybe disabled so an interrupt can not interrupt a handler. • These handlers then must execute to completion before another interrupt into the handler is allowed. Non re-entrant. • E.g. Kbd handler interrupted by the Kbd is bad. • key sequences can then arrive out of order. • However, kbd handler can be interrupted by a bad address handler. • Safe operation • Generally this is done on a case by case basis.
Disabling Interrupts • 1 - Set (make it 0) bit zero of the status register to disable all interrupts. • 2 - Set an individual bit or bits in the Interrupt Mask to disable one or more.
Priority Interrupts • Interrupts have priorities. • Lower interrupts, higher priority. • Typically the timer has highest. • Sends an interrupt at regular intervals so OS always has control. • Prevents processes from hogging the CPU. • Kbd is 2nd. • Always good for the user to have control. • Handlers are executed based on priority.
Interrupts on PCsprior to 2000 IRQ 0 - Timer IRQ 8 - Real Time Clock IRQ 1 - Keyboard IRQ 9 - Open (Redirected IRQ 2) IRQ 2 - Cascade IRQ 10 - Open (Triggers IRQs 8 to 15) IRQ 3 - Com 2, Com 4 IRQ 11 - Open IRQ 4 - Com 1, Com 3 IRQ 12 - Open IRQ 5 - Open IRQ 13 - Math Coprocessor (Lpt2:, if present) IRQ 6 - Floppy Disk IRQ 14 - Fixed Disk IRQ 7 - LPT1: IRQ 15 - Open