11.

I/O Systems
11.1 Basic Issues in Device Management
11.2 A Hierarchical Model
11.3 I/O Devices
11.4 Device Drivers
– Memory-Mapped vs Explicit Device Interfaces
– Programmed I/O with Polling
– Programmed I/O with Interrupts
– Direct Memory Access (DMA)
11.5 Device Management
– Buffering
– Error Handling
– Disk Scheduling
– Device Sharing
11.6 The Abstract I/O Interface
ICS 143 1
 Basic issues
• I/O devices:
– communication devices
• input only (keyboard, mouse, joystick)
• output only (printer, display)
• input/output (network card)
– storage devices
• input/output (disk, tape)
• input only (CD-ROM)

ICS 143 2
 Basic issues
• main tasks of I/O system:
– present logical (abstract) view of devices
• hide details of hardware interface
• hide error handling
– facilitate efficient use
• overlap CPU and I/O
– support sharing of devices
• protection when device is shared (disk)
• scheduling when exclusive access needed (printer)

ICS 143 3
 A hierarchical model of I/O
Figure 11-1
• abstract I/O interface
– block devices, character devices, network
• device-independent software
– buffering, scheduling, caching
• device-dependent software
– I/O drivers (supplied by device manufacturer)

ICS 143 4
 I/O devices
• display monitors
– character or graphics oriented
Figure 11-2

– different data rates:

• 25 x 80 characters vs 800 x 600 x 256
• 30-60 times/sec

ICS 143 5
 I/O devices
• keyboards
– most common: QWERTY
– very low data rate (<10 char/sec)

• pointing devices
– mouse (optical, optical-mechanical)
– trackball
– joystick
– low data rate (hundreds of bytes/sec)

ICS 143 6
 I/O devices
• printers
– line printers, dot-matrix, ink-jet, laser
– low data rates
– character-oriented
• scanners
– digitize picture into bit map (similar to video
RAM)
– low data rates

ICS 143 7
 I/O devices
• floppy disks
– surface, tracks/surface, sectors/track,
bytes/sector
– all sectors numbered sequentially 0..(n-1):
logical disk

Figure 11-3 (a,b)

ICS 143 8
 I/O devices
• floppy disks
– track skew
• account for seek to next track to minimize latency
Figure 11-3 (c)
– double-sided floppy
• tracks with same diameter: cylinder
• number all sector within cylinder consecutively to
minimize seek
Figure 11-3 (d)

ICS 143 9
 I/O devices
• hard disks
– multiple surfaces
Figure 11-4
– higher densities and data rates than floppy
floppy hard disk
bytes/sec 512 512-4096
sec/track 9, 15, 18, 36 100-400
tracks/surf 40, 80, 160 1000-10,000
# surf 1-2 2-24
seek 30-100 ms 5-12 ms
rotation 400-700 rpm 3600-10,000 rpm

ICS 143 10
 I/O devices
• optical disks
– CD-ROM, CD-R (WORM), CD-RW
– originally designed for music
– data stored as continuous spiral, subdivided into
sectors
– constant linear speed (200-530 rpm)
– higher storage capacity than magnetic disks:
0.66 GB/surface

ICS 143 11
 I/O devices
• data transfer rates of disks
– sustained: continuous data delivery
– peek: transfer once r/w head is in place
• depends on rotation speed and data density
• 1 revolution passes over all sectors of 1 track
– Example: 7200 rpm, 100 sect/track, 512 B/sect
• 7200 rpm: 60,000/7200=8.3 ms/rev
• 8.3/100 = 0.083 ms/sector
• 512 bytes transferred in 0.083 ms: ~6MB/s

ICS 143 12
 I/O devices
• magnetic tapes (reel or cartridge)
– large storage capacity (GB)
– data transfer rate: ~ 2 MB/sec
• networks (interface card)
– Ethernet, token ring, slotted ring
• controller implements protocol to accept, transmit,
receive packets
– modem
• convert between analog and digital (phone lines)
• character-oriented (like printer and keyboard)

ICS 143 13
 Device drivers
• accept command from application
– get/put char, read/write block, send/rec. packet
• interact with device controller (HW) to
carry out command
• typical device controller interface: set of
registers
Figure 11-6
• Example: serial or parallel port on PC
– generic driver reads/writes chars to registers
ICS 143 14
 Device drivers
• memory-mapped vs explicit dev interface
– similar idea to memory-mapped files
Figure 11-7
– explicit: special I/O instruction:
io_store cpu_reg, dev_no, dev_reg
– memory-mapped: regular CPU instruction:
store cpu_reg, n
(n is a memory address)

ICS 143 15
 Programmed I/O with polling
• CPU is responsible for
– moving every char to/from controller buffer
– detecting when I/O operation completed
• protocol to input a character:
Figure 11-8

ICS 143 16
 Programmed I/O with polling
• driver operation to input sequence of chars
i = 0;
do { write_reg(opcode, read);
while (busy_flag == true) {…};
mm_in_area[i] = data_buffer;
increment i;
compute;
} while (…)

• what to do while waiting?

– idle (busy wait)
– some other computation
• how frequently to poll? -- Figure 11-9
– give up CPU
• device may remain unused for a long time
ICS 143 17
 Programmed I/O with interrupts
• CPU is responsible for
– moving chars to/from controller buffer, but
• interrupt signal informs CPU when I/O
operation completes
• protocol to input a character:
Figure 11-10

ICS 143 18
 Programmed I/O with interrupts
• compare polling with interrupts:
i = 0;
do { write_reg(opcode, read);
while (busy_flag == true) {…};
mm_in_area[i] = data_buffer;
increment i;
compute;
} while (…)
i = 0;
do { write_reg(opcode, read);
block to wait for interrupt;
mm_in_area[i] = data_buffer;
increment i;
compute;
} while (…)
ICS 143 19
 Programmed I/O with interrupts
• Example: keyboard driver
i = 0;
do { block to wait for interrupt;
mm_in_area[i] = data_buffer;
increment i;
compute(mm_in_area[]);
} while (data_buffer != ENTER)

• timing of interrupt-driven I/O

Figure 11-11
– more OS overhead but better device utilization

ICS 143 20
 DMA
• CPU only initiates operation
• DMA controller transfers data directly
to/from main memory
• interrupt when transfer completed
• protocol to input data using DMA:
Figure 11-12

ICS 143 21
 DMA
• driver operation to input sequence of chars
write_reg(mm_buf, m);
write_reg(count, n);
write_reg(opcode, read);
block to wait for interrupt;
– writing opcode triggers DMA controller
– DMA controller issues interrupt after n chars in
memory
• I/O processor (channel)
– extended DMA controller
– executes I/O program in own memory

ICS 143 22
 Device management
• device-independent techniques
• reasons for buffering
– allows asynchronous operation of producers
and consumers
– allows different granularities of data
– consumer or producer can be swapped out
while waiting for buffer fill/empty

ICS 143 23
 Device management
• single buffer operation
Figure 11-14 (a)
• double buffer (buffer swapping)
Figure 11-14 (b)
– increases overlap
– ideal when:
time to fill = time to empty = constant
– when times differ, benefits diminish

ICS 143 24
 Device management
• circular buffer
– when average times to fill and empty are
comparable but vary over time: circular buffer
absorbs bursts
– producer and consumer each use an index
• nextin gives position of next input
• nextout gives position of next output
• both are incremented modulo n at end of operation

ICS 143 25
 Device management
• error handling
– persistent vs transient; SW vs HW
– persistent SW error
• repair/reinstall program
– other errors: build in defense mechanisms
– Examples:
• transient SW errors: error correcting codes,
retransmission
• transient HW errors: retry disk seek/read/write
• persistent HW errors: redundancy in storage media

ICS 143 26
 Device management
• bad block detection and handling
– block may be defective as a manufacturing fault
or during use (a common problem)
– parity bit is used to detect faulty block
– the controller bypasses faulty block by
renumbering
– a spare block is used instead
– two possible remappings:
Figure 11-17
• more work but contiguity of allocation preserved

ICS 143 27
 Device management
• stable storage
– some applications cannot tolerate any loss of
data (even temporarily)
– stable storage protocols:
• use 2 independent disks, A and B
• write: write to A; if successful, write to B
• read: read from A and B; if A!=B, go to recovery
• recovery from media failure: A or B contains correct
data; remap failed disk block
• recovery from crash: if before writing A, B is
correct; if after writing A, A is correct; recover

ICS 143 28
 Device management
• RAID (redundant array of independent
disks)
– increased performance through parallel access
– increased reliability through redundant data
– maintain exact replicas of all disks
Figure 11-19(a)
• most reliable but wasteful
– maintain only partial recovery information (e.g.
error correcting codes)
Figure 11-19(b)
ICS 143 29
 Device management
• disk scheduling
– minimize seek time and rotational delay
– requests from different processes arrive
concurrently:
• scheduler must attempt to preserve locality
– rotational delay:
• order requests to blocks on each track in the
direction of rotation: access in one rotation
• proceed with next track on same cylinder

ICS 143 30
 Device management
• minimizing seek time: more difficult
– r/w arm can move in two directions
– minimize total travel distance
– guarantee fairness
– FIFO: simple, fair, but inefficient
Figure 11-20 (a)
– SSTF: most efficient but prone to starvation
Figure 11-20 (b)
– Scan: fair, acceptable performance
Figure 11-20 (c)

ICS 143 31