Assignment 5: Keyboard and Simple Shell

Written by Philip Levis, updated by Julie Zelenski

Due: Tuesday, February 23 at 11:59 pm PT

$ make run -p build/uart_shell.bin
Found serial device: /dev/cu.SLAB_USBtoUART
Sending `uart_shell.bin` (22128 bytes): .........................................................................
Successfully sent!

Welcome to the CS107E reference shell.
Remember to type on your PS/2 keyboard!
Pi> help
   help:     <[cmd]> print a list of commands or description of cmd
   echo:     <...> echo user input to the screen
   reboot:   reboot the Raspberry Pi back to the bootloader
   peek:     [address] print contents of memory at address
   poke:     [address] [value] store value at address
Pi> help reboot
   reboot:   reboot the Raspberry Pi back to the bootloader
Pi> echo cs107e rocks my sox
cs107e rocks my sox 
Pi> echo I can use poke to turn on the green ACT LED!     
I can use poke to turn on the green ACT LED! 
Pi> poke 0x20200020 0x8000


For this week’s assignment, you will implement a PS/2 keyboard driver and implement a simple command-line shell. You will then be able to type commands and execute them on your Pi. Neat!

In completing this assignment you will have:

  • written code that interfaces with an input device. When you next download a device driver, you will think "I have a decent idea how that code operates"
  • seen the design of a complex interface into hierarchical levels and appreciated its benefits
  • implemented a simple command-line interpreter
  • explored the use of C function pointers for callbacks and command dispatch

This is a fun assignment, and brings us back to using physical devices and making them do cool things. These additions are the first steps toward turning your humble little Raspberry Pi into a standalone personal computer.

Get starter files

To ensure that your courseware files are up to date, do a pull in the repo.

$ cd ~/cs107e_home/
$ git pull

Change to your local assignments repo, switch to the dev branch, and update your repo with the starter code by pulling from the remote repo:

$ cd ~/cs107e_home/assignments
$ git checkout dev
$ git pull --allow-unrelated-histories starter-code assign5-starter

The final setup step is to change the symbolic link to point Makefile to the assign5/makefile.

$ cd ~/cs107e_home/assignments
$ ln -sf makefiles/assign5.makefile Makefile

You can edit the MY_MODULES list in the assign5.makefile to choose which modules of yours to build on. (See instructions for use of MY_MODULES in assignment 3.)

The source files added to your repo for assignment 5 are:

  • src/lib/ps2.c
    • library module for ps2 device
  • src/lib/keyboard.c
    • library module for keyboard
  • src/lib/shell.c
    • library module for shell
  • src/tests/test_keyboard_shell.c
    • test program with your unit tests
  • src/apps/uart_shell.c
    • application program that runs your shell, reading input from the PS2/ keyboard and printing output to the uart. You will use this program unchanged.

In this rest of this writeup, we refer to these files by their basename, e.g. just shell.c or test_keyboard_shell.c, but keep in mind each file is located in its respective subdirectory depending on the type of source file (application, library, or test).

The make run target builds and runs the sample application build/uart_shell.bin. You can use this target to test the full integration of your ps2, keyboard, and shell modules. The make test target builds and run the test program build/test_keyboard_shell.bin. This test program is where you will add all of your unit tests. You will make heavy use of this target throughout your development. You can run the debugger in simulator mode on the corresponding ELF file, e.g. arm-none-eabi-gdb build/test_keyboard_shell.elf.

Core functionality

PS2 Keyboard driver

1) Review interface (ps2.h and keyboard.h)

The PS2 keyboard driver is divided over the two modules ps2 and keyboard. Start by reviewing the header files (available in $CS107E/include or browse here). The functions of the ps2 module are:

  • ps2_device_t *ps2_new(unsigned int clock_gpio, unsigned int data_gpio)
  • unsigned char ps2_read(ps2_device_t *dev)

And keyboard functions:

  • void keyboard_init(unsigned int clock_gpio, unsigned int data_gpio)
  • unsigned char keyboard_read_next(void)
  • key_event_t keyboard_read_event(void)
  • key_action_t keyboard_read_sequence(void)
  • unsigned char keyboard_read_scancode(void)

The design of the keyboard driver is worth a pause to understand and appreciate. All of the nitty-gritty details of the protocol are encapsulated within the ps2 module. The keyboard module layers on that to implement the logic to process scancodes into sequences, events, and typed characters. Within the keyboard module, the functionality is arranged hierarchically, each routine building on the next. The bottom level routine reads a raw scancode (by delegating to the ps2 module), the next level gathers a sequence of scancodes into one logical key action, the higher level routines translate those actions into key events and typed characters.

The layered interface cleanly supports the needs of different clients. A client that simply wants typed characters might need only the top level keyboard_read_next; a client that reacts to up and down events also accesses the mid level keyboard_read_event.

The hierarchical design also eases the job of the implementor. Each level focuses on a discrete part of the operation and delegates tasks above and below to other functions. This makes each function simpler to implement and test. Your implementation plan of attack is to start at the bottom and work your way upward.

2) Read and validate scancodes (ps2_read)

In lab 5, you got a start on implementing ps2_read function. Copy that work into ps2.c now. The entire PS2 protocol rests on this cornerstone, so your first task is to finish it off and ensure it is reliable and robust.

A PS/2 scancode is an 11-bit packet organized in 8-odd-1 format. The first bit, the start bit, is low. The next 8 bits are data bits, least significant bit first. The following bit is a parity bit. The PS/2 protocol uses odd parity, which means that there should be an odd number of 1s among the data and parity bits. The 11th and final bit is the stop bit, which is always high.

PS/2 Packet Format

The return value from ps2_read is the 8 data bits from a well-formed packet. It is important to detect and recover from transmission errors. Check the value of each start, parity, and stop bit. If you detect an erroneous bit, discard the partial scancode and retry reading a new scancode from the beginning. Discard as many invalid tries as necessary until you receive a valid scancode.

In a similar vein, a dropped bit or discarded partial read could cause your driver to become de-synchronized. When that happens your driver can get stuck, trying to read a scancode byte starting mid-packet and waiting for further bits to arrive that aren't forthcoming. One way to resynchronize is using a simple timeout reset. Use your timer module to note if the current clock edge occurred more than 3ms after the previous one, and if so, reset the state and assume the current clock edge is for a start bit. This small effort provides additional robustness to combat flaky connections and hardware blips.

The struct ps2_device_t at the top of the ps2.c is used to track the state associated with a given PS2 device. The struct defined in the starter code has just two fields (clock and data gpios). If you encounter a need to track additional state for the device, simply add new fields to the struct.

Timing is everything! The timing of the PS/2 protocol has to be strictly followed. The keyboard sends the bits rapid-fire and you must catch each bit as it arrives. Once your driver sees the falling clock edge for the start bit, it needs to stay on task to read each subsequent bit. There is not time between clock edges to complete a call to a complex function like printf. Save any debug printing for after you read the entire sequence.

The first test in test_keyboard_shell.c will read and echo scancodes. Use this test to verify your basic scancode functionality.

3) Gather sequence into key action (keyboard_read_sequence)

The next level function keyboard_read_sequence gathers a sequence of scancodes into one logical key action. A key action is the press or release of a single key.

When you press a key, the PS/2 keyboard sends the scancode for that key. When you release the key, it sends a two-byte sequence: 0xF0 (the "break" code) followed by the key's scancode. For example, typing z will cause the keyboard to send 0x1A and releasing z will cause the keyboard to send 0xF0, 0x1A.

The press or release of an extended key sends a sequence with the extra byte 0xE0 inserted at the front. For example, pressing the right Control key sends the sequence 0xE0, 0x14 and releasing sends 0xE0, 0xF0, 0x14.

The keyboard_read_sequence function reads the sequence (1, 2 or 3 bytes depending on context) and translates it into a key_action_t struct which reports the type of action (press or release) and which key was involved.

Use the test in test_keyboard_shell.c that reads sequences to verify the operation of this function before moving on.

4) Process key events (keyboard_read_event)

There are additional type and constant definitions used starting at this level. Review keyboard.h for the definitions of keyboard_modifiers_t and key_event_t and ps2_keys.h for the definition of ps2_key_t and keycode constants.

The mid level routine keyboard_read_event processes key actions into key events. It calls keyboard_read_sequence to get a key_action_t and packages the action into a key_event_t struct which includes the state of keyboard modifiers and the PS/2 key that was acted upon.

The modifiers field of a key_event_t reports which modifier keys are in effect. The state of all modifier keys is compactly represented using a bit set. The keyboard_modifiers_t enumeration type designates a particular bit for each modifier key. If a bit is set in modifiers, this indicates the corresponding modifier key is currently held down or in the active state. If the bit is clear, it indicates the modifier is inactive.

The PS/2 protocol does not provide a way to ask the keyboard which modifiers are active, instead your driver must track the modifier state itself. A simple approach is a static variable in your keyboard module that you update in response to modifier key events. The Shift, Control, and Alt modifiers are active iff the modifier key is currently down. Caps Lock operates differently in that its setting is "sticky". A press makes Caps Lock active and that persists until a subsequent press inverts the state.

To identify which characters can be produced by a given key, we provide an array to use as a lookup table. Review the definition of the ps2_keys array in the source file $CS107E/src/ps2_keys.c The array is indexed by scancode. The array element at each index is a struct. For example, the A key generates scancode 0x1C. The array element ps2_keys[0x1c] holds the struct {'a', 'A'} which is the unmodified and modified character produced by this key.

Use the functions in test_keyboard_shell.c to verify your processing of key events before moving on.

5) Produce ASCII characters (keyboard_read_next)

You now have all of the pieces needed to implement the final top-level routine keyboard_read_next. This function calls keyboard_read_event to get the next key press event and produces the character corresponding to the key that was typed.

The return value is the ASCII character produced by an ordinary key or a value designated for a special key such as Escape or F9. The character produced by a key is determined by its ps2_key_t entry in the lookup table.

A ps2_key_t has two fields for each key, ch and other_ch, which correspond to the unmodified and modified character produced by the key. The A key produces { 'a', 'A' }. The Four key produces { '4', '$' }. Keys such as Tab that are unchanged by the modifier have the same character for ch and other_ch, e.g. {'\t', '\t'}.

Your keyboard should handle all keys shown in this keyboard diagram. (The number pad and movement keys are not shown and do not need to be handled by your keyboard driver.)

PS/2 Scancodes

Typing an ordinary key produces an ASCII character. The ordinary keys are:

  • All letters, digits, and punctuation keys
  • Whitespace keys (Space, Tab, Return)

Typing a special key produces its designated value. These values are greater than 0x90 to distinguish from ASCII values. The special keys are:

  • Escape
  • Function keys F1-F12
  • Backspace key (sometimes marked with ← or ⌫)

Press or release of a modifier key changes the event modifiers. No character or code is produced. The modifier keys are:

  • Shift, Caps Lock, Alt, Control

A change in modifiers can affect the character produced by future typed keys. The keyboard translation layer does not produce modified characters based on state of Alt or Control, only for Shift and Caps Lock. When the Shift modifier is active, other_ch is produced when typing a key that has an other_ch entry. If Caps Lock is active, other_ch is produced only for the alphabetic keys. Caps Lock has no effect on digits, punctuation, and other keys. If Shift and Caps Lock applied together, Shift "wins", e.g. other_ch is produced. (Caps Lock and Shift together do not invert letters to lowercase).

If you are using a Mac, Keyboard Viewer is a handy tool for visualizing the character produced for a given key combination. Try it out! If you are still unsure how to handle a particular case, experiment with our reference implementation of the keyboard using the test application from lab.

Simple shell

With a keyboard as input device, your Pi has gone interactive! The simple shell application allows the user to enter commands and control the Pi without needing to plug another computer into it.

The video below is a demonstration of our reference shell. The user is typing on a PS/2 keyboard connected to the Pi and the shell output is displaying over uart to a Mac laptop.

1) Review shell interface and starter code

A shell, such as bash or zsh, is a program that operates as a command-line interpreter. The program sits in a loop, reading a command typed by the user and then executing it.

The starter code for shell_run demonstrates the standard read-eval-print loop that is at the heart of an interpreter. Here it is in pseudocode:

loop forever
    display prompt
    read line of input from user
    evaluate command (parse and execute)

The public functions you will implement in shell are:

  • void shell_readline(char buf[], size_t bufsize)
  • int shell_evaluate(const char *line)

The shell_readline function reads a command typed by the user. The shell_evaluate executes that command. Review the documentation for these operations in the header file shell.h.

The client can configure the input and output for the shell. The shell_init function takes two function pointer arguments to be supplied by the client, one for input and one for output. Whenever the shell wants to read the next character entered by the user, it calls the client's input function. When it needs to display output, it calls the client's output function.

The apps/uart_shell.c application program initializes the shell with shell_init(keyboard_read_next, printf); this call configures the shell to read characters from the PS2 keyboard and send output to the serial uart interface.

The shell stores the client's input function pointer in the static variable shell_read and the output function into the variable shell_printf. Thereafter the shell calls shell_read whenever it needs to read input from the user and calls shell_printf whenever it needs to write output. This applies to all shell output, whether it be the shell prompt, displaying the result from a command, or responding with an error message to an invalid request.

When testing your shell in assignment 5, you will likely always use keyboard_read_next as the input function and printf as the output, but the flexibility you are building in now lays the groundwork for different choices in the future. In assignment 6, you'll write a console_printf function that draws to a HDMI monitor. With a change of one argument in the call to shell_init(keyboard_read_next, console_printf) your shell will display output on the graphical console instead of writing to the uart — nifty!

2) Read line

shell_readline calls shell_read to read a character entered by the user and gathers the line of input into a buffer. The user indicates the end of the line by typing Return (\n).

Handling backspace adds a slight twist. When the user types Backspace (\b), the shell deletes the last character typed on the current line. Removing it from the buffer is simple enough, but how to un-display it? If you output a backspace character, e.g. shell_printf("%c", '\b') , it moves the cursor backwards one position. If you back up, output a space, and then back up again, you will have effectively "erased" a character. (Wacky, but it works!)

There are two error conditions that shell_readline should detect:

  • disallow typing more characters than fit in the buffer
  • disallow backspacing through the shell prompt or to previous line

Reject the attempt and call the provided shell_bell function to get an audio/visual beep.

shell_readline is a great way to exercise your shiny new keyboard driver!

3) Parse command line

shell_evaluate first takes the line entered by the user and parses it into a command and arguments. The parsing job is all about string manipulation, which is right up your alley after assign 3.

  • Divide the line into an array of tokens. A token consists of a sequence of non-space chars.
  • Ignore/skip all whitespace in between tokens as well as leading and trailing whitespace. Whitespace includes space, tab, and newline.
  • The first token is the name of the command to execute, the subsequent tokens are the arguments to the command.

When tokenizing, be sure to take advantage of the nifty functions you implemented in your strings and malloc modules. They will be helpful!

4) Execute command

Now that you have the command name and arguments, you're ready to evaluate it.

  • Look up the function pointer for the command by name. The command table associates a command name string with its function pointer.
    • If no matching command is found, output message error: no such command 'binky'.
  • Call the function pointer, passing the array of tokens and the count of tokens. The first element in the array is the command name, the subsequent elements are the arguments to the command.
  • The return value of the command is used as the return value for shell_evaluate.

There is a command table started in shell.c. You will modify the table as you add commands. Each entry in the table is a command_t struct as defined in shell_commands.h. A command function pointer takes two parameters: argv is an array of char *, i.e. an array of strings, and argc is the count of elements in the argv array. A command function returns an int to indicate success or failure. The result is 0 if the command executed successfully, or nonzero otherwise.

Your shell has five commands:

  • int cmd_echo (int argc, const char *argv[])
  • int cmd_help (int argc, const char *argv[])
  • int cmd_reboot (int argc, const char *argv[])
  • int cmd_peek (int argc, const char *argv[])
  • int cmd_poke (int argc, const char *argv[])

The documentation shell_commands.h explains each of the commands and the expected behavior.

We've implemented cmd_echo as an example. This command simply echoes its arguments:

Pi> echo Hello, world!
Hello, world!

The additional commands you are to implement are:

  • help

    Without any arguments, help prints a list of all available commands along with their description in the following format:

    Pi> help
    cmd1: description for cmd1
    cmd2: description for cmd2

    If an argument is given, help prints the description for that command, or an error message if the command doesn't exist:

    Pi> help reboot
    reboot:  reboot the Raspberry Pi back to the bootloader
    Pi> help please
    error: no such command `please`.
  • reboot

    This command restarts your Pi by sending a uart_putchar(EOT) to end communication with your latop, then calling the pi_reboot function from the pi module of libpi. See ya back at the bootloader!

  • peek

    This command takes one argument: [address]. It prints the 4-byte value stored at memory address address.

    Try using peek to get the contents at address 0x8000. This memory location stores the first instruction of your program. The first instruction of the start sequence is mov sp, #0x8000000, encoded as e3a0d302.

    Pi> peek 0x8000
    0x00008000:   e3a0d302

    Hint: the strtonum function from your strings module is handy for converting strings to numbers. If the address argument is missing or cannot be converted, peek prints an error message:

    Pi> peek
    error: peek expects 1 argument [address]
    Pi> peek bob
    error: peek cannot convert 'bob'

    Recall that the ARM architecture is not keen to load/store on an unaligned address. A 4-byte value can only be read starting from an address that is a multiple of 4. If the user asks to peek (or poke) at an unaligned address, respond with an error message:

    Pi> peek 7
    error: peek address must be 4-byte aligned

    Peek displays the contents in memory at an address. To test the peek command, use peek on a location in the text or data section to see the encoded instruction or value of a global variable. Use arm-none-eabi-nm build/uart_shell.elf to get the symbol address to know which address to peek. You can also peek at the location for the system timer or gpio FSEL or LEV registers to read values of a peripheral register.

  • poke

    This command takes two arguments: [address] [value]. The poke function stores value into the memory at address.

    Here is an example of using poke to write to at address 0x8000. (This memory location stores the first instruction of your program. Since that code has already executed and will not be re-entered, we can overwrite it without causing problems for the executing program):

    Pi> peek 0x8000
    0x00008000:   e3a0d302
    Pi> poke 0x8000 1
    Pi> peek 0x8000
    0x00008000:  00000001
    Pi> poke 0x8000 0
    Pi> peek 0x8000
    0x00008000:  00000000

    strtonum will again be handy. If either argument is missing or cannot be converted, poke prints an error message:

    Pi> poke 0x8000
    error: poke expects 2 arguments [address] [value]
    Pi> poke fred 5
    error: poke cannot convert 'fred'
    Pi> poke 0x8000 wilma
    error: poke cannot convert 'wilma'

    You can now turn a GPIO pin on and off by entering shell commands!

    Pi> poke 0x20200010 0x200000
    Pi> poke 0x20200020 0x8000
    Pi> poke 0x2020002C 0x8000

    Review the BCM2835 manual to see what is stored at these addresses. What do the above commands do? Hint: the ACT LED on the Pi is GPIO pin 47.

    Take care when testing poke. Unlike peek which can read from any address without causing harm, poke is writing to the address and changing the contents of live memory. Depending on what you are changing, this could interfere with the executing program. Before poking an address, figure out what is stored there and what effect your update will have. Once you have confirmed that poke works for select addresses, it is reasonable to extrapolate that it will also handle the rest without testing on every single address from 0 to 2^32 😅

    Check out the Wikipedia article on peek and poke if you're curious to learn about their historical origins.

For all of Your shell commands, be sure that all your output goes through shell_printf and please try to make the output exactly match the format and wording of the output and error messages given above. We know this request is nitpicky, but this enables automated comparison of shell output which translates to fewer staff hours going into manual grading and more time to hang out with y'all in office hours.

Testing and debugging

As usual, the effort you put into writing good tests will be evaluated along with your code submission. An interactive program such as this one adds new challenges for testing. Your inventive solutions to overcoming these challenges are welcome!

The given code in test_keyboard_shell.c program has functions that test each layer of the keyboard module. These functions simply echo the data returned by the keyboard module. You must manually verify the correctness of the output.

The test_keyboard_assert function demonstrates an example approach for an assert-based test where you coordinate with the user to provide the keyboard input.

The shell module is intended to be run interactively and does not lend itself well to assert-based testing. This doesn't mean you should eschew testing it, but you will have to be more creative in how you proceed. It may help to testing shell_evaluate separately from shell_readline. Try calling shell_evaluate on a fixed command and manually observe that the output is as expected.

The function shell_readline can get messy if trying to test your shell and keyboard at same time. There is some sample code in test_keyboard_shell.c that demonstrates a testing function that returns the next character from a fixed sequence. This function can be used as input function for the shell in place of keyboard_read_next as a way to test shell_readline independent of keyboard.

Extension: editing and history

The extension consists of two parts. You should do both parts.

  1. Command-line editing

    Implement the left and right arrow keys to move the cursor within the current line and allow inserting and deleting characters at the point of the cursor. What other editing features might be nice to have: Ctrl-a and Ctrl-e to move the cursor to the first and last character of the line? Ctrl-u to delete the entire line? Implement your favorite vim/emacs/editor must-haves!

  2. Command history

    Number the commands entered starting from 1 and maintain a rolling history of the last 10 commands entered. Change the prompt to include the command number of the current line. Add a history command that displays the history of recent commands, each prefixed with its command number. See the example below:

    [1] Pi> help echo
    echo: <...> echo user input to the screen
    [2] Pi> echo cs107e rocks
    cs107e rocks
    [3] Pi> history
      1 help echo
      2 echo cs107e rocks
      3 history

    Implement the up and down arrow keys to access commands from the history. Typing the up arrow key changes the current line to display the command from the history that is previous to the one on the current line. Typing a down array changes the current line to display the command that ran after the one on the current line, or whatever the user had typed until he/she typed up. Use the shell_beep when the user tries to move beyond either end of the history.

    What other history features might be handy? !! to repeat the last command? !ze to repeat the most recent command matching the specified prefix ze?

    If you didn't already know that your regular shell includes editing and history features like these, now is a great time to pick up a few new tricks to help boost your productivity!

What other neat tricks can you teach your Pi now that you have a an interactive shell? Our reference shell throws in a few cute freebies such as gpio to turn on and off gpio pins and the simple calculator from lab5. If you wrote the disassemble extension for assign3 (or you want to do it now!), a nifty option is to integrate a disassemble [addr] command into your shell. The fun never stops!

In your, list the editing features and shell commands supported by your fancy extended shell so we'll know how to try it out when grading.


The deliverables for assign5-submission are:

  • implementations of the ps2.c keyboard.c and shell.c library modules
  • comprehensive tests for all modules in test_keyboard_shell.c
  • (possibly empty)

Submit the finished version of your assignment by tagging assign5-submission and making a pull request. The steps to follow are given in the Assignment 0 writeup.


To grade this assignment, we will:

  • Verify that your submission builds correctly, with no warnings. Clean build always!
  • Run automated tests on your ps2/keyboard and shell modules
    • Take care! Our automated testing requires that your shell has absolute consistency in calling shell_printf for all shell output. Double-check that you are compliant. Be sure to remove or comment out all use of printf and debugging output.
  • Go over the test cases you added to test_keyboard_shell.c and evaluate for thoughtfulness and completeness in coverage.
  • Review your code and provide feedback on your design and style choices.

Our highest priority tests will focus on the core features for this assignment:

  • Essential functionality of your library modules
    • ps2
      • read well-formed scancode
      • discard of malformed (wrong start/stop/parity or timeout), resynchronize
    • keyboard
      • read events from all keys
      • handling of modifiers, including caps lock
    • shell
      • user-entered input (including handling of backspace)
      • parse and execute command

The additional tests of lower priority will examine less critical features, edge cases, and robustness. Make sure you thoroughly tested for a variety of scenarios!

nice job!