The learning objectives for Deepfried dd are:
- Working with files
- fseek() and fread()
The year is 3072. Other than being the sum of two consecutive powers of two, nothing was particularly special about this year. That is, until you came across an ancient laptop washed up on the sides of Boneyard Creek. Opening the laptop, you were delighted to find the block I and CS 125 stickers still firmly attached to the keyboard. Of course, you’d heard the stories. Before humanity had found a way to implant knowledge in each other via electromagnetic gimmicks, so-called “students” used to attend “classes” where “professors” disperesed their knowledge in oral and visual form. CS 125 had been such a class - oh, the wonder it must have been! The stickers alone, ancient relics of a past society, went for millions of dollars these days. Looking around to make sure no one else had laid eyes upon your lottery ticket, you hurried away to your apartment to have a closer look at the laptop. After some research on how ancient laptops were powered (turns out quantum resonance charging wasn’t discovered until 2048!), you quickly splice together a compatible laptop “charger” that would make any fire marshall wince, and finally hit the one thousand year old power button. After a few moments, you see a password prompt. Thinking for a moment on anything that could unite two CS @ Illinois nerds across a millenium, you smile and enter “imaginebeinganECEmajor”. Almost unsurprisingly, the laptop unlocks. And inside it, you find something even more enticing than the CS 125 sticker - four years worth of ancient homework assignments, practice exams, and even the infamous MPs of old. Almost eight hours pass before you look at the clock again, having quenched your insatiable curiosity for the ways of ancient societies. How did such a primitive world build such advanced exercises for their school children? And what was “PraireLearn”? Thousands of questions go through your head, but before you can answer you realize what you must do first: make a backup of this quaint laptop’s disks, so you can meticulously dissect its contents later on your own computer without threat of burning an entire building down. Luckily, you have a USB-C flash drive you purchased from a museum gift store, but how do you go about making a copy of the laptop’s entire disk? There’s no tool you could possibly download, since surely the networking stack won’t be able to connect with any modern network. Luckily, you notice that the laptop has a C IDE already installed. Somehow, C survived all the way from the 1980s to the present day. Almost gleefully, you realize you’ll have to code a C utility on this ancient relic of a laptop in order to make an image of the drive - just like the CS students of old would have done. You put on your hacking gloves, and get to work.
dd is a command-line utility used to copy data to and from files. Since Linux treats many external devices (including USB drives) as files, this makes
dd very powerful. For example, the tool can be used to create a backup image of your hard drive and store it as a file which can be uploaded to cloud storage.
dd could also be used to directly clone one drive to another, write a bootable iso image to a USB drive, and much more.
For example, the following command would write an .img file (
if) to a USB drive represented as /dev/disk4 (
of), in chunks of 4 MB blocks (
dd if=/path/to/bootable.img of=/dev/disk4 bs=4M
See the man pages for dd here, and example usage here. We suggest you get a feel of
dd by using it to copy a file from one folder to another. Be careful, however! dd can easily be used to (accidentally or on purpose) to corrupt and entirely wipe disk drives and partitions, so make sure you know exactly what a
dd command is going to do before running it! We recommend you make a testing folder on your VM and only use
dd with paths pointing to that folder, so there’s virtually no chance of overwriting something you don’t want to.
Note that Linux will automatically prevent you (usually) from writing to physical devices unless you run
dd as root (i.e. via sudo).
Implementing dd #
For this assignment, you will be implementing the
dd utility in C. Your
dd implementation will copy data from an input file to an output file in a manner specified by its arguments.
A block is simply a unit measuring the number of bytes that are read or written at one time. For example, modern hard drives have a sector (block) size of 4 kB - reads or writes to the disk can only address 4 kB portions at a time. If you write a 64 kB file to the disk, it will be broken down into 16 writes of 4 kB each. Because
dd is a file manipulation tool, it supports reading/writing with a configurable block size. For example, if we run
./dd -i input_file -o output_file -b 128, we are telling
dd to copy
output_file, 128 bytes at a time. Your code should write the first 128 bytes of
output_file, then the next 128, and so on and so forth in a loop until the
input_file is exhausted (as opposed to copying the entire file in one go). The arguments of
dd along with more example usage can be found below.
You must implement the following arguments from the real
dd. Since it is $CURRENT_YEAR, we won’t be using
dd’s style of arguments (
if=file, etc), and instead use the standard style (
-i file, etc). You should use
getopt to parse these arguments, just as you did in the Shell MP.
-i <file>: input file (defaults to stdin)
-o <file>: output file (defaults to stdout)
- You should create this file if does not already exist.
-b <size>: block size, the number of bytes copied at a time (defaults to 512)
-c <count>: total number of blocks copied (defaults to the entire file)
-p <count>: number of blocks to skip at the start of the input file (defaults to 0)
-k <count>: number of blocks to skip at the start of the output file (defaults to 0)
- The documentation on the
fopenmay be useful here.
- The documentation on the
- For any other arguments, you should exit with code 1.
getoptwill automatically print an error message for you.
Your code will be compiled into an executable and run via the command line.
WARNING: Note that some of these arguments refer to blocks, not bytes.
If implemented optimally, there is no need to specially handle the case where the input file defaults to
stdin, instead of a “real” file. Keep in mind that you don’t need to know the size of the entire input file in order to copy the full thing:
feof is a useful function. When copying from
dd should write bytes until the user enters Control+D (i.e. end of file) into their terminal. You can see this functionality by running the real
dd: simply run
dd of=my_file in your terminal, write some text, and press Control+D.
my_file will then contain the text you typed into the terminal.
fseek is not meant to be used on streams such as stdin and stdout! As such, we will not be testing the -p or -k functionality when the input or output files are set to stdin or stdout respectively.
You must print a status report after dd finishes, similar to the real dd. An example is below:
$ ./dd [...parameters] 1182465+1 records in 1182465+1 records out 605422080 bytes copied, 4.354 s, 139034.891 kB/s
In addition, your dd should be able to give a “live” status report. You should catch SIGUSR1 and print a status report when the signal is triggered:
$ ./dd -i my_file.txt -o my_file2.txt & # start as a background process $ kill -n SIGUSR1 <pid of dd> # send SIGUSR1 to dd 1182465+0 records in 1182465+0 records out 605422080 bytes copied, 4.354 s, 139034.891 kB/s
Use the functions provided in
format.h to print these reports. You should use
clock_gettime to time the execution of
printf (among other functions) is not safe to call in a signal handler, since it is not reentrant. Ensure your signal handler does not call these functions, including any function in
format.h, and instead indicates to your program to print this status report elsewhere.
The following command should copy 32 blocks of size 4 kB (a total of 128 kB) from
output.dat, skipping 2 blocks from the start of
input.dat and 10 blocks from the start of
$ ./dd -i input.dat -o output.dat -b 4096 -c 32 -p 2 -k 10 32+0 records in 32+0 records out 131072 bytes copied, 0.000 s, 919441.220 kB/s
This command should dump the output of
echo "Hello, World!" into a file called
$ echo "Hello World" | ./dd -o output.dat 0+1 records in 0+1 records out 12 bytes copied, 0.000 s, 1861.039 kB/s
This command should write about 2 GB of random data (from
/dev/urandom) into a file called
random.bin, in chunks of 1000 bytes:
$ ./dd -i /dev/urandom -o random.bin -b 1000 -c 200000 200000+0 records in 200000+0 records out 200000000 bytes copied, 1.459 s, 137094.439 kB/s
If the input or output file given to your
dd is invalid, use the functions in
format.h to print the corresponding error and exit with return code
1. Any other invalid inputs (e.g. negative block sizes) are considered undefined behavior.
Though it is helpful to write tests that call any functions you write in
dd.c, because your code will be run as a command line utility, we recommend testing in the command line as well. You can assemble a series of calls to your
dd executable in a bash script, and use
md5sum along with spot checks to ensure correct functionality. For example, the following shell script would print nothing if your dd implementation is correct:
# create a random 32 MB file using the real dd dd if=/dev/urandom of=test_file.img bs=4M count=8 # copy to my_test_file using my own implementation of dd ./dd -i test_file.img -o my_test_file.img # print the differences between the two files, if any diff test_file.img my_test_file.img
We included a Makefile target to automatically generate a few files of different sizes in the
test_files directory. You can build it by running
make test-data inside the assignment directory on your VM.
Your code will be auto-graded on all of the parameters listed above. We will also be testing your status reports, both while dd is running (via
SIGUSR1) and after it completes. Note that a handful of tests depend on various aspects of your progress report working.
WARNING: Make sure to remove all print statements from your code (except calls to functions in format.h) before running the autograder!
WARNING: You will fail the assignment if you use use
exec, or similar.