Lab3: Adding a Reptilian System Call

COP4600
Lab3: Adding a Reptilian System Call
COP4600 – Operating Systems
Prologue
You work for a top-secret shadow government organization dedicated to the rise of the Silurian
overlords. You, as a faithful member of the Lizard Legion, are part of the team charged with improving
data storage and handling, particularly tracking metadata – that is, data about data – within the
organization’s computer systems. You have been tasked to build a coded message subsystem under the
guise of process logging for kernels running in “Sky Skink”, the cloud computing system. Naturally, the
Legion uses the superior Reptilian operating system distribution.
Overview
In this project, you will implement a system call in Reptilian along with three static library functions that
allow the system call to be invoked from a C API. These custom system calls will get and set a custom
process log level that will sit atop the standard Linux kernel’s diagnostic message logging system
(dmesg) and allow processes to submit log entries along with a log level. If the log level for the message
is more severe (lower than) the current log level, the message will be added to the kernel log. Log levels
and names will correspond to those in the Linux kernel. We, as your benevolent lizard overlords, will
provide a program that exercises and demonstrates the new calls. You will write a README.md file to
explain your code. (Our masters will be most pleased.) You will submit the project via Canvas so as not
to invite suspicion.
NOTE: Take snapshots of your VM! You will probably brick your machine at some point during
this or other projects, and you will not want to start from scratch. No, seriously – take snapshots!
Table 1. Kernel Log Levels and Corresponding Process Log Levels
Kernel Level Name Description # Process Level Name
KERN_EMERG Emergency / Crash Imminent (no process logging) 0 PROC_OVERRIDE
KERN_ALERT Immediate Action Required 1 PROC_ALERT
KERN_CRIT Critical/Serious Failure Occurred 2 PROC_CRITICAL
KERN_ERR Error Condition Occurred 3 PROC_ERROR
KERN_WARNING Warning; recoverable, but may indicate problems 4 PROC_WARNING
KERN_NOTICE Notable, but not serious (e.g., security events) 5 PROC_NOTICE
KERN_INFO Informational (e.g. initialization / shutdown) 6 PROC_INFO
KERN_DEBUG Debug messages 7 PROC_DEBUG
Structure
The project is broken into four main parts:
1) Create a kernel-wide process log level attribute.
2) Create system calls that allow a process to get or set the process log level of the system.
3) Create system call that allows a process to add a process log message at a defined log level.
4) Create static library functions that allow the system calls to be invoked via a C API.
→ →
Figure 1: A system call invoked from a user program
While exact implementation may vary, the library functions must match the signatures laid out in this
document, and the system calls must apply the security model properly. Logged messages have format
“$log_level_name [$executable, $pid]: $message”, e.g.:
PROC_ERR [bacon_pancakes, 21]: Life is scary & dark. That is why we must find the
light.
System Call
The system will have a single, kernel-wide process log level which should initialize on boot in the
kernel and must be stored persistently (until shutdown / reboot). The rules for logging are as follows:
1) Any process can read (get) the process log level.
2) Any process may send a process log to the kernel.
3) Only a process running as the superuser may write (set) the process log level.
4) If a message’s log level is higher than the process log level, the message is ignored.
5) If a message’s log level is lower than or equal to the process log level, the message will be
logged.
6) The system-wide process log level should be initialized to zero (0) – i.e., override logging only.
7) Log levels can have values between 0-7 (3-bit unsigned integer). Invalid level results in call
failure.
8) Any successfully logged message should be logged with the corresponding kernel log level.
System calls are called via syscall(call_num, param1, param2). To log a message, the call
should be syscall(PROC_LOG_CALL, msg, level). Call parameters are limited to no more than
two!
Static Library
You will create a static library to invoke the system calls in a directory named process_log. This will
include a header, process_log.h (prototypes and level symbols), and static library file named
libprocess_log.a. You will also need to provide a Makefile for the library. All sources must be
contained within the process_log directory. Please note, these file names must match exactly!
You will create a tarred gzip file of the process_log directory with name process_log.tar.gz.
When testing, we will decompress the archive, enter the process_log directory, and build. All
functions enumerated below must be made available by including "process_log.h". See Submission
for details.
Library Functions
int get_proc_log_level()
Invokes system call which reads system-wide process log level. Returns the process log level on
success, and -1 otherwise.
int set_proc_log_level(int new_level)
Invokes system call which attempts to change the system-wide process log level to new_level.
Returns new_level on success, and -1 otherwise. On failure, log level should be unchanged.
int proc_log_message(int level, char *message)
Invokes system call to log a message for this process. If logged, the message should appear in dmesg
logs at the corresponding kernel level. Returns -1 for invalid log level, and level otherwise.
Harness Functions
In addition to the standard library functions, you will implement testing harness functions. The
testing harness functions are used to verify security of the system calls from the system library (and
are required for full credit on this assignment). We will call these functions to retrieve the
information needed to make a system call. We will then call the system call within our own program.
This ensures that no security checks are being done in the user-level library.
System call parameter retrieval data should be returned as a pointer to an int array of 2-4 values
that can be used to make the system call (which can be cast from other types). It has this format:
{ call_number, num_parameters [, parameter1] [, parameter2] }
e.g.: { 42, 2, 867, 5309 }  syscall(42, 867, 5309)
Figure 2: Harness functions can directly invoke system calls without the library functions.
These test harness elements must be implemented to test your security model:
#define PROC_LOG_CALL <number
Definition for the system call number for proc_log_message (see System Call); should be in
header.
int* retrieve_set_level_params(int new_level)
Returns an int array of 2-4 values that can be used to make the set-process-log-level system call.
int* retrieve_get_level_params(
Returns an int array of 2-4 values that can be used to make the get-process-log-level system call.
int interpret_set_level_result(int ret_value)
After making the system call, we will pass the syscall return value to this function call. It should
return set_proc_log_level’s interpretation of the system call completing with return value
ret_value.
int interpret_get_level_result(int ret_value)
After making the system call, we will pass the syscall return value to this function call. It should
return get_proc_log_level’s interpretation of the system call completing with return value
ret_value.
int interpret_log_message_result(int ret_value)
After making the system call, we will pass the syscall return value to this function call. It should
return proc_log_message’s interpretation of the system call completing with return value
ret_value.
Note that there is no retrieve function for log_message as its system call format is defined above.
Submissions
You will submit the following at the end of this project (3 separate files):
● README (README.md) in Markdown file format on Canvas
● Kernel Patch File (lab3.diff) on Canvas
● Compressed tar archive (process_log.tar.gz) for process_log library on Canvas
README
Your README will explain how you implemented the new system call in the kernel, including what
changes were made to which files and why for each. It will describe how testing was performed and
any known bugs. The README should be aimed at a technical reader with little knowledge of
operating systems concepts, and should be saved as a .md. The README should be no more than 700
words, cover all relevant aspects of the project, and be organized and formatted professionally – this is
not a memo!
You may include more sections as you see fit, but we require that at least the following three sections
be included in your README:
● Description: a specification of what your project accomplishes once installed
● Design: explanations of non-trivial design decisions you made
● Verification: an explanation of how to verify the project’s functionality after installation
Patch File
The patch file will include all changes to all files in a single patch. Applying the patches and remaking
the necessary parts of Reptilian, then rebooting and then building the test code (which we will also
copy over) should compile the test program.
Your project will be tested by applying the patch while in /usr/rep/src/reptilian-kernel:
$ git apply lab3.diff
$ make && sudo make install && sudo make modules_install
Compressed Archive (process_log.tar.gz)
Your compressed tar file should have the following directory/file structure:
process_log.tar.gz
process_log.tar
process_log (directory)
process_log.h
Makefile
(Other source files)
To build the library, we will execute these commands (from a non-kernel-source directory):
$ tar zxvf process_log.tar.gz
$ cd process_log
$ make
$ cd ..
To link against the library, we will execute this command:
$ cc -o program_name sourcefile.c -L ./process_log -lprocess_log
Please test your library build and linking before submission! If your library does not compile it will
result in zero credit (0, none, goose-egg) for the library portion of the project.