Testing an LKM(Loadable Kernel Module)[1] support based Scheduler. The LKM Scheduler primarily works on the principle of load and go. The scheduler is initiated by compiling the source code associated with the LKM and loading the kernel modules using the scripts. Once the compiling and loading is complete, the user processes are handed out an option to use the existing OS scheduler or to use the LKM based scheduler.
- The user processes initially writes its process id to the file
/proc/process_sched_addwhich corresponds to the kernel moduleprocess_set. This procedure completes the registration of a process to the LKM Scheduler. - The LKM based scheduler is executed internally via the kernel module
process_scheduler. The module executes a work queue construct for every time quanta. - The
process_setandprocess_schedulermodules are coupled through the kernel moduleprocess_queue. Theprocess_queuemodule handles the internal details of all the processes associated with the LKM Scheduler. It stores the process info as simple link list nodes. - Various interfaces are defined within the
process_queueto perform add, remove, get_first, print operations on the queue. The scheduler performs an add and remove based on the context switch operation being triggered for every time quanta. - On every time quanta, the scheduler pushes the currently executing PID to the
process_queueviaadd_to_processinterface. And change its execution from Running to wait viataskbased interfaces. Once the currently executing process is added successfully into the queue, the process in the front of the queue is selected. The selected process state is changed to running and also removed from the queue.
- Linux OS with kernel version > 4.0 with LKM support enabled.
makegcc
- process_set.c - source code for setting a process to the custom scheduler.
- process_scheduler.c - source code for the custom scheduler
- process_queue.c - source code for the process queue maintainance.
- test_pr.c - test process for custom scheduler execution.
- Makefile - For compiling various source code related to the scheduler LKM.
- insmod_scr.sh - LKM insertion script.
- rmmod_scr.sh - LKM removal script.
- Open a terminal and go to the location where the git cloning was performed.
- Compile the LKM by running the Makefile. Run
maketo compile the source code related to LKM. If needed a clean build, runmake cleanallbefore defaultmake. - After successful compilation of Makefile, you would witness .ko files in the folder. These files are kernel object files for LKM related aspects.
- Now run the script insmod_scr.sh using makefile
make insmodor you can runmake loadwhich would compile and load the kernel modules. The execution of above script will only insert the kernel modules process_set, process_scheduler and process_queue to set of kernel modules. After this step the LKM is loaded. You can rundmesgto verify if the insertion was successful or not. - Now compile the test_pr.c source file. This source code can be compiled before the LKM compilation process. But you cannot execute it before the above LKM is loaded. Compilation is done by runnning
make comp_testor you can compile and run using commandmake test. - After successful compilation, open two new terminals in the same location and run the test_pr.out in those terminals with the command
./test_pr.outormake test. Note: the test_pr.out will execute infinitely, therefore you will need to terminate it manually. - Now you can witness the effect of the scheduler within seconds. Currently the scheme used is static round robin scheme with time quantum of 3 secs(default value). You can modify the time quantum through the option
time_quantumand then the value(in secs). For example, to load tq value we provide the value as:insmod process_scheduler.ko time_quantum=4say 4 is the new time quantum. - Finally if you are done using the LKM and you need to remove it run the command
make unloadwhich would unload the kernel modules and clean them or run the scriptmake rmmodwhich would only remove the kernel module but not clean them.