Last week, I sent first patch to upstream and it was accepted in mm tree :). I am very happy about that. You can see it here.
Subsystem maintainers are very careful and a lot of people review the codes. Out side of staging directory, and before the internship, my first patch is about y2038 project. When I sent patch for y2038, a lot of developers reviewed and suggested something. I have recently seen the patch here: http://lwn.net/Articles/620870/, it was my first experience :).
Todays, I work with linux kernel mm community, this makes me very excited and happy :). I gained some experiences in this process and learnt how can I be sure with my changes. This is most important case for coding. When I talked with my mentor, every time I reported different thing :) and said "oh this prevents collapsing pages into a thp!". Because I was testing wrong. Finally I could find what was the problem.
For test results, I look /var/log/kern.log, it is very large file so I split it like that: "split -n 5" and look newly created small files and log time stamps is important. To be sure with my changes print out virtual memory address area for my test programs.
/proc/pid/smaps shows whole vmas for the process. pr_info("vm_start = %04lx\n", vma->vm_start); is enough to see begining address of the vma. Sometimes I need to see which process run this function, I print out current->pid. If I know what happened in every step, I find my faults very easy. I have to do something like that, because other processes will log about their huge pages in kern.log and I shouldn't confuse which process logged the results. To examining kern.log was big scale thing for me.
Before sending patch I need to be careful and check something for my patch. Also keeping focus on the issues is important. Working on kernel needs to pay attention more accorrding to other projects which I got experiences with them when I was student.
After this patch, I will work on zero pages and discover new things :).