On Thu, 01 Apr 2021 20:20:04 +0530, Navin P said:
After printing the task->start_boottime as 53rd field, i found multiple process containing same value due to a race. The race exists for p->start_time as well.
Well, yeah. There's no existing reason that the kernel shouldn't be able to spawn processes on two CPUs in parallel (and initialize the start_time field to the same value along the way). There should be as little synchronizing locking as possible, and if you need to do something that requires locking, you should probably piggyback it on top of a code section that already has the appropriate locks in place.
What are the other ways that can be used to fix this problem ?
Is there a particular reason why your code *needs* to have different unique values for the start_time and start_boottime fields? If your code needs something unique per-process, the tuple (start_time, PID) *will* be unique system-wide. Or just PID. In other words, what problem are you trying to solve by making the start_time unique?