为什么 for_each_process 不显示每个任务？

Question

I am trying to loop through every process in a /proc utility I'm writing (a kernel module in /fs/proc ). The problem is, I am only seeing processes in the root namespace. I'm trying to use the macro for_each_process() from sched.h .

I can type ps in a shell and see plenty of processes, but my for_each_process() loop doesn't see them. What gives?

Note: I am wondering if it has something to do with rcu_read_lock() ? I'm afraid to put an rcu_read_lock() and I don't know where it should go. The trouble is, the documentation I read seems to say that in a preemptive kernel (mine is), it is illegal to sleep inside of rcu_read_lock() . I need to call down_read(mmap_sem) which I am afraid will sleep. So that means I can't use rcu_read_lock() ?

Answer 1

It should show you all the processes. I have written code like this.

struct task_struct *task;

for_each_process(p) {
    printk("Task %s (pid = %d)\n",p->comm, task_pid_nr(p));
}

This is printing all the processes. I suspect your proc_read function. Can you paste your proc_read function over here?

Answer 2

I think for_each_process() just gives you the thread group leader.

This iterates over all task_struct variables (defined in sched.h).

struct task_struct *g, *p;
do_each_thread(g, p) {
    //do something with p
} while_each_thread(g, p);

Answer 3

This example prints out a long list, looks like it might be printing almost all of them....thought I have not counted them

http://tuxthink.blogspot.in/2011/03/using-foreachprocess-in-proc-entry.html

Answer 4

for_each_process is a kernel layer function, and it's really not how you're supposed to loop over processes in unix. Something a lot simpler, such as this (python, easily implemented in other languages), might be the solution you want.

#Print each process id
import re,os
for fn in os.listdir( '/proc' ):
    if re.match('\d+', fn):
        print 'process %s'%fn