On Mon, Oct 19, 2009 at 4:08 PM, Leena M. <kernmonk@xxxxxxxxx> wrote:
> Hi All ,
> I have sprinkled printk's in each and every function in VFS and
> Filesystem which I want to understand.
> I want to trace the functions which are executed when I execute a system
> call in user space for my process.
> I want to print the trace only for my process , so i want to somehow pass
> the the pid from the user space and
> use struct task_struct . Can I do it though the proc filesystem or is there
> any other idea to do this.
A quick idea would be to implement something like /proc/tracethispid
which read integer. Then you do i.e "echo 1111 > /proc/tracethispid"
and store this integer in a kernel variable.
Use this variable inside a condition checking: (not sure if "current"
is still called like 2.6.x used to be)
if unlikely(current->pid = target_pid)
{
printk("tracing function xyz"); /* FIXME: I forgot the gcc
function to print current functio name, maybe it's __function__
}
here, "unlikely" is used to accelerate the whole function call,
because IMHO not all the time your we hit the function when the
backing process's pid is our target pid, right?
just 2 cents idea...
--
regards,
Mulyadi Santosa
Freelance Linux trainer and consultant
blog: the-hydra.blogspot.com
training: mulyaditraining.blogspot.com
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
