mjhjmtu

tgid and pid are the same concept for any process or for any lightweight process.

In /proc/$pid/status, tgid and pid are distinct fields. Are tgid and pid ever different for a process or lightweight process?

Thanks.

When looking at /proc/$pid/status, then the Tgid: and Pid: fields will always match, since they're the same for a process or for the main thread of a process.

The reason why there are two separate fields is that the same code is used to produce /proc/$pid/task/$tid/status, in which Tgid: and Pid: may differ from each other. (More specifically, Tgid: will match $pid and Pid: will match $tid in the file name template used above.)

The naming is a bit confusing, mainly because threading support was only added to the Linux kernel later on and, at the time, the scheduler code was modified to reuse the logic that used to schedule processes so it would now schedule threads. This resulted in reusing the concept of "pids" to identify individual threads. So, in effect, from the kernel's point of view, "pid" is still used for threads and "tgid" was introduced for processes. But from userspace you still want the PID to identify a process, therefore userspace utilities such as ps, etc. will map kernel's "tgid" to PID and kernel's "pid" to "tid" (thread id.)

TGid and Pid are different for threads spawned from the process leader.

eg looking at my polkitd process:

% ps -Lp 2642 
 PID LWP TTY TIME CMD
 2642 2642 ? 00:00:18 polkitd
 2642 2680 ? 00:00:00 gmain
 2642 2683 ? 00:00:30 gdbus
 2642 2685 ? 00:00:00 polkitd
 2642 2687 ? 00:00:00 JS GC Helper
 2642 2688 ? 00:00:00 JS Sour~ Thread
 2642 2692 ? 00:00:00 polkitd

If I look at process 2680:

% egrep ^'(Tgid|Pid):' /proc/2680/status
Tgid: 2642
Pid: 2680