Hi Tetsuo ... > Hello. > > sys_getpid() returns current->tgid for both 2.4 and 2.6 kernels. > http://lxr.linux.no/source/kernel/timer.c?v=2.4.28#L753 > http://lxr.linux.no/source/kernel/timer.c?v=2.6.11#L967 > > Then, why not use current->tgid instead of current->pid > in proc_self_readlink()/proc_self_follow_link() for 2.4 kernels? > http://lxr.linux.no/source/fs/proc/base.c?v=2.4.28#L1017 > http://lxr.linux.no/source/fs/proc/base.c?v=2.6.11#L1589 Interesting analysis! Allow me to share my opinion. We know that tgid is equal to pid, but whenever new process is created using CLONE_THREAD, they are no longer equal. tgid is same for all process under same process/thread group while pid is still unique. This way, we satisfy POSIX rule that getpid returns same pid for all thread under same group. But, in case of /proc/self, we must point to distinct ID, no matter it is created using CLONE_THREAD or not. So, task_struct->pid is the answer IMHO. I haven't checked this on 2.6.x though... Comments? regards Mulyadi -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
