On Mon, 7 Jan 2019 at 01:55, David Miller wrote:
>
> From: Taehee Yoo
> Date: Sun, 6 Jan 2019 14:34:52 +0900
>
> > How about adding a new PF_UMH flag for task_struct->flags to identify
> > UMH process?
> > By using this flag, the exit_umh() can avoid unnecessary lookups.
>
> Yes, that might be mo
From: Taehee Yoo
Date: Sun, 6 Jan 2019 14:34:52 +0900
> How about adding a new PF_UMH flag for task_struct->flags to identify
> UMH process?
> By using this flag, the exit_umh() can avoid unnecessary lookups.
Yes, that might be more efficient and eliminate the high cost for
non-UMH tasks.
On Sun, 6 Jan 2019 at 07:10, David Miller wrote:
>
> From: Taehee Yoo
> Date: Mon, 31 Dec 2018 01:31:43 +0900
>
> > +void exit_umh(struct task_struct *tsk)
> > +{
> > + struct umh_info *info;
> > + pid_t pid = tsk->pid;
> > +
> > + mutex_lock(&umh_list_lock);
> > + list_for_each_e
From: Taehee Yoo
Date: Mon, 31 Dec 2018 01:31:43 +0900
> +void exit_umh(struct task_struct *tsk)
> +{
> + struct umh_info *info;
> + pid_t pid = tsk->pid;
> +
> + mutex_lock(&umh_list_lock);
> + list_for_each_entry(info, &umh_list, list) {
So this is probably too expensive of a c
A UMH process which is created by the fork_usermode_blob() such as
bpfilter needs to release members of the umh_info when process is
terminated.
But the do_exit() does not release members of the umh_info. hence module
which uses UMH needs own code to detect whether UMH process is
terminated or not.
5 matches
Mail list logo