Re: [RFC,PATCH 2/2] change __wake_up_parent() to use filteredwakeup

From: Oleg Nesterov
Date: Mon Jun 29 2009 - 02:53:53 EST


On 06/25, Ratan Nalumasu wrote:
>
> I understood it in the _past_, but somehow forgotten the details--I have a
> vague recollection that there was some special handling for the group leader
> stuff.
> However, to reconfirm, we believe that the
> following condition in eligible_child() is good, right:
> ===
> if ((wo->wo_flags & __WNOTHREAD) && wo->child_wait.private !=
> p->parent)
> return 0;
> ===
>
> I will run it on my test machines and see if everything looks good.

OK, thanks.

The only problem it uses ->parent, this conflicts with other out-of-tree
ptrace changes...

Roland, do you think we should do this change now or later?

If now, then we can also do another nice optimization, and it is for free.
See the untested patch below. It conflicts with those patches too, but
in both cases fixups are trivial.

Oleg.

------------------------------------------------------------------------
[PATCH] do not place sub-threads on task_struct->children list

Currently we add sub-threads to ->real_parent->children list. This
buys nothing but slows down do_wait().

With this patch ->children contains only main threads (group leaders).
The only complication is that forget_original_parent() should iterate
over sub-threads by hand. (note that we can move reparent_thread()
outside of while_each_thread() loop and simplify the code further).

>From now do_wait_thread() can never see task_detached() && !EXIT_DEAD
tasks, we can remove this check. (note that now we can unify
do_wait_thread() and ptrace_do_wait()).

This change can confuse the optimistic search inmm_update_next_owner(),
but this is fixable and minor.

Perhaps badness() and oom_kill_process() should be updated, but they
should be fixed in any case.

Do you see any problems?

---

kernel/fork.c | 2 +-
kernel/exit.c | 36 +++++++++++++++++-------------------
2 files changed, 18 insertions(+), 20 deletions(-)

--- WAIT/kernel/fork.c~3_CHILDREN_NO_THREADS 2009-06-19 01:12:47.000000000 +0200
+++ WAIT/kernel/fork.c 2009-06-29 04:30:34.000000000 +0200
@@ -1245,7 +1245,6 @@ static struct task_struct *copy_process(
}

if (likely(p->pid)) {
- list_add_tail(&p->sibling, &p->real_parent->children);
tracehook_finish_clone(p, clone_flags, trace);

if (thread_group_leader(p)) {
@@ -1257,6 +1256,7 @@ static struct task_struct *copy_process(
p->signal->tty = tty_kref_get(current->signal->tty);
attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
attach_pid(p, PIDTYPE_SID, task_session(current));
+ list_add_tail(&p->sibling, &p->real_parent->children);
list_add_tail_rcu(&p->tasks, &init_task.tasks);
__get_cpu_var(process_counts)++;
}
--- WAIT/kernel/exit.c~3_CHILDREN_NO_THREADS 2009-06-24 17:36:28.000000000 +0200
+++ WAIT/kernel/exit.c 2009-06-29 04:47:21.000000000 +0200
@@ -67,11 +67,11 @@ static void __unhash_process(struct task
detach_pid(p, PIDTYPE_PGID);
detach_pid(p, PIDTYPE_SID);

+ list_del_init(&p->sibling);
list_del_rcu(&p->tasks);
__get_cpu_var(process_counts)--;
}
list_del_rcu(&p->thread_group);
- list_del_init(&p->sibling);
}

/*
@@ -742,10 +742,10 @@ static void reparent_thread(struct task_
if (p->pdeath_signal)
group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);

- list_move_tail(&p->sibling, &p->real_parent->children);
-
if (task_detached(p))
return;
+
+ list_move_tail(&p->sibling, &p->real_parent->children);
/*
* If this is a threaded reparent there is no need to
* notify anyone anything has happened.
@@ -771,7 +771,7 @@ static void reparent_thread(struct task_

static void forget_original_parent(struct task_struct *father)
{
- struct task_struct *p, *n, *reaper;
+ struct task_struct *g, *p, *n, *reaper;
LIST_HEAD(dead_children);

exit_ptrace(father);
@@ -779,13 +779,16 @@ static void forget_original_parent(struc
write_lock_irq(&tasklist_lock);
reaper = find_new_reaper(father);

- list_for_each_entry_safe(p, n, &father->children, sibling) {
- p->real_parent = reaper;
- if (p->parent == father) {
- BUG_ON(task_ptrace(p));
- p->parent = p->real_parent;
- }
- reparent_thread(father, p, &dead_children);
+ list_for_each_entry_safe(g, n, &father->children, sibling) {
+ p = g;
+ do {
+ p->real_parent = reaper;
+ if (p->parent == father) {
+ BUG_ON(task_ptrace(p));
+ p->parent = p->real_parent;
+ }
+ reparent_thread(father, p, &dead_children);
+ } while_each_thread(g, p);
}
write_unlock_irq(&tasklist_lock);

@@ -1533,14 +1536,9 @@ static int do_wait_thread(struct wait_op
struct task_struct *p;

list_for_each_entry(p, &tsk->children, sibling) {
- /*
- * Do not consider detached threads.
- */
- if (!task_detached(p)) {
- int ret = wait_consider_task(wo, tsk, 0, p);
- if (ret)
- return ret;
- }
+ int ret = wait_consider_task(wo, tsk, 0, p);
+ if (ret)
+ return ret;
}

return 0;

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/