Having these functions and their documentation split out and somewhere
makes it harder, not easier, to follow what's going on.
Inline them directly where charge moving is prepared and finished, and
put an explanation right next to it.
Signed-off-by: Johannes Weiner <[email protected]>
---
mm/memcontrol.c | 40 ++++++++++++----------------------------
1 file changed, 12 insertions(+), 28 deletions(-)
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 3cd4f1e0bfb3..5b5c784dc39d 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1447,32 +1447,6 @@ int mem_cgroup_swappiness(struct mem_cgroup *memcg)
}
/*
- * memcg->moving_account is used for checking possibility that some thread is
- * calling move_account(). When a thread on CPU-A starts moving pages under
- * a memcg, other threads should check memcg->moving_account under
- * rcu_read_lock(), like this:
- *
- * CPU-A CPU-B
- * rcu_read_lock()
- * memcg->moving_account+1 if (memcg->mocing_account)
- * take heavy locks.
- * synchronize_rcu() update something.
- * rcu_read_unlock()
- * start move here.
- */
-
-static void mem_cgroup_start_move(struct mem_cgroup *memcg)
-{
- atomic_inc(&memcg->moving_account);
- synchronize_rcu();
-}
-
-static void mem_cgroup_end_move(struct mem_cgroup *memcg)
-{
- atomic_dec(&memcg->moving_account);
-}
-
-/*
* A routine for checking "mem" is under move_account() or not.
*
* Checking a cgroup is mc.from or mc.to or under hierarchy of
@@ -5431,7 +5405,8 @@ static void mem_cgroup_clear_mc(void)
mc.from = NULL;
mc.to = NULL;
spin_unlock(&mc.lock);
- mem_cgroup_end_move(from);
+
+ atomic_dec(&memcg->moving_account);
}
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
@@ -5464,7 +5439,16 @@ static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
VM_BUG_ON(mc.precharge);
VM_BUG_ON(mc.moved_charge);
VM_BUG_ON(mc.moved_swap);
- mem_cgroup_start_move(from);
+
+ /*
+ * Signal mem_cgroup_begin_page_stat() to take
+ * the memcg's move_lock while we're moving
+ * its pages to another memcg. Then wait for
+ * already started RCU-only updates to finish.
+ */
+ atomic_inc(&memcg->moving_account);
+ synchronize_rcu();
+
spin_lock(&mc.lock);
mc.from = from;
mc.to = memcg;
--
2.1.2
On Thu 23-10-14 10:42:07, Johannes Weiner wrote:
> Having these functions and their documentation split out and somewhere
> makes it harder, not easier, to follow what's going on.
>
> Inline them directly where charge moving is prepared and finished, and
> put an explanation right next to it.
I do not see the open coded version much more readable or maintainable to be
honest. mem_cgroup_{start,end}_move are a good markers of the transaction.
But I do not have strong opinion about this. The preparation and
move_lock parts are so far from each other that either way would be
non-trivial to follow. That being said, I do not really care much.
Your comment is better and more clear though.
> Signed-off-by: Johannes Weiner <[email protected]>
> ---
> mm/memcontrol.c | 40 ++++++++++++----------------------------
> 1 file changed, 12 insertions(+), 28 deletions(-)
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 3cd4f1e0bfb3..5b5c784dc39d 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -1447,32 +1447,6 @@ int mem_cgroup_swappiness(struct mem_cgroup *memcg)
> }
>
> /*
> - * memcg->moving_account is used for checking possibility that some thread is
> - * calling move_account(). When a thread on CPU-A starts moving pages under
> - * a memcg, other threads should check memcg->moving_account under
> - * rcu_read_lock(), like this:
> - *
> - * CPU-A CPU-B
> - * rcu_read_lock()
> - * memcg->moving_account+1 if (memcg->mocing_account)
> - * take heavy locks.
> - * synchronize_rcu() update something.
> - * rcu_read_unlock()
> - * start move here.
> - */
> -
> -static void mem_cgroup_start_move(struct mem_cgroup *memcg)
> -{
> - atomic_inc(&memcg->moving_account);
> - synchronize_rcu();
> -}
> -
> -static void mem_cgroup_end_move(struct mem_cgroup *memcg)
> -{
> - atomic_dec(&memcg->moving_account);
> -}
> -
> -/*
> * A routine for checking "mem" is under move_account() or not.
> *
> * Checking a cgroup is mc.from or mc.to or under hierarchy of
> @@ -5431,7 +5405,8 @@ static void mem_cgroup_clear_mc(void)
> mc.from = NULL;
> mc.to = NULL;
> spin_unlock(&mc.lock);
> - mem_cgroup_end_move(from);
> +
> + atomic_dec(&memcg->moving_account);
> }
>
> static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
> @@ -5464,7 +5439,16 @@ static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
> VM_BUG_ON(mc.precharge);
> VM_BUG_ON(mc.moved_charge);
> VM_BUG_ON(mc.moved_swap);
> - mem_cgroup_start_move(from);
> +
> + /*
> + * Signal mem_cgroup_begin_page_stat() to take
> + * the memcg's move_lock while we're moving
> + * its pages to another memcg. Then wait for
> + * already started RCU-only updates to finish.
> + */
> + atomic_inc(&memcg->moving_account);
> + synchronize_rcu();
> +
> spin_lock(&mc.lock);
> mc.from = from;
> mc.to = memcg;
> --
> 2.1.2
>
--
Michal Hocko
SUSE Labs
On Thu, Oct 23, 2014 at 05:56:07PM +0200, Michal Hocko wrote:
> On Thu 23-10-14 10:42:07, Johannes Weiner wrote:
> > Having these functions and their documentation split out and somewhere
> > makes it harder, not easier, to follow what's going on.
> >
> > Inline them directly where charge moving is prepared and finished, and
> > put an explanation right next to it.
>
> I do not see the open coded version much more readable or maintainable to be
> honest. mem_cgroup_{start,end}_move are a good markers of the transaction.
What transaction, though? The names are really non-descript and they
actively hide what's going on. mem_cgroup_start_move() could mean
anything that prepares for moving and it doesn't seem out of place in
can_attach(). atomic_inc(&memcg->moving_account) on the other hand is
much more specific and nicely shows that we are currently forcing the
page stat update slow path way too early. There is no reason to make
it take the move_lock while we are still counting rss and precharging,
we are not actually moving charges and flipping pc->mem_cgroups yet.
[ I already have a patch to relocate it into mem_cgroup_move_charge(),
but let's let the dust in -mm settle a bit first. :-) ]