If the rcu_min_cached_objs is set to zero at boot params, the
krcp->page_cache_work will not be triggered to fill page cache,
and the put_cached_bnode() also does not fill page cache, this
also means the krcp->bkvcache is always empty, so not need to
acquire unnecessary krcp->lock to get page from krcp->bkvcache,
this commit therefore make drain_page_cache() return directly if
the rcu_min_cached_objs is zero.
Signed-off-by: Zqiang <[email protected]>
---
kernel/rcu/tree.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 93faf9f5c58f..00620a2117de 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2924,6 +2924,9 @@ drain_page_cache(struct kfree_rcu_cpu *krcp)
struct llist_node *page_list, *pos, *n;
int freed = 0;
+ if (!rcu_min_cached_objs)
+ return 0;
+
raw_spin_lock_irqsave(&krcp->lock, flags);
page_list = llist_del_all(&krcp->bkvcache);
WRITE_ONCE(krcp->nr_bkv_objs, 0);
--
2.32.0
On Tue, Apr 18, 2023 at 08:27:02PM +0800, Zqiang wrote:
> If the rcu_min_cached_objs is set to zero at boot params, the
> krcp->page_cache_work will not be triggered to fill page cache,
> and the put_cached_bnode() also does not fill page cache, this
> also means the krcp->bkvcache is always empty, so not need to
> acquire unnecessary krcp->lock to get page from krcp->bkvcache,
> this commit therefore make drain_page_cache() return directly if
> the rcu_min_cached_objs is zero.
>
> Signed-off-by: Zqiang <[email protected]>
> ---
> kernel/rcu/tree.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 93faf9f5c58f..00620a2117de 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -2924,6 +2924,9 @@ drain_page_cache(struct kfree_rcu_cpu *krcp)
> struct llist_node *page_list, *pos, *n;
> int freed = 0;
>
> + if (!rcu_min_cached_objs)
> + return 0;
> +
> raw_spin_lock_irqsave(&krcp->lock, flags);
> page_list = llist_del_all(&krcp->bkvcache);
> WRITE_ONCE(krcp->nr_bkv_objs, 0);
> --
> 2.32.0
>
Reviewed-by: Uladzislau Rezki (Sony) <[email protected]>
--
Uladzislau Rezki
On Thu, Apr 20, 2023 at 01:06:27PM +0200, Uladzislau Rezki wrote:
> On Tue, Apr 18, 2023 at 08:27:02PM +0800, Zqiang wrote:
> > If the rcu_min_cached_objs is set to zero at boot params, the
> > krcp->page_cache_work will not be triggered to fill page cache,
> > and the put_cached_bnode() also does not fill page cache, this
> > also means the krcp->bkvcache is always empty, so not need to
> > acquire unnecessary krcp->lock to get page from krcp->bkvcache,
> > this commit therefore make drain_page_cache() return directly if
> > the rcu_min_cached_objs is zero.
> >
> > Signed-off-by: Zqiang <[email protected]>
> > ---
> > kernel/rcu/tree.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > index 93faf9f5c58f..00620a2117de 100644
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -2924,6 +2924,9 @@ drain_page_cache(struct kfree_rcu_cpu *krcp)
> > struct llist_node *page_list, *pos, *n;
> > int freed = 0;
> >
> > + if (!rcu_min_cached_objs)
> > + return 0;
> > +
> > raw_spin_lock_irqsave(&krcp->lock, flags);
> > page_list = llist_del_all(&krcp->bkvcache);
> > WRITE_ONCE(krcp->nr_bkv_objs, 0);
> > --
> > 2.32.0
> >
> Reviewed-by: Uladzislau Rezki (Sony) <[email protected]>
Queued for further testing and review, thank you both!
Thanx, Paul