[PATCHi v4] rcu/kvfree: Prevents cache growing when the backoff_page_cache_fill is set

[PATCHi v4] rcu/kvfree: Prevents cache growing when the backoff_page_cache_fill is set

[Zqiang]

Currently, in kfree_rcu_shrink_scan(), the drain_page_cache() is
executed before kfree_rcu_monitor() to drain page cache, if the bnode
structure's->gp_snap has done, the kvfree_rcu_bulk() will fill the
page cache again in kfree_rcu_monitor(), this commit add a check
for krcp structure's->backoff_page_cache_fill in put_cached_bnode(),
if the krcp structure's->backoff_page_cache_fill is set, prevent page
cache growing and start fill page from krcp->nr_bkv_objs, if the
krcp->nr_bkv_objs is greater or equal to nr_pages, not alloc page,
avoid unnecessary __get_free_page() calls.

Signed-off-by: Zqiang <qiang1.zhang@intel.com>
---
 kernel/rcu/tree.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 41daae3239b5..d5639025bd1a 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2908,6 +2908,8 @@ static inline bool
 put_cached_bnode(struct kfree_rcu_cpu *krcp,
 	struct kvfree_rcu_bulk_data *bnode)
 {
+	if (atomic_read(&krcp->backoff_page_cache_fill))
+		return false;
 	// Check the limit.
 	if (krcp->nr_bkv_objs >= rcu_min_cached_objs)
 		return false;
@@ -3223,7 +3225,7 @@ static void fill_page_cache_func(struct work_struct *work)
 	nr_pages = atomic_read(&krcp->backoff_page_cache_fill) ?
 		1 : rcu_min_cached_objs;
 
-	for (i = 0; i < nr_pages; i++) {
+	for (i = krcp->nr_bkv_objs; i < nr_pages; i++) {
 		bnode = (struct kvfree_rcu_bulk_data *)
 			__get_free_page(GFP_KERNEL | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
 
-- 
2.32.0

Re: [PATCHi v4] rcu/kvfree: Prevents cache growing when the backoff_page_cache_fill is set

[Paul E. McKenney]

On Tue, Apr 11, 2023 at 10:27:33PM +0800, Zqiang wrote:
> Currently, in kfree_rcu_shrink_scan(), the drain_page_cache() is
> executed before kfree_rcu_monitor() to drain page cache, if the bnode
> structure's->gp_snap has done, the kvfree_rcu_bulk() will fill the
> page cache again in kfree_rcu_monitor(), this commit add a check
> for krcp structure's->backoff_page_cache_fill in put_cached_bnode(),
> if the krcp structure's->backoff_page_cache_fill is set, prevent page
> cache growing and start fill page from krcp->nr_bkv_objs, if the
> krcp->nr_bkv_objs is greater or equal to nr_pages, not alloc page,
> avoid unnecessary __get_free_page() calls.
> 
> Signed-off-by: Zqiang <qiang1.zhang@intel.com>
> ---
>  kernel/rcu/tree.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 41daae3239b5..d5639025bd1a 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -2908,6 +2908,8 @@ static inline bool
>  put_cached_bnode(struct kfree_rcu_cpu *krcp,
>  	struct kvfree_rcu_bulk_data *bnode)
>  {
> +	if (atomic_read(&krcp->backoff_page_cache_fill))
> +		return false;

This one is still missing the logic that allows that cache to hold
one page during low memory?  Or is that being handled somewhere else?

							Thanx, Paul

>  	// Check the limit.
>  	if (krcp->nr_bkv_objs >= rcu_min_cached_objs)
>  		return false;
> @@ -3223,7 +3225,7 @@ static void fill_page_cache_func(struct work_struct *work)
>  	nr_pages = atomic_read(&krcp->backoff_page_cache_fill) ?
>  		1 : rcu_min_cached_objs;
>  
> -	for (i = 0; i < nr_pages; i++) {
> +	for (i = krcp->nr_bkv_objs; i < nr_pages; i++) {
>  		bnode = (struct kvfree_rcu_bulk_data *)
>  			__get_free_page(GFP_KERNEL | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
>  
> -- 
> 2.32.0
> 

RE: [PATCHi v4] rcu/kvfree: Prevents cache growing when the backoff_page_cache_fill is set

[Zhang, Qiang1]

> Currently, in kfree_rcu_shrink_scan(), the drain_page_cache() is
> executed before kfree_rcu_monitor() to drain page cache, if the bnode
> structure's->gp_snap has done, the kvfree_rcu_bulk() will fill the
> page cache again in kfree_rcu_monitor(), this commit add a check
> for krcp structure's->backoff_page_cache_fill in put_cached_bnode(),
> if the krcp structure's->backoff_page_cache_fill is set, prevent page
> cache growing and start fill page from krcp->nr_bkv_objs, if the
> krcp->nr_bkv_objs is greater or equal to nr_pages, not alloc page,
> avoid unnecessary __get_free_page() calls.
> 
> Signed-off-by: Zqiang <qiang1.zhang@intel.com>
> ---
>  kernel/rcu/tree.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 41daae3239b5..d5639025bd1a 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -2908,6 +2908,8 @@ static inline bool
>  put_cached_bnode(struct kfree_rcu_cpu *krcp,
>  	struct kvfree_rcu_bulk_data *bnode)
>  {
> +	if (atomic_read(&krcp->backoff_page_cache_fill))
> +		return false;
>
>This one is still missing the logic that allows that cache to hold
>one page during low memory?  Or is that being handled somewhere else?
>

Sorry, please ignore this patch,  I will resend as suggested by Ulad,
thanks for review😊.



>
>							Thanx, Paul
>
>  	// Check the limit.
>  	if (krcp->nr_bkv_objs >= rcu_min_cached_objs)
>  		return false;
> @@ -3223,7 +3225,7 @@ static void fill_page_cache_func(struct work_struct *work)
>  	nr_pages = atomic_read(&krcp->backoff_page_cache_fill) ?
>  		1 : rcu_min_cached_objs;
>  
> -	for (i = 0; i < nr_pages; i++) {
> +	for (i = krcp->nr_bkv_objs; i < nr_pages; i++) {
>  		bnode = (struct kvfree_rcu_bulk_data *)
>  			__get_free_page(GFP_KERNEL | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
>  
> -- 
> 2.32.0
>