* [PATCH v2] kvfree_rcu: Release page cache under memory pressure
@ 2021-01-29 8:04 qiang.zhang
[not found] ` <20210129141953.GA29827@pc638.lan>
0 siblings, 1 reply; 3+ messages in thread
From: qiang.zhang @ 2021-01-29 8:04 UTC (permalink / raw)
To: urezki; +Cc: paulmck, joel, rcu, linux-kernel
From: Zqiang <qiang.zhang@windriver.com>
Add free per-cpu existing krcp's page cache operation, when
the system is under memory pressure.
Signed-off-by: Zqiang <qiang.zhang@windriver.com>
---
kernel/rcu/tree.c | 25 +++++++++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index c1ae1e52f638..ec098910d80b 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3571,17 +3571,40 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
}
EXPORT_SYMBOL_GPL(kvfree_call_rcu);
+static int free_krc_page_cache(struct kfree_rcu_cpu *krcp)
+{
+ unsigned long flags;
+ struct kvfree_rcu_bulk_data *bnode;
+ int i;
+
+ for (i = 0; i < rcu_min_cached_objs; i++) {
+ raw_spin_lock_irqsave(&krcp->lock, flags);
+ bnode = get_cached_bnode(krcp);
+ raw_spin_unlock_irqrestore(&krcp->lock, flags);
+ if (!bnode)
+ break;
+ free_page((unsigned long)bnode);
+ }
+
+ return i;
+}
+
static unsigned long
kfree_rcu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
int cpu;
unsigned long count = 0;
+ unsigned long flags;
/* Snapshot count of all CPUs */
for_each_possible_cpu(cpu) {
struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu);
count += READ_ONCE(krcp->count);
+
+ raw_spin_lock_irqsave(&krcp->lock, flags);
+ count += krcp->nr_bkv_objs;
+ raw_spin_unlock_irqrestore(&krcp->lock, flags);
}
return count;
@@ -3598,6 +3621,8 @@ kfree_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu);
count = krcp->count;
+ count += free_krc_page_cache(krcp);
+
raw_spin_lock_irqsave(&krcp->lock, flags);
if (krcp->monitor_todo)
kfree_rcu_drain_unlock(krcp, flags);
--
2.17.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* 回复: [PATCH v2] kvfree_rcu: Release page cache under memory pressure
[not found] ` <20210129141953.GA29827@pc638.lan>
@ 2021-01-30 6:47 ` Zhang, Qiang
2021-01-30 11:21 ` Uladzislau Rezki
0 siblings, 1 reply; 3+ messages in thread
From: Zhang, Qiang @ 2021-01-30 6:47 UTC (permalink / raw)
To: Uladzislau Rezki; +Cc: paulmck, joel, rcu, linux-kernel
________________________________________
发件人: Uladzislau Rezki <urezki@gmail.com>
发送时间: 2021年1月29日 22:19
收件人: Zhang, Qiang
抄送: urezki@gmail.com; paulmck@kernel.org; joel@joelfernandes.org; rcu@vger.kernel.org; linux-kernel@vger.kernel.org
主题: Re: [PATCH v2] kvfree_rcu: Release page cache under memory pressure
[Please note: This e-mail is from an EXTERNAL e-mail address]
On Fri, Jan 29, 2021 at 04:04:42PM +0800, qiang.zhang@windriver.com wrote:
> From: Zqiang <qiang.zhang@windriver.com>
>
> Add free per-cpu existing krcp's page cache operation, when
> the system is under memory pressure.
>
> Signed-off-by: Zqiang <qiang.zhang@windriver.com>
> ---
> kernel/rcu/tree.c | 25 +++++++++++++++++++++++++
> 1 file changed, 25 insertions(+)
>
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index c1ae1e52f638..ec098910d80b 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -3571,17 +3571,40 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> }
> EXPORT_SYMBOL_GPL(kvfree_call_rcu);
>
> +static int free_krc_page_cache(struct kfree_rcu_cpu *krcp)
> +{
> + unsigned long flags;
> + struct kvfree_rcu_bulk_data *bnode;
> + int i;
> +
> + for (i = 0; i < rcu_min_cached_objs; i++) {
> + raw_spin_lock_irqsave(&krcp->lock, flags);
>I am not sure why we should disable IRQs. I think it can be >avoided.
Suppose in multi CPU system, the kfree_rcu_shrink_scan function is runing on CPU2,
and we just traverse to CPU2, and then call free_krc_page_cache function,
if not disable irq, a interrupt may be occurs on CPU2 after the CPU2 corresponds to krcp variable 's lock be acquired, if the interrupt or softirq handler function to call kvfree_rcu function, in this function , acquire CPU2 corresponds to krcp variable 's lock , will happen deadlock.
Or in single CPU scenario.
> + bnode = get_cached_bnode(krcp);
> + raw_spin_unlock_irqrestore(&krcp->lock, flags);
> + if (!bnode)
> + break;
> + free_page((unsigned long)bnode);
> + }
> +
> + return i;
> +}
>Also i forgot to add in my previous comment to this path. Can we >access
>to page cache once and then do the drain work? I mean if we had >100 objects
>in the cache we would need to access to a krcp->lock 100 times.
>
>What about something like below:
>
><snip>
>static int free_krc_page_cache(struct kfree_rcu_cpu *krcp)
>{
> struct llist_node *page_list, *pos, *n;
> int freed = 0;
>
> raw_spin_lock(&krcp->lock);
> page_list = llist_del_all(&krcp->bkvcache);
> krcp->nr_bkv_objs = 0;
> raw_spin_unlock(&krcp->lock);
>
> llist_for_each_safe(pos, n, page_list) {
> free_page((unsigned long) pos);
> freed++;
> }
>
> return freed;
>}
><snip>
this change looks better.
Thanks
Qiang
> +
> static unsigned long
> kfree_rcu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
> {
> int cpu;
> unsigned long count = 0;
> + unsigned long flags;
>
> /* Snapshot count of all CPUs */
> for_each_possible_cpu(cpu) {
> struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu);
>
> count += READ_ONCE(krcp->count);
> +
> + raw_spin_lock_irqsave(&krcp->lock, flags);
> + count += krcp->nr_bkv_objs;
> + raw_spin_unlock_irqrestore(&krcp->lock, flags);
>Should we disable irqs?
>
> return count;
> @@ -3598,6 +3621,8 @@ kfree_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
> struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu);
>
> count = krcp->count;
> + count += free_krc_page_cache(krcp);
> +
> raw_spin_lock_irqsave(&krcp->lock, flags);
> if (krcp->monitor_todo)
> kfree_rcu_drain_unlock(krcp, flags);
> --
> 2.17.1
Thanks!
--
Vlad Rezki
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: 回复: [PATCH v2] kvfree_rcu: Release page cache under memory pressure
2021-01-30 6:47 ` 回复: " Zhang, Qiang
@ 2021-01-30 11:21 ` Uladzislau Rezki
0 siblings, 0 replies; 3+ messages in thread
From: Uladzislau Rezki @ 2021-01-30 11:21 UTC (permalink / raw)
To: Zhang, Qiang; +Cc: Uladzislau Rezki, paulmck, joel, rcu, linux-kernel
On Sat, Jan 30, 2021 at 06:47:31AM +0000, Zhang, Qiang wrote:
>
>
> ________________________________________
> 发件人: Uladzislau Rezki <urezki@gmail.com>
> 发送时间: 2021年1月29日 22:19
> 收件人: Zhang, Qiang
> 抄送: urezki@gmail.com; paulmck@kernel.org; joel@joelfernandes.org; rcu@vger.kernel.org; linux-kernel@vger.kernel.org
> 主题: Re: [PATCH v2] kvfree_rcu: Release page cache under memory pressure
>
> [Please note: This e-mail is from an EXTERNAL e-mail address]
>
> On Fri, Jan 29, 2021 at 04:04:42PM +0800, qiang.zhang@windriver.com wrote:
> > From: Zqiang <qiang.zhang@windriver.com>
> >
> > Add free per-cpu existing krcp's page cache operation, when
> > the system is under memory pressure.
> >
> > Signed-off-by: Zqiang <qiang.zhang@windriver.com>
> > ---
> > kernel/rcu/tree.c | 25 +++++++++++++++++++++++++
> > 1 file changed, 25 insertions(+)
> >
> > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > index c1ae1e52f638..ec098910d80b 100644
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -3571,17 +3571,40 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > }
> > EXPORT_SYMBOL_GPL(kvfree_call_rcu);
> >
> > +static int free_krc_page_cache(struct kfree_rcu_cpu *krcp)
> > +{
> > + unsigned long flags;
> > + struct kvfree_rcu_bulk_data *bnode;
> > + int i;
> > +
> > + for (i = 0; i < rcu_min_cached_objs; i++) {
> > + raw_spin_lock_irqsave(&krcp->lock, flags);
> >I am not sure why we should disable IRQs. I think it can be >avoided.
>
> Suppose in multi CPU system, the kfree_rcu_shrink_scan function is runing on CPU2,
> and we just traverse to CPU2, and then call free_krc_page_cache function,
> if not disable irq, a interrupt may be occurs on CPU2 after the CPU2 corresponds to krcp variable 's lock be acquired, if the interrupt or softirq handler function to call kvfree_rcu function, in this function , acquire CPU2 corresponds to krcp variable 's lock , will happen deadlock.
> Or in single CPU scenario.
>
Right. Deadlock scenario. It went away from my head during writing that :)
Thanks!
--
Vlad Rezki
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-01-30 11:22 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-29 8:04 [PATCH v2] kvfree_rcu: Release page cache under memory pressure qiang.zhang
[not found] ` <20210129141953.GA29827@pc638.lan>
2021-01-30 6:47 ` 回复: " Zhang, Qiang
2021-01-30 11:21 ` Uladzislau Rezki
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).