2021-05-11 13:19:40

by Nikolay Borisov

Subject: [PATCH] percpu_ref: Don't opencode percpu_ref_is_dying

Signed-off-by: Nikolay Borisov <[email protected]>
---
lib/percpu-refcount.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
index a1071cdefb5a..af9302141bcf 100644
--- a/lib/percpu-refcount.c
+++ b/lib/percpu-refcount.c
@@ -275,7 +275,7 @@ static void __percpu_ref_switch_mode(struct percpu_ref *ref,
wait_event_lock_irq(percpu_ref_switch_waitq, !data->confirm_switch,
percpu_ref_switch_lock);

- if (data->force_atomic || (ref->percpu_count_ptr & __PERCPU_REF_DEAD))
+ if (data->force_atomic || percpu_ref_is_dying(ref))
__percpu_ref_switch_to_atomic(ref, confirm_switch);
else
__percpu_ref_switch_to_percpu(ref);
@@ -385,7 +385,7 @@ void percpu_ref_kill_and_confirm(struct percpu_ref *ref,

spin_lock_irqsave(&percpu_ref_switch_lock, flags);

- WARN_ONCE(ref->percpu_count_ptr & __PERCPU_REF_DEAD,
+ WARN_ONCE(percpu_ref_is_dying(ref),
"%s called more than once on %ps!", __func__,
ref->data->release);

@@ -465,7 +465,7 @@ void percpu_ref_resurrect(struct percpu_ref *ref)

spin_lock_irqsave(&percpu_ref_switch_lock, flags);

- WARN_ON_ONCE(!(ref->percpu_count_ptr & __PERCPU_REF_DEAD));
+ WARN_ON_ONCE(!percpu_ref_is_dying(ref));
WARN_ON_ONCE(__ref_is_percpu(ref, &percpu_count));

ref->percpu_count_ptr &= ~__PERCPU_REF_DEAD;
--
2.25.1

2021-05-11 13:52:26

by Dennis Zhou

Subject: Re: [PATCH] percpu_ref: Don't opencode percpu_ref_is_dying

Hello,

On Tue, May 11, 2021 at 04:16:33PM +0300, Nikolay Borisov wrote:
> Signed-off-by: Nikolay Borisov <[email protected]>
> ---
> lib/percpu-refcount.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
> index a1071cdefb5a..af9302141bcf 100644
> --- a/lib/percpu-refcount.c
> +++ b/lib/percpu-refcount.c
> @@ -275,7 +275,7 @@ static void __percpu_ref_switch_mode(struct percpu_ref *ref,
> wait_event_lock_irq(percpu_ref_switch_waitq, !data->confirm_switch,
> percpu_ref_switch_lock);
>
> - if (data->force_atomic || (ref->percpu_count_ptr & __PERCPU_REF_DEAD))
> + if (data->force_atomic || percpu_ref_is_dying(ref))
> __percpu_ref_switch_to_atomic(ref, confirm_switch);
> else
> __percpu_ref_switch_to_percpu(ref);
> @@ -385,7 +385,7 @@ void percpu_ref_kill_and_confirm(struct percpu_ref *ref,
>
> spin_lock_irqsave(&percpu_ref_switch_lock, flags);
>
> - WARN_ONCE(ref->percpu_count_ptr & __PERCPU_REF_DEAD,
> + WARN_ONCE(percpu_ref_is_dying(ref),
> "%s called more than once on %ps!", __func__,
> ref->data->release);
>
> @@ -465,7 +465,7 @@ void percpu_ref_resurrect(struct percpu_ref *ref)
>
> spin_lock_irqsave(&percpu_ref_switch_lock, flags);
>
> - WARN_ON_ONCE(!(ref->percpu_count_ptr & __PERCPU_REF_DEAD));
> + WARN_ON_ONCE(!percpu_ref_is_dying(ref));
> WARN_ON_ONCE(__ref_is_percpu(ref, &percpu_count));
>
> ref->percpu_count_ptr &= ~__PERCPU_REF_DEAD;
> --
> 2.25.1
>

If no one has any objections to me officially taking percpu_ref, I'll
pick this up.

Thanks,
Dennis

2021-05-13 04:49:28

by Dennis Zhou

Subject: Re: [PATCH] percpu_ref: Don't opencode percpu_ref_is_dying

On Tue, May 11, 2021 at 01:49:23PM +0000, Dennis Zhou wrote:
> Hello,
>
> On Tue, May 11, 2021 at 04:16:33PM +0300, Nikolay Borisov wrote:
> > Signed-off-by: Nikolay Borisov <[email protected]>
> > ---
> > lib/percpu-refcount.c | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
> > index a1071cdefb5a..af9302141bcf 100644
> > --- a/lib/percpu-refcount.c
> > +++ b/lib/percpu-refcount.c
> > @@ -275,7 +275,7 @@ static void __percpu_ref_switch_mode(struct percpu_ref *ref,
> > wait_event_lock_irq(percpu_ref_switch_waitq, !data->confirm_switch,
> > percpu_ref_switch_lock);
> >
> > - if (data->force_atomic || (ref->percpu_count_ptr & __PERCPU_REF_DEAD))
> > + if (data->force_atomic || percpu_ref_is_dying(ref))
> > __percpu_ref_switch_to_atomic(ref, confirm_switch);
> > else
> > __percpu_ref_switch_to_percpu(ref);
> > @@ -385,7 +385,7 @@ void percpu_ref_kill_and_confirm(struct percpu_ref *ref,
> >
> > spin_lock_irqsave(&percpu_ref_switch_lock, flags);
> >
> > - WARN_ONCE(ref->percpu_count_ptr & __PERCPU_REF_DEAD,
> > + WARN_ONCE(percpu_ref_is_dying(ref),
> > "%s called more than once on %ps!", __func__,
> > ref->data->release);
> >
> > @@ -465,7 +465,7 @@ void percpu_ref_resurrect(struct percpu_ref *ref)
> >
> > spin_lock_irqsave(&percpu_ref_switch_lock, flags);
> >
> > - WARN_ON_ONCE(!(ref->percpu_count_ptr & __PERCPU_REF_DEAD));
> > + WARN_ON_ONCE(!percpu_ref_is_dying(ref));
> > WARN_ON_ONCE(__ref_is_percpu(ref, &percpu_count));
> >
> > ref->percpu_count_ptr &= ~__PERCPU_REF_DEAD;
> > --
> > 2.25.1
> >
>
> If no one has any objections to me officially taking percpu_ref, I'll
> pick this up.
>
> Thanks,
> Dennis

I've applied this to for-5.13-fixes.

Thanks,
Dennis