2020-01-29 18:27:20

by Davidlohr Bueso

Subject: [PATCH 7/8] usb, cdc-acm: optimize barrier usage for Rmw atomic bitops

We can avoid the unnecessary barrier on non LL/SC architectures,
such as x86. Instead, use the smp_mb__after_atomic() call.

Signed-off-by: Davidlohr Bueso <[email protected]>
---
drivers/usb/class/cdc-acm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/usb/class/cdc-acm.c b/drivers/usb/class/cdc-acm.c
index 62f4fb9b362f..0bc4b558f387 100644
--- a/drivers/usb/class/cdc-acm.c
+++ b/drivers/usb/class/cdc-acm.c
@@ -842,7 +842,7 @@ static void acm_tty_unthrottle(struct tty_struct *tty)
clear_bit(ACM_THROTTLED, &acm->flags);

/* Matches the smp_mb__after_atomic() in acm_read_bulk_callback(). */
- smp_mb();
+ smp_mb__after_atomic();

acm_submit_read_urbs(acm, GFP_KERNEL);
}
--
2.16.4

2020-01-30 10:08:25

by Johan Hovold

Subject: Re: [PATCH 7/8] usb, cdc-acm: optimize barrier usage for Rmw atomic bitops

On Wed, Jan 29, 2020 at 10:15:45AM -0800, Davidlohr Bueso wrote:
> We can avoid the unnecessary barrier on non LL/SC architectures,
> such as x86. Instead, use the smp_mb__after_atomic() call.
>
> Signed-off-by: Davidlohr Bueso <[email protected]>

Reviewed-by: Johan Hovold <[email protected]>

> ---
> drivers/usb/class/cdc-acm.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/usb/class/cdc-acm.c b/drivers/usb/class/cdc-acm.c
> index 62f4fb9b362f..0bc4b558f387 100644
> --- a/drivers/usb/class/cdc-acm.c
> +++ b/drivers/usb/class/cdc-acm.c
> @@ -842,7 +842,7 @@ static void acm_tty_unthrottle(struct tty_struct *tty)
> clear_bit(ACM_THROTTLED, &acm->flags);
>
> /* Matches the smp_mb__after_atomic() in acm_read_bulk_callback(). */
> - smp_mb();
> + smp_mb__after_atomic();
>
> acm_submit_read_urbs(acm, GFP_KERNEL);
> }

We have another barrier in USB serial which can also be relaxed this
way. I'll post a patch shortly.

Johan

2020-03-23 17:31:58

by Davidlohr Bueso

Subject: Re: [PATCH 7/8] usb, cdc-acm: optimize barrier usage for Rmw atomic bitops

ping? Can these please be picked up?

On Wed, 29 Jan 2020, Davidlohr Bueso wrote:

>We can avoid the unnecessary barrier on non LL/SC architectures,
>such as x86. Instead, use the smp_mb__after_atomic() call.
>
>Signed-off-by: Davidlohr Bueso <[email protected]>
>---
> drivers/usb/class/cdc-acm.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
>diff --git a/drivers/usb/class/cdc-acm.c b/drivers/usb/class/cdc-acm.c
>index 62f4fb9b362f..0bc4b558f387 100644
>--- a/drivers/usb/class/cdc-acm.c
>+++ b/drivers/usb/class/cdc-acm.c
>@@ -842,7 +842,7 @@ static void acm_tty_unthrottle(struct tty_struct *tty)
> clear_bit(ACM_THROTTLED, &acm->flags);
>
> /* Matches the smp_mb__after_atomic() in acm_read_bulk_callback(). */
>- smp_mb();
>+ smp_mb__after_atomic();
>
> acm_submit_read_urbs(acm, GFP_KERNEL);
> }
>--
>2.16.4
>