[PATCH] Ipv6: Delete rcu_read_lock

[PATCH] Ipv6: Delete rcu_read_lock

[rongqing.li]

From: Roy.Li <rongqing.li@windriver.com>

Delete rcu_read_lock, it is redundant.

Signed-off-by: Roy.Li <rongqing.li@windriver.com>
---
 net/ipv6/ndisc.c |    3 ---
 1 files changed, 0 insertions(+), 3 deletions(-)

diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
index 9da6e02..dd633ff 100644
--- a/net/ipv6/ndisc.c
+++ b/net/ipv6/ndisc.c
@@ -370,17 +370,14 @@ static int ndisc_constructor(struct neighbour *neigh)
 	struct neigh_parms *parms;
 	int is_multicast = ipv6_addr_is_multicast(addr);
 
-	rcu_read_lock();
 	in6_dev = in6_dev_get(dev);
 	if (in6_dev == NULL) {
-		rcu_read_unlock();
 		return -EINVAL;
 	}
 
 	parms = in6_dev->nd_parms;
 	__neigh_parms_put(neigh->parms);
 	neigh->parms = neigh_parms_clone(parms);
-	rcu_read_unlock();
 
 	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
 	if (!dev->header_ops) {
-- 
1.7.1

Re: [PATCH] Ipv6: Delete rcu_read_lock

[Yan, Zheng]

On Thu, Sep 29, 2011 at 3:46 PM,  <rongqing.li@windriver.com> wrote:
> From: Roy.Li <rongqing.li@windriver.com>
>
> Delete rcu_read_lock, it is redundant.
>
> Signed-off-by: Roy.Li <rongqing.li@windriver.com>
> ---
>  net/ipv6/ndisc.c |    3 ---
>  1 files changed, 0 insertions(+), 3 deletions(-)
>
> diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
> index 9da6e02..dd633ff 100644
> --- a/net/ipv6/ndisc.c
> +++ b/net/ipv6/ndisc.c
> @@ -370,17 +370,14 @@ static int ndisc_constructor(struct neighbour *neigh)
>        struct neigh_parms *parms;
>        int is_multicast = ipv6_addr_is_multicast(addr);
>
> -       rcu_read_lock();
>        in6_dev = in6_dev_get(dev);
>        if (in6_dev == NULL) {
> -               rcu_read_unlock();
>                return -EINVAL;
>        }
>
>        parms = in6_dev->nd_parms;
>        __neigh_parms_put(neigh->parms);
>        neigh->parms = neigh_parms_clone(parms);
> -       rcu_read_unlock();
>
>        neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
>        if (!dev->header_ops) {
> --

I think the rcu_read_lock protects in6_dev->nd_parms.

Yan, Zheng

Re: [PATCH] Ipv6: Delete rcu_read_lock

[Rongqing Li]

On 09/29/2011 04:16 PM, Yan, Zheng wrote:
> I think the rcu_read_lock protects in6_dev->nd_parms.
>
It can not protect in6_dev->nd_parms.

rcu_read_lock protects the data which is accessed by rcu_XXX,


-Roy

> Yan, Zheng

-- 
Best Reagrds,
Roy | RongQing Li

Re: [PATCH] Ipv6: Delete rcu_read_lock

[Eric Dumazet]

Le jeudi 29 septembre 2011 à 15:46 +0800, rongqing.li@windriver.com a
écrit :
> From: Roy.Li <rongqing.li@windriver.com>
> 
> Delete rcu_read_lock, it is redundant.
> 

A good changelog and patch title are required.

Something like :

[PATCH net-next] ipv6: remove a rcu_read_lock in ndisc_constructor

in6_dev_get(dev) takes a reference on struct inet6_dev, we dont need rcu
locking in ndisc_constructor()



-----

Or, you could try to really use RCU and not take a reference (ie, use 
__in6_dev_get() instead of in6_dev_get(dev).


Untested patch :

diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
index 9da6e02..a4f4472 100644
--- a/net/ipv6/ndisc.c
+++ b/net/ipv6/ndisc.c
@@ -371,7 +371,7 @@ static int ndisc_constructor(struct neighbour *neigh)
 	int is_multicast = ipv6_addr_is_multicast(addr);
 
 	rcu_read_lock();
-	in6_dev = in6_dev_get(dev);
+	in6_dev = __in6_dev_get(dev);
 	if (in6_dev == NULL) {
 		rcu_read_unlock();
 		return -EINVAL;
@@ -380,7 +380,6 @@ static int ndisc_constructor(struct neighbour *neigh)
 	parms = in6_dev->nd_parms;
 	__neigh_parms_put(neigh->parms);
 	neigh->parms = neigh_parms_clone(parms);
-	rcu_read_unlock();
 
 	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
 	if (!dev->header_ops) {
@@ -409,7 +408,7 @@ static int ndisc_constructor(struct neighbour *neigh)
 		else
 			neigh->output = neigh->ops->output;
 	}
-	in6_dev_put(in6_dev);
+	rcu_read_unlock();
 	return 0;
 }
 

Re: [PATCH] Ipv6: Delete rcu_read_lock

[Rongqing Li]

On 09/29/2011 04:44 PM, Eric Dumazet wrote:
> A good changelog and patch title are required.
>
> Something like :
>
> [PATCH net-next] ipv6: remove a rcu_read_lock in ndisc_constructor
>
> in6_dev_get(dev) takes a reference on struct inet6_dev, we dont need rcu
> locking in ndisc_constructor()
>
>

Thank you, I will resend it.

-Roy

Re: [PATCH] Ipv6: Delete rcu_read_lock

[Yan, Zheng]

On Thu, Sep 29, 2011 at 4:25 PM, Rongqing Li <rongqing.li@windriver.com> wrote:
> On 09/29/2011 04:16 PM, Yan, Zheng wrote:
>>
>> I think the rcu_read_lock protects in6_dev->nd_parms.
>>
> It can not protect in6_dev->nd_parms.
>
> rcu_read_lock protects the data which is accessed by rcu_XXX,
>

I'm confused, why does neigh_parms_release() use call_rcu() to release nd_parms.

Re: [PATCH] Ipv6: Delete rcu_read_lock

[Rongqing Li]

On 09/29/2011 04:52 PM, Yan, Zheng wrote:
> On Thu, Sep 29, 2011 at 4:25 PM, Rongqing Li<rongqing.li@windriver.com>  wrote:
>> On 09/29/2011 04:16 PM, Yan, Zheng wrote:
>>>
>>> I think the rcu_read_lock protects in6_dev->nd_parms.
>>>
>> It can not protect in6_dev->nd_parms.
>>
>> rcu_read_lock protects the data which is accessed by rcu_XXX,
>>
>
> I'm confused, why does neigh_parms_release() use call_rcu() to release nd_parms.
>

Same as idev, there is a reference on struct neigh_parms
which protects them

-Roy