[PATCH 0/2] SYSCTL: export root handling routines

[PATCH 0/2] SYSCTL: export root handling routines

[Stanislav Kinsbursky]

These rouitines are required to make SUNRPC sysctl's network-namespace-aware.

The following series consists of:

---

Stanislav Kinsbursky (2):
      SYSCTL: root unregister routine introduced
      SYSCTL: export root register and unregister routines


 kernel/sysctl.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

[PATCH 1/2] SYSCTL: root unregister routine introduced

[Stanislav Kinsbursky]

This routine is required for SUNRPC sysctl's, which are going to be allocated,
processed and destroyed per network namespace context.
IOW, new sysctl root will be registered on network namespace creation and
thus have to unregistered before network namespace destruction.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>

---
 kernel/sysctl.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index ae27196..21e68c1 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -1701,6 +1701,13 @@ void register_sysctl_root(struct ctl_table_root *root)
 	spin_unlock(&sysctl_lock);
 }
 
+void unregister_sysctl_root(struct ctl_table_root *root)
+{
+	spin_lock(&sysctl_lock);
+	list_del(&root->root_list);
+	spin_unlock(&sysctl_lock);
+}
+
 /*
  * sysctl_perm does NOT grant the superuser all rights automatically, because
  * some sysctl variables are readonly even to root.

[PATCH 2/2] SYSCTL: export root register and unregister routines

[Stanislav Kinsbursky]

These routines will be used in SUNPRC module and thus have to be exported.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>

---
 kernel/sysctl.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 21e68c1..0ee5b73 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -1700,6 +1700,7 @@ void register_sysctl_root(struct ctl_table_root *root)
 	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
 	spin_unlock(&sysctl_lock);
 }
+EXPORT_SYMBOL_GPL(register_sysctl_root);
 
 void unregister_sysctl_root(struct ctl_table_root *root)
 {
@@ -1707,6 +1708,7 @@ void unregister_sysctl_root(struct ctl_table_root *root)
 	list_del(&root->root_list);
 	spin_unlock(&sysctl_lock);
 }
+EXPORT_SYMBOL_GPL(unregister_sysctl_root);
 
 /*
  * sysctl_perm does NOT grant the superuser all rights automatically, because

Re: [PATCH 1/2] SYSCTL: root unregister routine introduced

[Andrew Morton]

On Mon, 12 Dec 2011 21:50:00 +0300
Stanislav Kinsbursky <skinsbursky@parallels.com> wrote:

> This routine is required for SUNRPC sysctl's, which are going to be allocated,
> processed and destroyed per network namespace context.
> IOW, new sysctl root will be registered on network namespace creation and
> thus have to unregistered before network namespace destruction.
> 

It's a bit suspicious that such a mature subsystem as sysctl newly
needs its internals exported like this.  Either a) the net namespaces
work is doing something which hasn't been done before or b) it is doing
something wrong.

So, please explain further so we can confirm that it is a) and not b).

> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -1701,6 +1701,13 @@ void register_sysctl_root(struct ctl_table_root *root)
>  	spin_unlock(&sysctl_lock);
>  }
>  
> +void unregister_sysctl_root(struct ctl_table_root *root)
> +{
> +	spin_lock(&sysctl_lock);
> +	list_del(&root->root_list);
> +	spin_unlock(&sysctl_lock);
> +}
> +

This requires the addition of a declaration to include/linux/sysctl.h.

Once that is done and review is complete, I'd suggest that these two
patches be joined into a single patch, and that patch become part of
whatever patch series it is which needs them.

Re: [PATCH 1/2] SYSCTL: root unregister routine introduced

[Stanislav Kinsbursky]

13.12.2011 02:52, Andrew Morton пишет:
> On Mon, 12 Dec 2011 21:50:00 +0300
> Stanislav Kinsbursky<skinsbursky@parallels.com>  wrote:
>
>> This routine is required for SUNRPC sysctl's, which are going to be allocated,
>> processed and destroyed per network namespace context.
>> IOW, new sysctl root will be registered on network namespace creation and
>> thus have to unregistered before network namespace destruction.
>>
>
> It's a bit suspicious that such a mature subsystem as sysctl newly
> needs its internals exported like this.  Either a) the net namespaces
> work is doing something which hasn't been done before or b) it is doing
> something wrong.
>
> So, please explain further so we can confirm that it is a) and not b).
>

Hello, Andrew.
The goal is to provide an ability to control and modify data by sysctl's in 
network namespace context. This is done by "net" sysctl's.
But there are two more issues to solve:
1) Sysctl's have to be in /proc/sys/sunrpc
2) Sysctl's content should be accessible from creator's network context (not 
current user ones's).

>> --- a/kernel/sysctl.c
>> +++ b/kernel/sysctl.c
>> @@ -1701,6 +1701,13 @@ void register_sysctl_root(struct ctl_table_root *root)
>>   	spin_unlock(&sysctl_lock);
>>   }
>>
>> +void unregister_sysctl_root(struct ctl_table_root *root)
>> +{
>> +	spin_lock(&sysctl_lock);
>> +	list_del(&root->root_list);
>> +	spin_unlock(&sysctl_lock);
>> +}
>> +
>
> This requires the addition of a declaration to include/linux/sysctl.h.
>
> Once that is done and review is complete, I'd suggest that these two
> patches be joined into a single patch, and that patch become part of
> whatever patch series it is which needs them.
>

Ok, I'll do so.

-- 
Best regards,
Stanislav Kinsbursky

Re: [PATCH 1/2] SYSCTL: root unregister routine introduced

[Eric W. Biederman]

Stanislav Kinsbursky <skinsbursky@parallels.com> writes:

> 13.12.2011 02:52, Andrew Morton пишет:
>> On Mon, 12 Dec 2011 21:50:00 +0300
>> Stanislav Kinsbursky<skinsbursky@parallels.com>  wrote:
>>
>>> This routine is required for SUNRPC sysctl's, which are going to be allocated,
>>> processed and destroyed per network namespace context.
>>> IOW, new sysctl root will be registered on network namespace creation and
>>> thus have to unregistered before network namespace destruction.
>>>
>>
>> It's a bit suspicious that such a mature subsystem as sysctl newly
>> needs its internals exported like this.  Either a) the net namespaces
>> work is doing something which hasn't been done before or b) it is doing
>> something wrong.
>>
>> So, please explain further so we can confirm that it is a) and not b).
>>
>
> Hello, Andrew.
> The goal is to provide an ability to control and modify data by sysctl's in
> network namespace context. This is done by "net" sysctl's.
> But there are two more issues to solve:
> 1) Sysctl's have to be in /proc/sys/sunrpc

The sysctl root has nothing to with what directory the files show up in,
so this should not be an issue.

> 2) Sysctl's content should be accessible from creator's network context (not
> current user ones's).

Making the sunrpc sysctls per network namespace would seem to address
this.    I don't see why you would need a new root to handle this case.

Eric