net 2.6.25

net 2.6.25

[Denis V. Lunev]

Hello, guys!

David has been opened net-2.6.25 tree yesterday, so we could send a new
portion of NETNS there. If there are no objections, I'll send first six
patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
Unix sockets virtualization + some our /proc cleanups.

Regards,
	Den

Re: net 2.6.25

[Eric W. Biederman]

"Denis V. Lunev" <den-3ImXcnM4P+0@public.gmane.org> writes:

> Hello, guys!
>
> David has been opened net-2.6.25 tree yesterday, so we could send a new
> portion of NETNS there. If there are no objections, I'll send first six
> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
> Unix sockets virtualization + some our /proc cleanups.

I think I can pretty much agree on that set of patches.  However I do strongly
object to the form they are in at: http://git.openvz.org/?p=linux-2.6-netns

All of my subject lines (which are the primary summary of the patch) have
been lost.  So it is much more difficult to see what the patch is doing.

I will get some sleep and see what I can sort out.

Eric

Re: net 2.6.25

[Denis V. Lunev]

Will you be happy, if I'll re-merge all of them from scratch preserving
original subjects from the letters?

Regards,
	Den

Eric W. Biederman wrote:
> "Denis V. Lunev" <den-3ImXcnM4P+0@public.gmane.org> writes:
> 
>> Hello, guys!
>>
>> David has been opened net-2.6.25 tree yesterday, so we could send a new
>> portion of NETNS there. If there are no objections, I'll send first six
>> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
>> Unix sockets virtualization + some our /proc cleanups.
> 
> I think I can pretty much agree on that set of patches.  However I do strongly
> object to the form they are in at: http://git.openvz.org/?p=linux-2.6-netns
> 
> All of my subject lines (which are the primary summary of the patch) have
> been lost.  So it is much more difficult to see what the patch is doing.
> 
> I will get some sleep and see what I can sort out.
> 
> Eric
> 

Re: net 2.6.25

[Daniel Lezcano]

Denis V. Lunev wrote:
> Hello, guys!
> 
> David has been opened net-2.6.25 tree yesterday, so we could send a new
> portion of NETNS there. If there are no objections, I'll send first six
> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
> Unix sockets virtualization + some our /proc cleanups.

Did Pavel had time to look at RCU looking mechanism to replace 
rtnl_lock/rtnl_unlock to protect the netns list ?

ps: isn't the openvz mailing list already subscribed to containers@

Re: net 2.6.25

[Denis V. Lunev]

Daniel Lezcano wrote:
> Denis V. Lunev wrote:
>> Hello, guys!
>>
>> David has been opened net-2.6.25 tree yesterday, so we could send a new
>> portion of NETNS there. If there are no objections, I'll send first six
>> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
>> Unix sockets virtualization + some our /proc cleanups.
> 
> Did Pavel had time to look at RCU looking mechanism to replace
> rtnl_lock/rtnl_unlock to protect the netns list ?
not yet

> ps: isn't the openvz mailing list already subscribed to containers@
hmm :) Pavel say yes :) trying...

Re: [Devel] Re: net 2.6.25

[Pavel Emelyanov]

Daniel Lezcano wrote:
> Denis V. Lunev wrote:
>> Hello, guys!
>>
>> David has been opened net-2.6.25 tree yesterday, so we could send a new
>> portion of NETNS there. If there are no objections, I'll send first six
>> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
>> Unix sockets virtualization + some our /proc cleanups.
> 
> Did Pavel had time to look at RCU looking mechanism to replace 
> rtnl_lock/rtnl_unlock to protect the netns list ?

Unfortunately, I had not. Andrew has released the rc2-mm1 tree, so
I'm now trying to prepare all the patches I wanted for this mm.

I will revisit this problem, but Denis said, that the problem we're
trying to solve with the RCU protection can be solved w/o it, no?

> ps: isn't the openvz mailing list already subscribed to containers@

devel-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org is.

> _______________________________________________
> Devel mailing list
> Devel-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org
> https://openvz.org/mailman/listinfo/devel
> 

Re: [Devel] Re: net 2.6.25

[Daniel Lezcano]

Pavel Emelyanov wrote:
> Daniel Lezcano wrote:
>> Denis V. Lunev wrote:
>>> Hello, guys!
>>>
>>> David has been opened net-2.6.25 tree yesterday, so we could send a new
>>> portion of NETNS there. If there are no objections, I'll send first six
>>> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
>>> Unix sockets virtualization + some our /proc cleanups.
>> Did Pavel had time to look at RCU looking mechanism to replace 
>> rtnl_lock/rtnl_unlock to protect the netns list ?
> 
> Unfortunately, I had not. Andrew has released the rc2-mm1 tree, so
> I'm now trying to prepare all the patches I wanted for this mm.
> 
> I will revisit this problem, but Denis said, that the problem we're
> trying to solve with the RCU protection can be solved w/o it, no?

Yes, if we create a timer per network namespace for ipv6 garbage 
collecting, we can avoid to use a rcu, in this case.

I am doing that right now, but I am afraid we can fall into another case 
where we need to browse the network list and I thought having the RCU 
will bring more flexibility.

But, we can live without rcu and go ahead with ipv6 anyway :)

Re: net 2.6.25

[Eric W. Biederman]

"Denis V. Lunev" <den-3ImXcnM4P+0@public.gmane.org> writes:

> Daniel Lezcano wrote:
>> Denis V. Lunev wrote:
>>> Hello, guys!
>>>
>>> David has been opened net-2.6.25 tree yesterday, so we could send a new
>>> portion of NETNS there. If there are no objections, I'll send first six
>>> patches from OpenVZ GIT tomorrow, i.e. RTNL infrastructure, Packet &
>>> Unix sockets virtualization + some our /proc cleanups.
>> 
>> Did Pavel had time to look at RCU looking mechanism to replace
>> rtnl_lock/rtnl_unlock to protect the netns list ?
> not yet
>
>> ps: isn't the openvz mailing list already subscribed to containers@
> hmm :) Pavel say yes :) trying...

Well honestly I think I would rather have a spin lock then an rcu thing.

It may not be a big deal because we already have to check the count and
if it is zero ignore the namespace, so rcu may not add complications but
changing something that already has interesting usage and lifetime rules
to rcu sends up warning signs for me.

Eric

Re: net 2.6.25

[Eric W. Biederman]

"Denis V. Lunev" <den-3ImXcnM4P+0@public.gmane.org> writes:

> Will you be happy, if I'll re-merge all of them from scratch preserving
> original subjects from the letters?

Yes that is my big issue.

Eric

Re: [Devel] Re: net 2.6.25

[Kir Kolyshkin]

Pavel Emelyanov wrote:
> Daniel Lezcano wrote:
>   
>> Denis V. Lunev wrote:
>>     
>
>> ps: isn't the openvz mailing list already subscribed to containers@
>>     
>
> devel-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org is.
>
>   

That shouldn't stop you from posting to both -- mailman is smart enough 
to cope with two copies of the same message, so there will be no dupes.