Re: Network namespaces a path to mergable code.

Andrey Savochkin wrote:
On Tue, Jun 27, 2006 at 11:20:40AM -0600, Eric W. Biederman wrote:

Thinking about this I am going to suggest a slightly different direction
for get a patchset we can merge.

First we concentrate on the fundamentals.
- How we mark a device as belonging to a specific network namespace.
- How we mark a socket as belonging to a specific network namespace.


I agree with the direction of your thoughts.
I was trying to do a similar thing, define clear steps in network
namespace merging.

My first patchset covers devices but not sockets.
The only difference from what you're suggesting is ipv4 routing.
For me, it is not less important than devices and sockets. May be even
more important, since routing exposes design deficiencies less obvious at
socket level.


It sounds then like it would be a good start to have general socket
namespaces, if it would merge more easily - perhaps then network device
namespaces would fall into place more easily.

AIUI socket namespaces are also necessary for situations where you want
containers to share IP addresses. AIUI, PlanetLab do something like this
with a module atop of VServer already (but read
http://openvz.org/pipermail/devel/2006-June/000666.html for a proper
explanation from Mark Huang)

As part of the fundamentals we add a patch to the generic socket code
that by default will disable it for protocol families that do not indicate
support for handling network namespaces, on a non-default network namespace.


Fine

Can you summarize you objections against my way of handling devices, please?

There were many objections, the major one being the patch was too large for certainty of adequate review.

Quoting what I perceived as a summary from Eric:
When I went through this, my patchset just added an explicit
continue if the devices was not in the appropriate namespace.
I actually prefer the multiple list implementation but at
the same time I think it is harder to get a clean implementation
out of it.


You offered to re-do the patch without separate lists - I suggest that
this go ahead. No-one should really care; splitting it out into separate
lists can then be considered a performance optimization for later.

And what was the typo you referred to in your letter to Kirill Korotaev?

I think this is the comment he refers to:
These hunks should use for_each_netdev(ifp);


Both quotes are from http://lkml.org/lkml/2006/6/26/147

Though, in Kirill's defense, it seems a bit strange to expect him to
raise a fault that was just raised by Eric, in a reply to the message
where he raised it.

Sam.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Relevant Pages

  • [PATCH 1/1] IPN: Inter Process Networking
    ... +IPN is an Inter Process Communication service. ... +interface and protocols used for networking. ... +to a "network". ... +creates a communication socket. ...
    (Linux-Kernel)
  • Re: [PATCH 1/1] IPN: Inter Process Networking
    ... +IPN is an Inter Process Communication service. ... +interface and protocols used for networking. ... +to a "network". ... +creates a communication socket. ...
    (Linux-Kernel)
  • Re: Home Networking
    ... has 4 network sockets) in the roofspace and then wire a CAT5 network ... NTE5 master socket, so I don't need a filter at each phone outlet. ... A number of different suppliers offer such equipment, including Solwise whose range I use including their Homeplug ADSL Modem Router. ...
    (uk.telecom.broadband)
  • Re: multithreaded dialog application
    ... multiple threads, a rewrite of the piece of crap in the MSDN. ... the socket messages are posted to the main thread to be processed. ... Nearly all the socket messages are simply passed on to a child dialog ... If you have only one network interface then there ...
    (microsoft.public.vc.mfc)
  • Re: [patch 2/6] [Network namespace] Network device sharing by view
    ... the device names so that each guest can have a separate ... Example as a prefix: guest0-eth0. ... We really want the fundamental rule that a network device ... is tied to a single namespace, and that a socket is tied ...
    (Linux-Kernel)