Re: [Propose] Isolate core_pattern in mnt namespace.

[Dongsheng Yang]

On 12/22/2015 05:52 AM, Eric W. Biederman wrote:
> Dongsheng Yang <yangds.fnst@xxxxxxxxxxxxxx> writes:
>
> > On 12/20/2015 05:47 PM, Eric W. Biederman wrote:
> > > Dongsheng Yang <yangds.fnst@xxxxxxxxxxxxxx> writes:
> > >
> > > > On 12/20/2015 10:37 AM, Al Viro wrote:
> > > > > On Sun, Dec 20, 2015 at 10:14:29AM +0800, Dongsheng Yang wrote:
> > > > > > On 12/17/2015 07:23 PM, Dongsheng Yang wrote:
> > > > > > > Hi guys,
> > > > > > >        We are working on making core dump behaviour isolated in
> > > > > > > container. But the problem is, the /proc/sys/kernel/core_pattern
> > > > > > > is a kernel wide setting, not belongs to a container.
> > > > > > >
> > > > > > >        So we want to add core_pattern into mnt namespace. What
> > > > > > > do you think about it?
> > > > > >
> > > > > > Hi Eric,
> > > > > > 	I found your patch about "net: Implement the per network namespace
> > > > > > sysctl infrastructure", I want to do the similar thing
> > > > > > in mnt namespace. Is that suggested way?
> > > > >
> > > > > Why mnt namespace and not something else?
> > > >
> > > > Hi Al,
> > > >
> > > > Well, because core_pattern indicates the path to store core file.
> > > > In different mnt namespace, we would like to change the path with
> > > > different value.
> > > >
> > > > In addition, Let's considering other namespaces:
> > > > UTS ns: contains informations of kernel and arch, not proper for core_pattern.
> > > > IPC ns: communication informations, not proper for core_pattern
> > > > PID ns: core_pattern is not related with pid
> > > > net ns: obviousely no.
> > > > user ns: not proper too.
> > > >
> > > > Then I believe it's better to do this in mnt namespace. of course,
> > > > core_pattern is just one example. After this infrastructure finished,
> > > > we can implement more sysctls as per-mnt if necessary, I think.
> > > >
> > > > Al, what do you think about this idea?
> > >
> > > The hard part is not the sysctl.  The hard part is starting the usermode
> > > helper, in an environment that it can deal with.  The mount namespace
> > > really provides you with no help there.
> >
> > Do you mean the core dump helper? But I think I don't want to touch it
> > in my development. I think I can use non-pipe way to get what I want,
> > Let me try to explain what I want here.
> >
> > (1). introduce a --core-path option in docker run command to specify the
> > path in host to store core file in one container.
> > E.g: docker run --core-path=/core/test --name=test IMAGE
> >
> > (2). When the container starting, docker attach a volume to it, similar
> > with "-v /core/test:/var/lib/docker/coredump". That means, the path of
> > /var/lib/docker/coredump in container is a link to /core/test in host.
> >
> > (3). Set the /proc/sys/kernel/core_pattern in container as
> > "/var/lib/docker/coredump". But that should not affect the core_pattern
> > in host or other containers.
> >
> > Then I think I can collect the core files from each container and save
> > them in the paths where I want.
>
> For your case that sounds like it would work.  Unfortunately for this to
> be generally applicable and to let the OS in the contianer control it's
> fate the core dump pattern needs to be supported.
>
> Otherwise something clever in userspace that can be written now should
> be sufficient to fill the gap.  There is enough information for the user
> mode helper to implement the policy you would like today.

Hi Eric,

To make sure I understand your point correctlyï
	Do you mean we can write a userspace helper in host such as
/usr/libexec/docker-pipe to get what I want?

Yes, I would say, for my case, it would work. This helper can get the
dump data from containers and dispatch them to different path such
as /var/lib/docker/cores/<ContainerID>/.

But there would be two problems in this solution.
(1). It may affect core dump on host. Normally, other processes in
 host would not be happy to use a helper of docker-pipe for themselves.
 But host have to share the core_pattern with containers, can't config
 it by itself.

(2). If there are some containers don't want to pass the core files
to host, they can't set the core_pattern in this solution.

IMO, we can get core files on host currently, by either non-pipe way I
described above or the pipe way you suggested. But the problem is both
of these methods would affect the core_pattern on host and other
containers.

So, I think the key point here is just isolating the core dump related
sysctl in mnt namespace.

Thanx
Yang
> Eric
>
>
> .



--
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/