just a question: why do we keep the fs (struct_fs)
outside of nsproxy?

TIA,
Herbert

_______________________________________________
Containers mailing list
Containers@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/containers

Quoting Herbert Poetzl (herbert@13thfloor.at):
> 
> just a question: why do we keep the fs (struct_fs)
> outside of nsproxy?

Good question.  So we have a mounts namespace, and you
would consider the per-process fs root to be an fs
namespace?  Practically, it would mean that chroot
and pivot_mount would create a new nsproxy, but i guess
that's not a real problem.

It might force us to stop our current lazy checks for
'current->nsproxy==&init_nsproxy', since the pivot_mount
in early boot would make that not true.

-serge
_______________________________________________
Containers mailing list
Containers@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/containers

Quoting Herbert Poetzl (herbert@13thfloor.at):
> On Fri, Dec 08, 2006 at 08:40:59AM -0600, Serge E. Hallyn wrote:
> > Quoting Herbert Poetzl (herbert@13thfloor.at):
> > >
> > > just a question: why do we keep the fs (struct_fs)
> > > outside of nsproxy?
> >
> > Good question.  So we have a mounts namespace, and you
> > would consider the per-process fs root to be an fs
> > namespace?  Practically, it would mean that chroot
> > and pivot_mount would create a new nsproxy, but i guess
> > that's not a real problem.
> >
> > It might force us to stop our current lazy checks for
> > 'current->nsproxy==&init_nsproxy', since the pivot_mount
> > in early boot would make that not true.
> 
> well, IMHO those are broken anyway, I can imagine

Yeah I wasn't defending them by calling them lazy  :)

> a number of applications using private namespaces
> (the old ones) without running in 'containers'

Do you have a patch to move the fs_struct into nsproxy?  I'd be
interested in running some benchmarks with and without such a
patch to see the effect of dereferencing the nsproxy so frequently.

-serge
_______________________________________________
Containers mailing list
Containers@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/containers

On Fri, Dec 08, 2006 at 08:40:59AM -0600, Serge E. Hallyn wrote:
> Quoting Herbert Poetzl (herbert@13thfloor.at):
> > 
> > just a question: why do we keep the fs (struct_fs)
> > outside of nsproxy?
> 
> Good question.  So we have a mounts namespace, and you
> would consider the per-process fs root to be an fs
> namespace?  Practically, it would mean that chroot
> and pivot_mount would create a new nsproxy, but i guess
> that's not a real problem.
> 
> It might force us to stop our current lazy checks for
> 'current->nsproxy==&init_nsproxy', since the pivot_mount
> in early boot would make that not true.

well, IMHO those are broken anyway, I can imagine 
a number of applications using private namespaces
(the old ones) without running in 'containers'

best,
Herbert

> -serge
_______________________________________________
Containers mailing list
Containers@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/containers