[CRIU] [PATCH 4/6] mnt: add --enable-external-masters option
Tycho Andersen
tycho.andersen at canonical.com
Wed Apr 8 12:39:28 PDT 2015
On Wed, Apr 08, 2015 at 08:42:55AM -0600, Tycho Andersen wrote:
> On Wed, Apr 08, 2015 at 04:18:18PM +0300, Andrew Vagin wrote:
> > On Tue, Apr 07, 2015 at 11:37:19PM +0000, Tycho Andersen wrote:
> > > This option enables external (slave) bind mounts to be resolved.
> > >
> > > Signed-off-by: Tycho Andersen <tycho.andersen at canonical.com>
> > > ---
> > > crtools.c | 11 +++++++++++
> > > include/cr_options.h | 1 +
> > > mount.c | 3 +++
> > > 3 files changed, 15 insertions(+)
> > >
> > > diff --git a/crtools.c b/crtools.c
> > > index fe77f39..095c86e 100644
> > > --- a/crtools.c
> > > +++ b/crtools.c
> > > @@ -205,6 +205,7 @@ int main(int argc, char *argv[], char *envp[])
> > > { "feature", required_argument, 0, 1063 },
> > > { "skip-mnt", required_argument, 0, 1064},
> > > { "enable-external-sharing", no_argument, 0, 1065 },
> > > + { "enable-external-masters", no_argument, 0, 1066 },
> > > { },
> > > };
> > >
> > > @@ -425,6 +426,9 @@ int main(int argc, char *argv[], char *envp[])
> > > case 1065:
> > > opts.enable_external_sharing = true;
> > > break;
> > > + case 1066:
> > > + opts.enable_external_masters = true;
> > > + break;
> > > case 'M':
> > > {
> > > char *aux;
> > > @@ -461,6 +465,11 @@ int main(int argc, char *argv[], char *envp[])
> > > return 1;
> > > }
> > >
> > > + if (!opts.autodetect_ext_mounts && (opts.enable_external_masters || opts.enable_external_sharing)) {
> > > + pr_msg("must specify --ext-mount-map auto with --enable-external-{sharing|masters}");
> > > + return 1;
> > > + }
> > > +
> > > if (work_dir == NULL)
> > > work_dir = imgs_dir;
> > >
> > > @@ -655,6 +664,8 @@ usage:
> > > " attempt to autodetect external mount mapings\n"
> > > " --enable-external-sharing\n"
> > > " allow autoresolving mounts with external sharing\n"
> > > +" --enable-external-masters\n"
> > > +" allow autoresolving mounts with external masters\n"
> > > " --manage-cgroups dump or restore cgroups the process is in\n"
> > > " --cgroup-root [controller:]/newroot\n"
> > > " change the root cgroup the controller will be\n"
> > > diff --git a/include/cr_options.h b/include/cr_options.h
> > > index 29be2c8..28bed45 100644
> > > --- a/include/cr_options.h
> > > +++ b/include/cr_options.h
> > > @@ -64,6 +64,7 @@ struct cr_options {
> > > struct list_head new_cgroup_roots;
> > > bool autodetect_ext_mounts;
> > > bool enable_external_sharing;
> > > + bool enable_external_masters;
> > > bool aufs; /* auto-deteced, not via cli */
> > > };
> > >
> > > diff --git a/mount.c b/mount.c
> > > index 8835a81..7d565d8 100644
> > > --- a/mount.c
> > > +++ b/mount.c
> > > @@ -691,6 +691,9 @@ static int resolve_external_mounts(struct mount_info *info)
> > > if (opts.enable_external_sharing && pm->shared_id == m->shared_id)
> > > match = true;
> > >
> > > + if (opts.enable_external_masters && pm->shared_id == m->master_id)
> > > + match = true;
> >
> > if think this code should be inverted:
> > if (!opts.enable_external_masters || pm->shared_id != m->master_id)
> > match = false.
> >
> > In your case, we can meet pm with another shared_id and set it as
> > external, but it looks wrong. Do I miss something?
>
> Are the peer group ids globally unique across all namespaces? If so,
> then I think this should be correct (I was working under the
> assumption that they were, mostly because when I make a new bind mount
> in a namespace and make it shared, it ends up with the "next" logical
> id).
Hm, actually on second thought I think this is wrong; I think it
should behave exactly as the shared one, i.e. all this does is
test that the slavery is internal to the mount namespace vs. external.
Tycho
> Tycho
>
> > > +
> > > if (!match)
> > > continue;
> > >
> > > --
> > > 2.1.4
> > >
> > > _______________________________________________
> > > CRIU mailing list
> > > CRIU at openvz.org
> > > https://lists.openvz.org/mailman/listinfo/criu
More information about the CRIU
mailing list