On restore we add a temporary root to a mount point path. It's convinient
for restoring mount namespaces, but real paths are used for restoring
link-remap files.
v2: replace the offset field on a char * field
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The m->bind will be used to get the "lowest" mount. A mount is
called "lowest" in the chain of bind-mounts when its m->bind is
NULL. I.e. -- it's the mount of the root dir of an FS, all the
others are bind-mounted from it.
The propagate_siblings() is called for each new mount and sets
the bind field for others. A part of these other mounts can be
already mounted.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
For that mnt namespaces should be dumped after files.
v2: rework enumeration of namespaces in dump_mnt_namespaces()
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currently collect_shared is executed for each mount namespace separately,
so we miss the dependencies between namespaces.
For example, we have two namespaces, the first one contains a tmpfs mount
and the second one contains a non-root bind-mount of this tmpfs. Without
this patch this example can't be dumped.
On restore mnt_build_tree() is called for all namespaces at once, thus
this bug doesn't exist there.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
During the dump phase, /proc/cgroups is parsed to find co-mounted cgroups.
Then, for each task /proc/self/cgroup is parsed for the cgroups that it is a
member of, and that cgroup is traversed to find any child cgroups which may
also need restoring. Any cgroups not currently mounted will be temporarily
mounted and traversed. All of this information is persisted along with the
original cg_sets, which indicate which cgroups a task is a member of.
On restore, an initial phase creates all the cgroups which were saved. Tasks
are then restored into these cgroups via cg_sets as usual.
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The newinstance options isn't shown in mountinfo. Currently it is
detected in devpts_dump. It is added only for root mounts and it
isn't added for bind-mounts. So mounts_equal(a, b, true) returns false
for such mounts and criu doesn't understand that they should be
bind-mounted.
Reported-by: Tycho Andersen <tycho.andersen@canonical.com>
Cc: Serge Hallyn <serge.hallyn@ubuntu.com>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Serge E. Hallyn <serge.hallyn@ubuntu.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
A file system can be bind-mounted a few times and some of these mounts
can be non-root. We need to find one of root mounts and dump it.
v2: don't forget to check pm->dumped and pm->parent
don't dump a root file system, it's always external for now.
Reported-by: Saied Kazemi <saied@google.com>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
One file system can be mounted a few times, so mnt_id isn't unique for it.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
On dump one uses one or more --ext-mount-map option with A:B arguments.
A denotes a mountpoint (as seen from the target mount namespace) criu
dumps and B is the string that will be written into the image file
instead of the mountpoint's root.
On restore one uses the same --ext-mount-map option(s) with similar
A:B arguments, but this time criu treats A as string from the image's
root field (foobar in the example above) and B as the path in criu's
mount namespace the should be bind mounted into the mountpoint.
v3:
* Added documentation
* Added RPC bits
* Changed option name into --ext-mount-map
* Use colon as key and value separator
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
These are mounted by default in ubuntu containers, so criu should know about
them and remount them on restore.
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch consists of 3 unsplittable (from my POV) fixes.
1. Remove messy check from dump_one_mountpoint() -- we have
validate_mounts to check whether we can dump the tree
or not.
2. Other than being in the wron place the mentioned check
is wrong. Comparing of the length of the mp->source-s
makes no sense -- it should be mp->root, but even this
would be wrong...
3. ... instead, we should check for bind mount root path
being accessible from the target mount root path, i.e.
the bind->root should start with src->root.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Acked-by: Andrew Vagin <avagin@parallels.com>
The memcpy() in devpts_dump() just overwrites part of them.
Fix this and move the whole code into sub-routine for future.
v2: Fix off-by-one error spotted by Filipe.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Acked-by: Filipe Brandenburger <filbranden@google.com>
This reverts commit f55a0b912c2bb105513b5f6e08a024958968ab9a.
The collected mounts list a) points to nsid and b) should be
put on nsid's mntinfo_tree. Thus is _should_ be done by separate
calls.
Nowadays this routine is mainly used for getting an
fd, rather than keeping one for future reference.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Now mntns_collect_root() should be called each time when we need to get
a root of a specified namespace and we don't need to call it for
initializing the global variable.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
