When a child exited, SIGCHLD is sent to parent,
if SIGCHLD has handler without SA_RESTART, the current
syscall will be interrupted.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

 - Use sizeof instead of PATH_MAX constant
 - Place EOS exactly after read data end
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Andrew Vagin <avagin@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Revert and tune 9eda83b7. Andrey reminded me that we have
global instance of cr_options.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

On this stage crtools unlink old socket files, create ghost files and etc,
so we should be in a correct namespace.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The option is -r|--pivot-root and an argument is a path to new root.
A root task will make pivot_root. LXC CT does that, so we need that
for restoring.

v2: s/pivot-root/root/
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The pstree_item for helpers is allocated without rst_info.
Before this patch prepare_fd_pid was executed for such items and
touched rst_info.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The kernel patch is in 3.5 branch, so we can make this test
run automatically.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

It's not in zdtm.sh yet, since kernel patch is not yet
committed. When kernel is updated, the zdtm.sh will be too.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Only the fact of the fd presence, its flags and fown and the sigmask.
The sigpending state is tightly coupled with the task's sigpending
state which is not yet supported.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The list of mounts is used by inotify to restore a watch with open-by-handle. The problem
is that after the clean_mnt_ns is done the mntinfo list is no longer relevant.

Fix it by freeing the old one and initializing the new with the read-from-image stuff.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Only checks for lo, one simple ifaddr and two routes are handled. Sanity
test and nothing more.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Same here -- run ip tool and that's it.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Just run the ip addr save and ip addr restore respectively.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The thing is that the ip utililty is very likely to have support
for ifaddrs and routes dumping and restore (rtnl messages are
symmetrical wrt dump/restore and this can be easily dump in there).

Thus, the best we can do it just use this tool for that and carry
the "raw" images with ifaddrs and routes dump.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Only support the lo device. This is not final yet (much more
stuff is to be handled for a link) but is rather a skeleton
showing how to do it and letting us check the LXC container
early.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

If the target task sits in another netns the sockets will
not be able to interconnect unless we create the host-side
one in the same netns. Thus do it and then roll back the
namespace.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

This will be required for parasite transport socket creation -- it will
have to be created in a net ns we're putting parasite in and then we'll
have to restore it back to original to go on dumping.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

We'll need to check whether the socket is to be created in a net
namespace (since we can't detect this automatically yet). Thus the
namespaces flags are required there.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Raw images are supposed to be files with non-crtools contents.
E.g. tarballs with tmpfs contents or iptables-save output. Another
example is -- ip tool addr and route dumps we're about to merge
soon.

Having these type of images allows to use the open_image and the
cr_fdset engine easily.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

This type of message is also reported for NLM_F_ACK is requested
and no necessarilly indicates an error. The error code is in the
message body.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Will need it int netns dumping callbacks later.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

All we need is (should be) in rtnetlink.h
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The function that does socket collecting is actually a rtnl
request sending one. It will be usefull for netns dump/restore,
so move it to generic place.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

We need proc for restoring processes. The existent /proc may be not suitable.
E.g. If processes are in pidns.

crtools mounts procfs in a temporary directory, but it should be
umounted at the end. Before this patch crtools did that, but
it doesn't work if processes in a mount namespace.

Actually this logic can be simplified and this patch does that.
* create a tmp dir
* mount procfs
* open this directory and save a file descriptor.
* detach procfs
* remove the tmp dir
* access to proc via openat, statat and so on.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

When dumping a net namespace we'll have to create the transport
socket in the target namespace. In order to do this we'll have
to call setns swice -- before and after socket() call.

I don't want to do it for every single task to dump, let's cache
the socket instead.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Do init and transport sockets interconnections in one
switch to parasite.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Otherwise dump succeeds but is inconsistent.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

* s/pidns/ns/
* Add option --namespace mnt to "crtools dump"
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

A part of code has been copied from mountpoints.c

This code umounts all points except /proc and /
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

If we dump tasks with mntns, we should look at pathes from point of a mntns root.

Now we support a situation when a root of an init task has the same root as the
mntns root, because we have not another way to get a root of mntns.

A path to an unix socket is copied, because the origin copy will be gone
out from the function, where it was created.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

We need to check that opened files can be restored in mntns
[xemul: s/slabinfo/meminfo as in 9dc050fd]
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Otherwise a parent hangs if a child died, because
it reads from a status pipe.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>