This test uses systemtap to replace random parasite command on
invalide command. The parasite code received this command and should
rollback itself. CRIU didn't recived ack, should understand that
something wrong and exits. The systemtap inverts exit code of criu.

"zdtm.sh -d" is used for executing tests and testing that the state of
processes don't corrupted.

criu should be compiled without optimizations (make DEBUG=1)

v2: randomize a command, which will be replaced
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

The packet can be retransmited after dumping the tcp connect.  The first
one is that the connection is blocked for only one direction. The second
one is that TCP timers continue work during dumping and they can send
packets. tcp_timestamp is saved for each tcp connections and then it’s
restored. So if a packet is sent after dumping, its timestamps is
saved by another side and this timestamp is sent back in the next packet
as the tsecr parameter. If this packet is received after restoring, it
looks like a packets from the future.

https://bugzilla.openvz.org/show_bug.cgi?id=2676Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

 /dev/null

Clear the umask when creating /dev/null and restore the umask
after the creation.
Mountpoints test unlinks and makes new /dev/null inside by mknod.
But in some linux, there is default umask, e.g., as 022. It means that
/dev/null created directly cannot be wrote by group and other users.
It is not reasonable because /dev/null should be accessable to all
users.
Signed-off-by: Yicheng Qin <yichengq@google.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

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

The difficulty is that this code is required in both -- pie
and non-pie contexts.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Will put more things there.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

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

for that the mount point is bind-mounted in a temporary place.

v2: * check, that the fs you get access to at the end is _really_ the
      one you wanted to
    * use switch_ns/restore_ns helpers
v3: reuse code of __open_mountpoint and a few small cleanups
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

When CRIU wants to dump content, it checks that nothing is overmounted.
The list of children for such mounts must be empty, but these lists are
filled during constructing a tree of mounts.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

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

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

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

All scripts are executed for all actions, so if one of them failed for
any actions, crtools returns with non-zero code.
E.g: ./zdtm.sh -d ns/static/env00
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

page-server is forked and the child is running in background, so a bash
variable $! contains PID of the last command, which is executed in
background.

Currently it's tcpdump and zdtm.sh is waiting it and never returns back.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

v2:
 - Use special opcode 32 to distinguish error cases
 - Use --scripts for snapshot testings
v3:
 - test for both 32 and 0 for error code
v4: (by avagin@)
 - restore from the last snapshot
v5:
 - update comment

https://bugzilla.openvz.org/show_bug.cgi?id=2583Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

This is usefull if one needs to do some final action before checkpoint
is complete. For example in case of online migration one may provide
a script which would check the restore procedure on remote note
ended without errors, thus the script returns zero code and criu
simply kills running instance of application.

In turn, if migration failed, the script can return nonzero code
and criu won't kill the application but continue its execution
instead.

https://bugzilla.openvz.org/show_bug.cgi?id=2583Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Sometimes these tests don't pass successfully and we need more
information to investigate the problem.

https://bugzilla.openvz.org/show_bug.cgi?id=2676Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Add target lib to Makefile for criu library.
Created directory lib and add there Makefile and criu.c - skelettons for criu library.
Signed-off-by: Ruslan Kupreev <kupruser@gmail.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

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

The version, that might not wait for ack is always called with
"async" flag set. Cleanup things according to this.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

... fixing erroneous goto on error path :)
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

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

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

There are parts dumping which is common to thread and task,
and this stuff is represented by parasite_dump_thread structure.

Merge this into parasite_dump_misc and facror out dumping code.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

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

There are 3 users of one -- remote syscall, thread dump and
daemon run.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

PTRACE_SETSIGMASK takes pointer on mask of blocked signals,
but sigmask is already pointer.
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

If a kernel supports PTRACE_SETSIGMASK, criu don't need to execute
PARASITE_CMD_INIT and PARASITE_CMD_DAEMONIZE, because the frist command
is used only for blocking signals. If criu crashes between these
commands, a process state will be corrupted, because all signals remain
blocked.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

We use two commands to get task registers safely. The first command
blocked signals, then crtools dumped registers and all per-thread data
and the the second command unblocks signals.

Currently signals can be blocked with help SETSIGMASK, so we need only
one command to dump per-thread data.
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>

Currenly we can restore blocking signals with help ptrace,
so if registers are restored after command, a process state
will not be corrupted if criu crashes.

Same as with previous patch -- at the end there will be only
two trap commands, so this is OK.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

Pending signals should be saved, so signals should be blocked.
Signals are blocked for EACH command, because a chance of destroying a
process state should be a small as possible.

At the end there will only two "trapped" commands -- to dump thread
and to start daemon in parasite, so this doesn't add significant
overheads.

If crtools is killed between two commands, a dumped process will run
continue.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

It will be used in a few places
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>

The initilization stage is not good for that, because it can fail.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

tcp_repair_off implicitly modifies SO_REUSEADDR option
inside the kernel (thanks avagin@ for pointing this
feature out) thus if we are to rollback and restore
the former settings of socket -- don't forget to
repair this particular one.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
CC: 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>

Window probe is sent during disabling repair mode on a socket, so
network must be unlocked in this moment.

https://bugzilla.openvz.org/show_bug.cgi?id=2670

v2: don't fail after unlocking network
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

It never returns errors.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>

This patch fixes a stack corruption caused by the routine
since it pushes the registers R4, R5, R6, and R7
onto the stack that results in a restorer crash
while restoring a zombie (https://bugzilla.openvz.org/show_bug.cgi?id=2663)
Reported-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>