This values will be doubled in kernel to account for
"struct sk_buff" etc. overhead.
Currently criu restores snd and rcv buffer limits incorrectly,
they become bigger on each iteration.
$ ../crit show dump/zdtm/live/static/socket-tcp/6299/1/inetsk.img | grep buf
"so_sndbuf": 2626560,
"so_rcvbuf": 1060720,
"so_sndbuf": 16384,
"so_rcvbuf": 87380,
$ ../crit show dump/zdtm/live/static/socket-tcp/6299/2/inetsk.img | grep buf
"so_sndbuf": 5253120,
"so_rcvbuf": 2121440,
"so_sndbuf": 32768,
"so_rcvbuf": 174760,
$ ../crit show dump/zdtm/live/static/socket-tcp/6299/3/inetsk.img | grep buf
"so_sndbuf": 10506240,
"so_rcvbuf": 4242880,
"so_sndbuf": 65536,
"so_rcvbuf": 349520,
With-help-of: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Actually make use of the ns->type field and remove all getpid()'s
and other strange/inconsistent checks.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We'll use this in the next patch to correctly write sysctls.
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
There was a problem with diag modules, that were auto-loaded on dump and
put some data into uevent sockets thus causing the dump to fail on that
socket. The solution that time was to call modprobe in advance to make
sure the diag modules are there.
This turned out to be not nice solution as modprobe may fail if the module
is missing and the requred functionality is built-in.
To handle this for diag stuff it is mush faster and more reliable to ping
the kernel with simple diag request and check that it reqched the required
module.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
In the next patch we will need to care about the exact error reported by
the kernel, so add the error callback for this.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
See the comment below for an explanation of what is going on. We will
ultimately need to handle dumping the netlink data, but I think it is good to
prevent injecting events into the stream during a dump. So we pre-load the
modules, even though it isn't very pretty.
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The setns() syscall (called by switch_ns()) can be extremely
slow. If we call it two or more times from the same task the
kernel will synchonously go on a very slow routine called
synchronize_rcu() trying to put a reference on old namespaces.
To avoid doing this more than once I propose to create all
per-ns sockets in one place with one setns call. In this
patch there's on nl diag socket used to collect other sockets
is created this way.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We want to have buffered images to speed up dump and,
slightly, restore. Right now we use plan file descriptors
to write and read images to/from. Making them buffered
cannot be gracefully done on plain fds, so introduce
a new class.
This will also help if (when?) we will want to do more
complex changes with images, e.g. store them all in one
file or send them directly to the network.
For now the cr_img just contains one int _fd variable.
This patch chages the prototype of open_image() to
return struct cr_img *, pb_(read|write)* to accept one
and fixes the compilation of the rest of the code :)
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Now we supports sub-mntns, so root_ns_mask sounds more correct than
current_ns_mask.
v2: typo fix
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We have a mess of uintX_t and uX usage. Drop off uintX_t ones.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Andrew Vagin <avagin@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
crtools.h is too heavy to be included in many sources
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
SO_SNDBUF and SO_RCVBUF sets a double input value, because
"Linux reserves half of te socket buffer for metadata."
So if a process is suspended/restored many times, a socket buffer
size is doubled on each iteration and in a one moment it is overflowed.
Very likely the program hangs in such situation, because the socket with
negative buffer size is unusable.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Since *all* of them just call do_dump_gen_file with proper ops,
just call one directly. Compacts the code.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Absense of packet and netlink diag is OK at dump -- we can be on
older kernel and will fail later if _really_ need these sockets.
During check stage all diag is expected to be in place.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The idea is simple. If the collection of given type of sockets failed,
crtools can't be sure, that it's able to dump such sockets correctly.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The packet diag are not compiled in Fedora 19 kernel.
The netlink diag is going to be merged in 3.10.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
All info about bound sockets are got via socket diag interface.
All connected sockets are automatically bound.
For other sockets only protocol must be dumped, which is got
with help getsockopt.
A netlink sockets with pending data are not supported yet and
probably will not be supported in a near future.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It's no longer required to use this option -- two currently
supported cases (tasks on host and tasks in containers) can
be detected automatically. Keep this option for future.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
These are structs that (now) tie together ns string
and the CLONE_ flag. It's nice to have one (some code
becomes simpler) and will help us with auto-namespaces
detection.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currently fdinfo dumps for each task, so CR_FD_FDINFO is in cr_fdset.
A few tasks can share one fd table and the set of descriptors will be
dumped once and a image name will contain files_id instead of pid.
In this case CR_FD_FDINFO will go away from cr_fdset.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The SO_BINDTODEVICE getter is changed in the kernel (before
official release) to report not index, but name to be in
harmony with setter.
Fix crtools accordingly.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Sizes of send and recv buffers are set to maximum to restore queues,
after that sizes of buffers are restored.
O_NONBLOCK is set to a socket to prevent blocking during restore.
#2411
v2: do the same for unix sockets.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
One thing to note. The socket filter proggie is a set of struct-s
wuth 8 and 16 bits values in it. Protobuf doesn't support such thing
and it's too annoying to mess with yet another message for that.
Instead, I encode all this stuff into array of fixed64 fields to
handle endianity (yes, protobuf handles it, but each field is not
just 64-bit value, but a structure).
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The kernel SO_BINDTODEVICE option is not symmetrical --
set required device name, but get reports index. Thus
need the index to name resolver.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
