This patch switches reg_file_entry structure with RegFileEntry
protobuf entry. Note that there a few nits remains
- old fown_t structure doesn't scale well with FownEntry from
RegFileEntry, so to not mess with rest of crtools code
I've added a couple of opencoded copy operands, will clean
this aspect up once rest of code is converted
- the "show" procedure doesn't print fown for same reason
v2:
- rename fown_t to fown_entry in .proto
v3:
- don't print nil symbol on show
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
A short story -- there were a long conversation on which format should
be used to keep checkpointed data on disk image. We ended up in using
Google's Protocol Buffers (see https://developers.google.com/protocol-buffers/
for detailed description). Thus image entries should be convered to PB.
This patch converts fdinfo_entry to PB "message fdinfo_entry".
Build note: one should have protobuf and protobuf-c installed to be able
to build crtools.
- http://code.google.com/p/protobuf/
- http://code.google.com/p/protobuf-c/
Inspired-by: Pavel Emelianov <xemul@parallels.com>
Inspired-by: Kinsbursky Stanislav <skinsbursky@openvz.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
To not bloat util.h or image.h the protobuf.h introduced to handle all PB needs.
We represent our PB objects as pairs of records
| 4 bytes for object packed size
+---
| X bytes -- object itself
v3:
- Use plain read/write calls
- Use local buffers for small objects
v4:
- Add pb_pksize helper to calculate object
size with header
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Before this patch a transport socket was created on-demand for each operation
requiring it. It will be problematic, when we start supporting net namespaces.
That said -- create a transport sockets and interconnect them early.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
No need to keep it in header, sk_packet used solely in one C file.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This serves two things -- kills annoying data offset management
which makes protobuf switch too ugly. And removes the need to
seek the pipe data image, which in turn is the prerequisity for
streaming migration.
Memory usage is still optimal, since we still use splice on dump
path and on restore stage we F_GIFT pipe buffers to the kernel.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Make it in two stages.
First -- collect pipe data into hash without checking for pipe/fifo existance.
Next -- when pipe/fifo gets restored walk the hash and search where the pipe data
is.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The existing code checks for ID to be already dumped. But pipe and fifo may have the same ID
since they reside on different filesystems.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This routine checks for pipe data being dumped based on pipe ID, but the arg in questoin
is the pipe's end ID which makes code reading confusing.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We will need to use pointer here (to PB object) anyway
so better to make it in a separate patch for bisectability sake.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Pgid leader should become such before any other task tries
to enter this group (with setpgid). Thus we introduce a yet
another global sync point -- before it all pgid leaders call
setpgid after it all the others do it.
v2: wait while helpers restored pgid
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Don't push open cb onto reg_file_into, pass it into regfiles engine
as an argument.
Note: I haven't merged the fifo zdtms yet, thus this patch is untested.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Tested-by: Cyrill Gorcunov <gorcunov@openvz.org>
Checkpoint and restore of fifo is similar to
pipes c/r except the pipe end-points are named
file.
Because the fifo has a name we use regular files
facility for fifo path c/r.
Still there is a trick used to "open" fifo:
the opening procedure migh sleep if a fifo's peer
is not yet opened, so before doing a real open
we yield a fake open procedure (with O_RDWR flag)
which prevents us from sleeping even if peer
is not yet ready. Also we need writable fifo
end to restore data queued.
v2:
- add open/priv members to reg_file_info
- make open_fifo_fd to use open_fe_fd
- comment on pipe_id
- make sure the fifo data is not restored twice
v3:
- drop useless fixme comment and add sane one
v4:
- Use restore_data flag to escape data restore duplication
- Use S_ISREG for file contents copying
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
To reuse with fifo data restore.
v2:
- use image type in functions arguments instead of forcing
calling side to provide file descriptor
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
v2:
- Use do_dump_opt in dump_socket to not call
lookup_socket redundantly
v3:
- use getsockopt to check that unconnected
socket has no peername and it's not listening
- do test only socket protocol since family and
type will be called in dump_one_inet_fd.
v4:
- Move proto tests to can_dump_inet_sk
- Use SOL_TCP/TCP_INFO to gather info about tcp sockets
v5:
- hash unconnected sockets to run BUG_ON(already-dumped) logic
- no default value for sk->wqlen, use zero from xzalloc
- drop bogus xfree on error
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Namespace restore goes in two steps -- first, clean old mountpoints
(except root) that are left after clone(CLONE_NEWNS) call. Next,
repopulate the namespace with new mountpoints (except root) read
from the image file.
This patch adds the skeleton for the above -- reading the mountpoints
from current proc and image and calling for mount/umount. The real
action will come in further patches.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Build a tree of mountpoins that can be (u)mounted in a straight
(forward or backward) order without EBUSY errors.
The tree is built in two steps -- first create hierarchy based
on mount iDs. Next -- resort siblings in the path depth order.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Encode it in u32 since we have to check filesystem name anyway
(not everyone ie easily migratable).
Dump _any_ root fs as we need the root entry on restore though
do not mount it then -- it should be provided before crtools
restore start.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Dumping is straightforward -- just copy all the mount_info
fields into the new image file. Dump everything but fstype,
this one will come in a separate patch.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This structure will be used on restore and will be created
from the image, thus the name proc_ is not suitable.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
On the restore path this structure will be used and it
will be better to have them char * rather than char[64].
When scanning proc use the %ms specifier for this.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
In olden times, when people didn't know how to write,
they sent a courier with a letter.
Now we can write messages from parasite, so SET_PARASITE_RET may be dropped.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Looks good to me as well. Maybe we will refine messages to be able
to identify where exactly parasite failed (say line). But it better
to do in another patch.
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
- add argument names
- align members
- add extern kw
- group struct decl. on top
- make file_desc_ops::type being unsigned
integer similar to fdtype_ops::type
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
These are declared in files-reg.h, so get rid of
them and add files-reg.h inclusion where needed.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It's sign, that a parent has been changed sid after forking a child.
We should know a sid with which a process was born, because in a processes
chain, more then one process might change SID.
v2: fix names of variables
v3: prevent rewriting of born_sid
v4: Abort the restorer with error message if a born_sid can't be determing.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
They will be used for restoring sid. For example, if a session
group leader is absent, a helper process is created with this id
and it will die after restoring all other tasks.
Before this patch restore failed if anyone exited.
Now we should skip helpers, which exited successfully. It's a bit tricky.
All children are collected in sigchld_handler, but we have a point,
where we want to wait all helpers. For that waitpit is used and ECHLD
is ignored, because it signs that a helper exited and has been waited in
sigchld_handler.
v2: check that me isn't NULL in the sig handler
v3: move code about waiting helpers in a separate function
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
In the open_fdinfo we need to get a file_desc associated with the given
fdinfo_list_entry. This is done by searching the hash of descs, but this
can be speeded up by saving the desc pointer on the fdinfo at the time
of collecting them.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
because now it cares about tid_addr and thread id.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Instead of spreading regular files handling code over
files.c and cr-dump.c move it to files-reg.c.
This allows to extend regular file handling code in
future without disturbing other source files and
make code logically coupled, where files.c is mostly
a place for general file handling code unrelated to
file type specifics.
While mostly it was code tossing there is some change
I have to notice -- the structure ghost_file was previously
declared in two paces cr-dump.c and files.c, in cr-dump.c
it was used for dumping while in files.c it was used
for restore. The structure had different members.
So to resolve this conflict the ghost_file structure
was renamed to ghost_file_dumpee.
Nothing else is changed.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It's only required for making pipe_id, but making it is better to
be done in place and using the st_ino only.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
v2: rework this by using openat() and service fds for proc root.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Now pid is dumped from pid ns, it's gotted from parasite.
v2: fail if a zombie is in PIDNS
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It uses pid for create image file and real_pid for dumping ns-s.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
If we try to dump a process from another pid ns,
we have not another way to get its pid.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Add struct pid and use it everywhere. This struct contains
two fields: pid and real_pid.
real_pid is a pid outside of the target pid namespace.
pid is in the target pid namespace
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
