Instead of passing self-vma file path to restorer
code simply open it before restore_task call and
pass descriptor instead. This saves some memory.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
This allows us to get rid of open-coded "/proc/pid/X".
Based-on-patch-from: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
It basically reverts e189efc1763d9cae55e1cafd7aff7ffef6e47303
which was overdone one.
Reported-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Some process can share one struct file-s, we may find them by "object IDs".
A file descriptor is opened in one process and send to other via unix socket.
The procedure of restoring files contains four stages.
* Collect data about all file's descriptors
On this stage we find process which will restore a file descriptor and
create a list of processes, who should get this descriptor.
* Create datagrams unix sockets
If a file descriptor should be received, a unix socket is created
instead of it.
* Open file descriptors
A process with the least pid opens a file and sends this file
descriptors to all one who wait it.
* Receive file descriptors.
When we were thinking up this algoritm, we wanted to minimize a number
of context switches. A number of context switches is proportional of a
number of processes.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Restorer does really restore shared memory (including page
contents restore) only on master process, while all other
processes do open such memory area via map_files/ procfs entry
so that we might have a situation when shared VMA is present
in some particular core-%d.img file but it's not listed in
collected shmems array and find_shmem_by_pid will return NULL.
This is perfectly fine, be ready for that.
Another issue is that shared memory might look like
CR_FD_SHMEM: /home/cyrill/projects/kernel/crtools/shmem-2641.img
----------------------------------------
0x7f2200775000-0x7f2200776000 id 19664
0x7f2200776000-0x7f2200777000 id 19663
----------------------------------------
So vma area is [x;y) range and we should distinguish two
shmem lookup cases
- one when we search for page in shmem area
- second when we lookup shmem area in collected ranges
They both have a different lookup conditions so single
find_shmem splitted into two helpers find_shmem and
find_shmem_page as appropriate.
This patch finally fixes the three process asynchronious
shared memory updates test-case.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
The remapping of /proc path to shmem should be
done with pid of process which did mmap() call
initially.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Several places just need to open an image, thus the helper is OK to use.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
The next patch will remap shmems region and will use it in restorer.
Actually this patch moves shmem_nr to struct shmems.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
[gorcunov: Various tuneups]
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Currently it can only work with stream sockets, which have no skbs in queues
(listening or established -- both work OK).
The cpt part uses the sock_diag engine that was merged to Dave recently to
collect sockets. Then it dumps sockets by checking the filesystem ID of a
failed-to-open through /proc/pid/fd descriptors (sockets do not allow for
such tricks with opens through proc) against SOCKFS_TYPE.
The rst part is more tricky. Listen sockets are just restored, this is simple.
Connected sockets are restored like this:
1. One end establishes a listening anon socket at the desired descriptor;
2. The other end just creates a socket at the desired descriptor;
3. All sockets, that are to be connect()-ed call connect. Unix sockets
do not block connect() till the accept() time and thus we continue with...
4. ... all listening sockets call accept() and ... dup2 the new fd into the
accepting end.
There's a problem with this approach -- socket names are not preserved, but
looking into our OpenVZ implementation I think this is OK for existing apps.
What should be done next is:
1. Need to merge the file IDs patches in our tree and make Andrey to
support files sharing. This will solve the
sk = socket();
fork();
case. Currently it simply doesn't work :(
2. Need to add support for DGRAM sockets -- I wrote comment how to do it
in the can_dump_unix_sk()
3. Need to add support for in-flight connections
4. Implement support for UDP sockets (quite simple)
5. Implement support for listening TCP sockets (also not very complex)
6. Implement support for connected TCP scokets (hard one, Tejun's patches are not
very good for this from my POV)
Cyrill, plz, apply this patch and put the above descriptions onto wiki docs (do we
have the plans page yet?).
Andrey, plz, take care of unix sockets tests in zdtm. Most likely it won't work till
you do the shared files support for sockets.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
When opening an fd at a position where the current image fd sits
the latter one should be moved.
Introduce a helper for this.
Signed-off-by; Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Nobody cares the returned fd, so just make it 0/-1 and clean up the callers.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Instead of keeping all unrelated to
C/R procedure helpers in util.c move
logging related helpers to log.c.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
This as well implies we change paths to PATH_MAX size.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Pavel Emelianov <xemul@parallels.com>
We have a mess in our return codes:
some functions return negative values,
while others -- positive.
Sanitize the return codes and make error
condition to return negative values.
Reported-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Otherwise it may begin to restore a parent objects...
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
For this reason a function may return success in case of error.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
1. pipe() returns two descriptors with numbers (r: x, w: y).
We may want to restore it as (r: y, w: x).
2. Close a extra end, before try to restore a target end.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Even u64 address might consume not greater
than 20 bytes of string so 128 is enough and
safe here.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Since we use pure syscalls there is no
need to keep intermediate layer for signals.
Moreover mask entry moved at the end of the structure
so we will easily expand it if it'll be ever needed.
Note it breaks backward compatibility with older image
but since it's development stage it should be safe.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Andrew Vagin <avagin@parallels.com>
Probably all of you know about "lock inversion". There was a similar
problem on restoring pipes.
One process try to restore pipe1 and waits when another process attached
to it. In this time another process restores pipe2 and waits too.
I know two solves.
1. Open all pipes -> attach to them -> close unnecessary ends.
This method has a problem, if only one end belongs to the process. In
this cases another end occupies a descriptor, which may be needed to
another pipe.
2. Restore pipes in the same order. This patch does that.
A sorted list of pipe entries are constructed.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelianov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
This patch removed stange logic about minusers.
Now we calculate foreign users only.
The main idea of this patch, that we bind a pipe ends to correct
descriptos immediately.
When a process X creates a pipe, we have two cases:
1. Both ends belongs to this process. It's simple, we bind both ends
to their descriptors and we go further.
2. Only one end of a pipe belongs to this process. In this case we are
waiting all foreign users and we go futher.
In second case we should wait, because a end which doesn't belong to
this process occupies a file descriptor, which may be used by another
pipe.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelianov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
CWD is saved as file descriptor with number -1.
v2: use dump_regular_file
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Now I try to restore CWD and a relative path will be invalid.
Add new options -D to set image files directory.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
The standard descriptors may be redirected.
crtool dumplicates stderr in rlimit.maxfileno-1 and this descriptor
is inherited by all children and will be closed before sigreturn.
Known issues:
- The logging descriptor may be used by a target process and
a resume will fail.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Since we operate with syscalls directly we are
to convert signal's structures between image and
kernel formats, without intermediate glibc layer.
Note this involves chaging sa_entry::flags to u64
(since it's long int value in kernel).
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
sigactions are restored before memory and if a parent get a signal
which has a handler, it will get segmentation fault.
This problem will disappear, when we add a freezer, because
all processes will be resumed after restoring.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
v2:
- Pavel reported there is no need to lock/unlock
last-pid file in cycle, just lock it once before
threads creation and unlock at the end.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
