This is a precursor patch. Netlink request code is mostly the same for any
kind of sockets. Only 2 places differs: request creation and callback address.
This patch parametrizes netlink call function with a pointer to a request
structure, request size and receive callback, which replaces hard-coded unix
sockets parts.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Remove explicit parasite targets dependence on deps files and use them as
makefile parts.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Objects does not depend on dep files. Dep files provide rules to build object
files. Thus we need to include dep files.
This also leads make to rebuild dep files (if required) before an other
targets.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
v2: add FIXME for linking restorer-log.c and restorer.c by ld
I don't know how to do it now.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
The previouse patch doesn't change it to simplify review
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Before this patch the restorer's code is linked in crtools and then
we copied functions from it. In this case all function should
be inline and we can't use a global variables.
I suggest to make it like parasite. The restorer's code is isolated in
own file and will be copied wholly. The restorer's code is compiled as
position-independent code, so we can use functions and global variale
(E.g. to save descriptor for log messages).
v2: correct indentions in a separate patch
v3: introduce a variable restore_task_exec_start symmetrical to
restore_thread_exec_start
v4: don't give command in restorer_thread()
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
BUG_ON_HANDLER from util.h may be used in restorer.
E.g. #include <util.h>
#include <lock.h>
Now the process will be segfaulted and you can find a line
number in assemgler code. We will print messages a bit late.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
This patch prepares code to handle errors. In the near future
we will handle SIGCHLD. If a restore of one task fails, we will
send a signal to other for completing.
For this we should have ability to wait until all task wills be
restored. This patch does it.
v2: Don't wait children.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Sometime we want to have a shared mapping in restorer. E.g. A storage
for shared memory entries. This entries contains locks, which should
be released in restorer.
v2: fixed according to Pavel's comments
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
There is no need to use sys_ versions of libc functions
when we run in non relocated code. It's a leftover from
early testing time. Fix it.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
We allocate only one page for fdinfo_list_entries.
In the future we will be able to resize this memory.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Yet again -- this makes code easier to understand from my POV.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
This one is skipped at restore and leaves an open core file
in target task's fdtable.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Otherwise it pops up after restore in target task's fdtable.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Make it shorter and simpler and use one in open_image_ro_nocheck.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
create_pipe() may restore up to 3 descriptors. They may be both ends
of pipes and a target descriptor. The image fd may hold any of them.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
This reverts commit 46c613cc7d.
Andrey posted a proper fix for it. Moreover, the problem in first
place was initiated by a parasite application running during test
case, crtools knows nothing about.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
It's perfectly fine to reuse descriptors here,
since we use plain pipe() call and it migh choose
the descriptor which we need after.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Add the -l option to show the list of tests
Make the default zdtm/live/ prefix when running individual test
Now you can do it like
# sh ./zdtm.sh for running all tests
# sh ./zdtm.sh -l to list available tests
# sh ./zdtm.sh <test> to run test <test>, e.g. "static/pipe00"
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
This makes crtools correctly abort when working on wrong kernel.
Otherwise all the open files will have the same (garbage) ID and
the subsequent restore will result in broken app.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
The old code carried the path through the stack. Now we have
pstree_pid and handy helpers to get one.
Tested by pipes00 test from zdtm (it forks).
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
And do not use strcpy, better to stick with strncpy.
Moreover, to be on a safe side make proc internal
buffer big enough even for "(%16s)" format, it's
hardly possible that the kernel ever change stat
format but just to be on a safe side.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
