If a socket has been closed, it still has both addresses and
we need to dump them.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For that we restore all sockets properties and then
disable the repair mode before calling connect() and
the kernel will sent a syn packet and move the socket
into the sys-sent state.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
A socket is in one of half-closed states, if it sent a fin packet
or it received a fin packet.
CRIU plays with fin packets to restore half-closed states too.
When we need to sent a fin packet from a socket, we can call
shutdown(SHUT_WR). When a fin packet has to be restore in
a received queue, criu generate a fin packet and send it via
a raw ip socket.
A raw packet is sent with the SOCCR_MARK mark to be able
to not block it.
v2: remove the SOCCR_FLAGS_ACKED_FIN flag
introduce sets of bits for different actions with fin packets
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We will need to disable the repair mode before connect()
to restore syn-sent sockets. And even now connect() looks
like a part of internal logic of libsoccr.
travis-ci: success for series starting with [01/21] build: install libnet-dev
v2: remove libsoccr_set_sk_data_unbound() from soccr.h
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
These addresses will be used to restore half-closed sockets,
we will need to send a fake fin packet for that.
And in the next patch connect() will be moved into soccr.
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
loge() will start a message with "Error:" and
logd() will start a message with "Debug:".
logerr() will add strerror(errno) to the end of messages.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We get sockets from task_diag, then we block network and dump sockets.
A socket state can be changed between first two steps, so lets get
a socket state on the third step and dump it.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Without this patch when we change libsoccr, criu isn't rebuilt.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Originally the repair mode could be enable only for sockets
in closed and established states. Starting with the 4.10 kernel,
it is possible to enable the repair mode for all connected sockets
and now we can dump syn-sent and half closed sockets.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
It will be used to generate tcp packets for restoring half-closed sockets.
travis-ci: success for series starting with [01/21] build: install libnet-dev
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Some of the loge messages where missing a '\n'. For consistency this
adds '\n' to all messages which were missing it.
travis-ci: success for libsoccr debug/info messages (rev2)
Signed-off-by: Adrian Reber <areber@redhat.com>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Initialize logging function and log_level for libsoccr.
Otherwise all soccr messages are skipped.
travis-ci: success for libsoccr debug/info messages
Reported-by: Adrian Reber <adrian@lisas.de>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Otherwise there is a broken symlink which makes tags/etags unhappy.
travis-ci: success for soccr: Generate config.h dynamically (rev5)
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
[extended .gitignore, maked link relative to sources]
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Made a symlink to CRIU's generated config.h, played with
moving some code to soccr.h header.
Changed types from u32 to uint32_t, otherwise we should
include CRIU's "int.h" in soccr, but I'm lazy to add that.
Cc: Pavel Emelyanov <xemul@virtuozzo.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
TODO: Write proper man page/doc.
TODO: Document that the library spoils the reuseaddrs option on socket.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For now we only support ESTABLISHED state, so this is mostly
a preparation for the future extention.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
In TCP repair some stuff should be restored before calling
connect/bind on socker and some -- after it.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
As an API for dumping and restoring the data the structure
with __u32-s is used. Getting one (and restoring) also involve
the size of the structure so that both -- caller and library --
can know which "version" of it the other one is using.
As a starting point all the fields that tcp_entry carries are
used.
Also note, that only tcp connection info is handled by the
library. Stuff like addresses and ports, socket options and
other are left for the caller to care for.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
The calls put socket into a mode where it can be saved or
restored. Add take one out of this mode.
TODO: In criu we turn repair off (resume the socket) from
restorer blob that cannot use the library. Need to somehow
fix it.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
In order to get messages from libsoccr I propose to
introduce a printing callback that can be set by library
user and that will be called by libsoccr.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For now build the .a library not to produce criu wrappers.
Next version should include the .so library as well.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Function has unsigned int return value, and error is coded by 0, not -1.
travis-ci: success for ns: Fix error return value in __get_ns_id()
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
The man page says, it's not optional argument:
The application shall ensure that the infop argument
points to a siginfo_t structure.
Fixes:
Test output: ================================
17:40:26.128: 5: FAIL: helper_zombie_child.c:33: waitid (errno = 14 (Bad address))
17:40:26.128: 4: FAIL: helper_zombie_child.c:78: read (errno = 2 (No such file or directory))
17:40:26.129: 3: ERR: test.c:229: Test exited unexpectedly with code 1
Test output: ================================
15:30:49.021: 30: ERR: sigpending.c:213: waitid (errno = 14 (Bad address))
15:30:49.021: 29: ERR: test.c:229: Test exited unexpectedly with code 1
and etc.
travis-ci: success for 32-bit tests fixes
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For ZDTM purposes, poll.c uses delta of 5 seconds, which is fine, since
tests don't take a very long time to pass. However, in vzt-cpt, we start
a batch with multitude of tests and starting them all could take over an
hour. Hence, the test is bound to fail. An easy solution would be to
introduce an environment variable, which would allow to specify the
delta - ZDTM_DELTA. If it's unset (which is default behaviour in ZDTM) -
it would default to 5 seconds. However, if it's set (what we can do from
vzt-cpt) - then it will use delta which is set by our test.
travis-ci: success for test: add variable delta for poll.c test (rev2)
Signed-off-by: Vitaly Ostrosablin <vostrosablin@virtuozzo.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
CID 173737 (#1 of 2): Resource leak (RESOURCE_LEAK)
9. leaked_handle: Handle variable rmntns_root going out of scope leaks the handle.
travis-ci: success for files: don't leak a file descriptor
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
When splicing page server data from UNIX socket we may get
error (EAGAIN) from splice if no data is available on the
socket yet. This is because the SPLICE_F_NONBLOCK flag is
checked by af_unix.c in the kernel to decide whether or
not to do blocking read.
This is not symmetrical with TCP sockets, which only check
for the socket's O_NONBLOCK flag for the same decicion.
Dropping the SPLICE_F_NONBLOCK flag is not possible too, as
otherwise we'll block on the pipe when trying to put data
into it. Even if part of the data fits into it kernel would
block anyway untill full buffer is in. And there will be
no read() from the pipe, as it should happen one step later
in the same task.
So to untie this, we need to wait for the data explicitly
with poll().
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
A unix socket can be bound to itself, so we need only one socket
instead of a pair of pipe.
If we use a socket, we can peek a message and if it isn't for us
we don't need to send it back.
Currently a process reads message from pipe, if it isn't for him,
it sends it back and sleeps for a random timeout. In a worst case
each waiter will read and send a message back, before the right
one will get it.
travis-ci: success for series starting with [1/3] zdtm: call test_init() before doing anything
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
travis-ci: success for series starting with [1/3] zdtm: call test_init() before doing anything
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
test_init() has to be executed first, because it prepares a container.
travis-ci: success for series starting with [1/3] zdtm: call test_init() before doing anything
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
travis-ci: success for Fetch mmap_minimal_address and drop MAP_ANON
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We assume that mmap_minimal_address with non-default
setting is not widespread but it's not true. Instead
lets fetch this setting runtime.
travis-ci: success for Fetch mmap_minimal_address and drop MAP_ANON
Signed-off-by: Cyrill Gorcunov <gorcunov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
This one covers the following cases:
* socket is bound to an interface
* socket isn't bound to an interface
https://github.com/xemul/criu/issues/73
travis-ci: success for series starting with [1/2] socket: Add support for SOCK_PACKET sockets
Signed-off-by: Gleb Valin <the7winds@yandex.ru>
Signed-off-by: Eugene Batalov <eabatalov89@gmail.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Test seem to assume that zdtm lib directory is on PATH, but if it's not
- calling for parseargs.sh would fail (if cwd is static). Therefore,
this patch changes this to ../lib/parseargs.sh which should be fine both
for zdtm and vzt-cpt.
travis-ci: success for test: fix shell script tests (conntracks & route_rules)
Signed-off-by: Vitaly Ostrosablin <vostrosablin@virtuozzo.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Compiling zdtm on the latest Fedora rawhide gives errors like this:
ns.c: In function ‘prepare_mntns’:
ns.c:122:13: error: In the GNU C Library, "makedev" is defined
by <sys/sysmacros.h>. For historical compatibility, it is
currently defined by <sys/types.h> as well, but we plan to
remove this soon. To use "makedev", include <sys/sysmacros.h>
directly. If you did not intend to use a system-defined macro
"makedev", you should undefine it after including <sys/types.h>. [-Werror]
if (mknod("/dev/ptmx", 0666 | S_IFCHR, makedev(5, 2)) == 0) {
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
This commit fixes those errors.
travis-ci: success for zdtm: fix '"makedev" is defined by <sys/sysmacros.h>'
Signed-off-by: Adrian Reber <areber@redhat.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
When going daemon criu services ask cr_daemon() to keep
some descriptors open and the routine dup2-s them. The
problem is that the original descriptor is left where it
was and is leaked.
travis-ci: success for util: Do not leak dup-ed file descriptor
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
The idea of this test is to check non-root shared mounts
which lives in a few shared groups.
v2: fix the link description.
travis-ci: success for test: check non-root shared mounts (rev2)
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
The PagemapEntry and struct iovec are quite interchangeable.
PagemapEntry can be used directly in the few places that converted it to
struct iovec.
The only place that used iovec2pagemap can be switched to open-coded
assignments without really hurting code readability.
travis-ci: success for pagemap: minor cleanups (rev2)
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
CID 172196, issues/233.
The advance method can keep pr->pe NULL while the rest of
the code doesn't count on that.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Acked-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
We have pr->pe with the same data available widely, no need
to copy this thing twice.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Acked-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
The only external users of ->seek_page method are page-xfer and uffd and
both of them are anyway interested in the entire pagemap rather than a
particular page.
The added 'skip_zero' parameter allows proper detection of zero pagemaps by
uffd.
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
