travis-ci: success for core-aarch64: mark gregs and clear_tid_addr as (criu).hex (rev2)
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

For compel-ization it's MUCH easier to introduce single
constant, all the more so no VMAs can be shorter than
one page :)

travis-ci: success for parasite: Relax size checking for get_exec_start()
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Binfmt_misc may not exist on NS_ROOT, and exist on NS_OTHER that time.
So, add binfmt_misc only if no one mnt_ns has it.
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Since cr-time mount parent may be shared, and we do
not propagate the mount there, this patch excludes
them from validation.
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

There is a small race between exitings tasks and cgroup freezer:
it can report a pid of exiting process which we will consider
as a zombie and refuse to checkpoint. Lets do a trick instead:
if we meets such task just wait a bit and repeate a freezing.

v3: Use EAGAIN instead of EPERM since EPERM collide
    with general -1 error code.

travis-ci: success for freeze -- Wait for exiting tasks to complete (rev3)
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

It is already present in CLI and RPC, so libcriu should reflect it too.

travis-ci: success for lib: add inherit_fd
Signed-off-by: Ruslan Kuprieiev <rkuprieiev@cloudlinux.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

* make sure that tracefs is actually loaded, and try and access it after
  restore
* don't mount debugfs in /, leaving an ugly directory in /test
* clean up the debugfs mount point (with a hook, since the test can't
  unmount it due to perms)

travis-ci: success for test: robustify mnt_tracefs test
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

When this code is compiled by clang, it optimizes two byte writes into
one word write, like this:

>        fake_grow_down[0] = 'c';
>        *(fake_grow_down - 1) = 'b';
> 401b60:       66 41 c7 46 ff 62 63    movw   $0x6362,-0x1(%r14)

This is incorrect, as the stack is supposed to grow page by page,
so we need to touch one page then another, i.e. the order is important.

To fix, let's use volatile pointer. After this change, the following
(correct) code is generated:

>        *p-- = 'c';
>   401b60:       41 c6 06 63             movb   $0x63,(%r14)
>        *p = 'b';
>   401b64:       41 c6 46 ff 62          movb   $0x62,-0x1(%r14)

[v2: same fix for another similar place]

Cc: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Acked-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Move GCOV to travis environment, so now we can change it easily
on a per-build basis.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

I found that everything we test on travis-ci is built with -O0.
I believe this is the wrong thing to do as we test something that is
compiled not the way it would be compiled for any real use.

Drop enabling DEBUG for GCOV.

Signed-off-by: Kir Kolyshkin <kir at openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

It is easier to handle it this way
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

This "struct ctl_msg" initialization is not really needed, as it is
initalized right away in __parasite_daemon_wait_msg() function.

The real reason of removing this init is when compiled by clang
with -O0 etc (i.e. compiler flags set when GCOV=1 env var is present),
clang inserts a call to memset to initialize the structure. This,
of course, leads to "compel piegen" complaining:

	Unexpected undefined symbol: `memset'. External symbol in PIE?

One way of solving it would be to provide a memset() which will call
builtin_memset(), or a simple C implemenation. But as this is the only
place that triggers such a call to memset(), and this initialization
is not really needed, it's easier just to remove it.

Now, there might be more cases like this in the future, but for now
we are good.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

This 'struct timespec' initialization is non needed as we call
clock_gettime() right away which is supposed to fill it in.

The real reason of removing this init is when compiled by clang
with -O0 etc (i.e. compiler flags set when GCOV=1 env var is present),
clang inserts a call to memset to initialize the structure. This,
of course, leads to "compel piegen" complaining:

	Unexpected undefined symbol: `memset'. External symbol in PIE?

One way of solving it would be to provide a memset() which will call
builtin_memset(), or a simple C implemenation. But as this is the only
place that triggers such a call to memset(), and this initialization
is not really needed, it's easier just to remove it.

Now, there might be more cases like this in the future, but for now
we are good.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

CLang doesn't like explicit NULL pointer dereference:
> maps007.c:176:2: error: indirection of non-volatile null pointer will be deleted, not trap [-Werror,-Wnull-dereference]
>         *((int *) 0) = 0;
>         ^~~~~~~~~~~~
> maps007.c:176:2: note: consider using __builtin_trap() or qualifying pointer with 'volatile'
>

Apparently, we need to add volatile to make it happy.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains:
> cmdlinenv00.c:35:11: error: comparison of unsigned expression < 0 is always false [-Werror,-Wtautological-compare]
>                 if (ret < 0) {
>                     ~~~ ^ ~

The fix is to use ssize_t not size_t.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains:
> apparmor.c:86:20: error: too many arguments in call to 'checkprofile' [-Werror]
>         if (checkprofile(0) == 0)
>             ~~~~~~~~~~~~  ^
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains:
> clang -g -O2 -Wall -Werror -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0  -iquote ../lib/arch/x86/include -I../lib   rtc.c ../lib/libzdtmtst.a ../lib/libzdtmtst.a -o rtc
> rtc.c:50:7: error: taking the absolute value of unsigned type 'unsigned long' has no effect [-Werror,-Wabsolute-value]
>                 if (labs(delta - 1000000 / TEST_HZ ) > 100000) {
>                     ^

Declare delta as long.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains:
> clang -g -O2 -Wall -Werror -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0  -iquote ../lib/arch/x86/include -I../lib   groups.c ../lib/libzdtmtst.a ../lib/libzdtmtst.a -o groups
> groups.c:22:9: error: comparison of unsigned expression < 0 is always false [-Werror,-Wtautological-compare]
>         if (ng < 0) {
>             ~~ ^ ~

Declare ng as int.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains:
> clang -g -O2 -Wall -Werror -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0  -iquote ../lib/arch/x86/include -I../lib   vdso01.c ../lib/libzdtmtst.a ../lib/libzdtmtst.a -lrt -o vdso01
> vdso01.c:305:6: error: absolute value function 'abs' given an argument of type 'long' but has parameter of type 'int' which may cause truncation of value
>       [-Werror,-Wabsolute-value]
>         if (abs(ts1.tv_sec - ts2.tv_sec) > TIME_DELTA_SEC) {
>             ^
> vdso01.c:305:6: note: use function 'labs' instead

Let's use labs() indeed.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

When compiling with clang, it complains like this:

> clang -g -O2 -Wall -Werror -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0  -iquote ../lib/arch/x86/include -I../lib   inotify_system.c ../lib/libzdtmtst.a ../lib/libzdtmtst.a -o inotify_system
> inotify_system.c:323:50: error: comparison of unsigned expression < 0 is always false [-Werror,-Wtautological-compare]
>         if ((common_desc.inot < 0) || (common_desc.file < 0) || \
>                                        ~~~~~~~~~~~~~~~~ ^ ~
> inotify_system.c:324:21: error: comparison of unsigned expression < 0 is always false [-Werror,-Wtautological-compare]
>                         (common_desc.dir < 0) || (common_desc.link < 0)) {
>                          ~~~~~~~~~~~~~~~ ^ ~
> inotify_system.c:324:47: error: comparison of unsigned expression < 0 is always false [-Werror,-Wtautological-compare]
>                         (common_desc.dir < 0) || (common_desc.link < 0)) {
>                                                   ~~~~~~~~~~~~~~~~ ^ ~
> 3 errors generated.

Indeed these fields are uint32_t. As they are used to store the results of
inotify_init() and inotify_add_watch() function calls, the type should be int.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains a number of times on our inline asm code, like this:

> fpu00.c:13:6: error: ambiguous instructions require an explicit suffix
> (could be 'flds', 'fldl', or 'fldt')
>                           "fld  %0\n"
>                           ^
> <inline asm>:1:2: note: instantiated into assembly here
>         fld     -4(%rsp)
>         ^

As floats are used, the suffix should be 's' ("short", 32-bit, float).

Add it where needed.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

CLang doesn't like explicit NULL pointer dereference:

> zombie00.c:52:5: error: indirection of non-volatile null pointer will be
> deleted, not trap [-Werror,-Wnull-dereference]
>                                 *(int *)NULL = 0;
>                                 ^~~~~~~~~~~~
> zombie00.c:52:5: note: consider using __builtin_trap() or qualifying
> pointer with 'volatile'
> 1 error generated.

Apparently, we need to add volatile to make it happy.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

clang complains about an unused function:

> clang -g -O2 -Wall -Werror -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0
> -iquote ./arch/x86/include  -c -o parseargs.o parseargs.c
> parseargs.c:12:1: error: unused function '__check_help'
> [-Werror,-Wunused-function]
> TEST_OPTION(help, bool, "print help message and exit", 0);
> ^
> ./zdtmtst.h:71:2: note: expanded from macro 'TEST_OPTION'
>         param_check_##type(name, &(name));
> \
>         ^
> <scratch space>:46:1: note: expanded from here
> param_check_bool
> ^
> ./zdtmtst.h:84:35: note: expanded from macro 'param_check_bool'
> #define param_check_bool(name, p) __param_check(name, p, int)
>                                   ^
> ./zdtmtst.h:78:22: note: expanded from macro '__param_check'
>         static inline type *__check_##name(void) { return(p); }
>                             ^
> <scratch space>:47:1: note: expanded from here
> __check_help
> ^
> 1 error generated.
> <builtin>: recipe for target 'parseargs.o' failed

As far as I can tell, the functions __check_##name are generated in
order to check the argument type by the compiler.

For gcc, the "inline" keyword is enough to suppress the "unused function"
warning even when -Wunused-function is used. For clang, it's not the
case (see https://llvm.org/bugs/show_bug.cgi?id=22712).

A way to "use" the function is to use its name as a pointer and cast it
to void. This is what this patch does.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Commit 308741d0 removes all calls to path_length(), but the function
itself is still there. Now, clang complains that it is unused.

Remove it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

The following error is emitted by clang:

>   CC       criu/filesystems.o
> criu/filesystems.c:280:13: error: variable 'ret' is used uninitialized
>    whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
>         } else if (bme->extension) {
>                    ^~~~~~~~~~~~~~
> criu/filesystems.c:287:6: note: uninitialized use occurs here
>         if (ret > 0) {
>             ^~~
> criu/filesystems.c:280:9: note: remove the 'if' if its condition is
>    always true
>         } else if (bme->extension) {
>                ^~~~~~~~~~~~~~~~~~~~
> criu/filesystems.c:272:9: note: initialize the variable 'ret' to silence
>    this warning
>         int ret;
>                ^
>                 = 0
> 1 error generated.

This code was a result of commit 398e7d3.

If we look closely, this is a false alarm, as "else if (bme->extension)"
is always true as it was checked before. But this is not very clear,
and the issue with clangs still needs to be fixed.

There are many ways to do so:

1. Initialize ret to 0. This is what initial version of this patch did.

2. Remove the always-true condition, like this:

	-	} else if (bme->extension) {
	+	} else {

In my opinion this would hurt readability.

3. Change the code flow, improving readability at the same time.

I believe that #3 is what this patch does. In addition, it fixes
handling of a few corner cases:

- an overflow in snprintf();
- a case when bme->name is NULL (as it is used for strlen/snprintf,
  there is a potential for SIGSEGV);
- a case of ret == 0 (currently there is no code flow that
  results in ret being 0, so it's just for the future).

[v2: use linux kernel style for 'else' after a block]
[xemul: Fix // comments ]

Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Acked-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Adrian Reber <areber@redhat.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Not a lot this time, just a bunch of bugfixes and improvements.
A lot has happened in -dev branch around compel, hopefully the
next release will have it.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

It's typo fix. MS_STRICTATIME has to be added to other flags there.

travis-ci: success for mount: don't overwrite mount flags by MS_STRICTATIME
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

In particular, if the thing in criu's namespace is also a slave of
something in an external mount namespace (as it is in the tests), we won't
see the master, just the external slave. However, this also counts as an
"external master" because it is an external bind mount, just from something
that was itself a slave.

I'm not sure why the test broke, but it works on ubuntu 16.04 and fails on
16.10, so I suppose something with the debugfs mount configuration has
changed. I can look into it more if we're curious, but this patch fixes it
and should probably go in anyway.

Closes #223

travis-ci: success for mnt: allow external slaves too
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Currently a buffer is allocated on stack and its size is 1024.
Usually we use our static buffer to read proc files and its
size is 4096 (page size).

I know that this patch doesn't fix this problem completly,
it just does it less critical.

https://github.com/opencontainers/runc/issues/1070
travis-ci: success for proc_parse: use the static buffer to parse mountinfo
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

[root@fc24 criu]# python test/zdtm.py run -t zdtm/transition/shmem  --pre 3
=== Run 1/1 ================

======================== Run zdtm/transition/shmem in h ========================
cc -g -O2 -Wall -Werror -fno-strict-aliasing  -iquote ../lib/arch/x86/include -I../lib   shmem.c ../lib/libzdtmtst.a ../lib/libzdtmtst.a -o shmem
Start test
./shmem --pidfile=shmem.pid --outfile=shmem.out
Run criu pre-dump
Run criu pre-dump
Run criu pre-dump
Run criu dump
Run criu restore
Send the 15 signal to  33
Wait for zdtm/transition/shmem(33) to die for 0.100000
Test output: ================================
15:12:25.444:    33: FAIL: shmem.c:70: checksum mismatch: ea71000 109c9000

Cc: Eugene Batalov <eabatalov89@gmail.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Only real pid is used internally now.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Localize the processes_to_wait calculations in seize.c. In
order to distinguish dead tasks which has already been wait()-ed
from dead tasks that hasn't introduce internal 'zombie' state
for seize_wait_task().
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Don't do comparisons of creds insize core seizing routine. Instead,
pull the creds on the caller's stack and compare them there if
required.

At the same time more the proc_status_creds_dumpable() to the place
where it's only needed and make it static.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
[ldufour@linux.vnet.ibm.com: mostly rewrite to fix functional issue]
Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>

Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>