Glibc has BUG with process creation:
https://sourceware.org/bugzilla/show_bug.cgi?id=21386
It doesn't behave well when parent and child are from
different pid namespaces and have the same pid.
Use raw syscall without glibc's asserts as workaround.
Also, use raw syscall for getpid() in tests too,
as these two function go in the pair (glibc's getpid()
relies on glibc's fork()).
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
p + FAST_SIZE > buffer + length
In this sentence p + FAST_SIZE may be bigger than (1<<32),
and we will be in trouble.
$ gdb -c coredump test/zdtm/static/write_read01
(gdb) p p
$3 = (uint8_t *) 0xffffa89e
(gdb) p buffer
$4 = (uint8_t *) 0xfff06780
(gdb) p length
$5 = 1000000
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Currently test_daemon() sends a signal to a parent process,
when a test completed a prepartion stage. But it doesn't work,
if a child changed uid and gid, because it doesn't have enough
permissions in this case.
This patch reworks a synchronization scheme to use a futex
instead of sending signals.
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Debian is compiled for z900 - therefore we can't use the "laa"
instruction.
Fix this and replace "laa" by the older "cs" (compare and swap) loop.
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The function is not included into the library, so having its prototype
there was a shortcut. Move it to a separate include file.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
After the commit
02c763939c10 ("test/zdtm: unify common code")
CFLAGS with -D_GNU_SOURCE defined in the top Makefile
are being passed to tests Makefiles.
As _GNU_SOURCE is also defined by tests, that resulted in
zdtm tests build failures:
make[2]: Entering directory `/home/criu/test/zdtm/lib'
CC test.o
test.c:1:0: error: "_GNU_SOURCE" redefined [-Werror]
#define _GNU_SOURCE
^
<command-line>:0:0: note: this is the location of the previous definition
cc1: all warnings being treated as errors
make[2]: *** [test.o] Error 1
However, we didn't catch this in time by Travis-CI, as zdtm.py doesn't
do `make zdtm`, rather it does `make -C test/zdtm/{lib,static,transition}`.
By calling middle makefile this way, it doesn't have _GNU_SOURCE in
CFLAGS from top-Makefile.
I think the right thing to do here - is following CRIU's way:
rely on definition of _GNU_SOURCE by Makefiles.
This patch is almost fully generated with
find test/zdtm/ -name '*.c' -type f \
-exec sed -i '/define _GNU_SOURCE/{n;/^$/d;}' '{}' \; \
-exec sed -i '/define _GNU_SOURCE/d' '{}' \;
With an exception for adding -D_GNU_SOURCE in tests Makefile.inc for
keeping the same behaviour for zdtm.py.
Also changed utsname.c to use utsname::domainname, rather private
utsname::__domainname, as now it's uncovered (from sys/utsname.h):
> struct utsname
> {
...
> # ifdef __USE_GNU
> char domainname[_UTSNAME_DOMAIN_LENGTH];
> # else
> char __domainname[_UTSNAME_DOMAIN_LENGTH];
> # endif
Reported-by: Adrian Reber <areber@redhat.com>
Cc: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This commit expands the success with lib/Makefile to the rest of
ztdm/Makefiles. In particular, it moves the common part to Makefile.inc
and let {lib,static,transition}/Makefile use it.
This results in:
- dependencies being handled properly;
- reduction of makefiles length due to unification;
- possibly more proper dependencies and builds.
While at it, let's also:
- make rules for criu-rtc.so silent;
- set default CC=gcc;
- remove duplicate -m32 from CFLAGS.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
When optimizing our Travis runs, I noticed that test/zdtm/lib is built 3
times in a row -- first time the binaries, 2nd and 3rd time the deps,
in parallel (sic!).
I took a look at Makefile and found this, and some more problems,
like:
- dependencies are generated but not included
- cleandep does not depend on clean
- groups.c dependencies are not generated (or used)
- "make all" does not make groups binary
- overcomplicated and unused groups.{pid,out,cleanout} targets
- extra -f flag to rm
- etc.
This patch (hopefully) fixes all of the above. In addition, it provides
a few enhancements:
- enables "silent make" rules, so now lib/ is in line with the rest of zdtm
- adds "dep" target (not used, just for convenience)
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It is an rudiment of zdtm.sh. Now we don't add ps to a test root
and we want to avoid forks in tests where it isn't supposed to be
called.
travis-ci: success for zdtm: don't call ps from tests
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Unfortunately the kernel wakes up only one waiter even
if they waits with MSG_PEEK, so the introduced scheme doesn't work.
Reported-by: Mr Jenkins
travis-ci: success for Revert "zdtm: use a unix socket instead of a pipe to synchronizer processes"
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
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>
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>
Each test has a process which controls other processes and
if it dies the other processes have to be killed.
travis-ci: success for zdtm: exit a test if the "root" process died
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Since glibc is moving away from implicitly including sys/sysmacros.h
all the time via sys/types.h, include the header directly in more
places. This seems to cover most makedev/major/minor usage.
Signed-off-by: Yixun Lan <dlan@gentoo.org>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
This is useful so that hooks can do a $(cat $pidfile) or [ -f $pidfile ]
and rely on the result.
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
It-makes-sence-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
This one will contain pid of the task living in a sub-ns and
waiting for the rest of the test to get daemonized.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Acked-by: Tycho Andersen <tycho.andersen@canonical.com>
Travised-by: https://travis-ci.org/xemul/criu/builds/170726663
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>
Here is a race when someone umounted something and this operation
isn't propagated into our namespace.
CRIU | Another process
-----------------------------------------------------------------
pivot_root(".", put_root) |
mount(put_root, MS_REC|MS_PRIVATE) |
| umount /xxx/yyy
| umount /xxx -> EBUSY
umount(put_root)
We do this to not affect mounts in put_root, but we can mask
these mounts as slave and this will work for us and for external
users.
Reported-by: Mr Travis
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
It can hapen in tty tests, where we get SIGHUP.
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
...and fix misprints that weren't caught before.
I guess, I never fixed that much (possible) bugs by one commit, heh.
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
In the next patch, we'll introduce an option to allow for leaving zombie
processes in the pid ns for the test so that we can test the behavior of
zombies. Let's not reap everything after restore, since we'll reap the
restored zombies as well.
v2: restore the old behavior when in reap mode
CC: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Acked-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We'll use this variable in the next test to make sure the test suite
doesn't accidentally reap the zombie we want to leave around for the actual
test.
This is kind of ugly and there might be a better way to pass information to
the test's init, I'm open for suggestions :)
CC: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Acked-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
00:24:30.269: 44: ERR: test.c:208: Test exited on unexpected signal
It's a correct name for this function. For example, we use this name
here criu/arch/ppc64/include/asm/atomic.h
And you can find the same set of functions here:
test/zdtm/lib/arch/aarch64/include/asm/atomic.h
I think this code is more readable (understandable) than just
return t - 1.
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
CRIU processes save *.gcda and *.gcno files near source files,
so when we restore tests into another mount namespace,
we need to have access to sources from this namespace.
Cc: Sergey Bronnikov <sergeyb@openvz.org>
Reported-by: Sergey Bronnikov <sergeyb@openvz.org>
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For the caller there should be only one pidfile -- the one
that should be fired with TERM signal on stop. All the other
stuff is test's internal business.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Replace stack alignment magic constant with
__stack_aligned__ macro.
Also align stack for sigaltstack test case.
Signed-off-by: Vijaya Kumar K <vijayak@caviumnetworks.com>
Reviewed-by: Christopher Covington <cov@codeaurora.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
arm64 requires stack pointer to be aligned to 16 bytes.
Update all test cases that are using clone system call
to have 16 byte aligned stack pointer.
Signed-off-by: Vijaya Kumar K <vijayak@caviumnetworks.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
So here's the new test class that handles the test from
groups set. The class is inherited from zdtm_test one as
what it does is -- starts the pseudo-init in ns/uns flavors
and asks one to spawn() the sub-tests from the list.
All groups tests can only be run inside ns flavor, so if
the host flavor is asked, just the pseudo-init is spawned.
This is because using ns flavor is the easiest way to spawn
all the sub tests under this init (however, h flavor can be
supported by marking the pseudo-init as sub-reaper).
On stop this pseudo-init is signalled to stop, it in turn
stops all the sub-tests and then exits. When the pid
namespace destruction is complete, the sub-tests .out-s are
checked.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
When starting inside ns flavor the test_init() routine prepares
the binary to be run inside namespaces. In particular this routine
fork()-s an init, execve()-s one to pick up mappings and exe from
the new mntns and then fork()-s the test itself. In order to go
back to test_init() for test initialization the execve() is done
again, but it's actually not required and confuses the reader.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This one is to set up uids for userns, do ip l s lo up for netns
and do the prepare_mntns(). BTW, the latter's code is shifted one
tab left as this is where it should be.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The ns and uns flavors are MUCH better ways of launching tests in namespaces.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Acked-by: Andrew Vagin <avagin@virtuozzo.com>
Instead of doing some dirs/files/links in launcher and some in
libzdtm, make everything in the launcher. The library is then
just to set up the namespaces.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currently numbers for /dev/net/tun and /dev/rtc are hardcoded and
they changed after the last kernel update.
Reported-by: Mr Jenkins
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Log function pr_perror() automatically adds
- numeric errno value
- string errno value (aka strerror(errno), aka %m)
- a newline
So, its callers don't have to do all these things.
While at it, also
- fix or improve some message texts
- replace fnprintf() with pr_perror()
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Acked-by: Andrew Vagin <avagin@gmail.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
In (highly unlikely) case localtime() failed, test_msg() calls
pr_perror() which is a wrapper around test_msg() so we're
calling ourselves, which can lead to infinite recursion if
localtime() keeps failing. Use fprintf() here to avoid this.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Acked-by: Andrew Vagin <avagin@gmail.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
