[CRIU] [PATCH] zdtm: Fix fd01 cleanup

[Kirill Tkhai]

On 31.01.2018 21:20, Andrei Vagin wrote:
> On Tue, Jan 30, 2018 at 06:19:31PM +0300, Kirill Tkhai wrote:
>> waitpid() does not return child pid, when child has not exited.
>> So, we can't use it to find pids of children.
>>
>> Signed-off-by: Kirill Tkhai <ktkhai at virtuozzo.com>
>> ---
>>  test/zdtm/static/fd01.c |   17 ++++++++++++-----
>>  1 file changed, 12 insertions(+), 5 deletions(-)
>>
>> diff --git a/test/zdtm/static/fd01.c b/test/zdtm/static/fd01.c
>> index 7f0256a03..390d9d126 100644
>> --- a/test/zdtm/static/fd01.c
>> +++ b/test/zdtm/static/fd01.c
>> @@ -22,11 +22,19 @@ int main(int argc, char **argv)
>>  	unsigned int i, max_nr, flags;
>>  	int fd, status, ret;
>>  	struct rlimit rlim;
>> +	futex_t *futex;
>>  	char buf[16];
>>  	pid_t pid;
>>  
>>  	test_init(argc, argv);
>>  
>> +	futex = mmap(NULL, sizeof(*futex), PROT_WRITE | PROT_READ, MAP_ANONYMOUS|MAP_SHARED, -1, 0);
>> +	if (futex == MAP_FAILED) {
>> +		fail("mmap");
>> +		exit(1);
>> +	}
>> +	futex_init(futex);
>> +
>>  	fd = open("/proc/sys/fs/nr_open", O_RDONLY);
>>  	if (fd < 0) {
>>  		fail("Can't open /proc/sys/fs/nr_open");
>> @@ -86,7 +94,7 @@ int main(int argc, char **argv)
>>  			fail("fork");
>>  			exit(1);
>>  		} else if (!pid) {
>> -			pause();
>> +			futex_wait_while(futex, 0);
>>  			exit(0);
>>  		}
>>  	}
>> @@ -95,10 +103,9 @@ int main(int argc, char **argv)
>>  	test_waitsig();
>>  
>>  	/* Cleanup */
>> -	while ((pid = waitpid(-1, &status, WNOHANG)) > 0) {
>> -		if (kill(pid, SIGTERM) == 0)
>> -			waitpid(-1, &status, 0); /* Ignore errors */
>> -	}
>> +	futex_set_and_wake(futex, 1);
>> +	while (wait(&status) > 0) /* Ignore errors */
>> +		;
> 
> Pls, check an exit code from children, otherwise it isn't a test,
> because it nevel fails.

Ok, no problem. But the idea was to use zdtm.py debug hooks, which checks
for file descriptors in check_visible_state()...:

    print "%s: Old files lost: %s" % (pid, fold - fnew)
    print "%s: New files appeared: %s" % (pid, fnew - fold)

>In addition, it would be good if you add
>basic checks to be sure that file descriptors are restored correctly.

Sadly, but it won't be possible to check that in sane way, because
the behaviour of created children and files is almost "random".
I don't see a good way to check everything in the test having so many
entities. Maybe we better implement one more? And remain this test
for checks I wrote above.

Kirill