[CRIU] [PATCH REBASED 3/3] ns: Make pid_ns helpers as children of criu main process

Kirill Tkhai ktkhai at virtuozzo.com
Tue Jul 11 12:02:27 MSK 2017


On 11.07.2017 03:46, Andrei Vagin wrote:
> On Thu, Jul 06, 2017 at 02:14:17PM +0300, Kirill Tkhai wrote:
>> If a task, holding userns_sync_lock unexpectedly exits,
>> criu will hang on error path in restore_root_task(),
>> because it can't use usernsd to destroy them.
>>
>> Lets remove the intermediary: we'll create pid_ns helpers
>> as children of criu main task, and criu main task will
>> be able to use simple kill() to stop them.
>>
>> Signed-off-by: Kirill Tkhai <ktkhai at virtuozzo.com>
>> ---
>>  criu/cr-restore.c |    8 ++++++++
>>  criu/namespaces.c |   35 +++--------------------------------
>>  2 files changed, 11 insertions(+), 32 deletions(-)
>>
>> diff --git a/criu/cr-restore.c b/criu/cr-restore.c
>> index 965770f48..8d8349f32 100644
>> --- a/criu/cr-restore.c
>> +++ b/criu/cr-restore.c
>> @@ -1470,6 +1470,7 @@ static inline int fork_with_pid(struct pstree_item *item)
>>  static void sigchld_handler(int signal, siginfo_t *siginfo, void *data)
>>  {
>>  	int status, pid, exit;
>> +	struct ns_id *ns;
>>  
>>  	while (1) {
>>  		pid = waitpid(-1, &status, WNOHANG);
>> @@ -1489,6 +1490,13 @@ static void sigchld_handler(int signal, siginfo_t *siginfo, void *data)
>>  		break;
>>  	}
>>  
>> +	if (!current) {
>> +		ns = handle_pid_ns_helper_exit(pid);
> 
> Why is it out of the previous loop? What if two helpers exit together?

As we destroy the helpers with SIGCHLD blocked in do_destroy_pidns_helpers(),
a seeing of a one helper is already a error, and we should stop restore
in this case.

We mustn't see a helper in this function as we newer mustn't see any task
there. The helper handle_pid_ns_helper_exit() is only for printing debug
message to understand what happened.
 
>> +		if (ns)
>> +			pr_err("Unexpected pid ns helper: pid=%d, ns_id=%d\n",
>> +				pid, ns->id);
>> +	}
>> +
>>  	if (exit)
>>  		pr_err("%d exited, status=%d\n", pid, status);
>>  	else
> 		...
> 
> 	futex_abort_and_wake(&task_entries->nr_in_progress);
> 
> it means that we always abort "restore", doesn't it?

Yes, we break restore in this case.
 
>> diff --git a/criu/namespaces.c b/criu/namespaces.c
>> index 674b43820..fca125b6a 100644
>> --- a/criu/namespaces.c
>> +++ b/criu/namespaces.c
>> @@ -1603,26 +1603,6 @@ struct ns_id *handle_pid_ns_helper_exit(pid_t real_pid)
>>  	return ns;
>>  }
>>  
>> -static void usernsd_handler(int signal, siginfo_t *siginfo, void *data)
>> -{
>> -	pid_t pid = siginfo->si_pid;
>> -	struct ns_id *ns;
>> -	int status;
>> -
>> -	while (pid) {
>> -		pid = waitpid(-1, &status, WNOHANG);
>> -		if (pid <= 0)
>> -			return;
>> -
>> -		ns = handle_pid_ns_helper_exit(pid);
>> -		if (!ns)
>> -			pr_err("Spurious pid ns helper: pid=%d\n", pid);
>> -
>> -		pr_err("%d finished unexpected: status=%d\n", pid, status);
>> -		futex_abort_and_wake(&task_entries->nr_in_progress);
>> -	}
>> -}
>> -
>>  static int usernsd_recv_transport(void *arg, int fd, pid_t pid)
>>  {
>>  	if (install_service_fd(TRANSPORT_FD_OFF, fd) < 0) {
>> @@ -1666,11 +1646,6 @@ static int usernsd(int sk)
>>  {
>>  	pr_info("uns: Daemon started\n");
>>  
>> -	if (criu_signals_setup(usernsd_handler) < 0) {
>> -		pr_err("Can't setup handler\n");
>> -		return -1;
>> -	}
>> -
>>  	while (1) {
>>  		struct unsc_msg um;
>>  		static char msg[MAX_UNSFD_MSG_SIZE];
>> @@ -2725,7 +2700,7 @@ static int do_create_pid_ns_helper(void *arg, int sk, pid_t unused_pid)
>>  			unlock_last_pid();
>>  			goto restore_ns;
>>  		}
>> -	child = fork();
>> +	child = sys_clone_unified(CLONE_PARENT|SIGCHLD, NULL, NULL, NULL, 0);
>>  	if (!child)
>>  		exit(pid_ns_helper(ns, sk));
>>  	saved_errno = errno;
>> @@ -2787,7 +2762,7 @@ int create_pid_ns_helper(struct ns_id *ns)
>>  	return 0;
>>  }
>>  
>> -static int do_destroy_pid_ns_helpers(void *arg, int fd, pid_t unused)
>> +static int do_destroy_pid_ns_helpers(void)
>>  {
>>  	int status, sig_blocked = true, ret = 0;
>>  	sigset_t sig_mask;
>> @@ -2836,11 +2811,7 @@ int destroy_pid_ns_helpers(void)
>>  	if (!(root_ns_mask & CLONE_NEWPID))
>>  		return 0;
>>  
>> -	if (userns_call(do_destroy_pid_ns_helpers, 0, NULL, 0, -1) < 0) {
>> -		pr_err("Can't create pid_ns helper\n");
>> -		return -1;
>> -	}
>> -	return 0;
>> +	return do_destroy_pid_ns_helpers();
>>  }
>>  
>>  struct ns_desc pid_ns_desc = NS_DESC_ENTRY(CLONE_NEWPID, "pid", "pid_for_children");
>>


More information about the CRIU mailing list