[CRIU] [PATCH 1/2] restore: TASK_HELPERs live through restore stage

[Andrew Vagin]

Hi Tycho,

I highly recommend to execute tests. It's easy, you need to execute
only one command "make test".

ns/static/session00 fails with this patch.

[root at avagin-fc19-cr criu]# bash test/zdtm.sh ns/static/session00
================================= CRIU CHECK =================================
Looks good.
Execute zdtm/live/static/session00
./session00 --pidfile=session00.pid --outfile=session00.out
/root/git/criu/test
Dump 24400
Restore
Check results 24439
Waiting...
10:18:41.865:     4: FAIL: session00.c:213: The process with pid 16 returns 256
 (errno = 10 (Permission denied))
10:18:41.867:     4: ERR: session00.c:221: 10 isn't waited (errno = 0 (Permission denied))
Test: zdtm/live/static/session00, Result: FAIL
==================================== ERROR ====================================
Test: zdtm/live/static/session00, Namespace: 1
Dump log   : /root/git/criu/test/dump/static/session00/24400/1/dump.log
--------------------------------- grep Error ---------------------------------
------------------------------------- END -------------------------------------
Restore log: /root/git/criu/test/dump/static/session00/24400/1/restore.log
--------------------------------- grep Error ---------------------------------
------------------------------------- END -------------------------------------
Output file: /root/git/criu/test/zdtm/live/static/session00.out
------------------------------------------------------------------------------
10:18:41.865:     4: FAIL: session00.c:213: The process with pid 16 returns 256
 (errno = 10 (Permission denied))
10:18:41.867:     4: ERR: session00.c:221: 10 isn't waited (errno = 0 (Permission denied))
  PID   SID COMMAND
    1     1 session00
    4     4 session00
    7     7  \_ session00
    8     4  \_ session00
   11    11  \_ session00
   12     4  |   \_ session00
   13    13  \_ session00
   14    14      \_ session00
   15     4          \_ session00
    6     4 session00
   10     9 session00
   16     1 ps
9 return 256
Time to stop and check
  PID   SID COMMAND
    1     1 session00
    4     4 session00
   13    13  \_ session00
   14    14  |   \_ session00
   15     4  |       \_ session00
   16     4  \_ criu <defunct>
    7     7  \_ session00
    8     4  \_ session00
   11    11  \_ session00
   12     4      \_ session00
   10     9 session00
    6     4 session00
   17     1 ps
kill(4, SIGTERM)
------------------------------------- END -------------------------------------
================================= ERROR OVER =================================

On Fri, Sep 05, 2014 at 02:38:04PM -0500, Tycho Andersen wrote:
> In order to use TASK_HELPERS to open files from dead processes, they should
> persist through the end of the restore phase, since that is when the fds are
> set up.
> 
> Signed-off-by: Tycho Andersen <tycho.andersen at canonical.com>
> ---
>  cr-restore.c | 37 +++++++++++++++++++++++--------------
>  1 file changed, 23 insertions(+), 14 deletions(-)
> 
> diff --git a/cr-restore.c b/cr-restore.c
> index cefa78e..5f481b0 100644
> --- a/cr-restore.c
> +++ b/cr-restore.c
> @@ -92,6 +92,7 @@ static int prepare_restorer_blob(void);
>  static int prepare_rlimits(int pid, CoreEntry *core);
>  static int prepare_posix_timers(int pid, CoreEntry *core);
>  static int prepare_signals(int pid, CoreEntry *core);
> +static int restore_switch_stage(int next_stage);
>  
>  static int root_as_sibling;
>  
> @@ -765,35 +766,42 @@ err:
>  
>  static int restore_one_alive_task(int pid, CoreEntry *core)
>  {
> +	int ret = -1;
> +
>  	pr_info("Restoring resources\n");
>  
>  	rst_mem_switch_to_private();
>  
> -	if (pstree_wait_helpers())
> -		return -1;
> -
>  	if (prepare_fds(current))
> -		return -1;
> +		goto err;
>  
>  	if (prepare_file_locks(pid))
> -		return -1;
> +		goto err;
>  
>  	if (open_vmas(pid))
> -		return -1;
> +		goto err;
>  
>  	if (open_cores(pid, core))
> -		return -1;
> +		goto err;
>  
>  	if (prepare_signals(pid, core))
> -		return -1;
> +		goto err;
>  
>  	if (prepare_posix_timers(pid, core))
> -		return -1;
> +		goto err;
>  
>  	if (prepare_rlimits(pid, core) < 0)
> -		return -1;
> +		goto err;
> +
> +	if (sigreturn_restore(pid, core))
> +		goto err;
>  
> -	return sigreturn_restore(pid, core);
> +	ret = 0;
> +err:
> +	if (pstree_wait_helpers())
> +		ret = -1;
> +
> +	return ret;
>  }
>  
>  static void zombie_prepare_signals(void)
> @@ -930,9 +938,9 @@ static int restore_one_task(int pid, CoreEntry *core)
>  		ret = restore_one_alive_task(pid, core);
>  	else if (current->state == TASK_DEAD)
>  		ret = restore_one_zombie(pid, core);
> -	else if (current->state == TASK_HELPER)
> -		ret = 0;
> -	else {
> +	else if (current->state == TASK_HELPER) {
> +		ret = restore_finish_stage(CR_STATE_RESTORE);
> +	} else {
>  		pr_err("Unknown state in code %d\n", (int)core->tc->task_state);
>  		ret = -1;
>  	}
> @@ -1489,6 +1497,7 @@ static inline int stage_participants(int next_stage)
>  	case CR_STATE_FORKING:
>  		return task_entries->nr_tasks + task_entries->nr_helpers;
>  	case CR_STATE_RESTORE:
> +		return task_entries->nr_threads + task_entries->nr_helpers;
>  	case CR_STATE_RESTORE_SIGCHLD:
>  		return task_entries->nr_threads;
>  	case CR_STATE_RESTORE_CREDS:
> -- 
> 1.9.1
> 
> _______________________________________________
> CRIU mailing list
> CRIU at openvz.org
> https://lists.openvz.org/mailman/listinfo/criu