[Devel] Re: [PATCH user-cr v19-rc1] Revert "restart: coordinator in new pidns to always report status via pipe"

[Oren Laadan]

The problem is actually elsewhere: ckpt_coordinator_pidns() should
test for args->wait and wait for the coordinator (which became the
pidns's init) before exiting.

Will send a patch soon.

Oren.

Serge E. Hallyn wrote:
> Hi Oren,
> 
> commit 464ff47b6a1241410f432a55ce4bf93c8bd10861 isn't right.  I guess
> the problem is that ckpt_coordinator_status(), doesn't do a waitpid
> on the global_child_pid?  So if I do a 'restart -w --pids', then the
> restarted task proceeds just fine, but restart immediately returns.
> 
> Maybe the right answer is to modify ckpt_coordinator_status() to
> do a waitpid if ctx->args->wait, but just reverting this commit
> doesn't seem to have any ill effects - all my default tests pass.
> 
> This reverts commit 464ff47b6a1241410f432a55ce4bf93c8bd10861.
> ---
>  restart.c |   25 +++++++++++++------------
>  1 files changed, 13 insertions(+), 12 deletions(-)
> 
> diff --git a/restart.c b/restart.c
> index d5d069a..cfc359b 100644
> --- a/restart.c
> +++ b/restart.c
> @@ -994,12 +994,10 @@ static int ckpt_coordinator_pidns(struct ckpt_ctx *ctx)
>  	ckpt_dbg("forking coordinator in new pidns\n");
>  
>  	/*
> -	 * The coordinator report restart susccess/failure via pipe.
> -	 * (It cannot use return value, because the in the default
> -	 * --wait --copy-status case it is already used to report the
> -	 * root-task's return value).
> +	 * We won't wait for (collect) the coordinator, so we use a
> +	 * pipe instead for the coordinator to report success/failure.
>  	 */
> -	if (pipe(ctx->pipe_coord) < 0) {
> +	if (!ctx->args->wait && pipe(ctx->pipe_coord)) {
>  		perror("pipe");
>  		return -1;
>  	}
> @@ -1035,7 +1033,10 @@ static int ckpt_coordinator_pidns(struct ckpt_ctx *ctx)
>  		return -1;
>  
>  	ctx->args->copy_status = copy;
> -	return ckpt_coordinator_status(ctx);
> +	if (ctx->args->wait)
> +		return ckpt_collect_child(ctx);
> +	else
> +		return ckpt_coordinator_status(ctx);
>  }
>  #else
>  static int ckpt_coordinator_pidns(struct ckpt_ctx *ctx)
> @@ -1093,13 +1094,13 @@ static int ckpt_coordinator(struct ckpt_ctx *ctx)
>  		 * around and be reaper until all tasks are gone.
>  		 * Otherwise, container will die as soon as we exit.
>  		 */
> -
> -		/* Report success/failure to the parent */
> -		if (write(ctx->pipe_coord[1], &ret, sizeof(ret)) < 0) {
> -			perror("failed to report status");
> -			exit(1);
> +		if (!ctx->args->wait) {
> +			/* report status because parent won't wait for us */
> +			if (write(ctx->pipe_coord[1], &ret, sizeof(ret)) < 0) {
> +				perror("failed to report status");
> +				exit(1);
> +			}
>  		}
> -
>  		ret = ckpt_pretend_reaper(ctx);
>  	} else if (ctx->args->wait) {
>  		ret = ckpt_collect_child(ctx);
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers