[CRIU] Re: [PATCH cr] restore: use a currect stack for new processes

[Cyrill Gorcunov]

On Fri, Aug 24, 2012 at 10:27:44PM +0400, Andrey Vagin wrote:
>  
> -	stack = mmap(NULL, STACK_SIZE, PROT_WRITE | PROT_READ,
> -			MAP_PRIVATE | MAP_GROWSDOWN | MAP_ANONYMOUS, -1, 0);
> -	if (stack == MAP_FAILED) {
> -		pr_perror("Failed to map stack for the child");
> -		goto err;
> -	}
> -
>  	ca.item = item;
>  	ca.clone_flags = ns_clone_flags;
>  
> @@ -544,7 +537,10 @@ static inline int fork_with_pid(struct pstree_item *item, unsigned long ns_clone
>  		if (netns_pre_create())
>  			goto err_unlock;
>  
> -	ret = clone(restore_task_with_children, stack + STACK_SIZE,
> +	asm ("mov %%rsp, %0" : "=r" (stack));
> +	stack -= PAGE_SIZE;
> +
> +	ret = clone(restore_task_with_children, stack,
>  			ca.clone_flags | SIGCHLD, &ca);
>  

This looks a bit hackish for me but I can live with that ;)
I wonder what would happen if curren sp somewhere near stack bottom
and your "- page_size" move sp to some different VMA which doesn't
even belong to original stack VMA?

	Cyrill