Re: [v2 PATCH] mm: introduce arg_lock to protect arg_start|end and env_start|end in mm_struct

[Cyrill Gorcunov]

On Tue, Mar 27, 2018 at 07:00:56AM +0900, Tetsuo Handa wrote:
> 
> >             To be fair I would prefer to drop this old per-field
> > interface completely. This per-field interface was rather an ugly
> > solution from my side.
> 
> But this is userspace visible API and thus we cannot change.

Hi! We could deplrecate this API call for a couple of releases
and then if nobody complain we could rip it off completely.
There should not be many users I think, didn't heard that
someone except criu used it ever.

> > > Then, I wonder whether reading arg_start|end and env_start|end atomically makes
> > > sense. Just retry reading if arg_start > env_end or env_start > env_end is fine?
> > 
> > Tetsuo, let me re-read this code tomorrow, maybe I miss something obvious.
> > 
> 
> You are not missing my point. What I thought is
> 
> +retry:
> -	down_read(&mm->mmap_sem);
>  	arg_start = mm->arg_start;
>  	arg_end = mm->arg_end;
>  	env_start = mm->env_start;
>  	env_end = mm->env_end;
> -	up_read(&mm->mmap_sem);
>  
> -	BUG_ON(arg_start > arg_end);
> -	BUG_ON(env_start > env_end);
> +	if (unlikely(arg_start > arg_end || env_start > env_end)) {
> +		cond_resched();
> +		goto retry;
> +	}
> 
> for reading these fields.

I fear such contentional cycles are acceptable if only they
are guaranteed to finish eventually. Which doesn't look so
in the code above.

	Cyrill