Re: [PATCH v3] Fix rmmod/read/write races in /proc entries

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 8 Feb 2007 16:20:12 +0300 Alexey Dobriyan <adobriyan@openvz.org> wrote:

> +again:
>  	spin_lock(&proc_subdir_lock);
>  	for (p = &parent->subdir; *p; p=&(*p)->next ) {
>  		if (!proc_match(len, fn, *p))
>  			continue;
>  		de = *p;
> +
> +		/*
> +		 * Stop accepting new readers/writers. If you're dynamically
> +		 * allocating ->proc_fops, save a pointer somewhere.
> +		 */
> +		spin_lock(&de->pde_unload_lock);
> +		de->proc_fops = NULL;
> +		/* Wait until all readers/writers are done. */
> +		if (de->pde_users > 0) {
> +			spin_unlock(&de->pde_unload_lock);
> +			spin_unlock(&proc_subdir_lock);
> +			schedule();
> +			goto again;
> +		}
> +		spin_unlock(&de->pde_unload_lock);

aergh.  This will devolve into busy-wait-until-we-expire-our-timeslice.

Would be nicer to do this with a wait_for_completion().

I guess it doesn't happen very often - if another process happens to
be in the middle or a read or write syscall to that /proc file.