Re: [dm-devel] [PATCH v2 24/37] libmultipath: simplify get_free_id() assuming total ordering

[Benjamin Marzinski <bmarzins@redhat.com>]

On Mon, Sep 11, 2023 at 06:38:33PM +0200, mwilck@suse.com wrote:
> From: Martin Wilck <mwilck@suse.com>
> 
> If we can assume that the bindings array is totally ordered for every
> prefix, which the previous patch guarantees, the search for a free ID can be
> simplified.
> 
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
> Signed-off-by: Martin Wilck <mwilck@suse.com>
> ---
>  libmultipath/alias.c | 87 ++++++++++----------------------------------
>  1 file changed, 19 insertions(+), 68 deletions(-)
> 
> diff --git a/libmultipath/alias.c b/libmultipath/alias.c
> index af6565b..66e34e3 100644
> --- a/libmultipath/alias.c
> +++ b/libmultipath/alias.c
> @@ -356,83 +356,34 @@ int get_free_id(const Bindings *bindings, const char *prefix, const char *map_ww
>  {
>  	const struct binding *bdg;
>  	int i, id = 1;
> -	int biggest_id = 1;
> -	int smallest_bigger_id = INT_MAX;
>  
>  	vector_foreach_slot(bindings, bdg, i) {
>  		int curr_id = scan_devname(bdg->alias, prefix);
>  
> -		/*
> -		 * Find an unused index - explanation of the algorithm
> -		 *
> -		 * ID: 1 = mpatha, 2 = mpathb, ...
> -		 *
> -		 * We assume the bindings are unsorted. The only constraint
> -		 * is that no ID occurs more than once. IDs that occur in the
> -		 * bindings are called "used".
> -		 *
> -		 * We call the list 1,2,3,..., exactly in this order, the list
> -		 * of "expected" IDs. The variable "id" always holds the next
> -		 * "expected" ID, IOW the last "expected" ID encountered plus 1.
> -		 * Thus all IDs below "id" are known to be used. However, at the
> -		 * end of the loop, the value of "id" isn't necessarily unused.
> -		 *
> -		 * "smallest_bigger_id" is the smallest used ID that was
> -		 * encountered while it was larger than the next "expected" ID
> -		 * at that iteration. Let X be some used ID. If all IDs below X
> -		 * are used and encountered in the right sequence before X, "id"
> -		 * will be > X when the loop ends. Otherwise, X was encountered
> -		 * "out of order", the condition (X > id) holds when X is
> -		 * encountered, and "smallest_bigger_id" will be set to X; i.e.
> -		 * it will be less or equal than X when the loop ends.
> -		 *
> -		 * At the end of the loop, (id < smallest_bigger_id) means that
> -		 * the value of "id" had been encountered neither in order nor
> -		 * out of order, and is thus unused. (id >= smallest_bigger_id)
> -		 * means that "id"'s value is in use. In this case, we play safe
> -		 * and use "biggest_id + 1" as the next value to try.
> -		 *
> -		 * biggest_id is always > smallest_bigger_id, except in the
> -		 * "perfectly ordered" case.
> -		 */
> -		if (curr_id == id) {
> -			if (id < INT_MAX)
> -				id++;
> -			else {
> -				id = -1;
> -				break;
> -			}
> +		if (curr_id == -1)
> +			continue;
> +		if (id > curr_id) {
> +			condlog(0, "%s: ERROR: bindings are not sorted", __func__);
> +			return -1;
>  		}
> -		if (curr_id > biggest_id)
> -			biggest_id = curr_id;
> -
> -		if (curr_id > id && curr_id < smallest_bigger_id)
> -			smallest_bigger_id = curr_id;
> -	}
> -
> -	if (id >= smallest_bigger_id)
> -		id = biggest_id < INT_MAX ? biggest_id + 1 : -1;
> -
> -	if (id > 0) {
> -		while(id_already_taken(id, prefix, map_wwid)) {
> -			if (id == INT_MAX) {
> -				id = -1;
> -				break;
> -			}
> +		while (id < curr_id && id_already_taken(id, prefix, map_wwid))
>  			id++;
> -			if (id == smallest_bigger_id) {
> -				if (biggest_id == INT_MAX) {
> -					id = -1;
> -					break;
> -				}
> -				if (biggest_id >= smallest_bigger_id)
> -					id = biggest_id + 1;
> -			}
> -		}
> +		if (id < curr_id)
> +			return id;
> +		id++;
> +		if (id <= 0)
> +			break;
>  	}
>  
> -	if (id < 0)
> +	for (; id > 0; id++) {
> +		if (!id_already_taken(id, prefix, map_wwid))
> +			break;
> +	}
> +
> +	if (id <= 0) {
> +		id = -1;
>  		condlog(0, "no more available user_friendly_names");
> +	}
>  	return id;
>  }
>  
> -- 
> 2.42.0
--
dm-devel mailing list
dm-devel@redhat.com
https://listman.redhat.com/mailman/listinfo/dm-devel