Re: [PATCH v2] lib/btree.c: optimise the code by previously getpos function

[Andrew Morton]

On Thu, 11 May 2017 17:42:21 +0800 Leno Hou <lenohou@xxxxxxxxx> wrote:

> This patch optimized the code by previously getpos function call.
> Therefore, It's takes the convenience to understand logic of code.

I would rewrite this changelog to read

: Rework the getpos() helper function and use it to remove various
: open-coded implemetnations of its funtionality.


> ...
>
> @@ -466,7 +458,7 @@ static int btree_insert_level(struct btree_head *head, struct btree_geo *geo,
>  	/* two identical keys are not allowed */
>  	BUG_ON(pos < fill && keycmp(geo, node, pos, key) == 0);
>  
> -	if (fill == geo->no_pairs) {
> +	if (fill < 0) {
>  		/* need to split node */
>  		unsigned long *new;

The code here is a bit awkward.

: retry:
: 	node = find_level(head, geo, key, level);
: 	pos = getpos(geo, node, key);
: 	fill = getfill(geo, node, pos);
: 	/* two identical keys are not allowed */
: 	BUG_ON(pos < fill && keycmp(geo, node, pos, key) == 0);
: 
: 	if (fill < 0) {

If getpos() returns -ENOENT then we're passing -ENOENT into getfill(). 
That might happen to work OK (or it might go BUG) but it's ugly and
unobvious.

There's a similar issue in rebalance() and in btree_remove_level():
failed to update existing getpos() callers for the new getpos() return
value semantics.