Re: [PATCH v4 04/17] maple_tree: introduce mas_wr_store_type()

[Sid Kumar <sidhartha.kumar@oracle.com>]

On 9/24/24 9:04 PM, Wei Yang wrote:
> On Wed, Aug 14, 2024 at 12:19:31PM -0400, Sidhartha Kumar wrote:
>
> Sorry for a late reply, I just see this change.
>
>> +
>> +/*
>> + * mas_wr_store_type() - Set the store type for a given
>> + * store operation.
>> + * @wr_mas: The maple write state
>> + */
>> +static inline void mas_wr_store_type(struct ma_wr_state *wr_mas)
>> +{
>> +	struct ma_state *mas = wr_mas->mas;
>> +	unsigned char new_end;
>> +
>> +	if (unlikely(mas_is_none(mas) || mas_is_ptr(mas))) {
>> +		mas->store_type = wr_store_root;
>> +		return;
>> +	}
>> +
>> +	if (unlikely(!mas_wr_walk(wr_mas))) {
>> +		mas->store_type = wr_spanning_store;
>> +		return;
>> +	}
>> +
>> +	/* At this point, we are at the leaf node that needs to be altered. */
>> +	mas_wr_end_piv(wr_mas);
>> +	if (!wr_mas->entry)
>> +		mas_wr_extend_null(wr_mas);
>> +
>> +	new_end = mas_wr_new_end(wr_mas);
>> +	if ((wr_mas->r_min == mas->index) && (wr_mas->r_max == mas->last)) {
>> +		mas->store_type = wr_exact_fit;
>> +		return;
>> +	}
>> +
>> +	if (unlikely(!mas->index && mas->last == ULONG_MAX)) {
>> +		mas->store_type = wr_new_root;
>> +		return;
>> +	}
>> +
>> +	/* Potential spanning rebalance collapsing a node */
>> +	if (new_end < mt_min_slots[wr_mas->type]) {
>> +		if (!mte_is_root(mas->node)) {
>> +			mas->store_type = wr_rebalance;
>> +			return;
>> +		}
>> +		mas->store_type = wr_node_store;
>> +		return;
>> +	}
> After this check, we are sure new_end >= mt_min_slots[wr_mas->type].
>
>> +
>> +	if (new_end >= mt_slots[wr_mas->type]) {
>> +		mas->store_type = wr_split_store;
>> +		return;
>> +	}
>> +
>> +	if (!mt_in_rcu(mas->tree) && (mas->offset == mas->end)) {
>> +		mas->store_type = wr_append;
>> +		return;
>> +	}
>> +
>> +	if ((new_end == mas->end) && (!mt_in_rcu(mas->tree) ||
>> +		(wr_mas->offset_end - mas->offset == 1))) {
>> +		mas->store_type = wr_slot_store;
>> +		return;
>> +	}
>> +
>> +	if (mte_is_root(mas->node) || (new_end >= mt_min_slots[wr_mas->type]) ||
>> +		(mas->mas_flags & MA_STATE_BULK)) {
> The check (new_end >= mt_min_slots[wr_mas->type]) here seems always be true.
>
> So the if here seems not necessary. Do I miss something?

It is true that at this point new_end >= mt_min_slots[wr_mas->type] must 
be true but if we remove that check we won't catch this wr_node_store 
case if !mte_is_root() and !(mas->mas_flags & MA_STATE_BULK).

We could change the default store type to be wr_node_store and get rid 
of that whole if statement entirely.

This diff passes the tests:

diff --git a/lib/maple_tree.c b/lib/maple_tree.c index 
4f34e50c92b5..2ae0c4da9d74 100644 --- a/lib/maple_tree.c +++ 
b/lib/maple_tree.c @@ -4242,14 +4242,7 @@ static inline void 
mas_wr_store_type(struct ma_wr_state *wr_mas) return; } - if 
(mte_is_root(mas->node) || (new_end >= mt_min_slots[wr_mas->type]) || - 
(mas->mas_flags & MA_STATE_BULK)) { - mas->store_type = wr_node_store; - 
return; - } - - mas->store_type = wr_invalid; - MAS_WARN_ON(mas, 1); + 
mas->store_type = wr_node_store; }

do you think this makes sense?

Thanks,

Sid

>> +		mas->store_type = wr_node_store;
>> +		return;
>> +	}
>> +
>> +	mas->store_type = wr_invalid;
>> +	MAS_WARN_ON(mas, 1);
>> +}
>> +