On 05/09/25 03:08PM, Ira Weiny wrote:
Neeraj Kumar wrote:
On 19/08/25 01:56PM, Ira Weiny wrote:
>Neeraj Kumar wrote:
>> nd_mapping->labels maintains the list of labels present into LSA.
>> init_labels() prepares this list while adding new label into LSA
>> and updates nd_mapping->labels accordingly. During cxl region
>> creation nd_mapping->labels list and LSA was updated with one
>> region label. Therefore during new namespace label creation
>> pre-include the previously created region label, so increase
>> num_labels count by 1.
>
>Why does the count of the labels in the list not work?
>
>static int init_labels(struct nd_mapping *nd_mapping, int num_labels)
>{
> int i, old_num_labels = 0;
>...
> mutex_lock(&nd_mapping->lock);
> list_for_each_entry(label_ent, &nd_mapping->labels, list)
> old_num_labels++;
> mutex_unlock(&nd_mapping->lock);
>...
>
Hi Ira,
init_labels() allocates new label based on comparison with existing
count of the labels in the list and passed num_labels. If num_labels
is greater than count of the labels in the list then new label is
allocated and stored in list for later usage
I think I'm following better but shouldn't this hunk be included in the
code which creates the region label in the list?
Yes we can include this hunk in patch 5 where we are updating region
label. I will drop this commit and include it in patch 5.
I'm concerned that this '+ 1' out of the blue and will be confusing in the
future. Why can't count be kept up to date when the region label was
created and added?
Yes this '+1' is hardcoded and it creating confusion. I will fix this
with available region labels in the list.
What code (patch) added this region label?
Ira
Patch 5 add this region label using nd_pmem_region_label_update()
Regards,
Neeraj
