Re: [PATCH RFC 6/8] mm/slab: allow changing sheaf_capacity at runtime

["Harry Yoo (Oracle)" <harry@kernel.org>]

On 5/17/26 5:30 PM, Yeoreum Yun wrote:
> Hi Harry,

Hi Yeoreum,

>> @@ -9111,7 +9309,39 @@ static ssize_t sheaf_capacity_show(struct kmem_cache *s, char *buf)
>>   {
>>   	return sysfs_emit(buf, "%hu\n", s->sheaf_capacity);
>>   }
>> -SLAB_ATTR_RO(sheaf_capacity);
>> +static ssize_t sheaf_capacity_store(struct kmem_cache *s,
>> +				    const char *buf, size_t length)
>> +{
>> +	unsigned short capacity;
>> +	int err;
>> +
>> +	err = kstrtou16(buf, 10, &capacity);
>> +	if (err)
>> +		return err;
>> +
>> +	if (!cache_supports_sheaves(s))
>> +		return -EOPNOTSUPP;
>> +
>> +	cpus_read_lock();
>> +	mutex_lock(&slab_mutex);
> 
> This patchset looks good to me.

Thanks :)

> However, I’m not sure why we need slab_mutex here, as using only
> flush_lock seems sufficient

The main reason why I used slab_mutex was because both disabling sheaves
and shrinking the cache acquires flush_lock.

> Some refactoring might be required, but it seems better to remove
> the acquisition of slab_mutex here.

With some refactoring, we could teach those functions that the caller
already acquired the lock and use flush_lock instead of slab_mutex.

However, it won't improve things much... both are global locks, and
I doubt that it would cause visible latency issues when creating new 
slab caches.

I want to focus on refining the capacity change part for now.

-- 
Cheers,
Harry / Hyeonggon