Re: [PATCH RFC v2 09/20] slab: remove cpu (partial) slabs usage from allocation paths

[Vlastimil Babka <vbabka@suse.cz>]

On 1/14/26 07:07, Hao Li wrote:
>> @@ -4836,68 +4558,31 @@ static void *___slab_alloc(struct kmem_cache *s, gfp_t gfpflags, int node,
>>  	if (IS_ENABLED(CONFIG_SLUB_TINY) || kmem_cache_debug(s)) {
>>  		freelist = alloc_single_from_new_slab(s, slab, orig_size, gfpflags);
>>  
>> -		if (unlikely(!freelist)) {
>> -			/* This could cause an endless loop. Fail instead. */
>> -			if (!allow_spin)
>> -				return NULL;
>> -			goto new_objects;
>> +		if (likely(freelist)) {
>> +			goto success;
>>  		}
>> +	} else {
>> +		alloc_from_new_slab(s, slab, &freelist, 1, allow_spin);
> 
> IIUC, when CONFIG_SLUB_DEBUG is enabled, each successful new_slab() call
> should have a matching inc_slabs_node(), since __kmem_cache_shutdown()
> rely on the accounting done by inc_slabs_node(). Here
> alloc_single_from_new_slab() does call inc_slabs_node(), but
> alloc_from_new_slab() doesn't. Could this mismatch cause any issues?

Great spot, thanks a lot! Yes we should do inc_slabs_node() here.

>>  
>> -		if (s->flags & SLAB_STORE_USER)
>> -			set_track(s, freelist, TRACK_ALLOC, addr,
>> -				  gfpflags & ~(__GFP_DIRECT_RECLAIM));
>> -
>> -		return freelist;
>> -	}
>> -
>> -	/*
>> -	 * No other reference to the slab yet so we can
>> -	 * muck around with it freely without cmpxchg
>> -	 */
>> -	freelist = slab->freelist;
>> -	slab->freelist = NULL;
>> -	slab->inuse = slab->objects;
>> -	slab->frozen = 1;
>> -