netfilter: nft_set_pipapo: Use nested-BH locking for nft_pipapo_scratch

nft_pipapo_scratch is a per-CPU variable and relies on disabled BH for
its locking. Without per-CPU locking in local_bh_disable() on PREEMPT_RT
this data structure requires explicit locking.

Add a local_lock_t to the data structure and use local_lock_nested_bh() for
locking. This change adds only lockdep coverage and does not alter the
functional behaviour for !PREEMPT_RT.

Signed-off-by: Sebastian Andrzej Siewior <[e-mail]>
Signed-off-by: Florian Westphal <[e-mail]>

net/netfilter/nft_set_pipapo.c

···
 	scratch = *raw_cpu_ptr(m->scratch);
 	if (unlikely(!scratch))
 		goto out;
+	__local_lock_nested_bh(&scratch->bh_lock);
 
 	map_index = scratch->map_index;
 
···
 				  last);
 		if (b < 0) {
 			scratch->map_index = map_index;
+			__local_unlock_nested_bh(&scratch->bh_lock);
 			local_bh_enable();
 
 			return NULL;
···
 			 * *next* bitmap (not initial) for the next packet.
 			 */
 			scratch->map_index = map_index;
+			__local_unlock_nested_bh(&scratch->bh_lock);
 			local_bh_enable();
 			return e;
 		}
···
 		data += NFT_PIPAPO_GROUPS_PADDING(f);
 	}
 
+	__local_unlock_nested_bh(&scratch->bh_lock);
 out:
 	local_bh_enable();
 	return NULL;
···
 		}
 
 		pipapo_free_scratch(clone, i);
+		local_lock_init(&scratch->bh_lock);
 		*per_cpu_ptr(clone->scratch, i) = scratch;
 	}
 

net/netfilter/nft_set_pipapo.h

···
 
 /**
  * struct nft_pipapo_scratch - percpu data used for lookup and matching
+ * @bh_lock:    PREEMPT_RT local spinlock
  * @map_index:	Current working bitmap index, toggled between field matches
  * @__map:	store partial matching results during lookup
  */
 struct nft_pipapo_scratch {
+	local_lock_t bh_lock;
 	u8 map_index;
 	unsigned long __map[];
 };

net/netfilter/nft_set_pipapo_avx2.c

···
 	if (unlikely(!scratch))
 		return NULL;
 
+	__local_lock_nested_bh(&scratch->bh_lock);
 	map_index = scratch->map_index;
 	map = NFT_PIPAPO_LT_ALIGN(&scratch->__map[0]);
 	res  = map + (map_index ? m->bsize_max : 0);
···
 		if (ret < 0) {
 			scratch->map_index = map_index;
 			kernel_fpu_end();
+			__local_unlock_nested_bh(&scratch->bh_lock);
 			return NULL;
 		}
 
···
 
 			scratch->map_index = map_index;
 			kernel_fpu_end();
+			__local_unlock_nested_bh(&scratch->bh_lock);
 			return e;
 		}
 
···
 	}
 
 	kernel_fpu_end();
+	__local_unlock_nested_bh(&scratch->bh_lock);
 	return NULL;
 }