[PATCH nf-next v2 3/3] netfilter: nft_set_pipapo: Use nested-BH locking for nft_pipapo_scratch

[Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>]

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 <bigeasy@xxxxxxxxxxxxx>
---
 net/netfilter/nft_set_pipapo.c      |  5 +++++
 net/netfilter/nft_set_pipapo.h      |  1 +
 net/netfilter/nft_set_pipapo_avx2.c | 15 +++++++--------
 3 files changed, 13 insertions(+), 8 deletions(-)

diff --git a/net/netfilter/nft_set_pipapo.c b/net/netfilter/nft_set_pipapo.c
index 364b9abcce13b..5220a050c5025 100644
--- a/net/netfilter/nft_set_pipapo.c
+++ b/net/netfilter/nft_set_pipapo.c
@@ -429,6 +429,7 @@ static struct nft_pipapo_elem *pipapo_get(const struct nft_pipapo_match *m,
 	scratch = *raw_cpu_ptr(m->scratch);
 	if (unlikely(!scratch))
 		goto out;
+	__local_lock_nested_bh(&scratch->bh_lock);
 
 	map_index = scratch->map_index;
 
@@ -465,6 +466,7 @@ static struct nft_pipapo_elem *pipapo_get(const struct nft_pipapo_match *m,
 				  last);
 		if (b < 0) {
 			scratch->map_index = map_index;
+			__local_unlock_nested_bh(&scratch->bh_lock);
 			local_bh_enable();
 
 			return NULL;
@@ -484,6 +486,7 @@ static struct nft_pipapo_elem *pipapo_get(const struct nft_pipapo_match *m,
 			 * *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;
 		}
@@ -498,6 +501,7 @@ static struct nft_pipapo_elem *pipapo_get(const struct nft_pipapo_match *m,
 		data += NFT_PIPAPO_GROUPS_PADDING(f);
 	}
 
+	__local_unlock_nested_bh(&scratch->bh_lock);
 out:
 	local_bh_enable();
 	return NULL;
@@ -1180,6 +1184,7 @@ static int pipapo_realloc_scratch(struct nft_pipapo_match *clone,
 		}
 
 		pipapo_free_scratch(clone, i);
+		local_lock_init(&scratch->bh_lock);
 		*per_cpu_ptr(clone->scratch, i) = scratch;
 	}
 
diff --git a/net/netfilter/nft_set_pipapo.h b/net/netfilter/nft_set_pipapo.h
index 3655aa41fa949..4d9addea854c4 100644
--- a/net/netfilter/nft_set_pipapo.h
+++ b/net/netfilter/nft_set_pipapo.h
@@ -129,6 +129,7 @@ struct nft_pipapo_field {
  * @__map:	store partial matching results during lookup
  */
 struct nft_pipapo_scratch {
+	local_lock_t bh_lock;
 	u8 map_index;
 	unsigned long __map[];
 };
diff --git a/net/netfilter/nft_set_pipapo_avx2.c b/net/netfilter/nft_set_pipapo_avx2.c
index e907e48b474b6..4515aa0a49984 100644
--- a/net/netfilter/nft_set_pipapo_avx2.c
+++ b/net/netfilter/nft_set_pipapo_avx2.c
@@ -1170,20 +1170,18 @@ nft_pipapo_avx2_lookup(const struct net *net, const struct nft_set *set,
 	}
 
 	m = rcu_dereference(priv->match);
-
+	scratch = *raw_cpu_ptr(m->scratch);
+	if (unlikely(!scratch)) {
+		local_bh_enable();
+		return false;
+	}
+	__local_lock_nested_bh(&scratch->bh_lock);
 	/* Note that we don't need a valid MXCSR state for any of the
 	 * operations we use here, so pass 0 as mask and spare a LDMXCSR
 	 * instruction.
 	 */
 	kernel_fpu_begin_mask(0);
 
-	scratch = *raw_cpu_ptr(m->scratch);
-	if (unlikely(!scratch)) {
-		kernel_fpu_end();
-		local_bh_enable();
-		return NULL;
-	}
-
 	map_index = scratch->map_index;
 	map = NFT_PIPAPO_LT_ALIGN(&scratch->__map[0]);
 	res  = map + (map_index ? m->bsize_max : 0);
@@ -1262,6 +1260,7 @@ nft_pipapo_avx2_lookup(const struct net *net, const struct nft_set *set,
 	if (i % 2)
 		scratch->map_index = !map_index;
 	kernel_fpu_end();
+	__local_unlock_nested_bh(&scratch->bh_lock);
 	local_bh_enable();
 
 	return ext;
-- 
2.50.1