From 639c96315aba15fd372b951c31442b622f4eb3df Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 11 Oct 2017 17:31:13 +0800 Subject: [PATCH] net: better skb->sender_cpu and skb->napi_id cohabitation Orabug: 26953388 Orabug: 26591689 skb->sender_cpu and skb->napi_id share a common storage, and we had various bugs about this. We had to call skb_sender_cpu_clear() in some places to not leave a prior skb->napi_id and fool netdev_pick_tx() As suggested by Alexei, we could split the space so that these errors can not happen. 0 value being reserved as the common (not initialized) value, let's reserve [1 .. NR_CPUS] range for valid sender_cpu, and [NR_CPUS+1 .. ~0U] for valid napi_id. This will allow proper busy polling support over tunnels. Signed-off-by: Eric Dumazet Suggested-by: Alexei Starovoitov Acked-by: Alexei Starovoitov Signed-off-by: David S. Miller Signed-off-by: Ethan Zhao Reviewed-by: Jack Vogel --- include/linux/skbuff.h | 3 --- net/core/dev.c | 29 +++++++++++++---------------- net/core/flow_dissector.c | 4 +++- 3 files changed, 16 insertions(+), 20 deletions(-) diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index be7102b9cc16..4921d9aee585 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -956,9 +956,6 @@ static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) static inline void skb_sender_cpu_clear(struct sk_buff *skb) { -#ifdef CONFIG_XPS - skb->sender_cpu = 0; -#endif } #ifdef NET_SKBUFF_DATA_USES_OFFSET diff --git a/net/core/dev.c b/net/core/dev.c index 18bfb1659fe8..1346af1b16ba 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -180,7 +180,7 @@ EXPORT_SYMBOL(dev_base_lock); /* protects napi_hash addition/deletion and napi_gen_id */ static DEFINE_SPINLOCK(napi_hash_lock); -static unsigned int napi_gen_id; +static unsigned int napi_gen_id = NR_CPUS; static DEFINE_HASHTABLE(napi_hash, 8); static seqcount_t devnet_rename_seq; @@ -4519,25 +4519,22 @@ EXPORT_SYMBOL_GPL(napi_by_id); void napi_hash_add(struct napi_struct *napi) { - if (!test_and_set_bit(NAPI_STATE_HASHED, &napi->state)) { + if (test_and_set_bit(NAPI_STATE_HASHED, &napi->state)) + return; - spin_lock(&napi_hash_lock); + spin_lock(&napi_hash_lock); - /* 0 is not a valid id, we also skip an id that is taken - * we expect both events to be extremely rare - */ - napi->napi_id = 0; - while (!napi->napi_id) { - napi->napi_id = ++napi_gen_id; - if (napi_by_id(napi->napi_id)) - napi->napi_id = 0; - } + /* 0..NR_CPUS+1 range is reserved for sender_cpu use */ + do { + if (unlikely(++napi_gen_id < NR_CPUS + 1)) + napi_gen_id = NR_CPUS + 1; + } while (napi_by_id(napi_gen_id)); + napi->napi_id = napi_gen_id; - hlist_add_head_rcu(&napi->napi_hash_node, - &napi_hash[napi->napi_id % HASH_SIZE(napi_hash)]); + hlist_add_head_rcu(&napi->napi_hash_node, + &napi_hash[napi->napi_id % HASH_SIZE(napi_hash)]); - spin_unlock(&napi_hash_lock); - } + spin_unlock(&napi_hash_lock); } EXPORT_SYMBOL_GPL(napi_hash_add); diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c index 3556791fdc6e..a262b3780fc4 100644 --- a/net/core/flow_dissector.c +++ b/net/core/flow_dissector.c @@ -476,7 +476,9 @@ struct netdev_queue *netdev_pick_tx(struct net_device *dev, int queue_index = 0; #ifdef CONFIG_XPS - if (skb->sender_cpu == 0) + u32 sender_cpu = skb->sender_cpu - 1; + + if (sender_cpu >= (u32)NR_CPUS) skb->sender_cpu = raw_smp_processor_id() + 1; #endif -- 2.50.1