[PATCH v2 net-next 1/3] net: make softnet_data.defer_count an atomic
From: Eric Dumazet <edumazet@google.com>
Date: 2025-09-28 08:49:38
Subsystem:
networking drivers, networking [general], the rest · Maintainers:
Andrew Lunn, "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Linus Torvalds
This is preparation work to remove the softnet_data.defer_lock, as it is contended on hosts with large number of cores. Signed-off-by: Eric Dumazet <edumazet@google.com> Reviewed-by: Jason Xing <redacted> Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com> --- include/linux/netdevice.h | 2 +- net/core/dev.c | 2 +- net/core/skbuff.c | 6 ++---- 3 files changed, 4 insertions(+), 6 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 1b85454116f666ced61a1450d3f899940f499c05..27e3fa69253f694b98d32b6138cf491da5a8b824 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h@@ -3538,7 +3538,7 @@ struct softnet_data { /* Another possibly contended cache line */ spinlock_t defer_lock ____cacheline_aligned_in_smp; - int defer_count; + atomic_t defer_count; int defer_ipi_scheduled; struct sk_buff *defer_list; call_single_data_t defer_csd;
diff --git a/net/core/dev.c b/net/core/dev.c
index 8b54fdf0289ab223fc37d27a078536db37646b55..8566678d83444e8aacbfea4842878279cf28516f 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c@@ -6726,7 +6726,7 @@ static void skb_defer_free_flush(struct softnet_data *sd) spin_lock(&sd->defer_lock); skb = sd->defer_list; sd->defer_list = NULL; - sd->defer_count = 0; + atomic_set(&sd->defer_count, 0); spin_unlock(&sd->defer_lock); while (skb != NULL) {
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index daaf6da43cc9e199389c3afcd6621c177d247884..f91571f51c69ecf8c2fffed5f3a3cd33fd95828b 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c@@ -7201,14 +7201,12 @@ nodefer: kfree_skb_napi_cache(skb); sd = &per_cpu(softnet_data, cpu); defer_max = READ_ONCE(net_hotdata.sysctl_skb_defer_max); - if (READ_ONCE(sd->defer_count) >= defer_max) + if (atomic_read(&sd->defer_count) >= defer_max) goto nodefer; spin_lock_bh(&sd->defer_lock); /* Send an IPI every time queue reaches half capacity. */ - kick = sd->defer_count == (defer_max >> 1); - /* Paired with the READ_ONCE() few lines above */ - WRITE_ONCE(sd->defer_count, sd->defer_count + 1); + kick = (atomic_inc_return(&sd->defer_count) - 1) == (defer_max >> 1); skb->next = sd->defer_list; /* Paired with READ_ONCE() in skb_defer_free_flush() */
--
2.51.0.536.g15c5d4f767-goog