net: fully inline backlog_unlock_irq_restore()
Subsystems (2)
PATCH 1/1 net: fully inline backlog_unlock_irq_restore()
Some arches (like x86) do not inline spin_unlock_irqrestore().
backlog_unlock_irq_restore() is in RPS/RFS critical path, we prefer using spin_unlock() + local_irq_restore() for optimal performance.
Also change backlog_unlock_irq_restore() second argument to avoid a pointless dereference.
No difference in net/core/dev.o code size.
--- net/core/dev.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-)
net/core/dev.c +6 -7
@@ -246,12 +246,11 @@ static inline void backlog_lock_irq_disable(struct softnet_data *sd) }static inline void backlog_unlock_irq_restore (struct softnet_data *sd, -unsigned long *flags) +unsigned long flags) {if (IS_ENABLED (CONFIG_RPS ) ||use_backlog_threads ()) -spin_unlock_irqrestore (&sd->input_pkt_queue.lock, *flags); -else -local_irq_restore (*flags); +spin_unlock (&sd->input_pkt_queue.lock); +local_irq_restore (flags); }static inline void backlog_unlock_irq_enable (struct softnet_data *sd)
@@ -5247,7 +5246,7 @@ void kick_defer_list_purge(unsigned int cpu)if (!__test_and_set_bit (NAPI_STATE_SCHED , &sd->backlog.state))__napi_schedule_irqoff (&sd->backlog); -backlog_unlock_irq_restore (sd, &flags); +backlog_unlock_irq_restore (sd, flags); }else if (!cmpxchg (&sd->defer_ipi_scheduled,0 ,1 )) {smp_call_function_single_async (cpu, &sd->defer_csd);
@@ -5334,14 +5333,14 @@ static int enqueue_to_backlog(struct sk_buff *skb, int cpu, }__skb_queue_tail (&sd->input_pkt_queue, skb); tail =rps_input_queue_tail_incr (sd); -backlog_unlock_irq_restore (sd, &flags); +backlog_unlock_irq_restore (sd, flags);/* save the tail outside of the critical section */ rps_input_queue_tail_save (qtail, tail);return NET_RX_SUCCESS ; } -backlog_unlock_irq_restore (sd, &flags); +backlog_unlock_irq_restore (sd, flags); cpu_backlog_drop: reason =SKB_DROP_REASON_CPU_BACKLOG ;
Signature · 1 line
2.52.0.351.gbe84eed79e-goog
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 10 lines · On Mon, 5 Jan 2026 16:30:54 +0000 you wrote:
Some arches (like x86) do not inline
spin_unlock_irqrestore().
backlog_unlock_irq_restore()is in RPS/RFS critical path, we prefer usingspin_unlock()+local_irq_restore()for optimal performance.Also change
backlog_unlock_irq_restore()second argument to avoid a pointless dereference.[...]
Here is the summary with links:
- [net-next] net: fully inline backlog_unlock_irq_restore()
https://git.kernel.org/netdev/net-next/c/27a01c1969a5
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html