| Message ID | 20240221172032.78737-4-bigeasy@linutronix.de (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | net: Provide SMP threads for backlog NAPI | expand |
On Wed, 21 Feb 2024 18:00:13 +0100 Sebastian Andrzej Siewior wrote: > + if (use_backlog_threads()) { > + rps_lock_irqsave(sd, &flags); > + > + if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) > + napi_schedule_rps(sd); Why are you calling napi_schedule_rps() here? Just do __napi_schedule_irqoff(&sd->backlog); Then you can move the special case inside napi_schedule_rps() into the if (sd != mysd) block. > + rps_unlock_irq_restore(sd, &flags); Also not sure if the lock helpers should still be called RPS since they also protect state on non-RPS configs now.
On 2024-02-23 18:02:57 [-0800], Jakub Kicinski wrote: > On Wed, 21 Feb 2024 18:00:13 +0100 Sebastian Andrzej Siewior wrote: > > + if (use_backlog_threads()) { > > + rps_lock_irqsave(sd, &flags); > > + > > + if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) > > + napi_schedule_rps(sd); > > Why are you calling napi_schedule_rps() here? > Just do __napi_schedule_irqoff(&sd->backlog); Looking at it, __napi_schedule_irqoff() is enough here given that we are already in the use_backlog_threads() case. > Then you can move the special case inside napi_schedule_rps() > into the if (sd != mysd) block. Okay. > > + rps_unlock_irq_restore(sd, &flags); > > Also not sure if the lock helpers should still be called RPS since they > also protect state on non-RPS configs now. They protect the list in input_pkt_queue and the NAPI state. It is just in the !RPS case it is always CPU-local and the lock is avoided (while interrupts are still disabled/ enabled). What about input_queue_lock_irq_save() input_queue_lock_irq_disable() input_queue_lock_irq_restore() input_queue_lock_irq_enable() ? Sebastian
On Mon, 26 Feb 2024 12:59:22 +0100 Sebastian Andrzej Siewior wrote: > They protect the list in input_pkt_queue and the NAPI state. It is just > in the !RPS case it is always CPU-local and the lock is avoided (while > interrupts are still disabled/ enabled). > > What about > input_queue_lock_irq_save() > input_queue_lock_irq_disable() > input_queue_lock_irq_restore() > input_queue_lock_irq_enable() SGTM. Maybe I'd risk calling it backlog_lock_* but not sure others would agree.
On 2024-02-26 17:22:33 [-0800], Jakub Kicinski wrote: > On Mon, 26 Feb 2024 12:59:22 +0100 Sebastian Andrzej Siewior wrote: > > They protect the list in input_pkt_queue and the NAPI state. It is just > > in the !RPS case it is always CPU-local and the lock is avoided (while > > interrupts are still disabled/ enabled). > > > > What about > > input_queue_lock_irq_save() > > input_queue_lock_irq_disable() > > input_queue_lock_irq_restore() > > input_queue_lock_irq_enable() > > SGTM. Maybe I'd risk calling it backlog_lock_* but not sure others > would agree. Let me repost it with backlog_lock_* today evening (UTC) unless someone objects. Sebastian
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index f07c8374f29cb..0a7390f011be0 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -3368,6 +3368,7 @@ static inline void dev_xmit_recursion_dec(void) __this_cpu_dec(softnet_data.xmit.recursion); } +void kick_defer_list_purge(unsigned int cpu); void __netif_schedule(struct Qdisc *q); void netif_schedule_queue(struct netdev_queue *txq); diff --git a/net/core/dev.c b/net/core/dev.c index 6aa3547c03a4f..8b228861f29af 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -225,7 +225,7 @@ static bool use_backlog_threads(void) static inline void rps_lock_irqsave(struct softnet_data *sd, unsigned long *flags) { - if (IS_ENABLED(CONFIG_RPS)) + if (IS_ENABLED(CONFIG_RPS) || use_backlog_threads()) spin_lock_irqsave(&sd->input_pkt_queue.lock, *flags); else if (!IS_ENABLED(CONFIG_PREEMPT_RT)) local_irq_save(*flags); @@ -233,7 +233,7 @@ static inline void rps_lock_irqsave(struct softnet_data *sd, static inline void rps_lock_irq_disable(struct softnet_data *sd) { - if (IS_ENABLED(CONFIG_RPS)) + if (IS_ENABLED(CONFIG_RPS) || use_backlog_threads()) spin_lock_irq(&sd->input_pkt_queue.lock); else if (!IS_ENABLED(CONFIG_PREEMPT_RT)) local_irq_disable(); @@ -242,7 +242,7 @@ static inline void rps_lock_irq_disable(struct softnet_data *sd) static inline void rps_unlock_irq_restore(struct softnet_data *sd, unsigned long *flags) { - if (IS_ENABLED(CONFIG_RPS)) + if (IS_ENABLED(CONFIG_RPS) || use_backlog_threads()) spin_unlock_irqrestore(&sd->input_pkt_queue.lock, *flags); else if (!IS_ENABLED(CONFIG_PREEMPT_RT)) local_irq_restore(*flags); @@ -250,7 +250,7 @@ static inline void rps_unlock_irq_restore(struct softnet_data *sd, static inline void rps_unlock_irq_enable(struct softnet_data *sd) { - if (IS_ENABLED(CONFIG_RPS)) + if (IS_ENABLED(CONFIG_RPS) || use_backlog_threads()) spin_unlock_irq(&sd->input_pkt_queue.lock); else if (!IS_ENABLED(CONFIG_PREEMPT_RT)) local_irq_enable(); @@ -4735,6 +4735,26 @@ static void napi_schedule_rps(struct softnet_data *sd) __napi_schedule_irqoff(&mysd->backlog); } +void kick_defer_list_purge(unsigned int cpu) +{ + struct softnet_data *sd; + unsigned long flags; + + sd = &per_cpu(softnet_data, cpu); + + if (use_backlog_threads()) { + rps_lock_irqsave(sd, &flags); + + if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) + napi_schedule_rps(sd); + + rps_unlock_irq_restore(sd, &flags); + + } else if (!cmpxchg(&sd->defer_ipi_scheduled, 0, 1)) { + smp_call_function_single_async(cpu, &sd->defer_csd); + } +} + #ifdef CONFIG_NET_FLOW_LIMIT int netdev_flow_limit_table_len __read_mostly = (1 << 12); #endif diff --git a/net/core/skbuff.c b/net/core/skbuff.c index b9de3ee65ae64..427387ffd3c8a 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -7034,8 +7034,8 @@ nodefer: __kfree_skb(skb); /* Make sure to trigger NET_RX_SOFTIRQ on the remote CPU * if we are unlucky enough (this seems very unlikely). */ - if (unlikely(kick) && !cmpxchg(&sd->defer_ipi_scheduled, 0, 1)) - smp_call_function_single_async(cpu, &sd->defer_csd); + if (unlikely(kick)) + kick_defer_list_purge(cpu); } static void skb_splice_csum_page(struct sk_buff *skb, struct page *page,
The defer_list is a per-CPU list which is used to free skbs outside of the socket lock and on the CPU on which they have been allocated. The list is processed during NAPI callbacks so ideally the list is cleaned up. Should the amount of skbs on the list exceed a certain water mark then the softirq is triggered remotely on the target CPU by invoking a remote function call. The raise of the softirqs via a remote function call leads to waking the ksoftirqd on PREEMPT_RT which is undesired. The backlog-NAPI threads already provide the infrastructure which can be utilized to perform the cleanup of the defer_list. The NAPI state is updated with the input_pkt_queue.lock acquired. It order not to break the state, it is needed to also wake the backlog-NAPI thread with the lock held. This requires to acquire the use the lock in rps_lock_irq*() if the backlog-NAPI threads are used even with RPS disabled. Move the logic of remotely starting softirqs to clean up the defer_list into kick_defer_list_purge(). Make sure a lock is held in rps_lock_irq*() if backlog-NAPI threads are used. Schedule backlog-NAPI for defer_list cleanup if backlog-NAPI is available. Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> --- include/linux/netdevice.h | 1 + net/core/dev.c | 28 ++++++++++++++++++++++++---- net/core/skbuff.c | 4 ++-- 3 files changed, 27 insertions(+), 6 deletions(-)