A common task for most drivers is to remember the user's CPU affinity to its IRQs. On each netdev reset, the driver must then re-assign the user's setting to the IRQs.
Add CPU affinity mask to napi->config. To delegate the CPU affinity management to the core, drivers must: 1 - add a persistent napi config: netif_napi_add_config() 2 - bind an IRQ to the napi instance: netif_napi_set_irq() the core will then make sure to use re-assign affinity to the napi's IRQ. The default mask set to all IRQs is all online CPUs. Suggested-by: Jakub Kicinski <[email protected]> Signed-off-by: Ahmed Zaki <[email protected]> --- include/linux/netdevice.h | 6 ++++++ net/core/dev.c | 31 ++++++++++++++++++++++++++++++- 2 files changed, 36 insertions(+), 1 deletion(-) diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index b598de335d26..9bf91c3aca8d 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -350,9 +350,14 @@ struct napi_config { u64 gro_flush_timeout; u64 irq_suspend_timeout; u32 defer_hard_irqs; + cpumask_t affinity_mask; unsigned int napi_id; }; +enum { + NAPIF_F_IRQ_AFFINITY = BIT(0) +}; + /* * Structure for NAPI scheduling similar to tasklet but with weighting */ @@ -394,6 +399,7 @@ struct napi_struct { unsigned long irq_flags; int index; struct napi_config *config; + struct irq_affinity_notify affinity_notify; }; enum { diff --git a/net/core/dev.c b/net/core/dev.c index 6ef9eb401fb2..778ba27d2b83 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -6699,11 +6699,35 @@ void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, } EXPORT_SYMBOL(netif_queue_set_napi); +static void +netif_napi_affinity_notify(struct irq_affinity_notify *notify, + const cpumask_t *mask) +{ + struct napi_struct *napi = + container_of(notify, struct napi_struct, affinity_notify); + + if (napi->config) + cpumask_copy(&napi->config->affinity_mask, mask); +} + +static void +netif_napi_affinity_release(struct kref __always_unused *ref) +{ +} + static void napi_restore_config(struct napi_struct *n) { n->defer_hard_irqs = n->config->defer_hard_irqs; n->gro_flush_timeout = n->config->gro_flush_timeout; n->irq_suspend_timeout = n->config->irq_suspend_timeout; + + if (n->irq > 0 && n->irq_flags & NAPIF_F_IRQ_AFFINITY) { + n->affinity_notify.notify = netif_napi_affinity_notify; + n->affinity_notify.release = netif_napi_affinity_release; + irq_set_affinity_notifier(n->irq, &n->affinity_notify); + irq_set_affinity(n->irq, &n->config->affinity_mask); + } + /* a NAPI ID might be stored in the config, if so use it. if not, use * napi_hash_add to generate one for us. It will be saved to the config * in napi_disable. @@ -6720,6 +6744,8 @@ static void napi_save_config(struct napi_struct *n) n->config->gro_flush_timeout = n->gro_flush_timeout; n->config->irq_suspend_timeout = n->irq_suspend_timeout; n->config->napi_id = n->napi_id; + if (n->irq > 0 && n->irq_flags & NAPIF_F_IRQ_AFFINITY) + irq_set_affinity_notifier(n->irq, NULL); napi_hash_del(n); } @@ -11184,7 +11210,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, { struct net_device *dev; size_t napi_config_sz; - unsigned int maxqs; + unsigned int maxqs, i; BUG_ON(strlen(name) >= sizeof(dev->name)); @@ -11280,6 +11306,9 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, dev->napi_config = kvzalloc(napi_config_sz, GFP_KERNEL_ACCOUNT); if (!dev->napi_config) goto free_all; + for (i = 0; i < maxqs; i++) + cpumask_copy(&dev->napi_config[i].affinity_mask, + cpu_online_mask); strscpy(dev->name, name); dev->name_assign_type = name_assign_type; -- 2.47.0
