[PATCH net-next v2 0/8] net: napi: add CPU affinity to napi->config

[PATCH net-next v2 0/8] net: napi: add CPU affinity to napi->config

[Ahmed Zaki]

Move the IRQ affinity management to the napi struct. All drivers that are
already using netif_napi_set_irq() are modified to the new API. Except
mlx5 because it is implementing IRQ pools and moving to the new API does
not seem trivial.

Tested on bnxt, ice and idpf.
---
Opens: is cpu_online_mask the best default mask? drivers do this differently 

v2:
    - Also move the ARFS IRQ affinity management from drivers to core. Via
      netif_napi_set_irq(), drivers can ask the core to add the IRQ to the
      ARFS rmap (already allocated by the driver).

RFC -> v1:
    - https://lore.kernel.org/netdev/20241210002626.366878-1-ahmed.zaki@intel.com/
    - move static inline affinity functions to net/dev/core.c
    - add the new napi->irq_flags (patch 1)
    - add code changes to bnxt, mlx4 and ice.

Ahmed Zaki (8):
  net: napi: add irq_flags to napi struct
  net: allow ARFS rmap management in core
  lib: cpu_rmap: allow passing a notifier callback
  net: napi: add CPU affinity to napi->config
  bnxt: use napi's irq affinity
  ice: use napi's irq affinity
  idpf: use napi's irq affinity
  mlx4: use napi's irq affinity

 drivers/net/ethernet/amazon/ena/ena_netdev.c  | 21 ++---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 51 +++--------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  2 -
 drivers/net/ethernet/broadcom/tg3.c           |  2 +-
 drivers/net/ethernet/cisco/enic/enic_main.c   |  3 +-
 drivers/net/ethernet/google/gve/gve_utils.c   |  2 +-
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   |  2 +-
 drivers/net/ethernet/intel/e1000/e1000_main.c |  2 +-
 drivers/net/ethernet/intel/e1000e/netdev.c    |  2 +-
 drivers/net/ethernet/intel/ice/ice.h          |  3 -
 drivers/net/ethernet/intel/ice/ice_arfs.c     | 10 +--
 drivers/net/ethernet/intel/ice/ice_base.c     |  7 +-
 drivers/net/ethernet/intel/ice/ice_lib.c      | 14 +--
 drivers/net/ethernet/intel/ice/ice_main.c     | 44 ----------
 drivers/net/ethernet/intel/idpf/idpf_txrx.c   | 19 ++--
 drivers/net/ethernet/intel/idpf/idpf_txrx.h   |  6 +-
 drivers/net/ethernet/mellanox/mlx4/en_cq.c    |  8 +-
 .../net/ethernet/mellanox/mlx4/en_netdev.c    | 33 +------
 drivers/net/ethernet/mellanox/mlx4/eq.c       | 24 +----
 drivers/net/ethernet/mellanox/mlx4/main.c     | 42 +--------
 drivers/net/ethernet/mellanox/mlx4/mlx4.h     |  1 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h  |  1 -
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  2 +-
 .../net/ethernet/mellanox/mlx5/core/pci_irq.c |  2 +-
 drivers/net/ethernet/meta/fbnic/fbnic_txrx.c  |  3 +-
 drivers/net/ethernet/qlogic/qede/qede_main.c  | 28 +++---
 drivers/net/ethernet/sfc/falcon/efx.c         |  9 ++
 drivers/net/ethernet/sfc/falcon/nic.c         | 10 ---
 drivers/net/ethernet/sfc/nic.c                |  2 +-
 drivers/net/ethernet/sfc/siena/efx_channels.c |  9 ++
 drivers/net/ethernet/sfc/siena/nic.c          | 10 ---
 include/linux/cpu_rmap.h                      | 13 ++-
 include/linux/netdevice.h                     | 23 ++++-
 lib/cpu_rmap.c                                | 20 ++---
 net/core/dev.c                                | 87 ++++++++++++++++++-
 35 files changed, 215 insertions(+), 302 deletions(-)

-- 
2.43.0

[PATCH net-next v2 1/8] net: napi: add irq_flags to napi struct

[Ahmed Zaki]

Add irq_flags to the napi struct. This will allow the drivers to choose
how the core handles the IRQ assigned to the napi via
netif_napi_set_irq().

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/amazon/ena/ena_netdev.c      | 2 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt.c         | 2 +-
 drivers/net/ethernet/broadcom/tg3.c               | 2 +-
 drivers/net/ethernet/google/gve/gve_utils.c       | 2 +-
 drivers/net/ethernet/intel/e1000/e1000_main.c     | 2 +-
 drivers/net/ethernet/intel/e1000e/netdev.c        | 2 +-
 drivers/net/ethernet/intel/ice/ice_lib.c          | 2 +-
 drivers/net/ethernet/mellanox/mlx4/en_cq.c        | 4 ++--
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 2 +-
 drivers/net/ethernet/meta/fbnic/fbnic_txrx.c      | 3 ++-
 include/linux/netdevice.h                         | 6 ++----
 net/core/dev.c                                    | 9 ++++++++-
 12 files changed, 22 insertions(+), 16 deletions(-)

diff --git a/drivers/net/ethernet/amazon/ena/ena_netdev.c b/drivers/net/ethernet/amazon/ena/ena_netdev.c
index c1295dfad0d0..4898c8be78ad 100644
--- a/drivers/net/ethernet/amazon/ena/ena_netdev.c
+++ b/drivers/net/ethernet/amazon/ena/ena_netdev.c
@@ -1712,7 +1712,7 @@ static int ena_request_io_irq(struct ena_adapter *adapter)
 	for (i = 0; i < io_queue_count; i++) {
 		irq_idx = ENA_IO_IRQ_IDX(i);
 		irq = &adapter->irq_tbl[irq_idx];
-		netif_napi_set_irq(&adapter->ena_napi[i].napi, irq->vector);
+		netif_napi_set_irq(&adapter->ena_napi[i].napi, irq->vector, 0);
 	}
 
 	return rc;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index b86f980fa7ea..4763c6300bd3 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -11225,7 +11225,7 @@ static int bnxt_request_irq(struct bnxt *bp)
 		if (rc)
 			break;
 
-		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector);
+		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector, 0);
 		irq->requested = 1;
 
 		if (zalloc_cpumask_var(&irq->cpu_mask, GFP_KERNEL)) {
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 9cc8db10a8d6..0d6383804270 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -7447,7 +7447,7 @@ static void tg3_napi_init(struct tg3 *tp)
 	for (i = 0; i < tp->irq_cnt; i++) {
 		netif_napi_add(tp->dev, &tp->napi[i].napi,
 			       i ? tg3_poll_msix : tg3_poll);
-		netif_napi_set_irq(&tp->napi[i].napi, tp->napi[i].irq_vec);
+		netif_napi_set_irq(&tp->napi[i].napi, tp->napi[i].irq_vec, 0);
 	}
 }
 
diff --git a/drivers/net/ethernet/google/gve/gve_utils.c b/drivers/net/ethernet/google/gve/gve_utils.c
index 30fef100257e..2657e583f5c6 100644
--- a/drivers/net/ethernet/google/gve/gve_utils.c
+++ b/drivers/net/ethernet/google/gve/gve_utils.c
@@ -111,7 +111,7 @@ void gve_add_napi(struct gve_priv *priv, int ntfy_idx,
 	struct gve_notify_block *block = &priv->ntfy_blocks[ntfy_idx];
 
 	netif_napi_add(priv->dev, &block->napi, gve_poll);
-	netif_napi_set_irq(&block->napi, block->irq);
+	netif_napi_set_irq(&block->napi, block->irq, 0);
 }
 
 void gve_remove_napi(struct gve_priv *priv, int ntfy_idx)
diff --git a/drivers/net/ethernet/intel/e1000/e1000_main.c b/drivers/net/ethernet/intel/e1000/e1000_main.c
index 3f089c3d47b2..a83af159837a 100644
--- a/drivers/net/ethernet/intel/e1000/e1000_main.c
+++ b/drivers/net/ethernet/intel/e1000/e1000_main.c
@@ -1394,7 +1394,7 @@ int e1000_open(struct net_device *netdev)
 	/* From here on the code is the same as e1000_up() */
 	clear_bit(__E1000_DOWN, &adapter->flags);
 
-	netif_napi_set_irq(&adapter->napi, adapter->pdev->irq);
+	netif_napi_set_irq(&adapter->napi, adapter->pdev->irq, 0);
 	napi_enable(&adapter->napi);
 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, &adapter->napi);
 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, &adapter->napi);
diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c b/drivers/net/ethernet/intel/e1000e/netdev.c
index 286155efcedf..8fc5603ed962 100644
--- a/drivers/net/ethernet/intel/e1000e/netdev.c
+++ b/drivers/net/ethernet/intel/e1000e/netdev.c
@@ -4676,7 +4676,7 @@ int e1000e_open(struct net_device *netdev)
 	else
 		irq = adapter->pdev->irq;
 
-	netif_napi_set_irq(&adapter->napi, irq);
+	netif_napi_set_irq(&adapter->napi, irq, 0);
 	napi_enable(&adapter->napi);
 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, &adapter->napi);
 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, &adapter->napi);
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index a7d45a8ce7ac..ff91e70f596f 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -2735,7 +2735,7 @@ void ice_vsi_set_napi_queues(struct ice_vsi *vsi)
 	ice_for_each_q_vector(vsi, v_idx) {
 		struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
 
-		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq);
+		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq, 0);
 	}
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_cq.c b/drivers/net/ethernet/mellanox/mlx4/en_cq.c
index 0e92956e84cf..b8531283e3ac 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_cq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_cq.c
@@ -150,7 +150,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 	case TX:
 		cq->mcq.comp = mlx4_en_tx_irq;
 		netif_napi_add_tx(cq->dev, &cq->napi, mlx4_en_poll_tx_cq);
-		netif_napi_set_irq(&cq->napi, irq);
+		netif_napi_set_irq(&cq->napi, irq, 0);
 		napi_enable(&cq->napi);
 		netif_queue_set_napi(cq->dev, cq_idx, NETDEV_QUEUE_TYPE_TX, &cq->napi);
 		break;
@@ -158,7 +158,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 		cq->mcq.comp = mlx4_en_rx_irq;
 		netif_napi_add_config(cq->dev, &cq->napi, mlx4_en_poll_rx_cq,
 				      cq_idx);
-		netif_napi_set_irq(&cq->napi, irq);
+		netif_napi_set_irq(&cq->napi, irq, 0);
 		napi_enable(&cq->napi);
 		netif_queue_set_napi(cq->dev, cq_idx, NETDEV_QUEUE_TYPE_RX, &cq->napi);
 		break;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index dd16d73000c3..58b8313f4c5a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -2733,7 +2733,7 @@ static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
 	c->lag_port = mlx5e_enumerate_lag_port(mdev, ix);
 
 	netif_napi_add_config(netdev, &c->napi, mlx5e_napi_poll, ix);
-	netif_napi_set_irq(&c->napi, irq);
+	netif_napi_set_irq(&c->napi, irq, 0);
 
 	err = mlx5e_open_queues(c, params, cparam);
 	if (unlikely(err))
diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c b/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c
index b5050fabe8fe..6ca91ce85d48 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c
@@ -1227,7 +1227,8 @@ static int fbnic_alloc_napi_vector(struct fbnic_dev *fbd, struct fbnic_net *fbn,
 
 	/* Record IRQ to NAPI struct */
 	netif_napi_set_irq(&nv->napi,
-			   pci_irq_vector(to_pci_dev(fbd->dev), nv->v_idx));
+			   pci_irq_vector(to_pci_dev(fbd->dev), nv->v_idx),
+			   0);
 
 	/* Tie nv back to PCIe dev */
 	nv->dev = fbd->dev;
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 2593019ad5b1..ca91b6662bde 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -392,6 +392,7 @@ struct napi_struct {
 	struct list_head	dev_list;
 	struct hlist_node	napi_hash_node;
 	int			irq;
+	unsigned long		irq_flags;
 	int			index;
 	struct napi_config	*config;
 };
@@ -2671,10 +2672,7 @@ void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
 			  enum netdev_queue_type type,
 			  struct napi_struct *napi);
 
-static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
-{
-	napi->irq = irq;
-}
+void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags);
 
 /* Default NAPI poll() weight
  * Device drivers are strongly advised to not use bigger value
diff --git a/net/core/dev.c b/net/core/dev.c
index c7f3dea3e0eb..88a7d4b6e71b 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6705,6 +6705,13 @@ void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
 }
 EXPORT_SYMBOL(netif_queue_set_napi);
 
+void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags)
+{
+	napi->irq = irq;
+	napi->irq_flags = flags;
+}
+EXPORT_SYMBOL(netif_napi_set_irq);
+
 static void napi_restore_config(struct napi_struct *n)
 {
 	n->defer_hard_irqs = n->config->defer_hard_irqs;
@@ -6770,7 +6777,7 @@ void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
 	 */
 	if (dev->threaded && napi_kthread_create(napi))
 		dev->threaded = false;
-	netif_napi_set_irq(napi, -1);
+	netif_napi_set_irq(napi, -1, 0);
 }
 EXPORT_SYMBOL(netif_napi_add_weight);
 
-- 
2.43.0

Re: [PATCH net-next v2 1/8] net: napi: add irq_flags to napi struct

[Jakub Kicinski]

On Wed, 18 Dec 2024 09:58:36 -0700 Ahmed Zaki wrote:
> Add irq_flags to the napi struct. This will allow the drivers to choose
> how the core handles the IRQ assigned to the napi via
> netif_napi_set_irq().

I haven't read all the code, but I think the flag should be for the
netdev as a while, not NAPI by NAPI. In fact you can combine it with
allocating the map, too.

int netif_enable_cpu_rmap(dev, num_queues)
{
#ifdef CONFIG_RFS_ACCEL
	WARN_ON(dev->rx_cpu_rmap);

	dev->rx_cpu_rmap = alloc_irq_cpu_rmap(adapter->num_queues);
	if ...
	
	dev->rx_cpu_rmap_auto = 1;
	return 0;
#endif
}

void netif_disable_cpu_rmap(dev)
{
	dev->rx_cpu_rmap_auto = 0;
	free_irq_cpu_rmap(dev->rx_cpu_rmap);
}

Then in the NAPI code you just:

void netif_napi_set_irq(...)
{
	...

	if (napi->dev->rx_cpu_rmap_auto) {
		err = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq);
		...
	}
}

Re: [PATCH net-next v2 1/8] net: napi: add irq_flags to napi struct

[Ahmed Zaki]

On 2024-12-19 8:34 p.m., Jakub Kicinski wrote:
> On Wed, 18 Dec 2024 09:58:36 -0700 Ahmed Zaki wrote:
>> Add irq_flags to the napi struct. This will allow the drivers to choose
>> how the core handles the IRQ assigned to the napi via
>> netif_napi_set_irq().
> 
> I haven't read all the code, but I think the flag should be for the
> netdev as a while, not NAPI by NAPI. In fact you can combine it with
> allocating the map, too.
> 
> int netif_enable_cpu_rmap(dev, num_queues)

int netif_enable_cpu_rmap(dev, num_vectors)

> {
> #ifdef CONFIG_RFS_ACCEL
> 	WARN_ON(dev->rx_cpu_rmap);
> 
> 	dev->rx_cpu_rmap = alloc_irq_cpu_rmap(adapter->num_queues);
> 	if ...
> 	
> 	dev->rx_cpu_rmap_auto = 1;
> 	return 0;
> #endif
> }

I was trying to avoid adding an extra function, but since this will 
replace alloc_irq_cpu_rmap() I guess I can try. May be even use

dev->netdev_ops->ndo_rx_flow_steer

instead of dev->rx_cpu_rmap_auto.

I will keep the flag in patch 4 (NAPI_IRQ_AFFINITY) per NAPI since it is 
used in netif_napi_set_irq().

Thanks for the review.

[PATCH net-next v2 2/8] net: allow ARFS rmap management in core

[Ahmed Zaki]

Add a new napi->irq flag; NAPIF_IRQ_ARFS_RMAP. A driver can use the flag
when binding an irq to a napi:

 netif_napi_set_irq(napi, irq, NAPIF_IRQ_ARFS_RMAP)

and the core will update the ARFS rmap with the assigned irq affinity.

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/amazon/ena/ena_netdev.c  | 19 ++++---------
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 25 ++++++-----------
 drivers/net/ethernet/intel/ice/ice_arfs.c     | 10 +------
 drivers/net/ethernet/intel/ice/ice_lib.c      |  5 ++++
 drivers/net/ethernet/qlogic/qede/qede_main.c  | 28 +++++++++----------
 drivers/net/ethernet/sfc/falcon/efx.c         |  9 ++++++
 drivers/net/ethernet/sfc/falcon/nic.c         | 10 -------
 drivers/net/ethernet/sfc/siena/efx_channels.c |  9 ++++++
 drivers/net/ethernet/sfc/siena/nic.c          | 10 -------
 include/linux/netdevice.h                     | 12 ++++++++
 net/core/dev.c                                | 14 ++++++++++
 11 files changed, 77 insertions(+), 74 deletions(-)

diff --git a/drivers/net/ethernet/amazon/ena/ena_netdev.c b/drivers/net/ethernet/amazon/ena/ena_netdev.c
index 4898c8be78ad..752b1c61b610 100644
--- a/drivers/net/ethernet/amazon/ena/ena_netdev.c
+++ b/drivers/net/ethernet/amazon/ena/ena_netdev.c
@@ -165,23 +165,9 @@ int ena_xmit_common(struct ena_adapter *adapter,
 static int ena_init_rx_cpu_rmap(struct ena_adapter *adapter)
 {
 #ifdef CONFIG_RFS_ACCEL
-	u32 i;
-	int rc;
-
 	adapter->netdev->rx_cpu_rmap = alloc_irq_cpu_rmap(adapter->num_io_queues);
 	if (!adapter->netdev->rx_cpu_rmap)
 		return -ENOMEM;
-	for (i = 0; i < adapter->num_io_queues; i++) {
-		int irq_idx = ENA_IO_IRQ_IDX(i);
-
-		rc = irq_cpu_rmap_add(adapter->netdev->rx_cpu_rmap,
-				      pci_irq_vector(adapter->pdev, irq_idx));
-		if (rc) {
-			free_irq_cpu_rmap(adapter->netdev->rx_cpu_rmap);
-			adapter->netdev->rx_cpu_rmap = NULL;
-			return rc;
-		}
-	}
 #endif /* CONFIG_RFS_ACCEL */
 	return 0;
 }
@@ -1712,7 +1698,12 @@ static int ena_request_io_irq(struct ena_adapter *adapter)
 	for (i = 0; i < io_queue_count; i++) {
 		irq_idx = ENA_IO_IRQ_IDX(i);
 		irq = &adapter->irq_tbl[irq_idx];
+#ifdef CONFIG_RFS_ACCEL
+		netif_napi_set_irq(&adapter->ena_napi[i].napi, irq->vector,
+				   NAPIF_IRQ_ARFS_RMAP);
+#else
 		netif_napi_set_irq(&adapter->ena_napi[i].napi, irq->vector, 0);
+#endif
 	}
 
 	return rc;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 4763c6300bd3..ac729a25ba52 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -11192,11 +11192,8 @@ static void bnxt_free_irq(struct bnxt *bp)
 
 static int bnxt_request_irq(struct bnxt *bp)
 {
-	int i, j, rc = 0;
+	int i, rc = 0;
 	unsigned long flags = 0;
-#ifdef CONFIG_RFS_ACCEL
-	struct cpu_rmap *rmap;
-#endif
 
 	rc = bnxt_setup_int_mode(bp);
 	if (rc) {
@@ -11204,28 +11201,22 @@ static int bnxt_request_irq(struct bnxt *bp)
 			   rc);
 		return rc;
 	}
-#ifdef CONFIG_RFS_ACCEL
-	rmap = bp->dev->rx_cpu_rmap;
-#endif
-	for (i = 0, j = 0; i < bp->cp_nr_rings; i++) {
+
+	for (i = 0; i < bp->cp_nr_rings; i++) {
 		int map_idx = bnxt_cp_num_to_irq_num(bp, i);
 		struct bnxt_irq *irq = &bp->irq_tbl[map_idx];
 
-#ifdef CONFIG_RFS_ACCEL
-		if (rmap && bp->bnapi[i]->rx_ring) {
-			rc = irq_cpu_rmap_add(rmap, irq->vector);
-			if (rc)
-				netdev_warn(bp->dev, "failed adding irq rmap for ring %d\n",
-					    j);
-			j++;
-		}
-#endif
 		rc = request_irq(irq->vector, irq->handler, flags, irq->name,
 				 bp->bnapi[i]);
 		if (rc)
 			break;
 
+#ifdef CONFIG_RFS_ACCEL
+		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector,
+				   NAPIF_IRQ_ARFS_RMAP);
+#else
 		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector, 0);
+#endif
 		irq->requested = 1;
 
 		if (zalloc_cpumask_var(&irq->cpu_mask, GFP_KERNEL)) {
diff --git a/drivers/net/ethernet/intel/ice/ice_arfs.c b/drivers/net/ethernet/intel/ice/ice_arfs.c
index 7cee365cc7d1..54d51d218cae 100644
--- a/drivers/net/ethernet/intel/ice/ice_arfs.c
+++ b/drivers/net/ethernet/intel/ice/ice_arfs.c
@@ -590,14 +590,13 @@ void ice_free_cpu_rx_rmap(struct ice_vsi *vsi)
 }
 
 /**
- * ice_set_cpu_rx_rmap - setup CPU reverse map for each queue
+ * ice_set_cpu_rx_rmap - allocate CPU reverse map for a VSI
  * @vsi: the VSI to be forwarded to
  */
 int ice_set_cpu_rx_rmap(struct ice_vsi *vsi)
 {
 	struct net_device *netdev;
 	struct ice_pf *pf;
-	int i;
 
 	if (!vsi || vsi->type != ICE_VSI_PF)
 		return 0;
@@ -614,13 +613,6 @@ int ice_set_cpu_rx_rmap(struct ice_vsi *vsi)
 	if (unlikely(!netdev->rx_cpu_rmap))
 		return -EINVAL;
 
-	ice_for_each_q_vector(vsi, i)
-		if (irq_cpu_rmap_add(netdev->rx_cpu_rmap,
-				     vsi->q_vectors[i]->irq.virq)) {
-			ice_free_cpu_rx_rmap(vsi);
-			return -EINVAL;
-		}
-
 	return 0;
 }
 
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index ff91e70f596f..7c0b2d8e86ba 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -2735,7 +2735,12 @@ void ice_vsi_set_napi_queues(struct ice_vsi *vsi)
 	ice_for_each_q_vector(vsi, v_idx) {
 		struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
 
+#ifdef CONFIG_RFS_ACCEL
+		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq,
+				   NAPIF_IRQ_ARFS_RMAP);
+#else
 		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq, 0);
+#endif
 	}
 }
 
diff --git a/drivers/net/ethernet/qlogic/qede/qede_main.c b/drivers/net/ethernet/qlogic/qede/qede_main.c
index 99df00c30b8c..27c987435242 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_main.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_main.c
@@ -1944,12 +1944,24 @@ static void qede_napi_disable_remove(struct qede_dev *edev)
 
 static void qede_napi_add_enable(struct qede_dev *edev)
 {
+	struct qede_fastpath *fp = &edev->fp_array[i];
 	int i;
 
 	/* Add NAPI objects */
 	for_each_queue(i) {
-		netif_napi_add(edev->ndev, &edev->fp_array[i].napi, qede_poll);
-		napi_enable(&edev->fp_array[i].napi);
+		fp = &edev->fp_array[i];
+		netif_napi_add(edev->ndev, &fp->napi, qede_poll);
+		napi_enable(&fp->napi);
+		if (edev->ndev->rx_cpu_rmap && (fp->type & QEDE_FASTPATH_RX))
+#ifdef CONFIG_RFS_ACCEL
+			netif_napi_set_irq(&edev->fp_array[i].napi,
+					   edev->int_info.msix[i].vector,
+					   NAPIF_IRQ_ARFS_RMAP);
+#else
+			netif_napi_set_irq(&edev->fp_array[i].napi,
+					   edev->int_info.msix[i].vector,
+					   0);
+#endif
 	}
 }
 
@@ -1983,18 +1995,6 @@ static int qede_req_msix_irqs(struct qede_dev *edev)
 	}
 
 	for (i = 0; i < QEDE_QUEUE_CNT(edev); i++) {
-#ifdef CONFIG_RFS_ACCEL
-		struct qede_fastpath *fp = &edev->fp_array[i];
-
-		if (edev->ndev->rx_cpu_rmap && (fp->type & QEDE_FASTPATH_RX)) {
-			rc = irq_cpu_rmap_add(edev->ndev->rx_cpu_rmap,
-					      edev->int_info.msix[i].vector);
-			if (rc) {
-				DP_ERR(edev, "Failed to add CPU rmap\n");
-				qede_free_arfs(edev);
-			}
-		}
-#endif
 		rc = request_irq(edev->int_info.msix[i].vector,
 				 qede_msix_fp_int, 0, edev->fp_array[i].name,
 				 &edev->fp_array[i]);
diff --git a/drivers/net/ethernet/sfc/falcon/efx.c b/drivers/net/ethernet/sfc/falcon/efx.c
index b07f7e4e2877..8c2f850d4639 100644
--- a/drivers/net/ethernet/sfc/falcon/efx.c
+++ b/drivers/net/ethernet/sfc/falcon/efx.c
@@ -2004,6 +2004,15 @@ static void ef4_init_napi_channel(struct ef4_channel *channel)
 
 	channel->napi_dev = efx->net_dev;
 	netif_napi_add(channel->napi_dev, &channel->napi_str, ef4_poll);
+
+	if (efx->interrupt_mode == EF4_INT_MODE_MSIX &&
+	    channel->channel < efx->n_rx_channels)
+#ifdef CONFIG_RFS_ACCEL
+		netif_napi_set_irq(&channel->napi_str, channel->irq,
+				   NAPIF_IRQ_ARFS_RMAP);
+#else
+		netif_napi_set_irq(&channel->napi_str, channel->irq, 0);
+#endif
 }
 
 static void ef4_init_napi(struct ef4_nic *efx)
diff --git a/drivers/net/ethernet/sfc/falcon/nic.c b/drivers/net/ethernet/sfc/falcon/nic.c
index a6304686bc90..fa31d83e64e4 100644
--- a/drivers/net/ethernet/sfc/falcon/nic.c
+++ b/drivers/net/ethernet/sfc/falcon/nic.c
@@ -115,16 +115,6 @@ int ef4_nic_init_interrupt(struct ef4_nic *efx)
 			goto fail2;
 		}
 		++n_irqs;
-
-#ifdef CONFIG_RFS_ACCEL
-		if (efx->interrupt_mode == EF4_INT_MODE_MSIX &&
-		    channel->channel < efx->n_rx_channels) {
-			rc = irq_cpu_rmap_add(efx->net_dev->rx_cpu_rmap,
-					      channel->irq);
-			if (rc)
-				goto fail2;
-		}
-#endif
 	}
 
 	return 0;
diff --git a/drivers/net/ethernet/sfc/siena/efx_channels.c b/drivers/net/ethernet/sfc/siena/efx_channels.c
index d120b3c83ac0..6fed4f7b311f 100644
--- a/drivers/net/ethernet/sfc/siena/efx_channels.c
+++ b/drivers/net/ethernet/sfc/siena/efx_channels.c
@@ -1321,6 +1321,15 @@ static void efx_init_napi_channel(struct efx_channel *channel)
 
 	channel->napi_dev = efx->net_dev;
 	netif_napi_add(channel->napi_dev, &channel->napi_str, efx_poll);
+
+	if (efx->interrupt_mode == EFX_INT_MODE_MSIX &&
+	    channel->channel < efx->n_rx_channels)
+#ifdef CONFIG_RFS_ACCEL
+		netif_napi_set_irq(&channel->napi_str, channel->irq,
+				   NAPIF_IRQ_ARFS_RMAP);
+#else
+		netif_napi_set_irq(&channel->napi_str, channel->irq, 0);
+#endif
 }
 
 void efx_siena_init_napi(struct efx_nic *efx)
diff --git a/drivers/net/ethernet/sfc/siena/nic.c b/drivers/net/ethernet/sfc/siena/nic.c
index 32fce70085e3..28ef6222395e 100644
--- a/drivers/net/ethernet/sfc/siena/nic.c
+++ b/drivers/net/ethernet/sfc/siena/nic.c
@@ -117,16 +117,6 @@ int efx_siena_init_interrupt(struct efx_nic *efx)
 			goto fail2;
 		}
 		++n_irqs;
-
-#ifdef CONFIG_RFS_ACCEL
-		if (efx->interrupt_mode == EFX_INT_MODE_MSIX &&
-		    channel->channel < efx->n_rx_channels) {
-			rc = irq_cpu_rmap_add(efx->net_dev->rx_cpu_rmap,
-					      channel->irq);
-			if (rc)
-				goto fail2;
-		}
-#endif
 	}
 
 	efx->irqs_hooked = true;
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index ca91b6662bde..0df419052434 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -354,6 +354,18 @@ struct napi_config {
 	unsigned int napi_id;
 };
 
+enum {
+#ifdef CONFIG_RFS_ACCEL
+	NAPI_IRQ_ARFS_RMAP,		/* Core handles RMAP updates */
+#endif
+};
+
+enum {
+#ifdef CONFIG_RFS_ACCEL
+	NAPIF_IRQ_ARFS_RMAP		= BIT(NAPI_IRQ_ARFS_RMAP),
+#endif
+};
+
 /*
  * Structure for NAPI scheduling similar to tasklet but with weighting
  */
diff --git a/net/core/dev.c b/net/core/dev.c
index 88a7d4b6e71b..7c3abff48aea 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6707,8 +6707,22 @@ EXPORT_SYMBOL(netif_queue_set_napi);
 
 void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags)
 {
+	int  rc;
+
 	napi->irq = irq;
 	napi->irq_flags = flags;
+
+#ifdef CONFIG_RFS_ACCEL
+	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
+		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq);
+		if (rc) {
+			netdev_warn(napi->dev, "Unable to update ARFS map (%d).\n",
+				    rc);
+			free_irq_cpu_rmap(napi->dev->rx_cpu_rmap);
+			napi->dev->rx_cpu_rmap = NULL;
+		}
+	}
+#endif
 }
 EXPORT_SYMBOL(netif_napi_set_irq);
 
-- 
2.43.0

Re: [Intel-wired-lan] [PATCH net-next v2 2/8] net: allow ARFS rmap management in core

[kernel test robot]

Hi Ahmed,

kernel test robot noticed the following build warnings:

[auto build test WARNING on net-next/main]

url:    https://github.com/intel-lab-lkp/linux/commits/Ahmed-Zaki/net-napi-add-irq_flags-to-napi-struct/20241219-010125
base:   net-next/main
patch link:    https://lore.kernel.org/r/20241218165843.744647-3-ahmed.zaki%40intel.com
patch subject: [Intel-wired-lan] [PATCH net-next v2 2/8] net: allow ARFS rmap management in core
config: arm-randconfig-002-20241219 (https://download.01.org/0day-ci/archive/20241219/202412190318.OR90xHNu-lkp@intel.com/config)
compiler: arm-linux-gnueabi-gcc (GCC) 14.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20241219/202412190318.OR90xHNu-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202412190318.OR90xHNu-lkp@intel.com/

All warnings (new ones prefixed by >>):

   In file included from net/core/dev.c:92:
   include/linux/netdevice.h:361:1: error: empty enum is invalid
     361 | };
         | ^
   include/linux/netdevice.h:367:1: error: empty enum is invalid
     367 | };
         | ^
   net/core/dev.c: In function 'netif_napi_set_irq':
>> net/core/dev.c:6710:14: warning: unused variable 'rc' [-Wunused-variable]
    6710 |         int  rc;
         |              ^~


vim +/rc +6710 net/core/dev.c

  6707	
  6708	void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags)
  6709	{
> 6710		int  rc;
  6711	
  6712		napi->irq = irq;
  6713		napi->irq_flags = flags;
  6714	

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

[PATCH net-next v2 3/8] lib: cpu_rmap: allow passing a notifier callback

[Ahmed Zaki]

Allow the rmap users to pass a notifier callback function that can be
called instead of irq_cpu_rmap_notify().

Two modifications are made:
   * make struct irg_glue visible in cpu_rmap.h
   * pass a new "void* data" parameter that can be used by the cb
     function.

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/cisco/enic/enic_main.c   |  3 ++-
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   |  2 +-
 drivers/net/ethernet/mellanox/mlx4/eq.c       |  2 +-
 .../net/ethernet/mellanox/mlx5/core/pci_irq.c |  2 +-
 drivers/net/ethernet/sfc/nic.c                |  2 +-
 include/linux/cpu_rmap.h                      | 13 +++++++++++-
 lib/cpu_rmap.c                                | 20 +++++++++----------
 7 files changed, 28 insertions(+), 16 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c
index 9913952ccb42..e384b975b8af 100644
--- a/drivers/net/ethernet/cisco/enic/enic_main.c
+++ b/drivers/net/ethernet/cisco/enic/enic_main.c
@@ -1657,7 +1657,8 @@ static void enic_set_rx_cpu_rmap(struct enic *enic)
 			return;
 		for (i = 0; i < enic->rq_count; i++) {
 			res = irq_cpu_rmap_add(enic->netdev->rx_cpu_rmap,
-					       enic->msix_entry[i].vector);
+					       enic->msix_entry[i].vector,
+					       NULL, NULL);
 			if (unlikely(res)) {
 				enic_free_rx_cpu_rmap(enic);
 				return;
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 43377a7b2426..3f732516c8ee 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -697,7 +697,7 @@ static int hns3_set_rx_cpu_rmap(struct net_device *netdev)
 	for (i = 0; i < priv->vector_num; i++) {
 		tqp_vector = &priv->tqp_vector[i];
 		ret = irq_cpu_rmap_add(netdev->rx_cpu_rmap,
-				       tqp_vector->vector_irq);
+				       tqp_vector->vector_irq, NULL, NULL);
 		if (ret) {
 			hns3_free_rx_cpu_rmap(netdev);
 			return ret;
diff --git a/drivers/net/ethernet/mellanox/mlx4/eq.c b/drivers/net/ethernet/mellanox/mlx4/eq.c
index 9572a45f6143..d768a6a828c4 100644
--- a/drivers/net/ethernet/mellanox/mlx4/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/eq.c
@@ -1243,7 +1243,7 @@ int mlx4_init_eq_table(struct mlx4_dev *dev)
 				}
 
 				err = irq_cpu_rmap_add(
-					info->rmap, eq->irq);
+					info->rmap, eq->irq, NULL, NULL);
 				if (err)
 					mlx4_warn(dev, "Failed adding irq rmap\n");
 			}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
index 7db9cab9bedf..4f2c4631aecb 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
@@ -285,7 +285,7 @@ struct mlx5_irq *mlx5_irq_alloc(struct mlx5_irq_pool *pool, int i,
 
 	if (i && rmap && *rmap) {
 #ifdef CONFIG_RFS_ACCEL
-		err = irq_cpu_rmap_add(*rmap, irq->map.virq);
+		err = irq_cpu_rmap_add(*rmap, irq->map.virq, NULL, NULL);
 		if (err)
 			goto err_irq_rmap;
 #endif
diff --git a/drivers/net/ethernet/sfc/nic.c b/drivers/net/ethernet/sfc/nic.c
index 80aa5e9c732a..e7c6c3002826 100644
--- a/drivers/net/ethernet/sfc/nic.c
+++ b/drivers/net/ethernet/sfc/nic.c
@@ -122,7 +122,7 @@ int efx_nic_init_interrupt(struct efx_nic *efx)
 		if (efx->interrupt_mode == EFX_INT_MODE_MSIX &&
 		    channel->channel < efx->n_rx_channels) {
 			rc = irq_cpu_rmap_add(efx->net_dev->rx_cpu_rmap,
-					      channel->irq);
+					      channel->irq, NULL, NULL);
 			if (rc)
 				goto fail2;
 		}
diff --git a/include/linux/cpu_rmap.h b/include/linux/cpu_rmap.h
index 20b5729903d7..48f89d19bdb9 100644
--- a/include/linux/cpu_rmap.h
+++ b/include/linux/cpu_rmap.h
@@ -11,6 +11,15 @@
 #include <linux/gfp.h>
 #include <linux/slab.h>
 #include <linux/kref.h>
+#include <linux/interrupt.h>
+
+/* Glue between IRQ affinity notifiers and CPU rmaps */
+struct irq_glue {
+	struct irq_affinity_notify notify;
+	struct cpu_rmap *rmap;
+	void *data;
+	u16 index;
+};
 
 /**
  * struct cpu_rmap - CPU affinity reverse-map
@@ -61,6 +70,8 @@ static inline struct cpu_rmap *alloc_irq_cpu_rmap(unsigned int size)
 extern void free_irq_cpu_rmap(struct cpu_rmap *rmap);
 
 int irq_cpu_rmap_remove(struct cpu_rmap *rmap, int irq);
-extern int irq_cpu_rmap_add(struct cpu_rmap *rmap, int irq);
+extern int irq_cpu_rmap_add(struct cpu_rmap *rmap, int irq, void *data,
+			    void (*notify)(struct irq_affinity_notify *notify,
+					   const cpumask_t *mask));
 
 #endif /* __LINUX_CPU_RMAP_H */
diff --git a/lib/cpu_rmap.c b/lib/cpu_rmap.c
index 4c348670da31..0c9c1078143d 100644
--- a/lib/cpu_rmap.c
+++ b/lib/cpu_rmap.c
@@ -220,14 +220,6 @@ int cpu_rmap_update(struct cpu_rmap *rmap, u16 index,
 }
 EXPORT_SYMBOL(cpu_rmap_update);
 
-/* Glue between IRQ affinity notifiers and CPU rmaps */
-
-struct irq_glue {
-	struct irq_affinity_notify notify;
-	struct cpu_rmap *rmap;
-	u16 index;
-};
-
 /**
  * free_irq_cpu_rmap - free a CPU affinity reverse-map used for IRQs
  * @rmap: Reverse-map allocated with alloc_irq_cpu_map(), or %NULL
@@ -300,6 +292,8 @@ EXPORT_SYMBOL(irq_cpu_rmap_remove);
  * irq_cpu_rmap_add - add an IRQ to a CPU affinity reverse-map
  * @rmap: The reverse-map
  * @irq: The IRQ number
+ * @data: Generic data
+ * @notify: Callback function to update the CPU-IRQ rmap
  *
  * This adds an IRQ affinity notifier that will update the reverse-map
  * automatically.
@@ -307,16 +301,22 @@ EXPORT_SYMBOL(irq_cpu_rmap_remove);
  * Must be called in process context, after the IRQ is allocated but
  * before it is bound with request_irq().
  */
-int irq_cpu_rmap_add(struct cpu_rmap *rmap, int irq)
+int irq_cpu_rmap_add(struct cpu_rmap *rmap, int irq, void *data,
+		     void (*notify)(struct irq_affinity_notify *notify,
+				    const cpumask_t *mask))
 {
 	struct irq_glue *glue = kzalloc(sizeof(*glue), GFP_KERNEL);
 	int rc;
 
 	if (!glue)
 		return -ENOMEM;
-	glue->notify.notify = irq_cpu_rmap_notify;
+
+	if (!notify)
+		notify = irq_cpu_rmap_notify;
+	glue->notify.notify = notify;
 	glue->notify.release = irq_cpu_rmap_release;
 	glue->rmap = rmap;
+	glue->data = data;
 	cpu_rmap_get(rmap);
 	rc = cpu_rmap_add(rmap, glue);
 	if (rc < 0)
-- 
2.43.0

[PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Ahmed Zaki]

A common task for most drivers is to remember the user-set CPU affinity
to its IRQs. On each netdev reset, the driver should 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() with the new
     flag NAPIF_IRQ_AFFINITY

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 <kuba@kernel.org>
Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 include/linux/netdevice.h |  5 +++
 net/core/dev.c            | 66 +++++++++++++++++++++++++++++++++++++--
 2 files changed, 69 insertions(+), 2 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 0df419052434..4fa047fad8fb 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -351,6 +351,7 @@ struct napi_config {
 	u64 gro_flush_timeout;
 	u64 irq_suspend_timeout;
 	u32 defer_hard_irqs;
+	cpumask_t affinity_mask;
 	unsigned int napi_id;
 };
 
@@ -358,12 +359,16 @@ enum {
 #ifdef CONFIG_RFS_ACCEL
 	NAPI_IRQ_ARFS_RMAP,		/* Core handles RMAP updates */
 #endif
+	NAPI_IRQ_AFFINITY,		/* Core manages IRQ affinity */
+	NAPI_IRQ_NORMAP			/* Set by core (internal) */
 };
 
 enum {
 #ifdef CONFIG_RFS_ACCEL
 	NAPIF_IRQ_ARFS_RMAP		= BIT(NAPI_IRQ_ARFS_RMAP),
 #endif
+	NAPIF_IRQ_AFFINITY		= BIT(NAPI_IRQ_AFFINITY),
+	NAPIF_IRQ_NORMAP		= BIT(NAPI_IRQ_NORMAP),
 };
 
 /*
diff --git a/net/core/dev.c b/net/core/dev.c
index 7c3abff48aea..84745cea03a7 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6705,8 +6705,44 @@ void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
 }
 EXPORT_SYMBOL(netif_queue_set_napi);
 
+static void
+netif_irq_cpu_rmap_notify(struct irq_affinity_notify *notify,
+			  const cpumask_t *mask)
+{
+	struct irq_glue *glue =
+		container_of(notify, struct irq_glue, notify);
+	struct napi_struct *napi = glue->data;
+	unsigned int flags;
+	int rc;
+
+	flags = napi->irq_flags;
+
+	if (napi->config && flags & NAPIF_IRQ_AFFINITY)
+		cpumask_copy(&napi->config->affinity_mask, mask);
+
+#ifdef CONFIG_RFS_ACCEL
+	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
+		rc = cpu_rmap_update(glue->rmap, glue->index, mask);
+		if (rc)
+			pr_warn("%s: update failed: %d\n",
+				__func__, rc);
+	}
+#endif
+}
+
+static void
+netif_napi_affinity_release(struct kref __always_unused *ref)
+{
+	struct irq_glue *glue =
+		container_of(ref, struct irq_glue, notify.kref);
+
+	kfree(glue);
+}
+
 void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags)
 {
+	struct irq_glue *glue = NULL;
+	bool glue_created;
 	int  rc;
 
 	napi->irq = irq;
@@ -6714,15 +6750,29 @@ void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags)
 
 #ifdef CONFIG_RFS_ACCEL
 	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
-		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq);
+		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq, napi,
+				      netif_irq_cpu_rmap_notify);
 		if (rc) {
 			netdev_warn(napi->dev, "Unable to update ARFS map (%d).\n",
 				    rc);
 			free_irq_cpu_rmap(napi->dev->rx_cpu_rmap);
 			napi->dev->rx_cpu_rmap = NULL;
+		} else {
+			glue_created = true;
 		}
 	}
 #endif
+
+	if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
+		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
+		if (!glue)
+			return;
+		glue->notify.notify = netif_irq_cpu_rmap_notify;
+		glue->notify.release = netif_napi_affinity_release;
+		glue->data = napi;
+		glue->rmap = NULL;
+		napi->irq_flags |= NAPIF_IRQ_NORMAP;
+	}
 }
 EXPORT_SYMBOL(netif_napi_set_irq);
 
@@ -6731,6 +6781,10 @@ 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_IRQ_AFFINITY)
+		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.
@@ -6747,6 +6801,11 @@ 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_IRQ_AFFINITY | NAPIF_IRQ_NORMAP))
+		irq_set_affinity_notifier(n->irq, NULL);
+
 	napi_hash_del(n);
 }
 
@@ -11211,7 +11270,7 @@ 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));
 
@@ -11307,6 +11366,9 @@ 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;
-- 
2.43.0

Re: [Intel-wired-lan] [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[kernel test robot]

Hi Ahmed,

kernel test robot noticed the following build warnings:

[auto build test WARNING on net-next/main]

url:    https://github.com/intel-lab-lkp/linux/commits/Ahmed-Zaki/net-napi-add-irq_flags-to-napi-struct/20241219-010125
base:   net-next/main
patch link:    https://lore.kernel.org/r/20241218165843.744647-5-ahmed.zaki%40intel.com
patch subject: [Intel-wired-lan] [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config
config: arm-randconfig-001-20241219 (https://download.01.org/0day-ci/archive/20241219/202412190421.N2xtn20H-lkp@intel.com/config)
compiler: clang version 18.1.8 (https://github.com/llvm/llvm-project 3b5b5c1ec4a3095ab096dd780e84d7ab81f3d7ff)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20241219/202412190421.N2xtn20H-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202412190421.N2xtn20H-lkp@intel.com/

All warnings (new ones prefixed by >>):

   net/core/dev.c:6716:6: warning: unused variable 'rc' [-Wunused-variable]
    6716 |         int rc;
         |             ^~
   net/core/dev.c:6746:7: warning: unused variable 'rc' [-Wunused-variable]
    6746 |         int  rc;
         |              ^~
>> net/core/dev.c:6766:7: warning: variable 'glue_created' is uninitialized when used here [-Wuninitialized]
    6766 |         if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
         |              ^~~~~~~~~~~~
   net/core/dev.c:6745:19: note: initialize the variable 'glue_created' to silence this warning
    6745 |         bool glue_created;
         |                          ^
         |                           = 0
   3 warnings generated.


vim +/glue_created +6766 net/core/dev.c

  6765	
> 6766		if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
  6767			glue = kzalloc(sizeof(*glue), GFP_KERNEL);
  6768			if (!glue)
  6769				return;
  6770			glue->notify.notify = netif_irq_cpu_rmap_notify;
  6771			glue->notify.release = netif_napi_affinity_release;
  6772			glue->data = napi;
  6773			glue->rmap = NULL;
  6774			napi->irq_flags |= NAPIF_IRQ_NORMAP;
  6775		}
  6776	}
  6777	EXPORT_SYMBOL(netif_napi_set_irq);
  6778	

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [Intel-wired-lan] [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[kernel test robot]

Hi Ahmed,

kernel test robot noticed the following build warnings:

[auto build test WARNING on net-next/main]

url:    https://github.com/intel-lab-lkp/linux/commits/Ahmed-Zaki/net-napi-add-irq_flags-to-napi-struct/20241219-010125
base:   net-next/main
patch link:    https://lore.kernel.org/r/20241218165843.744647-5-ahmed.zaki%40intel.com
patch subject: [Intel-wired-lan] [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config
config: riscv-randconfig-001-20241219 (https://download.01.org/0day-ci/archive/20241219/202412190454.nwvp3hU2-lkp@intel.com/config)
compiler: clang version 16.0.6 (https://github.com/llvm/llvm-project 7cbf1a2591520c2491aa35339f227775f4d3adf6)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20241219/202412190454.nwvp3hU2-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202412190454.nwvp3hU2-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> net/core/dev.c:6755:7: warning: variable 'glue_created' is used uninitialized whenever 'if' condition is true [-Wsometimes-uninitialized]
                   if (rc) {
                       ^~
   net/core/dev.c:6766:7: note: uninitialized use occurs here
           if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
                ^~~~~~~~~~~~
   net/core/dev.c:6755:3: note: remove the 'if' if its condition is always false
                   if (rc) {
                   ^~~~~~~~~
>> net/core/dev.c:6752:6: warning: variable 'glue_created' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
           if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   net/core/dev.c:6766:7: note: uninitialized use occurs here
           if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
                ^~~~~~~~~~~~
   net/core/dev.c:6752:2: note: remove the 'if' if its condition is always true
           if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> net/core/dev.c:6752:6: warning: variable 'glue_created' is used uninitialized whenever '&&' condition is false [-Wsometimes-uninitialized]
           if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
               ^~~~~~~~~~~~~~~~~~~~~~
   net/core/dev.c:6766:7: note: uninitialized use occurs here
           if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
                ^~~~~~~~~~~~
   net/core/dev.c:6752:6: note: remove the '&&' if its condition is always true
           if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
               ^~~~~~~~~~~~~~~~~~~~~~~~~
   net/core/dev.c:6745:19: note: initialize the variable 'glue_created' to silence this warning
           bool glue_created;
                            ^
                             = 0
   net/core/dev.c:4176:1: warning: unused function 'sch_handle_ingress' [-Wunused-function]
   sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret,
   ^
   net/core/dev.c:4183:1: warning: unused function 'sch_handle_egress' [-Wunused-function]
   sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
   ^
   net/core/dev.c:5440:19: warning: unused function 'nf_ingress' [-Wunused-function]
   static inline int nf_ingress(struct sk_buff *skb, struct packet_type **pt_prev,
                     ^
   6 warnings generated.

Kconfig warnings: (for reference only)
   WARNING: unmet direct dependencies detected for FB_IOMEM_HELPERS
   Depends on [n]: HAS_IOMEM [=y] && FB_CORE [=n]
   Selected by [m]:
   - DRM_XE_DISPLAY [=y] && HAS_IOMEM [=y] && DRM [=m] && DRM_XE [=m] && DRM_XE [=m]=m [=m] && HAS_IOPORT [=y]


vim +6755 net/core/dev.c

8e5191fb19bffce Ahmed Zaki 2024-12-18  6741  
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6742  void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long flags)
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6743  {
8e5191fb19bffce Ahmed Zaki 2024-12-18  6744  	struct irq_glue *glue = NULL;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6745  	bool glue_created;
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6746  	int  rc;
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6747  
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6748  	napi->irq = irq;
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6749  	napi->irq_flags = flags;
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6750  
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6751  #ifdef CONFIG_RFS_ACCEL
a274d2669a73ef7 Ahmed Zaki 2024-12-18 @6752  	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
8e5191fb19bffce Ahmed Zaki 2024-12-18  6753  		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq, napi,
8e5191fb19bffce Ahmed Zaki 2024-12-18  6754  				      netif_irq_cpu_rmap_notify);
a274d2669a73ef7 Ahmed Zaki 2024-12-18 @6755  		if (rc) {
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6756  			netdev_warn(napi->dev, "Unable to update ARFS map (%d).\n",
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6757  				    rc);
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6758  			free_irq_cpu_rmap(napi->dev->rx_cpu_rmap);
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6759  			napi->dev->rx_cpu_rmap = NULL;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6760  		} else {
8e5191fb19bffce Ahmed Zaki 2024-12-18  6761  			glue_created = true;
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6762  		}
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6763  	}
a274d2669a73ef7 Ahmed Zaki 2024-12-18  6764  #endif
8e5191fb19bffce Ahmed Zaki 2024-12-18  6765  
8e5191fb19bffce Ahmed Zaki 2024-12-18  6766  	if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
8e5191fb19bffce Ahmed Zaki 2024-12-18  6767  		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
8e5191fb19bffce Ahmed Zaki 2024-12-18  6768  		if (!glue)
8e5191fb19bffce Ahmed Zaki 2024-12-18  6769  			return;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6770  		glue->notify.notify = netif_irq_cpu_rmap_notify;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6771  		glue->notify.release = netif_napi_affinity_release;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6772  		glue->data = napi;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6773  		glue->rmap = NULL;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6774  		napi->irq_flags |= NAPIF_IRQ_NORMAP;
8e5191fb19bffce Ahmed Zaki 2024-12-18  6775  	}
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6776  }
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6777  EXPORT_SYMBOL(netif_napi_set_irq);
001dc6db21f4bfe Ahmed Zaki 2024-12-18  6778  

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Jakub Kicinski]

On Wed, 18 Dec 2024 09:58:39 -0700 Ahmed Zaki wrote:
> +	if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
> +		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
> +		if (!glue)
> +			return;
> +		glue->notify.notify = netif_irq_cpu_rmap_notify;
> +		glue->notify.release = netif_napi_affinity_release;
> +		glue->data = napi;
> +		glue->rmap = NULL;
> +		napi->irq_flags |= NAPIF_IRQ_NORMAP;

Why allocate the glue? is it not possible to add the fields:

	struct irq_affinity_notify notify;
	u16 index;

to struct napi_struct ?

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Ahmed Zaki]

On 2024-12-19 8:42 p.m., Jakub Kicinski wrote:
> On Wed, 18 Dec 2024 09:58:39 -0700 Ahmed Zaki wrote:
>> +	if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
>> +		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
>> +		if (!glue)
>> +			return;
>> +		glue->notify.notify = netif_irq_cpu_rmap_notify;
>> +		glue->notify.release = netif_napi_affinity_release;
>> +		glue->data = napi;
>> +		glue->rmap = NULL;
>> +		napi->irq_flags |= NAPIF_IRQ_NORMAP;
> 
> Why allocate the glue? is it not possible to add the fields:
> 
> 	struct irq_affinity_notify notify;
> 	u16 index;
> 
> to struct napi_struct ?

In the first branch of "if", the cb function netif_irq_cpu_rmap_notify() 
is also passed to irq_cpu_rmap_add() where the irq notifier is embedded 
in "struct irq_glue".

I think this cannot be changed as long as some drivers are directly 
calling irq_cpu_rmap_add() instead of the proposed API.

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Jakub Kicinski]

On Fri, 20 Dec 2024 07:51:09 -0700 Ahmed Zaki wrote:
> On 2024-12-19 8:42 p.m., Jakub Kicinski wrote:
> > On Wed, 18 Dec 2024 09:58:39 -0700 Ahmed Zaki wrote:  
> >> +	if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
> >> +		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
> >> +		if (!glue)
> >> +			return;
> >> +		glue->notify.notify = netif_irq_cpu_rmap_notify;
> >> +		glue->notify.release = netif_napi_affinity_release;
> >> +		glue->data = napi;
> >> +		glue->rmap = NULL;
> >> +		napi->irq_flags |= NAPIF_IRQ_NORMAP;  
> > 
> > Why allocate the glue? is it not possible to add the fields:
> > 
> > 	struct irq_affinity_notify notify;
> > 	u16 index;
> > 
> > to struct napi_struct ?  
> 
> In the first branch of "if", the cb function netif_irq_cpu_rmap_notify() 
> is also passed to irq_cpu_rmap_add() where the irq notifier is embedded 
> in "struct irq_glue".

I don't understand what you're trying to say, could you rephrase?

> I think this cannot be changed as long as some drivers are directly 
> calling irq_cpu_rmap_add() instead of the proposed API.

Drivers which are not converted shouldn't matter if we have our own
notifier and call cpu_rmap_update() directly, no?

Drivers which are converted should not call irq_cpu_rmap_add().

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Ahmed Zaki]

On 2024-12-20 10:23 a.m., Jakub Kicinski wrote:
> On Fri, 20 Dec 2024 07:51:09 -0700 Ahmed Zaki wrote:
>> On 2024-12-19 8:42 p.m., Jakub Kicinski wrote:
>>> On Wed, 18 Dec 2024 09:58:39 -0700 Ahmed Zaki wrote:
>>>> +	if (!glue_created && flags & NAPIF_IRQ_AFFINITY) {
>>>> +		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
>>>> +		if (!glue)
>>>> +			return;
>>>> +		glue->notify.notify = netif_irq_cpu_rmap_notify;
>>>> +		glue->notify.release = netif_napi_affinity_release;
>>>> +		glue->data = napi;
>>>> +		glue->rmap = NULL;
>>>> +		napi->irq_flags |= NAPIF_IRQ_NORMAP;
>>>
>>> Why allocate the glue? is it not possible to add the fields:
>>>
>>> 	struct irq_affinity_notify notify;
>>> 	u16 index;
>>>
>>> to struct napi_struct ?
>>
>> In the first branch of "if", the cb function netif_irq_cpu_rmap_notify()
>> is also passed to irq_cpu_rmap_add() where the irq notifier is embedded
>> in "struct irq_glue".
> 
> I don't understand what you're trying to say, could you rephrase?

Sure. After this patch, we have (simplified):

void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long 
flags)
  {
	struct irq_glue *glue = NULL;
  	int  rc;

  	napi->irq = irq;

  #ifdef CONFIG_RFS_ACCEL
  	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq, napi,
				      netif_irq_cpu_rmap_notify);
		.
		.
		.
  	}
  #endif

	if (flags & NAPIF_IRQ_AFFINITY) {
		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
		if (!glue)
			return;
		glue->notify.notify = netif_irq_cpu_rmap_notify;
		glue->notify.release = netif_napi_affinity_release;
		.
		.
	}
  }


Both branches assign the new cb function "netif_irq_cpu_rmap_notify()" 
as the new IRQ notifier, but the first branch calls irq_cpu_rmap_add() 
where the notifier is embedded in "struct irq_glue". So the cb function 
needs to assume the notifier is inside irq_glue, so the second "if" 
branch needs to do the same.


> 
>> I think this cannot be changed as long as some drivers are directly
>> calling irq_cpu_rmap_add() instead of the proposed API.
> 
> Drivers which are not converted shouldn't matter if we have our own
> notifier and call cpu_rmap_update() directly, no?
> 

Only dependency is that irq_cpu_rmap_add() puts notifier inside irq_glue.

> Drivers which are converted should not call irq_cpu_rmap_add().

Correct, they don't.

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Jakub Kicinski]

On Fri, 20 Dec 2024 12:15:33 -0700 Ahmed Zaki wrote:
> > I don't understand what you're trying to say, could you rephrase?  
> 
> Sure. After this patch, we have (simplified):
> 
> void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long 
> flags)
>   {
> 	struct irq_glue *glue = NULL;
>   	int  rc;
> 
>   	napi->irq = irq;
> 
>   #ifdef CONFIG_RFS_ACCEL
>   	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
> 		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq, napi,
> 				      netif_irq_cpu_rmap_notify);
> 		.
> 		.
> 		.
>   	}
>   #endif
> 
> 	if (flags & NAPIF_IRQ_AFFINITY) {
> 		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
> 		if (!glue)
> 			return;
> 		glue->notify.notify = netif_irq_cpu_rmap_notify;
> 		glue->notify.release = netif_napi_affinity_release;
> 		.
> 		.
> 	}
>   }
> 
> 
> Both branches assign the new cb function "netif_irq_cpu_rmap_notify()" 
> as the new IRQ notifier, but the first branch calls irq_cpu_rmap_add() 
> where the notifier is embedded in "struct irq_glue". So the cb function 
> needs to assume the notifier is inside irq_glue, so the second "if" 
> branch needs to do the same.

First off, I'm still a bit confused why you think the flags should be
per NAPI call and not set at init time, once.
Perhaps rename netif_enable_cpu_rmap() suggested earlier to something
more generic (netif_enable_irq_tracking()?) and pass the flags there?
Or is there a driver which wants to vary the flags per NAPI instance?

Then you can probably register a single unified handler, and inside
that handler check if the device wanted to have rmap or just affinity?

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Ahmed Zaki]

On 2024-12-20 12:37 p.m., Jakub Kicinski wrote:
> On Fri, 20 Dec 2024 12:15:33 -0700 Ahmed Zaki wrote:
>>> I don't understand what you're trying to say, could you rephrase?
>>
>> Sure. After this patch, we have (simplified):
>>
>> void netif_napi_set_irq(struct napi_struct *napi, int irq, unsigned long
>> flags)
>>    {
>> 	struct irq_glue *glue = NULL;
>>    	int  rc;
>>
>>    	napi->irq = irq;
>>
>>    #ifdef CONFIG_RFS_ACCEL
>>    	if (napi->dev->rx_cpu_rmap && flags & NAPIF_IRQ_ARFS_RMAP) {
>> 		rc = irq_cpu_rmap_add(napi->dev->rx_cpu_rmap, irq, napi,
>> 				      netif_irq_cpu_rmap_notify);
>> 		.
>> 		.
>> 		.
>>    	}
>>    #endif
>>
>> 	if (flags & NAPIF_IRQ_AFFINITY) {
>> 		glue = kzalloc(sizeof(*glue), GFP_KERNEL);
>> 		if (!glue)
>> 			return;
>> 		glue->notify.notify = netif_irq_cpu_rmap_notify;
>> 		glue->notify.release = netif_napi_affinity_release;
>> 		.
>> 		.
>> 	}
>>    }
>>
>>
>> Both branches assign the new cb function "netif_irq_cpu_rmap_notify()"
>> as the new IRQ notifier, but the first branch calls irq_cpu_rmap_add()
>> where the notifier is embedded in "struct irq_glue". So the cb function
>> needs to assume the notifier is inside irq_glue, so the second "if"
>> branch needs to do the same.
> 
> First off, I'm still a bit confused why you think the flags should be
> per NAPI call and not set at init time, once.
> Perhaps rename netif_enable_cpu_rmap() suggested earlier to something
> more generic (netif_enable_irq_tracking()?) and pass the flags there?
> Or is there a driver which wants to vary the flags per NAPI instance?
> 

set_irq() just seemed like natural choice since it is already called for 
each IRQ. I was also trying to avoid adding a new function. But sure I 
can do that and move the flags to netdev.

> Then you can probably register a single unified handler, and inside
> that handler check if the device wanted to have rmap or just affinity?

This is what is in this patch already, all drivers following new 
approach will have netif_irq_cpu_rmap_notify() as their IRQ notifier.

IIUC, your goal is to have the notifier inside napi, not irq_glue. For 
this, we'll have to have our own version of irq_cpu_rmap_add() (for the 
above reason).

sounds OK?

Re: [PATCH net-next v2 4/8] net: napi: add CPU affinity to napi->config

[Jakub Kicinski]

On Fri, 20 Dec 2024 13:14:48 -0700 Ahmed Zaki wrote:
> > Then you can probably register a single unified handler, and inside
> > that handler check if the device wanted to have rmap or just affinity?  
> 
> This is what is in this patch already, all drivers following new 
> approach will have netif_irq_cpu_rmap_notify() as their IRQ notifier.
> 
> IIUC, your goal is to have the notifier inside napi, not irq_glue. For 
> this, we'll have to have our own version of irq_cpu_rmap_add() (for the 
> above reason).
> 
> sounds OK?

Yes.

[PATCH net-next v2 5/8] bnxt: use napi's irq affinity

[Ahmed Zaki]

Delete the driver CPU affinity info and use the core's napi config
instead.

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 28 ++++-------------------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  2 --
 2 files changed, 4 insertions(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index ac729a25ba52..f68f07686105 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -11177,14 +11177,8 @@ static void bnxt_free_irq(struct bnxt *bp)
 		int map_idx = bnxt_cp_num_to_irq_num(bp, i);
 
 		irq = &bp->irq_tbl[map_idx];
-		if (irq->requested) {
-			if (irq->have_cpumask) {
-				irq_update_affinity_hint(irq->vector, NULL);
-				free_cpumask_var(irq->cpu_mask);
-				irq->have_cpumask = 0;
-			}
+		if (irq->requested)
 			free_irq(irq->vector, bp->bnapi[i]);
-		}
 
 		irq->requested = 0;
 	}
@@ -11213,26 +11207,12 @@ static int bnxt_request_irq(struct bnxt *bp)
 
 #ifdef CONFIG_RFS_ACCEL
 		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector,
-				   NAPIF_IRQ_ARFS_RMAP);
+				   NAPIF_IRQ_AFFINITY | NAPIF_IRQ_ARFS_RMAP);
 #else
-		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector, 0);
+		netif_napi_set_irq(&bp->bnapi[i]->napi, irq->vector,
+				   NAPIF_IRQ_AFFINITY);
 #endif
 		irq->requested = 1;
-
-		if (zalloc_cpumask_var(&irq->cpu_mask, GFP_KERNEL)) {
-			int numa_node = dev_to_node(&bp->pdev->dev);
-
-			irq->have_cpumask = 1;
-			cpumask_set_cpu(cpumask_local_spread(i, numa_node),
-					irq->cpu_mask);
-			rc = irq_update_affinity_hint(irq->vector, irq->cpu_mask);
-			if (rc) {
-				netdev_warn(bp->dev,
-					    "Update affinity hint failed, IRQ = %d\n",
-					    irq->vector);
-				break;
-			}
-		}
 	}
 	return rc;
 }
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 7df7a2233307..8a97c1fb2083 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -1228,9 +1228,7 @@ struct bnxt_irq {
 	irq_handler_t	handler;
 	unsigned int	vector;
 	u8		requested:1;
-	u8		have_cpumask:1;
 	char		name[IFNAMSIZ + BNXT_IRQ_NAME_EXTRA];
-	cpumask_var_t	cpu_mask;
 };
 
 #define HWRM_RING_ALLOC_TX	0x1
-- 
2.43.0

[PATCH net-next v2 6/8] ice: use napi's irq affinity

[Ahmed Zaki]

Delete the driver CPU affinity info and use the core's napi config
instead.

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h      |  3 --
 drivers/net/ethernet/intel/ice/ice_base.c |  7 ++--
 drivers/net/ethernet/intel/ice/ice_lib.c  | 11 ++----
 drivers/net/ethernet/intel/ice/ice_main.c | 44 -----------------------
 4 files changed, 5 insertions(+), 60 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 2f5d6f974185..0db665a6b38a 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -477,9 +477,6 @@ struct ice_q_vector {
 	struct ice_ring_container rx;
 	struct ice_ring_container tx;
 
-	cpumask_t affinity_mask;
-	struct irq_affinity_notify affinity_notify;
-
 	struct ice_channel *ch;
 
 	char name[ICE_INT_NAME_STR_LEN];
diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
index b2af8e3586f7..86cf715de00f 100644
--- a/drivers/net/ethernet/intel/ice/ice_base.c
+++ b/drivers/net/ethernet/intel/ice/ice_base.c
@@ -147,10 +147,6 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
 	q_vector->reg_idx = q_vector->irq.index;
 	q_vector->vf_reg_idx = q_vector->irq.index;
 
-	/* only set affinity_mask if the CPU is online */
-	if (cpu_online(v_idx))
-		cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
-
 	/* This will not be called in the driver load path because the netdev
 	 * will not be created yet. All other cases with register the NAPI
 	 * handler here (i.e. resume, reset/rebuild, etc.)
@@ -276,7 +272,8 @@ static void ice_cfg_xps_tx_ring(struct ice_tx_ring *ring)
 	if (test_and_set_bit(ICE_TX_XPS_INIT_DONE, ring->xps_state))
 		return;
 
-	netif_set_xps_queue(ring->netdev, &ring->q_vector->affinity_mask,
+	netif_set_xps_queue(ring->netdev,
+			    &ring->q_vector->napi.config->affinity_mask,
 			    ring->q_index);
 }
 
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 7c0b2d8e86ba..75cf5ce447cc 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -2589,12 +2589,6 @@ void ice_vsi_free_irq(struct ice_vsi *vsi)
 		      vsi->q_vectors[i]->num_ring_rx))
 			continue;
 
-		/* clear the affinity notifier in the IRQ descriptor */
-		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
-			irq_set_affinity_notifier(irq_num, NULL);
-
-		/* clear the affinity_hint in the IRQ descriptor */
-		irq_update_affinity_hint(irq_num, NULL);
 		synchronize_irq(irq_num);
 		devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
 	}
@@ -2737,9 +2731,10 @@ void ice_vsi_set_napi_queues(struct ice_vsi *vsi)
 
 #ifdef CONFIG_RFS_ACCEL
 		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq,
-				   NAPIF_IRQ_ARFS_RMAP);
+				   NAPIF_IRQ_AFFINITY | NAPIF_IRQ_ARFS_RMAP);
 #else
-		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq, 0);
+		netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq,
+				   NAPIF_IRQ_AFFINITY);
 #endif
 	}
 }
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 0ab35607e5d5..33118a435cd2 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -2504,34 +2504,6 @@ int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
 	return 0;
 }
 
-/**
- * ice_irq_affinity_notify - Callback for affinity changes
- * @notify: context as to what irq was changed
- * @mask: the new affinity mask
- *
- * This is a callback function used by the irq_set_affinity_notifier function
- * so that we may register to receive changes to the irq affinity masks.
- */
-static void
-ice_irq_affinity_notify(struct irq_affinity_notify *notify,
-			const cpumask_t *mask)
-{
-	struct ice_q_vector *q_vector =
-		container_of(notify, struct ice_q_vector, affinity_notify);
-
-	cpumask_copy(&q_vector->affinity_mask, mask);
-}
-
-/**
- * ice_irq_affinity_release - Callback for affinity notifier release
- * @ref: internal core kernel usage
- *
- * This is a callback function used by the irq_set_affinity_notifier function
- * to inform the current notification subscriber that they will no longer
- * receive notifications.
- */
-static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
-
 /**
  * ice_vsi_ena_irq - Enable IRQ for the given VSI
  * @vsi: the VSI being configured
@@ -2595,19 +2567,6 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 				   err);
 			goto free_q_irqs;
 		}
-
-		/* register for affinity change notifications */
-		if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
-			struct irq_affinity_notify *affinity_notify;
-
-			affinity_notify = &q_vector->affinity_notify;
-			affinity_notify->notify = ice_irq_affinity_notify;
-			affinity_notify->release = ice_irq_affinity_release;
-			irq_set_affinity_notifier(irq_num, affinity_notify);
-		}
-
-		/* assign the mask for this irq */
-		irq_update_affinity_hint(irq_num, &q_vector->affinity_mask);
 	}
 
 	err = ice_set_cpu_rx_rmap(vsi);
@@ -2623,9 +2582,6 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 free_q_irqs:
 	while (vector--) {
 		irq_num = vsi->q_vectors[vector]->irq.virq;
-		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
-			irq_set_affinity_notifier(irq_num, NULL);
-		irq_update_affinity_hint(irq_num, NULL);
 		devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
 	}
 	return err;
-- 
2.43.0

[PATCH net-next v2 7/8] idpf: use napi's irq affinity

[Ahmed Zaki]

Delete the driver CPU affinity info and use the core's napi config
instead.

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 19 +++++--------------
 drivers/net/ethernet/intel/idpf/idpf_txrx.h |  6 ++----
 2 files changed, 7 insertions(+), 18 deletions(-)

diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.c b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
index 34f4118c7bc0..5728f0bfc610 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.c
@@ -3554,8 +3554,6 @@ void idpf_vport_intr_rel(struct idpf_vport *vport)
 		q_vector->tx = NULL;
 		kfree(q_vector->rx);
 		q_vector->rx = NULL;
-
-		free_cpumask_var(q_vector->affinity_mask);
 	}
 
 	kfree(vport->q_vectors);
@@ -3582,8 +3580,6 @@ static void idpf_vport_intr_rel_irq(struct idpf_vport *vport)
 		vidx = vport->q_vector_idxs[vector];
 		irq_num = adapter->msix_entries[vidx].vector;
 
-		/* clear the affinity_mask in the IRQ descriptor */
-		irq_set_affinity_hint(irq_num, NULL);
 		kfree(free_irq(irq_num, q_vector));
 	}
 }
@@ -3762,8 +3758,6 @@ static int idpf_vport_intr_req_irq(struct idpf_vport *vport)
 				   "Request_irq failed, error: %d\n", err);
 			goto free_q_irqs;
 		}
-		/* assign the mask for this irq */
-		irq_set_affinity_hint(irq_num, q_vector->affinity_mask);
 	}
 
 	return 0;
@@ -4184,12 +4178,12 @@ static void idpf_vport_intr_napi_add_all(struct idpf_vport *vport)
 
 	for (v_idx = 0; v_idx < vport->num_q_vectors; v_idx++) {
 		struct idpf_q_vector *q_vector = &vport->q_vectors[v_idx];
+		int irq_num = vport->adapter->msix_entries[v_idx].vector;
 
-		netif_napi_add(vport->netdev, &q_vector->napi, napi_poll);
-
-		/* only set affinity_mask if the CPU is online */
-		if (cpu_online(v_idx))
-			cpumask_set_cpu(v_idx, q_vector->affinity_mask);
+		netif_napi_add_config(vport->netdev, &q_vector->napi,
+				      napi_poll, v_idx);
+		netif_napi_set_irq(&q_vector->napi,
+				   irq_num, NAPIF_IRQ_AFFINITY);
 	}
 }
 
@@ -4233,9 +4227,6 @@ int idpf_vport_intr_alloc(struct idpf_vport *vport)
 		q_vector->rx_intr_mode = IDPF_ITR_DYNAMIC;
 		q_vector->rx_itr_idx = VIRTCHNL2_ITR_IDX_0;
 
-		if (!zalloc_cpumask_var(&q_vector->affinity_mask, GFP_KERNEL))
-			goto error;
-
 		q_vector->tx = kcalloc(txqs_per_vector, sizeof(*q_vector->tx),
 				       GFP_KERNEL);
 		if (!q_vector->tx)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index 9c1fe84108ed..5efb3402b378 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -397,7 +397,6 @@ struct idpf_intr_reg {
  * @rx_intr_mode: Dynamic ITR or not
  * @rx_itr_idx: RX ITR index
  * @v_idx: Vector index
- * @affinity_mask: CPU affinity mask
  */
 struct idpf_q_vector {
 	__cacheline_group_begin_aligned(read_mostly);
@@ -434,13 +433,12 @@ struct idpf_q_vector {
 	__cacheline_group_begin_aligned(cold);
 	u16 v_idx;
 
-	cpumask_var_t affinity_mask;
 	__cacheline_group_end_aligned(cold);
 };
 libeth_cacheline_set_assert(struct idpf_q_vector, 112,
 			    24 + sizeof(struct napi_struct) +
 			    2 * sizeof(struct dim),
-			    8 + sizeof(cpumask_var_t));
+			    8);
 
 struct idpf_rx_queue_stats {
 	u64_stats_t packets;
@@ -934,7 +932,7 @@ static inline int idpf_q_vector_to_mem(const struct idpf_q_vector *q_vector)
 	if (!q_vector)
 		return NUMA_NO_NODE;
 
-	cpu = cpumask_first(q_vector->affinity_mask);
+	cpu = cpumask_first(&q_vector->napi.config->affinity_mask);
 
 	return cpu < nr_cpu_ids ? cpu_to_mem(cpu) : NUMA_NO_NODE;
 }
-- 
2.43.0

[PATCH net-next v2 8/8] mlx4: use napi's irq affinity

[Ahmed Zaki]

Delete the driver CPU affinity info and use the core's napi config
instead.

Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
---
 drivers/net/ethernet/mellanox/mlx4/en_cq.c    |  8 ++--
 .../net/ethernet/mellanox/mlx4/en_netdev.c    | 33 +--------------
 drivers/net/ethernet/mellanox/mlx4/eq.c       | 22 ----------
 drivers/net/ethernet/mellanox/mlx4/main.c     | 42 ++-----------------
 drivers/net/ethernet/mellanox/mlx4/mlx4.h     |  1 -
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h  |  1 -
 6 files changed, 10 insertions(+), 97 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_cq.c b/drivers/net/ethernet/mellanox/mlx4/en_cq.c
index b8531283e3ac..46e28e0bfcd0 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_cq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_cq.c
@@ -90,6 +90,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 			int cq_idx)
 {
 	struct mlx4_en_dev *mdev = priv->mdev;
+	struct napi_config *napi_config;
 	int irq, err = 0;
 	int timestamp_en = 0;
 	bool assigned_eq = false;
@@ -100,11 +101,12 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 	*cq->mcq.set_ci_db = 0;
 	*cq->mcq.arm_db    = 0;
 	memset(cq->buf, 0, cq->buf_size);
+	napi_config = cq->napi.config;
 
 	if (cq->type == RX) {
 		if (!mlx4_is_eq_vector_valid(mdev->dev, priv->port,
 					     cq->vector)) {
-			cq->vector = cpumask_first(priv->rx_ring[cq->ring]->affinity_mask);
+			cq->vector = cpumask_first(&napi_config->affinity_mask);
 
 			err = mlx4_assign_eq(mdev->dev, priv->port,
 					     &cq->vector);
@@ -150,7 +152,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 	case TX:
 		cq->mcq.comp = mlx4_en_tx_irq;
 		netif_napi_add_tx(cq->dev, &cq->napi, mlx4_en_poll_tx_cq);
-		netif_napi_set_irq(&cq->napi, irq, 0);
+		netif_napi_set_irq(&cq->napi, irq, NAPIF_IRQ_AFFINITY);
 		napi_enable(&cq->napi);
 		netif_queue_set_napi(cq->dev, cq_idx, NETDEV_QUEUE_TYPE_TX, &cq->napi);
 		break;
@@ -158,7 +160,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
 		cq->mcq.comp = mlx4_en_rx_irq;
 		netif_napi_add_config(cq->dev, &cq->napi, mlx4_en_poll_rx_cq,
 				      cq_idx);
-		netif_napi_set_irq(&cq->napi, irq, 0);
+		netif_napi_set_irq(&cq->napi, irq, NAPIF_IRQ_AFFINITY);
 		napi_enable(&cq->napi);
 		netif_queue_set_napi(cq->dev, cq_idx, NETDEV_QUEUE_TYPE_RX, &cq->napi);
 		break;
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
index 281b34af0bb4..e4c2532b5909 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
@@ -1596,24 +1596,6 @@ static void mlx4_en_linkstate_work(struct work_struct *work)
 	mutex_unlock(&mdev->state_lock);
 }
 
-static int mlx4_en_init_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
-{
-	struct mlx4_en_rx_ring *ring = priv->rx_ring[ring_idx];
-	int numa_node = priv->mdev->dev->numa_node;
-
-	if (!zalloc_cpumask_var(&ring->affinity_mask, GFP_KERNEL))
-		return -ENOMEM;
-
-	cpumask_set_cpu(cpumask_local_spread(ring_idx, numa_node),
-			ring->affinity_mask);
-	return 0;
-}
-
-static void mlx4_en_free_affinity_hint(struct mlx4_en_priv *priv, int ring_idx)
-{
-	free_cpumask_var(priv->rx_ring[ring_idx]->affinity_mask);
-}
-
 static void mlx4_en_init_recycle_ring(struct mlx4_en_priv *priv,
 				      int tx_ring_idx)
 {
@@ -1663,16 +1645,9 @@ int mlx4_en_start_port(struct net_device *dev)
 	for (i = 0; i < priv->rx_ring_num; i++) {
 		cq = priv->rx_cq[i];
 
-		err = mlx4_en_init_affinity_hint(priv, i);
-		if (err) {
-			en_err(priv, "Failed preparing IRQ affinity hint\n");
-			goto cq_err;
-		}
-
 		err = mlx4_en_activate_cq(priv, cq, i);
 		if (err) {
 			en_err(priv, "Failed activating Rx CQ\n");
-			mlx4_en_free_affinity_hint(priv, i);
 			goto cq_err;
 		}
 
@@ -1688,7 +1663,6 @@ int mlx4_en_start_port(struct net_device *dev)
 		if (err) {
 			en_err(priv, "Failed setting cq moderation parameters\n");
 			mlx4_en_deactivate_cq(priv, cq);
-			mlx4_en_free_affinity_hint(priv, i);
 			goto cq_err;
 		}
 		mlx4_en_arm_cq(priv, cq);
@@ -1874,10 +1848,9 @@ int mlx4_en_start_port(struct net_device *dev)
 mac_err:
 	mlx4_en_put_qp(priv);
 cq_err:
-	while (rx_index--) {
+	while (rx_index--)
 		mlx4_en_deactivate_cq(priv, priv->rx_cq[rx_index]);
-		mlx4_en_free_affinity_hint(priv, rx_index);
-	}
+
 	for (i = 0; i < priv->rx_ring_num; i++)
 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
 
@@ -2011,8 +1984,6 @@ void mlx4_en_stop_port(struct net_device *dev, int detach)
 		napi_synchronize(&cq->napi);
 		mlx4_en_deactivate_rx_ring(priv, priv->rx_ring[i]);
 		mlx4_en_deactivate_cq(priv, cq);
-
-		mlx4_en_free_affinity_hint(priv, i);
 	}
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx4/eq.c b/drivers/net/ethernet/mellanox/mlx4/eq.c
index d768a6a828c4..b005eb697c64 100644
--- a/drivers/net/ethernet/mellanox/mlx4/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/eq.c
@@ -233,23 +233,6 @@ static void mlx4_slave_event(struct mlx4_dev *dev, int slave,
 	slave_event(dev, slave, eqe);
 }
 
-#if defined(CONFIG_SMP)
-static void mlx4_set_eq_affinity_hint(struct mlx4_priv *priv, int vec)
-{
-	int hint_err;
-	struct mlx4_dev *dev = &priv->dev;
-	struct mlx4_eq *eq = &priv->eq_table.eq[vec];
-
-	if (!cpumask_available(eq->affinity_mask) ||
-	    cpumask_empty(eq->affinity_mask))
-		return;
-
-	hint_err = irq_update_affinity_hint(eq->irq, eq->affinity_mask);
-	if (hint_err)
-		mlx4_warn(dev, "irq_update_affinity_hint failed, err %d\n", hint_err);
-}
-#endif
-
 int mlx4_gen_pkey_eqe(struct mlx4_dev *dev, int slave, u8 port)
 {
 	struct mlx4_eqe eqe;
@@ -1123,8 +1106,6 @@ static void mlx4_free_irqs(struct mlx4_dev *dev)
 
 	for (i = 0; i < dev->caps.num_comp_vectors + 1; ++i)
 		if (eq_table->eq[i].have_irq) {
-			free_cpumask_var(eq_table->eq[i].affinity_mask);
-			irq_update_affinity_hint(eq_table->eq[i].irq, NULL);
 			free_irq(eq_table->eq[i].irq, eq_table->eq + i);
 			eq_table->eq[i].have_irq = 0;
 		}
@@ -1516,9 +1497,6 @@ int mlx4_assign_eq(struct mlx4_dev *dev, u8 port, int *vector)
 			clear_bit(*prequested_vector, priv->msix_ctl.pool_bm);
 			*prequested_vector = -1;
 		} else {
-#if defined(CONFIG_SMP)
-			mlx4_set_eq_affinity_hint(priv, *prequested_vector);
-#endif
 			eq_set_ci(&priv->eq_table.eq[*prequested_vector], 1);
 			priv->eq_table.eq[*prequested_vector].have_irq = 1;
 		}
diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c
index febeadfdd5a5..5f88c297332f 100644
--- a/drivers/net/ethernet/mellanox/mlx4/main.c
+++ b/drivers/net/ethernet/mellanox/mlx4/main.c
@@ -2923,36 +2923,6 @@ static int mlx4_setup_hca(struct mlx4_dev *dev)
 	return err;
 }
 
-static int mlx4_init_affinity_hint(struct mlx4_dev *dev, int port, int eqn)
-{
-	int requested_cpu = 0;
-	struct mlx4_priv *priv = mlx4_priv(dev);
-	struct mlx4_eq *eq;
-	int off = 0;
-	int i;
-
-	if (eqn > dev->caps.num_comp_vectors)
-		return -EINVAL;
-
-	for (i = 1; i < port; i++)
-		off += mlx4_get_eqs_per_port(dev, i);
-
-	requested_cpu = eqn - off - !!(eqn > MLX4_EQ_ASYNC);
-
-	/* Meaning EQs are shared, and this call comes from the second port */
-	if (requested_cpu < 0)
-		return 0;
-
-	eq = &priv->eq_table.eq[eqn];
-
-	if (!zalloc_cpumask_var(&eq->affinity_mask, GFP_KERNEL))
-		return -ENOMEM;
-
-	cpumask_set_cpu(requested_cpu, eq->affinity_mask);
-
-	return 0;
-}
-
 static void mlx4_enable_msi_x(struct mlx4_dev *dev)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
@@ -2997,19 +2967,13 @@ static void mlx4_enable_msi_x(struct mlx4_dev *dev)
 			priv->eq_table.eq[i].irq =
 				entries[i + 1 - !!(i > MLX4_EQ_ASYNC)].vector;
 
-			if (MLX4_IS_LEGACY_EQ_MODE(dev->caps)) {
+			if (MLX4_IS_LEGACY_EQ_MODE(dev->caps))
 				bitmap_fill(priv->eq_table.eq[i].actv_ports.ports,
 					    dev->caps.num_ports);
-				/* We don't set affinity hint when there
-				 * aren't enough EQs
-				 */
-			} else {
+			else
 				set_bit(port,
 					priv->eq_table.eq[i].actv_ports.ports);
-				if (mlx4_init_affinity_hint(dev, port + 1, i))
-					mlx4_warn(dev, "Couldn't init hint cpumask for EQ %d\n",
-						  i);
-			}
+
 			/* We divide the Eqs evenly between the two ports.
 			 * (dev->caps.num_comp_vectors / dev->caps.num_ports)
 			 * refers to the number of Eqs per port
diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4.h b/drivers/net/ethernet/mellanox/mlx4/mlx4.h
index d7d856d1758a..66b1ebfd5816 100644
--- a/drivers/net/ethernet/mellanox/mlx4/mlx4.h
+++ b/drivers/net/ethernet/mellanox/mlx4/mlx4.h
@@ -403,7 +403,6 @@ struct mlx4_eq {
 	struct mlx4_eq_tasklet	tasklet_ctx;
 	struct mlx4_active_ports actv_ports;
 	u32			ref_count;
-	cpumask_var_t		affinity_mask;
 };
 
 struct mlx4_slave_eqe {
diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
index 28b70dcc652e..41594dd00636 100644
--- a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
+++ b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
@@ -357,7 +357,6 @@ struct mlx4_en_rx_ring {
 	unsigned long dropped;
 	unsigned long alloc_fail;
 	int hwtstamp_rx_filter;
-	cpumask_var_t affinity_mask;
 	struct xdp_rxq_info xdp_rxq;
 };
 
-- 
2.43.0

Re: [PATCH net-next v2 0/8] net: napi: add CPU affinity to napi->config

[Shay Drori]

On 18/12/2024 18:58, Ahmed Zaki wrote:
> External email: Use caution opening links or attachments
> 
> 
> Move the IRQ affinity management to the napi struct. All drivers that are
> already using netif_napi_set_irq() are modified to the new API. Except
> mlx5 because it is implementing IRQ pools and moving to the new API does
> not seem trivial.
> 
> Tested on bnxt, ice and idpf.
> ---
> Opens: is cpu_online_mask the best default mask? drivers do this differently

cpu_online_mask is not the best default mask for IRQ affinity management.
Here are two reasons:
- Performance Gains from Driver-Specific CPU Assignment: Many drivers
   assign different CPUs to each IRQ to optimize performance. This plays
a crucial role in CPU utilization.
- Impact of NUMA Node Distance on Traffic Performance: NUMA topology
   plays a crucial role in IRQ performance. Assigning IRQs to CPUs on
   the same NUMA node as the associated device minimizes latency caused
   by remote memory access.[1]

[1]
for more details on NUMA preference, you can look at commit 
2acda57736de1e486036b90a648e67a3599080a1


> 
> v2:
>      - Also move the ARFS IRQ affinity management from drivers to core. Via
>        netif_napi_set_irq(), drivers can ask the core to add the IRQ to the
>        ARFS rmap (already allocated by the driver).
> 
> RFC -> v1:
>      - https://lore.kernel.org/netdev/20241210002626.366878-1-ahmed.zaki@intel.com/
>      - move static inline affinity functions to net/dev/core.c
>      - add the new napi->irq_flags (patch 1)
>      - add code changes to bnxt, mlx4 and ice.
> 
> Ahmed Zaki (8):
>    net: napi: add irq_flags to napi struct
>    net: allow ARFS rmap management in core
>    lib: cpu_rmap: allow passing a notifier callback
>    net: napi: add CPU affinity to napi->config
>    bnxt: use napi's irq affinity
>    ice: use napi's irq affinity
>    idpf: use napi's irq affinity
>    mlx4: use napi's irq affinity
> 
>   drivers/net/ethernet/amazon/ena/ena_netdev.c  | 21 ++---
>   drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 51 +++--------
>   drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  2 -
>   drivers/net/ethernet/broadcom/tg3.c           |  2 +-
>   drivers/net/ethernet/cisco/enic/enic_main.c   |  3 +-
>   drivers/net/ethernet/google/gve/gve_utils.c   |  2 +-
>   .../net/ethernet/hisilicon/hns3/hns3_enet.c   |  2 +-
>   drivers/net/ethernet/intel/e1000/e1000_main.c |  2 +-
>   drivers/net/ethernet/intel/e1000e/netdev.c    |  2 +-
>   drivers/net/ethernet/intel/ice/ice.h          |  3 -
>   drivers/net/ethernet/intel/ice/ice_arfs.c     | 10 +--
>   drivers/net/ethernet/intel/ice/ice_base.c     |  7 +-
>   drivers/net/ethernet/intel/ice/ice_lib.c      | 14 +--
>   drivers/net/ethernet/intel/ice/ice_main.c     | 44 ----------
>   drivers/net/ethernet/intel/idpf/idpf_txrx.c   | 19 ++--
>   drivers/net/ethernet/intel/idpf/idpf_txrx.h   |  6 +-
>   drivers/net/ethernet/mellanox/mlx4/en_cq.c    |  8 +-
>   .../net/ethernet/mellanox/mlx4/en_netdev.c    | 33 +------
>   drivers/net/ethernet/mellanox/mlx4/eq.c       | 24 +----
>   drivers/net/ethernet/mellanox/mlx4/main.c     | 42 +--------
>   drivers/net/ethernet/mellanox/mlx4/mlx4.h     |  1 -
>   drivers/net/ethernet/mellanox/mlx4/mlx4_en.h  |  1 -
>   .../net/ethernet/mellanox/mlx5/core/en_main.c |  2 +-
>   .../net/ethernet/mellanox/mlx5/core/pci_irq.c |  2 +-
>   drivers/net/ethernet/meta/fbnic/fbnic_txrx.c  |  3 +-
>   drivers/net/ethernet/qlogic/qede/qede_main.c  | 28 +++---
>   drivers/net/ethernet/sfc/falcon/efx.c         |  9 ++
>   drivers/net/ethernet/sfc/falcon/nic.c         | 10 ---
>   drivers/net/ethernet/sfc/nic.c                |  2 +-
>   drivers/net/ethernet/sfc/siena/efx_channels.c |  9 ++
>   drivers/net/ethernet/sfc/siena/nic.c          | 10 ---
>   include/linux/cpu_rmap.h                      | 13 ++-
>   include/linux/netdevice.h                     | 23 ++++-
>   lib/cpu_rmap.c                                | 20 ++---
>   net/core/dev.c                                | 87 ++++++++++++++++++-
>   35 files changed, 215 insertions(+), 302 deletions(-)
> 
> --
> 2.43.0
> 

Re: [PATCH net-next v2 0/8] net: napi: add CPU affinity to napi->config

[Ahmed Zaki]

On 2024-12-22 2:23 a.m., Shay Drori wrote:
> 
> 
> On 18/12/2024 18:58, Ahmed Zaki wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> Move the IRQ affinity management to the napi struct. All drivers that are
>> already using netif_napi_set_irq() are modified to the new API. Except
>> mlx5 because it is implementing IRQ pools and moving to the new API does
>> not seem trivial.
>>
>> Tested on bnxt, ice and idpf.
>> ---
>> Opens: is cpu_online_mask the best default mask? drivers do this 
>> differently
> 
> cpu_online_mask is not the best default mask for IRQ affinity management.
> Here are two reasons:
> - Performance Gains from Driver-Specific CPU Assignment: Many drivers
>    assign different CPUs to each IRQ to optimize performance. This plays
> a crucial role in CPU utilization.
> - Impact of NUMA Node Distance on Traffic Performance: NUMA topology
>    plays a crucial role in IRQ performance. Assigning IRQs to CPUs on
>    the same NUMA node as the associated device minimizes latency caused
>    by remote memory access.[1]
> 
> [1]
> for more details on NUMA preference, you can look at commit 
> 2acda57736de1e486036b90a648e67a3599080a1
> 

Thanks for replying.

I will use cpumask_local_spread() (which now considers NUMA distances) 
in the next iteration.