[PATCH net-next v5 00/10] bng_en: add link management and statistics support

[PATCH net-next v5 00/10] bng_en: add link management and statistics support

[Bhargava Marreddy]

Hi,

This series enhances the bng_en driver by adding:
1. Link/PHY support
   a. Link query
   b. Async Link events
   c. Ethtool link set/get functionality
2. Hardware statistics reporting via ethtool -S

This version incorporates feedback received prior to splitting the
original series into two parts. 

 Addressed following comments from Andrew Lunn
   - Apply Rev-xmas fix in several places.
   - Correct ethtool-speed comment to reflect accurate behavior.
   https://lore.kernel.org/netdev/49930724-74b8-41fe-8f5c-482afc976b82@lunn.ch/
   https://lore.kernel.org/netdev/CANXQDtb5XuLKOOorCMYDUpVz6aFuQgvmQZ4pS6RJGkAgeM8n1A@mail.gmail.com/

 Addressed following comments from Jakub Kicinski
   - Took care of missing mutex_init()
   https://lore.kernel.org/netdev/20251127190855.69779791@kernel.org/#t
   https://lore.kernel.org/netdev/CANXQDtYy2JziaaVi=Cqt+gAJNt4NjFFs8NbhrS=RBV61ORwinA@mail.gmail.com/

 Addressed following comments from ALOK TIWARI
   - Add newline for clarity/formatting.
   https://lore.kernel.org/netdev/CANXQDtZbLhdwEtN+kKg1OVLV+uy-gsNfX+rU4MR77QPuO6_y9A@mail.gmail.com/

v5:
 Address Jakub Kicinski's comments by holding the instance lock across
 the entire service task (instead of flag-based locking) and checking
 the OPEN bit under the lock to resolve the close-path deadlock. 

v4:
 Addressed comments from Jakub Kicinski:
  - Explicitly size speed mask arrays to remove the need for manual
    sentinels in the hardware configuration.
  - Fix an off-by-one error in the queue index bound check.
  - Standardize ethtool -S per-ring stat naming to "rxq$id_$name"
    (e.g., rxq0_errors). Removed spaces and brackets to ensure
    compatibility with parsing scripts.
  - Exclude ethtool -S global counters (tx_err, tx_fifo_underruns,
    tx_total_collisions) already covered by rtnl_link_stats64.

v3:
 Addressed comments from Andrew Lunn:
  - Fix a comment and refactor bnge_hwrm_set_pause() 
    to improve code readability.

 Addressed comments from Jakub Kicinski:
  - Switch to RCU for stats synchronization and drop the 
    BNGE_STATE_READ_STATS state bit.

 Fix checkpatch warning by switching to kzalloc_node(). 

v2:
 Addressed comments from Andrew Lunn:
  - Add implementation for .get_pauseparam and .set_pauseparam to support
    a full flow control configuration, covering global autoneg, independent
    pause autoneg, and forced RX/TX pause functionality.
  - Move MAC, PHY, Control, Pause, and RMON statistics from
    .get_ethtool_stats to their respective structured ethtool
    callbacks (e.g., .get_eth_mac_stats, .get_rmon_stats).
 
 Addressed comments from Jakub Kicinski:
  - Fix potential out-of-bounds read in speed mask array by adding 
    the missing sentinel.
  - Simplify timer synchronization by removing the redundant intr_sem.
  - Replace link_lock with the netdev instance lock, as it is sufficient
    to handle the required synchronization.

v1: https://lore.kernel.org/all/20260223163641.755200-1-bhargava.marreddy@broadcom.com/

Bhargava Marreddy (10):
  bng_en: add per-PF workqueue, timer, and slow-path task
  bng_en: query PHY capabilities and report link status
  bng_en: add ethtool link settings, get_link, and nway_reset
  bng_en: implement ethtool pauseparam operations
  bng_en: add support for link async events
  bng_en: add HW stats infra and structured ethtool ops
  bng_en: periodically fetch and accumulate hardware statistics
  bng_en: implement ndo_get_stats64
  bng_en: implement netdev_stat_ops
  bng_en: add support for ethtool -S stats display

 drivers/net/ethernet/broadcom/bnge/Makefile   |    3 +-
 drivers/net/ethernet/broadcom/bnge/bnge.h     |   12 +
 .../net/ethernet/broadcom/bnge/bnge_core.c    |    1 +
 .../net/ethernet/broadcom/bnge/bnge_ethtool.c |  808 ++++++++++
 .../ethernet/broadcom/bnge/bnge_hwrm_lib.c    |  388 ++++-
 .../ethernet/broadcom/bnge/bnge_hwrm_lib.h    |    8 +
 .../net/ethernet/broadcom/bnge/bnge_link.c    | 1300 +++++++++++++++++
 .../net/ethernet/broadcom/bnge/bnge_link.h    |  204 +++
 .../net/ethernet/broadcom/bnge/bnge_netdev.c  |  632 +++++++-
 .../net/ethernet/broadcom/bnge/bnge_netdev.h  |  119 +-
 .../net/ethernet/broadcom/bnge/bnge_txrx.c    |   35 +-
 11 files changed, 3491 insertions(+), 19 deletions(-)
 create mode 100644 drivers/net/ethernet/broadcom/bnge/bnge_link.c
 create mode 100644 drivers/net/ethernet/broadcom/bnge/bnge_link.h

-- 
2.47.3

[PATCH net-next v5 01/10] bng_en: add per-PF workqueue, timer, and slow-path task

[Bhargava Marreddy]

Add a dedicated single-thread workqueue and a timer for each PF
to drive deferred slow-path work such as link event handling and
stats collection. The timer is stopped via timer_delete_sync()
when interrupts are disabled and restarted on open. The open and
close paths now start and drain these resources.

Signed-off-by: Bhargava Marreddy <bhargava.marreddy@broadcom.com>
Reviewed-by: Vikas Gupta <vikas.gupta@broadcom.com>
Reviewed-by: Ajit Kumar Khaparde <ajit.khaparde@broadcom.com>
---
 .../net/ethernet/broadcom/bnge/bnge_netdev.c  | 74 ++++++++++++++++++-
 .../net/ethernet/broadcom/bnge/bnge_netdev.h  | 12 ++-
 2 files changed, 84 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnge/bnge_netdev.c b/drivers/net/ethernet/broadcom/bnge/bnge_netdev.c
index a20dc3ca640c..edbb42efb70b 100644
--- a/drivers/net/ethernet/broadcom/bnge/bnge_netdev.c
+++ b/drivers/net/ethernet/broadcom/bnge/bnge_netdev.c
@@ -101,6 +101,36 @@ static int bnge_alloc_ring_stats(struct bnge_net *bn)
 	return rc;
 }
 
+static void bnge_timer(struct timer_list *t)
+{
+	struct bnge_net *bn = timer_container_of(bn, t, timer);
+	struct bnge_dev *bd = bn->bd;
+
+	if (!netif_running(bn->netdev) ||
+	    !test_bit(BNGE_STATE_OPEN, &bd->state))
+		return;
+
+	/* Periodic work added by later patches */
+
+	mod_timer(&bn->timer, jiffies + bn->current_interval);
+}
+
+static void bnge_sp_task(struct work_struct *work)
+{
+	struct bnge_net *bn = container_of(work, struct bnge_net, sp_task);
+	struct bnge_dev *bd = bn->bd;
+
+	netdev_lock(bn->netdev);
+	if (!test_bit(BNGE_STATE_OPEN, &bd->state)) {
+		netdev_unlock(bn->netdev);
+		return;
+	}
+
+	/* Event handling work added by later patches */
+
+	netdev_unlock(bn->netdev);
+}
+
 static void bnge_free_nq_desc_arr(struct bnge_nq_ring_info *nqr)
 {
 	struct bnge_ring_struct *ring = &nqr->ring_struct;
@@ -2507,6 +2537,9 @@ static int bnge_open_core(struct bnge_net *bn)
 	bnge_enable_int(bn);
 
 	bnge_tx_enable(bn);
+
+	mod_timer(&bn->timer, jiffies + bn->current_interval);
+
 	return 0;
 
 err_free_irq:
@@ -2542,6 +2575,8 @@ static void bnge_close_core(struct bnge_net *bn)
 	bnge_tx_disable(bn);
 
 	clear_bit(BNGE_STATE_OPEN, &bd->state);
+
+	timer_delete_sync(&bn->timer);
 	bnge_shutdown_nic(bn);
 	bnge_disable_napi(bn);
 	bnge_free_all_rings_bufs(bn);
@@ -2690,6 +2725,23 @@ static void bnge_init_ring_params(struct bnge_net *bn)
 	bn->netdev->cfg->hds_thresh = max(BNGE_DEFAULT_RX_COPYBREAK, rx_size);
 }
 
+static struct workqueue_struct *
+bnge_create_workqueue_thread(struct bnge_dev *bd, const char *thread_name)
+{
+	struct workqueue_struct *wq;
+	char *wq_name;
+
+	wq_name = kasprintf(GFP_KERNEL, "%s-%s", thread_name,
+			    dev_name(bd->dev));
+	if (!wq_name)
+		return NULL;
+
+	wq = create_singlethread_workqueue(wq_name);
+	kfree(wq_name);
+
+	return wq;
+}
+
 int bnge_netdev_alloc(struct bnge_dev *bd, int max_irqs)
 {
 	struct net_device *netdev;
@@ -2774,6 +2826,17 @@ int bnge_netdev_alloc(struct bnge_dev *bd, int max_irqs)
 	if (bd->tso_max_segs)
 		netif_set_tso_max_segs(netdev, bd->tso_max_segs);
 
+	INIT_WORK(&bn->sp_task, bnge_sp_task);
+	timer_setup(&bn->timer, bnge_timer, 0);
+	bn->current_interval = BNGE_TIMER_INTERVAL;
+
+	bn->bnge_pf_wq = bnge_create_workqueue_thread(bd, "bnge_pf_wq");
+	if (!bn->bnge_pf_wq) {
+		netdev_err(netdev, "Unable to create workqueue.\n");
+		rc = -ENOMEM;
+		goto err_netdev;
+	}
+
 	bn->rx_ring_size = BNGE_DEFAULT_RX_RING_SIZE;
 	bn->tx_ring_size = BNGE_DEFAULT_TX_RING_SIZE;
 	bn->rx_dir = DMA_FROM_DEVICE;
@@ -2789,11 +2852,13 @@ int bnge_netdev_alloc(struct bnge_dev *bd, int max_irqs)
 	rc = register_netdev(netdev);
 	if (rc) {
 		dev_err(bd->dev, "Register netdev failed rc: %d\n", rc);
-		goto err_netdev;
+		goto err_free_workq;
 	}
 
 	return 0;
 
+err_free_workq:
+	destroy_workqueue(bn->bnge_pf_wq);
 err_netdev:
 	free_netdev(netdev);
 	return rc;
@@ -2802,8 +2867,15 @@ int bnge_netdev_alloc(struct bnge_dev *bd, int max_irqs)
 void bnge_netdev_free(struct bnge_dev *bd)
 {
 	struct net_device *netdev = bd->netdev;
+	struct bnge_net *bn;
+
+	bn = netdev_priv(netdev);
 
 	unregister_netdev(netdev);
+
+	cancel_work_sync(&bn->sp_task);
+	destroy_workqueue(bn->bnge_pf_wq);
+
 	free_netdev(netdev);
 	bd->netdev = NULL;
 }
diff --git a/drivers/net/ethernet/broadcom/bnge/bnge_netdev.h b/drivers/net/ethernet/broadcom/bnge/bnge_netdev.h
index 70f1a7c24814..d2ccee725454 100644
--- a/drivers/net/ethernet/broadcom/bnge/bnge_netdev.h
+++ b/drivers/net/ethernet/broadcom/bnge/bnge_netdev.h
@@ -224,6 +224,12 @@ struct bnge_tpa_info {
 #define BNGE_NQ_HDL_TYPE(hdl)	(((hdl) & BNGE_NQ_HDL_TYPE_MASK) >>	\
 				 BNGE_NQ_HDL_TYPE_SHIFT)
 
+enum bnge_net_state {
+	BNGE_STATE_NAPI_DISABLED,
+};
+
+#define BNGE_TIMER_INTERVAL	HZ
+
 struct bnge_net {
 	struct bnge_dev		*bd;
 	struct net_device	*netdev;
@@ -281,13 +287,17 @@ struct bnge_net {
 	u32			stats_coal_ticks;
 
 	unsigned long		state;
-#define BNGE_STATE_NAPI_DISABLED	0
 
 	u32			msg_enable;
 	u16			max_tpa;
 	__be16			vxlan_port;
 	__be16			nge_port;
 	__be16			vxlan_gpe_port;
+
+	unsigned int		current_interval;
+	struct timer_list	timer;
+	struct workqueue_struct *bnge_pf_wq;
+	struct work_struct	sp_task;
 };
 
 #define BNGE_DEFAULT_RX_RING_SIZE	511
-- 
2.47.3