Merge: qed/qede: Fix scheduling while atomic

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/4930

JIRA: https://issues.redhat.com/browse/RHEL-9751

Statistics read through bond interface via sysfs causes
below bug and traces as it triggers the bonding module to
collect the slave device statistics while holding the spinlock,
beneath that qede->qed driver statistics flow gets scheduled out
due to usleep_range() used in PTT acquire logic

Fix this by collecting the statistics asynchronously from a periodic
delayed work scheduled at default stats coalescing interval and return
the recent copy of statisitcs from .ndo_get_stats64(), also add ability
to configure/retrieve stats coalescing interval using below commands -
    
ethtool -C ethx stats-block-usecs <val>
ethtool -c ethx
    
Signed-off-by: John Meneghini <jmeneghi@redhat.com>

Approved-by: Ewan D. Milne <emilne@redhat.com>
Approved-by: Maurizio Lombardi <mlombard@redhat.com>
Approved-by: Michal Schmidt <mschmidt@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: Lucas Zampieri <lzampier@redhat.com>

Author: zampierilucas

drivers/net/ethernet/qlogic/qed/qed_l2.c

@@ -1903,7 +1903,7 @@ void qed_get_vport_stats(struct qed_dev *cdev, struct qed_eth_stats *stats)
 {
 	u32 i;
 
-	if (!cdev) {
+	if (!cdev || cdev->recov_in_prog) {
 		memset(stats, 0, sizeof(*stats));
 		return;
 	}

drivers/net/ethernet/qlogic/qede/qede.h

@@ -271,6 +271,10 @@ struct qede_dev {
 #define QEDE_ERR_WARN			3
 
 	struct qede_dump_info		dump_info;
+	struct delayed_work		periodic_task;
+	unsigned long			stats_coal_ticks;
+	u32				stats_coal_usecs;
+	spinlock_t			stats_lock; /* lock for vport stats access */
 };
 
 enum QEDE_STATE {

drivers/net/ethernet/qlogic/qede/qede_ethtool.c

@@ -430,6 +430,8 @@ static void qede_get_ethtool_stats(struct net_device *dev,
 		}
 	}
 
+	spin_lock(&edev->stats_lock);
+
 	for (i = 0; i < QEDE_NUM_STATS; i++) {
 		if (qede_is_irrelevant_stat(edev, i))
 			continue;
@@ -439,6 +441,8 @@ static void qede_get_ethtool_stats(struct net_device *dev,
 		buf++;
 	}
 
+	spin_unlock(&edev->stats_lock);
+
 	__qede_unlock(edev);
 }
 
@@ -830,6 +834,7 @@ static int qede_get_coalesce(struct net_device *dev,
 
 	coal->rx_coalesce_usecs = rx_coal;
 	coal->tx_coalesce_usecs = tx_coal;
+	coal->stats_block_coalesce_usecs = edev->stats_coal_usecs;
 
 	return rc;
 }
@@ -843,6 +848,19 @@ int qede_set_coalesce(struct net_device *dev, struct ethtool_coalesce *coal,
 	int i, rc = 0;
 	u16 rxc, txc;
 
+	if (edev->stats_coal_usecs != coal->stats_block_coalesce_usecs) {
+		edev->stats_coal_usecs = coal->stats_block_coalesce_usecs;
+		if (edev->stats_coal_usecs) {
+			edev->stats_coal_ticks = usecs_to_jiffies(edev->stats_coal_usecs);
+			schedule_delayed_work(&edev->periodic_task, 0);
+
+			DP_INFO(edev, "Configured stats coal ticks=%lu jiffies\n",
+				edev->stats_coal_ticks);
+		} else {
+			cancel_delayed_work_sync(&edev->periodic_task);
+		}
+	}
+
 	if (!netif_running(dev)) {
 		DP_INFO(edev, "Interface is down\n");
 		return -EINVAL;
@@ -2255,7 +2273,8 @@ static int qede_get_per_coalesce(struct net_device *dev,
 }
 
 static const struct ethtool_ops qede_ethtool_ops = {
-	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS,
+	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS |
+					  ETHTOOL_COALESCE_STATS_BLOCK_USECS,
 	.get_link_ksettings		= qede_get_link_ksettings,
 	.set_link_ksettings		= qede_set_link_ksettings,
 	.get_drvinfo			= qede_get_drvinfo,
@@ -2306,7 +2325,8 @@ static const struct ethtool_ops qede_ethtool_ops = {
 };
 
 static const struct ethtool_ops qede_vf_ethtool_ops = {
-	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS,
+	.supported_coalesce_params	= ETHTOOL_COALESCE_USECS |
+					  ETHTOOL_COALESCE_STATS_BLOCK_USECS,
 	.get_link_ksettings		= qede_get_link_ksettings,
 	.get_drvinfo			= qede_get_drvinfo,
 	.get_msglevel			= qede_get_msglevel,

drivers/net/ethernet/qlogic/qede/qede_main.c

@@ -317,6 +317,8 @@ void qede_fill_by_demand_stats(struct qede_dev *edev)
 
 	edev->ops->get_vport_stats(edev->cdev, &stats);
 
+	spin_lock(&edev->stats_lock);
+
 	p_common->no_buff_discards = stats.common.no_buff_discards;
 	p_common->packet_too_big_discard = stats.common.packet_too_big_discard;
 	p_common->ttl0_discard = stats.common.ttl0_discard;
@@ -414,6 +416,8 @@ void qede_fill_by_demand_stats(struct qede_dev *edev)
 		p_ah->tx_1519_to_max_byte_packets =
 		    stats.ah.tx_1519_to_max_byte_packets;
 	}
+
+	spin_unlock(&edev->stats_lock);
 }
 
 static void qede_get_stats64(struct net_device *dev,
@@ -422,9 +426,10 @@ static void qede_get_stats64(struct net_device *dev,
 	struct qede_dev *edev = netdev_priv(dev);
 	struct qede_stats_common *p_common;
 
-	qede_fill_by_demand_stats(edev);
 	p_common = &edev->stats.common;
 
+	spin_lock(&edev->stats_lock);
+
 	stats->rx_packets = p_common->rx_ucast_pkts + p_common->rx_mcast_pkts +
 			    p_common->rx_bcast_pkts;
 	stats->tx_packets = p_common->tx_ucast_pkts + p_common->tx_mcast_pkts +
@@ -444,6 +449,8 @@ static void qede_get_stats64(struct net_device *dev,
 		stats->collisions = edev->stats.bb.tx_total_collisions;
 	stats->rx_crc_errors = p_common->rx_crc_errors;
 	stats->rx_frame_errors = p_common->rx_align_errors;
+
+	spin_unlock(&edev->stats_lock);
 }
 
 #ifdef CONFIG_QED_SRIOV
@@ -1073,6 +1080,23 @@ static void qede_unlock(struct qede_dev *edev)
 	rtnl_unlock();
 }
 
+static void qede_periodic_task(struct work_struct *work)
+{
+	struct qede_dev *edev = container_of(work, struct qede_dev,
+					     periodic_task.work);
+
+	qede_fill_by_demand_stats(edev);
+	schedule_delayed_work(&edev->periodic_task, edev->stats_coal_ticks);
+}
+
+static void qede_init_periodic_task(struct qede_dev *edev)
+{
+	INIT_DELAYED_WORK(&edev->periodic_task, qede_periodic_task);
+	spin_lock_init(&edev->stats_lock);
+	edev->stats_coal_usecs = USEC_PER_SEC;
+	edev->stats_coal_ticks = usecs_to_jiffies(USEC_PER_SEC);
+}
+
 static void qede_sp_task(struct work_struct *work)
 {
 	struct qede_dev *edev = container_of(work, struct qede_dev,
@@ -1092,6 +1116,7 @@ static void qede_sp_task(struct work_struct *work)
 	 */
 
 	if (test_and_clear_bit(QEDE_SP_RECOVERY, &edev->sp_flags)) {
+		cancel_delayed_work_sync(&edev->periodic_task);
 #ifdef CONFIG_QED_SRIOV
 		/* SRIOV must be disabled outside the lock to avoid a deadlock.
 		 * The recovery of the active VFs is currently not supported.
@@ -1282,6 +1307,7 @@ static int __qede_probe(struct pci_dev *pdev, u32 dp_module, u8 dp_level,
 		 */
 		INIT_DELAYED_WORK(&edev->sp_task, qede_sp_task);
 		mutex_init(&edev->qede_lock);
+		qede_init_periodic_task(edev);
 
 		rc = register_netdev(edev->ndev);
 		if (rc) {
@@ -1306,6 +1332,11 @@ static int __qede_probe(struct pci_dev *pdev, u32 dp_module, u8 dp_level,
 	edev->rx_copybreak = QEDE_RX_HDR_SIZE;
 
 	qede_log_probe(edev);
+
+	/* retain user config (for example - after recovery) */
+	if (edev->stats_coal_usecs)
+		schedule_delayed_work(&edev->periodic_task, 0);
+
 	return 0;
 
 err4:
@@ -1374,6 +1405,7 @@ static void __qede_remove(struct pci_dev *pdev, enum qede_remove_mode mode)
 		unregister_netdev(ndev);
 
 		cancel_delayed_work_sync(&edev->sp_task);
+		cancel_delayed_work_sync(&edev->periodic_task);
 
 		edev->ops->common->set_power_state(cdev, PCI_D0);