[PATCH 6.18.y 2/3] Reapply "bnxt_en: bring back rtnl_lock() in the bnxt_open() path"

[Sasha Levin <sashal@kernel.org>]

From: Jakub Kicinski <kuba@kernel.org>

[ Upstream commit b6197b386677ae5268d4702e23849d9ad53051ad ]

This reverts commit 850d9248d2eac662f869c766a598c877690c74e5.
This reapplies commit 325eb217e41f ("bnxt_en: bring back rtnl_lock()
in the bnxt_open() path").

Breno reports a lockdep warning in bnxt. During FW reset the driver
may end up calling netif_set_real_num_tx_queues() (if queue count
changes), so calls to bnxt_open() still require rtnl_lock.

  net/sched/sch_generic.c:1416 suspicious rcu_dereference_protected() usage!

   dev_qdisc_change_real_num_tx+0x54/0xe0
   netif_set_real_num_tx_queues+0x4ed/0xa80
   __bnxt_open_nic+0x9cb/0x3490
   bnxt_open+0x1cb/0x370
   bnxt_fw_reset_task+0x80d/0x1e80
   process_scheduled_works+0x9c1/0x13b0

The reverted commit was just an optimization / experiment
so let's go back to taking the lock.

Reported-by: Breno Leitao <leitao@debian.org>
Link: https://lore.kernel.org/ah726OtFX-Qw3U-R@gmail.com
Fixes: 850d9248d2ea ("Revert "bnxt_en: bring back rtnl_lock() in the bnxt_open() path"")
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260603195845.2574426-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Stable adaptation for dependency of 961e2a17c5e3559b3f8654d2daabdd25a42e770a:
Keep this branch's two-argument bnxt_ulp_start() calls, including error
notifications, and release RTNL before invoking them.

Reuse and rename the existing service-task lock/unlock helpers to add
RTNL coverage without adding functions. Inline their original netdev-only
operations in bnxt_fw_exception() and bnxt_fw_reset() so those paths keep
their existing locking and BNXT_STATE_IN_SP_TASK ordering. The RX reset
path then provides the bnxt_rtnl_unlock_sp() required by the target's
buffer-allocation failure handling.

Stable-dep-of: 961e2a17c5e3 ("bnxt_en: Handle buffer allocation failure in bnxt_rx_ring_reset()")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 44 ++++++++++++++---------
 1 file changed, 28 insertions(+), 16 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 9176b90f571fe..1555f337e3021 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -14271,31 +14271,32 @@ static void bnxt_timer(struct timer_list *t)
 	mod_timer(&bp->timer, jiffies + bp->current_interval);
 }
 
-static void bnxt_lock_sp(struct bnxt *bp)
+static void bnxt_rtnl_lock_sp(struct bnxt *bp)
 {
 	/* We are called from bnxt_sp_task which has BNXT_STATE_IN_SP_TASK
 	 * set.  If the device is being closed, bnxt_close() may be holding
-	 * netdev instance lock and waiting for BNXT_STATE_IN_SP_TASK to clear.
-	 * So we must clear BNXT_STATE_IN_SP_TASK before holding netdev
-	 * instance lock.
+	 * RTNL and the netdev instance lock while waiting for
+	 * BNXT_STATE_IN_SP_TASK to clear.  Clear it before taking either lock.
 	 */
 	clear_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
+	rtnl_lock();
 	netdev_lock(bp->dev);
 }
 
-static void bnxt_unlock_sp(struct bnxt *bp)
+static void bnxt_rtnl_unlock_sp(struct bnxt *bp)
 {
 	set_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
 	netdev_unlock(bp->dev);
+	rtnl_unlock();
 }
 
 /* Only called from bnxt_sp_task() */
 static void bnxt_reset(struct bnxt *bp, bool silent)
 {
-	bnxt_lock_sp(bp);
+	bnxt_rtnl_lock_sp(bp);
 	if (test_bit(BNXT_STATE_OPEN, &bp->state))
 		bnxt_reset_task(bp, silent);
-	bnxt_unlock_sp(bp);
+	bnxt_rtnl_unlock_sp(bp);
 }
 
 /* Only called from bnxt_sp_task() */
@@ -14303,9 +14304,9 @@ static void bnxt_rx_ring_reset(struct bnxt *bp)
 {
 	int i;
 
-	bnxt_lock_sp(bp);
+	bnxt_rtnl_lock_sp(bp);
 	if (!test_bit(BNXT_STATE_OPEN, &bp->state)) {
-		bnxt_unlock_sp(bp);
+		bnxt_rtnl_unlock_sp(bp);
 		return;
 	}
 	/* Disable and flush TPA before resetting the RX ring */
@@ -14344,7 +14345,7 @@ static void bnxt_rx_ring_reset(struct bnxt *bp)
 	}
 	if (bp->flags & BNXT_FLAG_TPA)
 		bnxt_set_tpa(bp, true);
-	bnxt_unlock_sp(bp);
+	bnxt_rtnl_unlock_sp(bp);
 }
 
 static void bnxt_fw_fatal_close(struct bnxt *bp)
@@ -14443,9 +14444,11 @@ void bnxt_fw_exception(struct bnxt *bp)
 	netdev_warn(bp->dev, "Detected firmware fatal condition, initiating reset\n");
 	set_bit(BNXT_STATE_FW_FATAL_COND, &bp->state);
 	bnxt_ulp_stop(bp);
-	bnxt_lock_sp(bp);
+	clear_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
+	netdev_lock(bp->dev);
 	bnxt_force_fw_reset(bp);
-	bnxt_unlock_sp(bp);
+	set_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
+	netdev_unlock(bp->dev);
 }
 
 /* Returns the number of registered VFs, or 1 if VF configuration is pending, or
@@ -14475,7 +14478,8 @@ static int bnxt_get_registered_vfs(struct bnxt *bp)
 void bnxt_fw_reset(struct bnxt *bp)
 {
 	bnxt_ulp_stop(bp);
-	bnxt_lock_sp(bp);
+	clear_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
+	netdev_lock(bp->dev);
 	if (test_bit(BNXT_STATE_OPEN, &bp->state) &&
 	    !test_bit(BNXT_STATE_IN_FW_RESET, &bp->state)) {
 		struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
@@ -14521,7 +14525,8 @@ void bnxt_fw_reset(struct bnxt *bp)
 		bnxt_queue_fw_reset_work(bp, tmo);
 	}
 fw_reset_exit:
-	bnxt_unlock_sp(bp);
+	set_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
+	netdev_unlock(bp->dev);
 }
 
 static void bnxt_chk_missed_irq(struct bnxt *bp)
@@ -15257,15 +15262,17 @@ static void bnxt_fw_reset_task(struct work_struct *work)
 		bp->fw_reset_state = BNXT_FW_RESET_STATE_OPENING;
 		fallthrough;
 	case BNXT_FW_RESET_STATE_OPENING:
-		while (!netdev_trylock(bp->dev)) {
+		while (!rtnl_trylock()) {
 			bnxt_queue_fw_reset_work(bp, HZ / 10);
 			return;
 		}
+		netdev_lock(bp->dev);
 		rc = bnxt_open(bp->dev);
 		if (rc) {
 			netdev_err(bp->dev, "bnxt_open() failed during FW reset\n");
 			bnxt_fw_reset_abort(bp, rc);
 			netdev_unlock(bp->dev);
+			rtnl_unlock();
 			goto ulp_start;
 		}
 
@@ -15285,6 +15292,7 @@ static void bnxt_fw_reset_task(struct work_struct *work)
 			bnxt_dl_health_fw_status_update(bp, true);
 		}
 		netdev_unlock(bp->dev);
+		rtnl_unlock();
 		bnxt_ulp_start(bp, 0);
 		bnxt_reenable_sriov(bp);
 		netdev_lock(bp->dev);
@@ -16273,7 +16281,7 @@ static int bnxt_queue_start(struct net_device *dev,
 		   rc);
 	napi_enable_locked(&bnapi->napi);
 	bnxt_db_nq_arm(bp, &cpr->cp_db, cpr->cp_raw_cons);
-	bnxt_reset_task(bp, true);
+	netif_close(dev);
 	return rc;
 }
 
@@ -17101,6 +17109,7 @@ static int bnxt_resume(struct device *device)
 	struct bnxt *bp = netdev_priv(dev);
 	int rc = 0;
 
+	rtnl_lock();
 	netdev_lock(dev);
 	rc = pci_enable_device(bp->pdev);
 	if (rc) {
@@ -17145,6 +17154,7 @@ static int bnxt_resume(struct device *device)
 
 resume_exit:
 	netdev_unlock(bp->dev);
+	rtnl_unlock();
 	bnxt_ulp_start(bp, rc);
 	if (!rc)
 		bnxt_reenable_sriov(bp);
@@ -17310,6 +17320,7 @@ static void bnxt_io_resume(struct pci_dev *pdev)
 	int err;
 
 	netdev_info(bp->dev, "PCI Slot Resume\n");
+	rtnl_lock();
 	netdev_lock(netdev);
 
 	err = bnxt_hwrm_func_qcaps(bp);
@@ -17327,6 +17338,7 @@ static void bnxt_io_resume(struct pci_dev *pdev)
 		netif_device_attach(netdev);
 
 	netdev_unlock(netdev);
+	rtnl_unlock();
 	bnxt_ulp_start(bp, err);
 	if (!err)
 		bnxt_reenable_sriov(bp);
-- 
2.53.0