[PATCH net-next 1/3] eth: bnxt: switch to netif_close

[PATCH net-next 1/3] eth: bnxt: switch to netif_close

[Stanislav Fomichev]

All (error) paths that call dev_close are already holding instance lock,
so switch to netif_close to avoid the deadlock.

Fixes: 004b5008016a ("eth: bnxt: remove most dependencies on RTNL")
Reported-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c         | 12 ++++++------
 drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c |  4 ++--
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 1a1e6da77777..e874530f1db2 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -12801,7 +12801,7 @@ int bnxt_open_nic(struct bnxt *bp, bool irq_re_init, bool link_re_init)
 		rc = __bnxt_open_nic(bp, irq_re_init, link_re_init);
 	if (rc) {
 		netdev_err(bp->dev, "nic open fail (rc: %x)\n", rc);
-		dev_close(bp->dev);
+		netif_close(bp->dev);
 	}
 	return rc;
 }
@@ -12839,7 +12839,7 @@ int bnxt_half_open_nic(struct bnxt *bp)
 half_open_err:
 	bnxt_free_skbs(bp);
 	bnxt_free_mem(bp, true);
-	dev_close(bp->dev);
+	netif_close(bp->dev);
 	return rc;
 }
 
@@ -14194,7 +14194,7 @@ void bnxt_fw_reset(struct bnxt *bp)
 			netdev_err(bp->dev, "Firmware reset aborted, rc = %d\n",
 				   n);
 			clear_bit(BNXT_STATE_IN_FW_RESET, &bp->state);
-			dev_close(bp->dev);
+			netif_close(bp->dev);
 			goto fw_reset_exit;
 		} else if (n > 0) {
 			u16 vf_tmo_dsecs = n * 10;
@@ -14809,7 +14809,7 @@ static void bnxt_fw_reset_abort(struct bnxt *bp, int rc)
 	if (bp->fw_reset_state != BNXT_FW_RESET_STATE_POLL_VF)
 		bnxt_dl_health_fw_status_update(bp, false);
 	bp->fw_reset_state = 0;
-	dev_close(bp->dev);
+	netif_close(bp->dev);
 }
 
 static void bnxt_fw_reset_task(struct work_struct *work)
@@ -16275,7 +16275,7 @@ int bnxt_restore_pf_fw_resources(struct bnxt *bp)
 
 	if (netif_running(bp->dev)) {
 		if (rc)
-			dev_close(bp->dev);
+			netif_close(bp->dev);
 		else
 			rc = bnxt_open_nic(bp, true, false);
 	}
@@ -16668,7 +16668,7 @@ static void bnxt_shutdown(struct pci_dev *pdev)
 		goto shutdown_exit;
 
 	if (netif_running(dev))
-		dev_close(dev);
+		netif_close(dev);
 
 	bnxt_ptp_clear(bp);
 	bnxt_clear_int_mode(bp);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c
index b06fcddfc81c..b6d6fcd105d7 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c
@@ -461,7 +461,7 @@ static int bnxt_dl_reload_down(struct devlink *dl, bool netns_change,
 		if (rc) {
 			NL_SET_ERR_MSG_MOD(extack, "Failed to deregister");
 			if (netif_running(bp->dev))
-				dev_close(bp->dev);
+				netif_close(bp->dev);
 			netdev_unlock(bp->dev);
 			rtnl_unlock();
 			break;
@@ -576,7 +576,7 @@ static int bnxt_dl_reload_up(struct devlink *dl, enum devlink_reload_action acti
 		*actions_performed |= BIT(action);
 	} else if (netif_running(bp->dev)) {
 		netdev_lock(bp->dev);
-		dev_close(bp->dev);
+		netif_close(bp->dev);
 		netdev_unlock(bp->dev);
 	}
 	rtnl_unlock();
-- 
2.48.1

[PATCH net-next 2/3] eth: bnxt: request unconditional ops lock

[Stanislav Fomichev]

netdev_lock_ops conditionally grabs instance lock when queue_mgmt_ops
is defined. However queue_mgmt_ops support is signaled via FW
so we can sometimes boot without queue_mgmt_ops being set.
This will result in bnxt running without instance lock which
the driver now heavily depends on. Set request_ops_lock to true
unconditionally to always request netdev instance lock.

Fixes: 004b5008016a ("eth: bnxt: remove most dependencies on RTNL")
Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index e874530f1db2..225d3e2db541 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -16612,6 +16612,7 @@ static int bnxt_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		bp->rss_cap |= BNXT_RSS_CAP_MULTI_RSS_CTX;
 	if (BNXT_SUPPORTS_QUEUE_API(bp))
 		dev->queue_mgmt_ops = &bnxt_queue_mgmt_ops;
+	dev->request_ops_lock = true;
 
 	rc = register_netdev(dev);
 	if (rc)
-- 
2.48.1

[PATCH net-next 3/3] eth: bnxt: add missing netdev lock management to bnxt_dl_reload_up

[Stanislav Fomichev]

bnxt_dl_reload_up is completely missing instance lock management
which can result in `devlink dev reload` leaving with instance
lock held. Add the missing calls.

Also add netdev_assert_locked to make it clear that the up() method
is running with the instance lock grabbed.

Fixes: 004b5008016a ("eth: bnxt: remove most dependencies on RTNL")
Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c
index b6d6fcd105d7..ea7f789be760 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c
@@ -518,6 +518,8 @@ static int bnxt_dl_reload_up(struct devlink *dl, enum devlink_reload_action acti
 	struct bnxt *bp = bnxt_get_bp_from_dl(dl);
 	int rc = 0;
 
+	netdev_assert_locked(bp->dev);
+
 	*actions_performed = 0;
 	switch (action) {
 	case DEVLINK_RELOAD_ACTION_DRIVER_REINIT: {
@@ -542,6 +544,7 @@ static int bnxt_dl_reload_up(struct devlink *dl, enum devlink_reload_action acti
 		if (!netif_running(bp->dev))
 			NL_SET_ERR_MSG_MOD(extack,
 					   "Device is closed, not waiting for reset notice that will never come");
+		netdev_unlock(bp->dev);
 		rtnl_unlock();
 		while (test_bit(BNXT_STATE_FW_ACTIVATE, &bp->state)) {
 			if (time_after(jiffies, timeout)) {
@@ -557,6 +560,7 @@ static int bnxt_dl_reload_up(struct devlink *dl, enum devlink_reload_action acti
 			msleep(50);
 		}
 		rtnl_lock();
+		netdev_lock(bp->dev);
 		if (!rc)
 			*actions_performed |= BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
 		clear_bit(BNXT_STATE_FW_ACTIVATE, &bp->state);
@@ -575,10 +579,9 @@ static int bnxt_dl_reload_up(struct devlink *dl, enum devlink_reload_action acti
 		}
 		*actions_performed |= BIT(action);
 	} else if (netif_running(bp->dev)) {
-		netdev_lock(bp->dev);
 		netif_close(bp->dev);
-		netdev_unlock(bp->dev);
 	}
+	netdev_unlock(bp->dev);
 	rtnl_unlock();
 	if (action == DEVLINK_RELOAD_ACTION_DRIVER_REINIT)
 		bnxt_ulp_start(bp, rc);
-- 
2.48.1

Re: [PATCH net-next 1/3] eth: bnxt: switch to netif_close

[kernel test robot]

Hi Stanislav,

kernel test robot noticed the following build errors:

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

url:    https://github.com/intel-lab-lkp/linux/commits/Stanislav-Fomichev/eth-bnxt-request-unconditional-ops-lock/20250308-091318
base:   net-next/main
patch link:    https://lore.kernel.org/r/20250308010840.910382-1-sdf%40fomichev.me
patch subject: [PATCH net-next 1/3] eth: bnxt: switch to netif_close
config: arm-randconfig-004-20250309 (https://download.01.org/0day-ci/archive/20250309/202503091014.T0oUWfdo-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/20250309/202503091014.T0oUWfdo-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/202503091014.T0oUWfdo-lkp@intel.com/

All errors (new ones prefixed by >>, old ones prefixed by <<):

WARNING: modpost: missing MODULE_DESCRIPTION() in arch/arm/probes/kprobes/test-kprobes.o
WARNING: modpost: missing MODULE_DESCRIPTION() in mm/kasan/kasan_test.o
WARNING: modpost: missing MODULE_DESCRIPTION() in lib/slub_kunit.o
>> ERROR: modpost: "netif_close" [drivers/net/ethernet/broadcom/bnxt/bnxt_en.ko] undefined!

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