[PATCH net 4/8] i40e: avoid deadlock when calling unregister_netdev()

[Tony Nguyen <anthony.l.nguyen@intel.com>]

From: Maciej Fijalkowski <maciej.fijalkowski@intel.com>

Sashiko reports:
***
> +err_netdev:
>  	if (vsi->netdev_registered) {
>  		vsi->netdev_registered = false;
>  		unregister_netdev(vsi->netdev);
Could this result in a deadlock when called during a device rebuild?
Looking at i40e_rebuild(), it explicitly acquires the RTNL lock before
proceeding:
drivers/net/ethernet/intel/i40e/i40e_main.c:i40e_rebuild() {
    ...
	if (!lock_acquired)
		rtnl_lock();
	ret = i40e_setup_pf_switch(pf, reinit, true);
    ...
}
If i40e_setup_pf_switch() calls i40e_vsi_reinit_setup() and takes this new
err_netdev path, unregister_netdev() will unconditionally attempt to acquire
rtnl_lock(), leading to a deadlock on the non-recursive mutex.
***

Use unregister_netdevice() when the rebuild path already holds RTNL, and
keep unregister_netdev() for callers that do not. This avoids both
recursive RTNL locking and dropping RTNL in the middle of the VSI unwind
path.

Fixes: bc7d338fbb3f ("i40e: reinit flow for the main VSI")
Reported-by: Sashiko AI Review <sashiko-bot@kernel.org>
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Tested-by: Sunitha Mekala <sunithax.d.mekala@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 65aa50330aac..5ea8731ece08 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -14245,13 +14245,15 @@ static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
 /**
  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
  * @vsi: pointer to the vsi.
+ * @lock_acquired: indicates whether or not the lock has been acquired
  *
  * This re-allocates a vsi's queue resources.
  *
  * Returns pointer to the successfully allocated and configured VSI sw struct
  * on success, otherwise returns NULL on failure.
  **/
-static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
+static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi,
+					      bool lock_acquired)
 {
 	struct i40e_vsi *main_vsi;
 	u16 alloc_queue_pairs;
@@ -14308,7 +14310,10 @@ static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
 err_netdev:
 	if (vsi->netdev_registered) {
 		vsi->netdev_registered = false;
-		unregister_netdev(vsi->netdev);
+		if (lock_acquired)
+			unregister_netdevice(vsi->netdev);
+		else
+			unregister_netdev(vsi->netdev);
 		free_netdev(vsi->netdev);
 		vsi->netdev = NULL;
 	}
@@ -15030,7 +15035,7 @@ static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit, bool lock_acqui
 			main_vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN,
 						  uplink_seid, 0);
 		else if (reinit)
-			main_vsi = i40e_vsi_reinit_setup(main_vsi);
+			main_vsi = i40e_vsi_reinit_setup(main_vsi, lock_acquired);
 		if (!main_vsi) {
 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
 			i40e_cloud_filter_exit(pf);
-- 
2.47.1