> -----Original Message-----
> From: Emil Tantilov <[email protected]>
> Sent: Wednesday, November 4, 2026 12:54 AM
> To: [email protected]
> Cc: Burra, Phani R <[email protected]>; Zaremba, Larysa
> <[email protected]>; Hay, Joshua A <[email protected]>;
> [email protected]; [email protected]; [email protected]
> Subject: [PATCH iwl-net v3 1/2] idpf: fix possible race on remove during
> a reset
> 
> Reset and remove can race, leaving NAPI registered and enabled:
> 
> modprobe idpf& sleep 1; ip link set eth0 up& rmmod idpf
> 
>   WARNING: net/core/dev.c:7699 at __netif_napi_del_locked+0x11a/0x130,
> CPU#30: rmmod/22393
>   ...
>   RIP: 0010:__netif_napi_del_locked+0x11a/0x130
>   ...
>   Call Trace:
>   <TASK>
>   free_netdev+0x110/0x2a0
>   idpf_vport_dealloc+0x452/0x460 [idpf]
>   idpf_deinit_task+0x51/0x70 [idpf]
>   idpf_vc_core_deinit+0x32/0x170 [idpf]
>   idpf_remove+0x40/0x200 [idpf]
>   pci_device_remove+0x40/0xa0
>   device_release_driver_internal+0x1a9/0x210
>   driver_detach+0x4b/0x90
>   bus_remove_driver+0x70/0x100
>   pci_unregister_driver+0x2e/0xb0
>   __do_sys_delete_module.constprop.0+0x190/0x2e0
>   do_syscall_64+0xc8/0x6b0
>   entry_SYSCALL_64_after_hwframe+0x76/0x7e
>   RIP: 0033:0x7fe628130beb
> 
> Make sure to call idpf_vport_stop() in idpf_stop(), irrespective of the
> IDPF_REMOVE_IN_PROG state, to allow a reset racing with remove to tear
> down NAPI in idpf_detach_and_close(), which runs under RTNL lock.
> 
> IDPF_VPORT_UP is modified under RTNL as idpf_vport_stop() and
> idpf_up_complete() either take the lock themselves or are reached from
> callbacks that do. Move the test inside the RTNL section so the test and
> the teardown are atomic with respect to a racing close, and assert the
> lock is held to keep the invariant documented and enforced.
> 
> Fixes: 2e281e1155fc ("idpf: detach and close netdevs while handling a
> reset")
> Signed-off-by: Emil Tantilov <[email protected]>
> ---
>  drivers/net/ethernet/intel/idpf/idpf_lib.c | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c
> b/drivers/net/ethernet/intel/idpf/idpf_lib.c
> index 2130bbef0c17..791ebff98507 100644
> --- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
> +++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
> @@ -1005,12 +1005,14 @@ static void idpf_vport_stop(struct idpf_vport
> *vport, bool rtnl)
>       struct idpf_queue_id_reg_info *chunks;
>       u32 vport_id = vport->vport_id;
> 
> -     if (!test_bit(IDPF_VPORT_UP, np->state))
> -             return;
> -
>       if (rtnl)
>               rtnl_lock();
> 
> +     /* IDPF_VPORT_UP should only be changed under RTNL */
> +     ASSERT_RTNL();
> +     if (!test_bit(IDPF_VPORT_UP, np->state))
> +             goto unlock;
> +
>       netif_carrier_off(vport->netdev);
>       netif_tx_disable(vport->netdev);
> 
> @@ -1036,6 +1038,7 @@ static void idpf_vport_stop(struct idpf_vport
> *vport, bool rtnl)
>       idpf_vport_intr_rel(rsrc);
>       clear_bit(IDPF_VPORT_UP, np->state);
> 
> +unlock:
>       if (rtnl)
>               rtnl_unlock();
>  }
> @@ -1052,12 +1055,8 @@ static void idpf_vport_stop(struct idpf_vport
> *vport, bool rtnl)
>   */
>  static int idpf_stop(struct net_device *netdev)  {
> -     struct idpf_netdev_priv *np = netdev_priv(netdev);
>       struct idpf_vport *vport;
> 
> -     if (test_bit(IDPF_REMOVE_IN_PROG, np->adapter->flags))
> -             return 0;
> -
>       idpf_vport_ctrl_lock(netdev);
>       vport = idpf_netdev_to_vport(netdev);
> 
> --
> 2.37.3

Reviewed-by: Aleksandr Loktionov <[email protected]>

Reply via email to