From c35c178fcdffa5f3bedf261a628769b9a52c2436 Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:19 +0200 Subject: [PATCH 1/9] ice: enforce RTNL assumption of queue NAPI manipulation Instead of making assumptions in comments move them into code. Be also more precise, RTNL must be locked only when there is NAPI, and we have VSIs w/o NAPI that call ice_vsi_clear_napi_queues() during rmmod. Signed-off-by: Przemek Kitszel Reviewed-by: Paul Menzel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_lib.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c index 4479c824561e..69cb0381c460 100644 --- a/drivers/net/ethernet/intel/ice/ice_lib.c +++ b/drivers/net/ethernet/intel/ice/ice_lib.c @@ -2769,7 +2769,6 @@ void ice_dis_vsi(struct ice_vsi *vsi, bool locked) * @vsi: VSI pointer * * Associate queue[s] with napi for all vectors. - * The caller must hold rtnl_lock. */ void ice_vsi_set_napi_queues(struct ice_vsi *vsi) { @@ -2779,6 +2778,7 @@ void ice_vsi_set_napi_queues(struct ice_vsi *vsi) if (!netdev) return; + ASSERT_RTNL(); ice_for_each_rxq(vsi, q_idx) netif_queue_set_napi(netdev, q_idx, NETDEV_QUEUE_TYPE_RX, &vsi->rx_rings[q_idx]->q_vector->napi); @@ -2799,7 +2799,6 @@ void ice_vsi_set_napi_queues(struct ice_vsi *vsi) * @vsi: VSI pointer * * Clear the association between all VSI queues queue[s] and napi. - * The caller must hold rtnl_lock. */ void ice_vsi_clear_napi_queues(struct ice_vsi *vsi) { @@ -2809,6 +2808,7 @@ void ice_vsi_clear_napi_queues(struct ice_vsi *vsi) if (!netdev) return; + ASSERT_RTNL(); /* Clear the NAPI's interrupt number */ ice_for_each_q_vector(vsi, v_idx) { struct ice_q_vector *q_vector = vsi->q_vectors[v_idx]; From 806c4f32a80627f8977fda53520afc41491d162f Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:20 +0200 Subject: [PATCH 2/9] ice: move service task start out of ice_init_pf() Move service task start out of ice_init_pf(). Do analogous with deinit. Service task is needed up to the very end of driver removal, later commit of the series will move it later on execution timeline. Signed-off-by: Przemek Kitszel Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice.h | 1 + drivers/net/ethernet/intel/ice/ice_main.c | 18 +++++++++++------- 2 files changed, 12 insertions(+), 7 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h index 22b8323ff0d0..0e58a58c23eb 100644 --- a/drivers/net/ethernet/intel/ice/ice.h +++ b/drivers/net/ethernet/intel/ice/ice.h @@ -1029,6 +1029,7 @@ int ice_open(struct net_device *netdev); int ice_open_internal(struct net_device *netdev); int ice_stop(struct net_device *netdev); void ice_service_task_schedule(struct ice_pf *pf); +void ice_start_service_task(struct ice_pf *pf); int ice_load(struct ice_pf *pf); void ice_unload(struct ice_pf *pf); void ice_adv_lnk_speed_maps_init(void); diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index ca95b8800bb3..f9e464b79bca 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -3951,7 +3951,6 @@ u16 ice_get_avail_rxq_count(struct ice_pf *pf) */ static void ice_deinit_pf(struct ice_pf *pf) { - ice_service_task_stop(pf); mutex_destroy(&pf->lag_mutex); mutex_destroy(&pf->adev_mutex); mutex_destroy(&pf->sw_mutex); @@ -4030,6 +4029,14 @@ static void ice_set_pf_caps(struct ice_pf *pf) pf->max_pf_rxqs = func_caps->common_cap.num_rxq; } +void ice_start_service_task(struct ice_pf *pf) +{ + timer_setup(&pf->serv_tmr, ice_service_timer, 0); + pf->serv_tmr_period = HZ; + INIT_WORK(&pf->serv_task, ice_service_task); + clear_bit(ICE_SERVICE_SCHED, pf->state); +} + /** * ice_init_pf - Initialize general software structures (struct ice_pf) * @pf: board private structure to initialize @@ -4049,12 +4056,6 @@ static int ice_init_pf(struct ice_pf *pf) init_waitqueue_head(&pf->reset_wait_queue); - /* setup service timer and periodic service task */ - timer_setup(&pf->serv_tmr, ice_service_timer, 0); - pf->serv_tmr_period = HZ; - INIT_WORK(&pf->serv_task, ice_service_task); - clear_bit(ICE_SERVICE_SCHED, pf->state); - mutex_init(&pf->avail_q_mutex); pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL); if (!pf->avail_txqs) @@ -4745,6 +4746,7 @@ int ice_init_dev(struct ice_pf *pf) ice_set_safe_mode_caps(hw); } + ice_start_service_task(pf); err = ice_init_pf(pf); if (err) { dev_err(dev, "ice_init_pf failed: %d\n", err); @@ -4791,6 +4793,7 @@ int ice_init_dev(struct ice_pf *pf) ice_clear_interrupt_scheme(pf); unroll_pf_init: ice_deinit_pf(pf); + ice_service_task_stop(pf); return err; } @@ -4799,6 +4802,7 @@ void ice_deinit_dev(struct ice_pf *pf) ice_free_irq_msix_misc(pf); ice_deinit_pf(pf); ice_deinit_hw(&pf->hw); + ice_service_task_stop(pf); /* Service task is already stopped, so call reset directly. */ ice_reset(&pf->hw, ICE_RESET_PFR); From 2fe18288fce6b421f1dc585bcb9dd3afa32d3ad9 Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:21 +0200 Subject: [PATCH 3/9] ice: move ice_init_interrupt_scheme() prior ice_init_pf() Move ice_init_interrupt_scheme() prior ice_init_pf(). To enable the move ice_set_pf_caps() was moved out from ice_init_pf() to the caller (ice_init_dev()), and placed prior to the irq scheme init. The move makes deinit order of ice_deinit_dev() and failure-path of ice_init_pf() match (at least in terms of not calling ice_clear_interrupt_scheme() and ice_deinit_pf() in opposite ways). The new order aligns with findings made by Jakub Buchocki in the commit 24b454bc354a ("ice: Fix ice module unload"). Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_main.c | 25 ++++++++++------------- 1 file changed, 11 insertions(+), 14 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index f9e464b79bca..e00c282a8c18 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -4043,8 +4043,6 @@ void ice_start_service_task(struct ice_pf *pf) */ static int ice_init_pf(struct ice_pf *pf) { - ice_set_pf_caps(pf); - mutex_init(&pf->sw_mutex); mutex_init(&pf->tc_mutex); mutex_init(&pf->adev_mutex); @@ -4746,11 +4744,18 @@ int ice_init_dev(struct ice_pf *pf) ice_set_safe_mode_caps(hw); } + ice_set_pf_caps(pf); + err = ice_init_interrupt_scheme(pf); + if (err) { + dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err); + return -EIO; + } + ice_start_service_task(pf); err = ice_init_pf(pf); if (err) { dev_err(dev, "ice_init_pf failed: %d\n", err); - return err; + goto unroll_irq_scheme_init; } pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port; @@ -4768,14 +4773,6 @@ int ice_init_dev(struct ice_pf *pf) pf->hw.udp_tunnel_nic.tables[1].tunnel_types = UDP_TUNNEL_TYPE_GENEVE; } - - err = ice_init_interrupt_scheme(pf); - if (err) { - dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err); - err = -EIO; - goto unroll_pf_init; - } - /* In case of MSIX we are going to setup the misc vector right here * to handle admin queue events etc. In case of legacy and MSI * the misc functionality and queue processing is combined in @@ -4784,16 +4781,16 @@ int ice_init_dev(struct ice_pf *pf) err = ice_req_irq_msix_misc(pf); if (err) { dev_err(dev, "setup of misc vector failed: %d\n", err); - goto unroll_irq_scheme_init; + goto unroll_pf_init; } return 0; -unroll_irq_scheme_init: - ice_clear_interrupt_scheme(pf); unroll_pf_init: ice_deinit_pf(pf); +unroll_irq_scheme_init: ice_service_task_stop(pf); + ice_clear_interrupt_scheme(pf); return err; } From 71430451f81bd6550e46d89b69103a111fc42982 Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:22 +0200 Subject: [PATCH 4/9] ice: ice_init_pf: destroy mutexes and xarrays on memory alloc failure Unroll actions of ice_init_pf() when it fails. ice_deinit_pf() happens to be perfect to call here. Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_main.c | 31 +++++++++-------------- 1 file changed, 12 insertions(+), 19 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index e00c282a8c18..09dee43e48aa 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -3951,6 +3951,8 @@ u16 ice_get_avail_rxq_count(struct ice_pf *pf) */ static void ice_deinit_pf(struct ice_pf *pf) { + /* note that we unroll also on ice_init_pf() failure here */ + mutex_destroy(&pf->lag_mutex); mutex_destroy(&pf->adev_mutex); mutex_destroy(&pf->sw_mutex); @@ -4055,25 +4057,6 @@ static int ice_init_pf(struct ice_pf *pf) init_waitqueue_head(&pf->reset_wait_queue); mutex_init(&pf->avail_q_mutex); - pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL); - if (!pf->avail_txqs) - return -ENOMEM; - - pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL); - if (!pf->avail_rxqs) { - bitmap_free(pf->avail_txqs); - pf->avail_txqs = NULL; - return -ENOMEM; - } - - pf->txtime_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL); - if (!pf->txtime_txqs) { - bitmap_free(pf->avail_txqs); - pf->avail_txqs = NULL; - bitmap_free(pf->avail_rxqs); - pf->avail_rxqs = NULL; - return -ENOMEM; - } mutex_init(&pf->vfs.table_lock); hash_init(pf->vfs.table); @@ -4086,7 +4069,17 @@ static int ice_init_pf(struct ice_pf *pf) xa_init(&pf->dyn_ports); xa_init(&pf->sf_nums); + pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL); + pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL); + pf->txtime_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL); + if (!pf->avail_txqs || !pf->avail_rxqs || !pf->txtime_txqs) + goto undo_init; + return 0; +undo_init: + /* deinit handles half-initialized pf just fine */ + ice_deinit_pf(pf); + return -ENOMEM; } /** From e3bf1cdde7471bab7fc20dd1a37c2cdb82d3f76b Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:23 +0200 Subject: [PATCH 5/9] ice: move udp_tunnel_nic and misc IRQ setup into ice_init_pf() Move udp_tunnel_nic setup and ice_req_irq_msix_misc() call into ice_init_pf(), remove some redundancy in the former while moving. Move ice_free_irq_msix_misc() call into ice_deinit_pf(), to mimic the above in terms of needed cleanup. Guard it via emptiness check, to keep the allowance of half-initialized pf being cleaned up. Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_main.c | 58 +++++++++++------------ 1 file changed, 28 insertions(+), 30 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index 09dee43e48aa..1691dda1067e 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -3978,6 +3978,9 @@ static void ice_deinit_pf(struct ice_pf *pf) if (pf->ptp.clock) ptp_clock_unregister(pf->ptp.clock); + if (!xa_empty(&pf->irq_tracker.entries)) + ice_free_irq_msix_misc(pf); + xa_destroy(&pf->dyn_ports); xa_destroy(&pf->sf_nums); } @@ -4045,6 +4048,11 @@ void ice_start_service_task(struct ice_pf *pf) */ static int ice_init_pf(struct ice_pf *pf) { + struct udp_tunnel_nic_info *udp_tunnel_nic = &pf->hw.udp_tunnel_nic; + struct device *dev = ice_pf_to_dev(pf); + struct ice_hw *hw = &pf->hw; + int err = -ENOMEM; + mutex_init(&pf->sw_mutex); mutex_init(&pf->tc_mutex); mutex_init(&pf->adev_mutex); @@ -4075,11 +4083,30 @@ static int ice_init_pf(struct ice_pf *pf) if (!pf->avail_txqs || !pf->avail_rxqs || !pf->txtime_txqs) goto undo_init; + udp_tunnel_nic->set_port = ice_udp_tunnel_set_port; + udp_tunnel_nic->unset_port = ice_udp_tunnel_unset_port; + udp_tunnel_nic->shared = &hw->udp_tunnel_shared; + udp_tunnel_nic->tables[0].n_entries = hw->tnl.valid_count[TNL_VXLAN]; + udp_tunnel_nic->tables[0].tunnel_types = UDP_TUNNEL_TYPE_VXLAN; + udp_tunnel_nic->tables[1].n_entries = hw->tnl.valid_count[TNL_GENEVE]; + udp_tunnel_nic->tables[1].tunnel_types = UDP_TUNNEL_TYPE_GENEVE; + + /* In case of MSIX we are going to setup the misc vector right here + * to handle admin queue events etc. In case of legacy and MSI + * the misc functionality and queue processing is combined in + * the same vector and that gets setup at open. + */ + err = ice_req_irq_msix_misc(pf); + if (err) { + dev_err(dev, "setup of misc vector failed: %d\n", err); + goto undo_init; + } + return 0; undo_init: /* deinit handles half-initialized pf just fine */ ice_deinit_pf(pf); - return -ENOMEM; + return err; } /** @@ -4751,36 +4778,8 @@ int ice_init_dev(struct ice_pf *pf) goto unroll_irq_scheme_init; } - pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port; - pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port; - pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared; - if (pf->hw.tnl.valid_count[TNL_VXLAN]) { - pf->hw.udp_tunnel_nic.tables[0].n_entries = - pf->hw.tnl.valid_count[TNL_VXLAN]; - pf->hw.udp_tunnel_nic.tables[0].tunnel_types = - UDP_TUNNEL_TYPE_VXLAN; - } - if (pf->hw.tnl.valid_count[TNL_GENEVE]) { - pf->hw.udp_tunnel_nic.tables[1].n_entries = - pf->hw.tnl.valid_count[TNL_GENEVE]; - pf->hw.udp_tunnel_nic.tables[1].tunnel_types = - UDP_TUNNEL_TYPE_GENEVE; - } - /* In case of MSIX we are going to setup the misc vector right here - * to handle admin queue events etc. In case of legacy and MSI - * the misc functionality and queue processing is combined in - * the same vector and that gets setup at open. - */ - err = ice_req_irq_msix_misc(pf); - if (err) { - dev_err(dev, "setup of misc vector failed: %d\n", err); - goto unroll_pf_init; - } - return 0; -unroll_pf_init: - ice_deinit_pf(pf); unroll_irq_scheme_init: ice_service_task_stop(pf); ice_clear_interrupt_scheme(pf); @@ -4789,7 +4788,6 @@ int ice_init_dev(struct ice_pf *pf) void ice_deinit_dev(struct ice_pf *pf) { - ice_free_irq_msix_misc(pf); ice_deinit_pf(pf); ice_deinit_hw(&pf->hw); ice_service_task_stop(pf); From ef825bdb4605742c4efc03f67d930e80c42f33cb Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:24 +0200 Subject: [PATCH 6/9] ice: move ice_init_pf() out of ice_init_dev() Move ice_init_pf() out of ice_init_dev(). Do the same for deinit counterpart. Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- .../net/ethernet/intel/ice/devlink/devlink.c | 16 ++++++++-- drivers/net/ethernet/intel/ice/ice.h | 2 ++ drivers/net/ethernet/intel/ice/ice_main.c | 32 +++++++++---------- 3 files changed, 31 insertions(+), 19 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/devlink/devlink.c b/drivers/net/ethernet/intel/ice/devlink/devlink.c index fb2de521731a..c354a03c950c 100644 --- a/drivers/net/ethernet/intel/ice/devlink/devlink.c +++ b/drivers/net/ethernet/intel/ice/devlink/devlink.c @@ -459,6 +459,7 @@ static void ice_devlink_reinit_down(struct ice_pf *pf) rtnl_lock(); ice_vsi_decfg(ice_get_main_vsi(pf)); rtnl_unlock(); + ice_deinit_pf(pf); ice_deinit_dev(pf); } @@ -1231,11 +1232,12 @@ static void ice_set_min_max_msix(struct ice_pf *pf) static int ice_devlink_reinit_up(struct ice_pf *pf) { struct ice_vsi *vsi = ice_get_main_vsi(pf); + struct device *dev = ice_pf_to_dev(pf); int err; err = ice_init_hw(&pf->hw); if (err) { - dev_err(ice_pf_to_dev(pf), "ice_init_hw failed: %d\n", err); + dev_err(dev, "ice_init_hw failed: %d\n", err); return err; } @@ -1246,13 +1248,19 @@ static int ice_devlink_reinit_up(struct ice_pf *pf) if (err) goto unroll_hw_init; + err = ice_init_pf(pf); + if (err) { + dev_err(dev, "ice_init_pf failed: %d\n", err); + goto unroll_dev_init; + } + vsi->flags = ICE_VSI_FLAG_INIT; rtnl_lock(); err = ice_vsi_cfg(vsi); rtnl_unlock(); if (err) - goto err_vsi_cfg; + goto unroll_pf_init; /* No need to take devl_lock, it's already taken by devlink API */ err = ice_load(pf); @@ -1265,7 +1273,9 @@ static int ice_devlink_reinit_up(struct ice_pf *pf) rtnl_lock(); ice_vsi_decfg(vsi); rtnl_unlock(); -err_vsi_cfg: +unroll_pf_init: + ice_deinit_pf(pf); +unroll_dev_init: ice_deinit_dev(pf); unroll_hw_init: ice_deinit_hw(&pf->hw); diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h index 0e58a58c23eb..9a1abd457337 100644 --- a/drivers/net/ethernet/intel/ice/ice.h +++ b/drivers/net/ethernet/intel/ice/ice.h @@ -1035,6 +1035,8 @@ void ice_unload(struct ice_pf *pf); void ice_adv_lnk_speed_maps_init(void); int ice_init_dev(struct ice_pf *pf); void ice_deinit_dev(struct ice_pf *pf); +int ice_init_pf(struct ice_pf *pf); +void ice_deinit_pf(struct ice_pf *pf); int ice_change_mtu(struct net_device *netdev, int new_mtu); void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue); int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp); diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index 1691dda1067e..a4acc42fabab 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -3949,7 +3949,7 @@ u16 ice_get_avail_rxq_count(struct ice_pf *pf) * ice_deinit_pf - Unrolls initialziations done by ice_init_pf * @pf: board private structure to initialize */ -static void ice_deinit_pf(struct ice_pf *pf) +void ice_deinit_pf(struct ice_pf *pf) { /* note that we unroll also on ice_init_pf() failure here */ @@ -4045,8 +4045,9 @@ void ice_start_service_task(struct ice_pf *pf) /** * ice_init_pf - Initialize general software structures (struct ice_pf) * @pf: board private structure to initialize + * Return: 0 on success, negative errno otherwise. */ -static int ice_init_pf(struct ice_pf *pf) +int ice_init_pf(struct ice_pf *pf) { struct udp_tunnel_nic_info *udp_tunnel_nic = &pf->hw.udp_tunnel_nic; struct device *dev = ice_pf_to_dev(pf); @@ -4772,23 +4773,12 @@ int ice_init_dev(struct ice_pf *pf) } ice_start_service_task(pf); - err = ice_init_pf(pf); - if (err) { - dev_err(dev, "ice_init_pf failed: %d\n", err); - goto unroll_irq_scheme_init; - } return 0; - -unroll_irq_scheme_init: - ice_service_task_stop(pf); - ice_clear_interrupt_scheme(pf); - return err; } void ice_deinit_dev(struct ice_pf *pf) { - ice_deinit_pf(pf); ice_deinit_hw(&pf->hw); ice_service_task_stop(pf); @@ -5030,21 +5020,28 @@ static void ice_deinit_devlink(struct ice_pf *pf) static int ice_init(struct ice_pf *pf) { + struct device *dev = ice_pf_to_dev(pf); int err; err = ice_init_dev(pf); if (err) return err; + err = ice_init_pf(pf); + if (err) { + dev_err(dev, "ice_init_pf failed: %d\n", err); + goto unroll_dev_init; + } + if (pf->hw.mac_type == ICE_MAC_E830) { err = pci_enable_ptm(pf->pdev, NULL); if (err) - dev_dbg(ice_pf_to_dev(pf), "PCIe PTM not supported by PCIe bus/controller\n"); + dev_dbg(dev, "PCIe PTM not supported by PCIe bus/controller\n"); } err = ice_alloc_vsis(pf); if (err) - goto err_alloc_vsis; + goto unroll_pf_init; err = ice_init_pf_sw(pf); if (err) @@ -5081,7 +5078,9 @@ static int ice_init(struct ice_pf *pf) ice_deinit_pf_sw(pf); err_init_pf_sw: ice_dealloc_vsis(pf); -err_alloc_vsis: +unroll_pf_init: + ice_deinit_pf(pf); +unroll_dev_init: ice_deinit_dev(pf); return err; } @@ -5093,6 +5092,7 @@ static void ice_deinit(struct ice_pf *pf) ice_deinit_pf_sw(pf); ice_dealloc_vsis(pf); + ice_deinit_pf(pf); ice_deinit_dev(pf); } From c2fb9398f73d41cb2b5da74ff505578525ee3fd8 Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:25 +0200 Subject: [PATCH 7/9] ice: extract ice_init_dev() from ice_init() Extract ice_init_dev() from ice_init(), to allow service task and IRQ scheme teardown to be put after clearing SW constructs in the subsequent commit. Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_main.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index a4acc42fabab..9a817c3c8b99 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -5023,14 +5023,10 @@ static int ice_init(struct ice_pf *pf) struct device *dev = ice_pf_to_dev(pf); int err; - err = ice_init_dev(pf); - if (err) - return err; - err = ice_init_pf(pf); if (err) { dev_err(dev, "ice_init_pf failed: %d\n", err); - goto unroll_dev_init; + return err; } if (pf->hw.mac_type == ICE_MAC_E830) { @@ -5080,8 +5076,6 @@ static int ice_init(struct ice_pf *pf) ice_dealloc_vsis(pf); unroll_pf_init: ice_deinit_pf(pf); -unroll_dev_init: - ice_deinit_dev(pf); return err; } @@ -5093,7 +5087,6 @@ static void ice_deinit(struct ice_pf *pf) ice_deinit_pf_sw(pf); ice_dealloc_vsis(pf); ice_deinit_pf(pf); - ice_deinit_dev(pf); } /** @@ -5323,10 +5316,14 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent) } pf->adapter = adapter; - err = ice_init(pf); + err = ice_init_dev(pf); if (err) goto unroll_adapter; + err = ice_init(pf); + if (err) + goto unroll_dev_init; + devl_lock(priv_to_devlink(pf)); err = ice_load(pf); if (err) @@ -5344,6 +5341,8 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent) unroll_init: devl_unlock(priv_to_devlink(pf)); ice_deinit(pf); +unroll_dev_init: + ice_deinit_dev(pf); unroll_adapter: ice_adapter_put(pdev); unroll_hw_init: @@ -5457,6 +5456,7 @@ static void ice_remove(struct pci_dev *pdev) devl_unlock(priv_to_devlink(pf)); ice_deinit(pf); + ice_deinit_dev(pf); ice_vsi_release_all(pf); ice_setup_mc_magic_wake(pf); From 8a37f9e2ff40f4a4fa8def22febefe4daf58e573 Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:26 +0200 Subject: [PATCH 8/9] ice: move ice_deinit_dev() to the end of deinit paths ice_deinit_dev() takes care of turning off adminq processing, which is much needed during driver teardown (remove, reset, error path). Move it to the very end where applicable. For example, ice_deinit_hw() called after adminq deinit slows rmmod on my two-card setup by about 60 seconds. ice_init_dev() and ice_deinit_dev() scopes were reduced by previous commits of the series, with a final touch of extracting ice_init_dev_hw() out now (there is no deinit counterpart). Note that removed ice_service_task_stop() call from ice_remove() is placed in the ice_deinit_dev() (and stopping twice makes no sense). Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- .../net/ethernet/intel/ice/devlink/devlink.c | 5 +++- drivers/net/ethernet/intel/ice/ice.h | 1 + drivers/net/ethernet/intel/ice/ice_common.c | 3 +++ drivers/net/ethernet/intel/ice/ice_main.c | 23 ++++++++++++------- 4 files changed, 23 insertions(+), 9 deletions(-) diff --git a/drivers/net/ethernet/intel/ice/devlink/devlink.c b/drivers/net/ethernet/intel/ice/devlink/devlink.c index c354a03c950c..938914abbe06 100644 --- a/drivers/net/ethernet/intel/ice/devlink/devlink.c +++ b/drivers/net/ethernet/intel/ice/devlink/devlink.c @@ -1233,6 +1233,7 @@ static int ice_devlink_reinit_up(struct ice_pf *pf) { struct ice_vsi *vsi = ice_get_main_vsi(pf); struct device *dev = ice_pf_to_dev(pf); + bool need_dev_deinit = false; int err; err = ice_init_hw(&pf->hw); @@ -1276,9 +1277,11 @@ static int ice_devlink_reinit_up(struct ice_pf *pf) unroll_pf_init: ice_deinit_pf(pf); unroll_dev_init: - ice_deinit_dev(pf); + need_dev_deinit = true; unroll_hw_init: ice_deinit_hw(&pf->hw); + if (need_dev_deinit) + ice_deinit_dev(pf); return err; } diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h index 9a1abd457337..9ee596773f34 100644 --- a/drivers/net/ethernet/intel/ice/ice.h +++ b/drivers/net/ethernet/intel/ice/ice.h @@ -1033,6 +1033,7 @@ void ice_start_service_task(struct ice_pf *pf); int ice_load(struct ice_pf *pf); void ice_unload(struct ice_pf *pf); void ice_adv_lnk_speed_maps_init(void); +void ice_init_dev_hw(struct ice_pf *pf); int ice_init_dev(struct ice_pf *pf); void ice_deinit_dev(struct ice_pf *pf); int ice_init_pf(struct ice_pf *pf); diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c index 2250426ec91b..b097cc8b175c 100644 --- a/drivers/net/ethernet/intel/ice/ice_common.c +++ b/drivers/net/ethernet/intel/ice/ice_common.c @@ -1161,6 +1161,9 @@ int ice_init_hw(struct ice_hw *hw) status = ice_init_hw_tbls(hw); if (status) goto err_unroll_fltr_mgmt_struct; + + ice_init_dev_hw(hw->back); + mutex_init(&hw->tnl_lock); ice_init_chk_recipe_reuse_support(hw); diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index 9a817c3c8b99..a1fe2d363adb 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -4742,9 +4742,8 @@ static void ice_decfg_netdev(struct ice_vsi *vsi) vsi->netdev = NULL; } -int ice_init_dev(struct ice_pf *pf) +void ice_init_dev_hw(struct ice_pf *pf) { - struct device *dev = ice_pf_to_dev(pf); struct ice_hw *hw = &pf->hw; int err; @@ -4764,6 +4763,12 @@ int ice_init_dev(struct ice_pf *pf) */ ice_set_safe_mode_caps(hw); } +} + +int ice_init_dev(struct ice_pf *pf) +{ + struct device *dev = ice_pf_to_dev(pf); + int err; ice_set_pf_caps(pf); err = ice_init_interrupt_scheme(pf); @@ -5220,6 +5225,7 @@ static int ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent) { struct device *dev = &pdev->dev; + bool need_dev_deinit = false; struct ice_adapter *adapter; struct ice_pf *pf; struct ice_hw *hw; @@ -5342,11 +5348,13 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent) devl_unlock(priv_to_devlink(pf)); ice_deinit(pf); unroll_dev_init: - ice_deinit_dev(pf); + need_dev_deinit = true; unroll_adapter: ice_adapter_put(pdev); unroll_hw_init: ice_deinit_hw(hw); + if (need_dev_deinit) + ice_deinit_dev(pf); return err; } @@ -5441,10 +5449,6 @@ static void ice_remove(struct pci_dev *pdev) ice_hwmon_exit(pf); - ice_service_task_stop(pf); - ice_aq_cancel_waiting_tasks(pf); - set_bit(ICE_DOWN, pf->state); - if (!ice_is_safe_mode(pf)) ice_remove_arfs(pf); @@ -5456,13 +5460,16 @@ static void ice_remove(struct pci_dev *pdev) devl_unlock(priv_to_devlink(pf)); ice_deinit(pf); - ice_deinit_dev(pf); ice_vsi_release_all(pf); ice_setup_mc_magic_wake(pf); ice_set_wake(pf); ice_adapter_put(pdev); + + ice_deinit_dev(pf); + ice_aq_cancel_waiting_tasks(pf); + set_bit(ICE_DOWN, pf->state); } /** From 1390b8b3d2bef9bfbb852fc735430798bfca36e7 Mon Sep 17 00:00:00 2001 From: Przemek Kitszel Date: Fri, 12 Sep 2025 15:06:27 +0200 Subject: [PATCH 9/9] ice: remove duplicate call to ice_deinit_hw() on error paths Current unwinding code on error paths of ice_devlink_reinit_up() and ice_probe() have manual call to ice_deinit_hw() (which is good, as there is also manual call to ice_hw_init() there), which is then duplicated (and was prior current series) in ice_deinit_dev(). Fix the above by removing ice_deinit_hw() from ice_deinit_dev(). Add a (now missing) call in ice_remove(). Reported-by: Jacob Keller Link: https://lore.kernel.org/intel-wired-lan/20250717-jk-ddp-safe-mode-issue-v1-1-e113b2baed79@intel.com/ Fixes: 4d3f59bfa2cd ("ice: split ice_init_hw() out from ice_init_dev()") Signed-off-by: Przemek Kitszel Reviewed-by: Aleksandr Loktionov Tested-by: Rinitha S (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index a1fe2d363adb..1de3da7b3907 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -4784,7 +4784,6 @@ int ice_init_dev(struct ice_pf *pf) void ice_deinit_dev(struct ice_pf *pf) { - ice_deinit_hw(&pf->hw); ice_service_task_stop(pf); /* Service task is already stopped, so call reset directly. */ @@ -5466,6 +5465,7 @@ static void ice_remove(struct pci_dev *pdev) ice_set_wake(pf); ice_adapter_put(pdev); + ice_deinit_hw(&pf->hw); ice_deinit_dev(pf); ice_aq_cancel_waiting_tasks(pf);