From b70bc066d77b460a63a8c3fb2ea0d811ce862a83 Mon Sep 17 00:00:00 2001 From: Wojciech Drewek Date: Fri, 17 Dec 2021 12:36:25 +0100 Subject: ice: Match on all profiles in slow-path In switchdev mode, slow-path rules need to match all protocols, in order to correctly redirect unfiltered or missed packets to the uplink. To set this up for the virtual function to uplink flow, the rule that redirects packets to the control VSI must have the tunnel type set to ICE_SW_TUN_AND_NON_TUN. As a result of that new tunnel type being set, ice_get_compat_fv_bitmap will select ICE_PROF_ALL. At that point all profiles would be selected for this rule, resulting in the desired behavior. Without this change slow-path would not work with tunnel protocols. Fixes: 8b032a55c1bd ("ice: low level support for tunnels") Signed-off-by: Wojciech Drewek Tested-by: Sandeep Penigalapati Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_eswitch.c | 1 + drivers/net/ethernet/intel/ice/ice_protocol_type.h | 1 + drivers/net/ethernet/intel/ice/ice_switch.c | 4 +++- 3 files changed, 5 insertions(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c index 864692b157b6..73edc24d81d5 100644 --- a/drivers/net/ethernet/intel/ice/ice_eswitch.c +++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c @@ -44,6 +44,7 @@ ice_eswitch_add_vf_mac_rule(struct ice_pf *pf, struct ice_vf *vf, const u8 *mac) ctrl_vsi->rxq_map[vf->vf_id]; rule_info.flags_info.act |= ICE_SINGLE_ACT_LB_ENABLE; rule_info.flags_info.act_valid = true; + rule_info.tun_type = ICE_SW_TUN_AND_NON_TUN; err = ice_add_adv_rule(hw, list, lkups_cnt, &rule_info, vf->repr->mac_rule); diff --git a/drivers/net/ethernet/intel/ice/ice_protocol_type.h b/drivers/net/ethernet/intel/ice/ice_protocol_type.h index dc1b0e9e6df5..695b6dd61dc2 100644 --- a/drivers/net/ethernet/intel/ice/ice_protocol_type.h +++ b/drivers/net/ethernet/intel/ice/ice_protocol_type.h @@ -47,6 +47,7 @@ enum ice_protocol_type { enum ice_sw_tunnel_type { ICE_NON_TUN = 0, + ICE_SW_TUN_AND_NON_TUN, ICE_SW_TUN_VXLAN, ICE_SW_TUN_GENEVE, ICE_SW_TUN_NVGRE, diff --git a/drivers/net/ethernet/intel/ice/ice_switch.c b/drivers/net/ethernet/intel/ice/ice_switch.c index 11ae0bee3590..475ec2afa210 100644 --- a/drivers/net/ethernet/intel/ice/ice_switch.c +++ b/drivers/net/ethernet/intel/ice/ice_switch.c @@ -4537,6 +4537,7 @@ ice_get_compat_fv_bitmap(struct ice_hw *hw, struct ice_adv_rule_info *rinfo, case ICE_SW_TUN_NVGRE: prof_type = ICE_PROF_TUN_GRE; break; + case ICE_SW_TUN_AND_NON_TUN: default: prof_type = ICE_PROF_ALL; break; @@ -5305,7 +5306,8 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups, if (status) goto err_ice_add_adv_rule; - if (rinfo->tun_type != ICE_NON_TUN) { + if (rinfo->tun_type != ICE_NON_TUN && + rinfo->tun_type != ICE_SW_TUN_AND_NON_TUN) { status = ice_fill_adv_packet_tun(hw, rinfo->tun_type, s_rule->pdata.lkup_tx_rx.hdr, pkt_offsets); -- cgit v1.2.3 From 932645c298c41aad64ef13016ff4c2034eef5aed Mon Sep 17 00:00:00 2001 From: Michal Swiatkowski Date: Mon, 3 Jan 2022 07:41:21 +0100 Subject: ice: fix setting l4 port flag when adding filter Accidentally filter flag for none encapsulated l4 port field is always set. Even if user wants to add encapsulated l4 port field. Remove this unnecessary flag setting. Fixes: 9e300987d4a81 ("ice: VXLAN and Geneve TC support") Signed-off-by: Michal Swiatkowski Tested-by: Sandeep Penigalapati Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_tc_lib.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_tc_lib.c b/drivers/net/ethernet/intel/ice/ice_tc_lib.c index e8aab664270a..65cf32eb4046 100644 --- a/drivers/net/ethernet/intel/ice/ice_tc_lib.c +++ b/drivers/net/ethernet/intel/ice/ice_tc_lib.c @@ -709,7 +709,7 @@ ice_tc_set_port(struct flow_match_ports match, fltr->flags |= ICE_TC_FLWR_FIELD_ENC_DEST_L4_PORT; else fltr->flags |= ICE_TC_FLWR_FIELD_DEST_L4_PORT; - fltr->flags |= ICE_TC_FLWR_FIELD_DEST_L4_PORT; + headers->l4_key.dst_port = match.key->dst; headers->l4_mask.dst_port = match.mask->dst; } @@ -718,7 +718,7 @@ ice_tc_set_port(struct flow_match_ports match, fltr->flags |= ICE_TC_FLWR_FIELD_ENC_SRC_L4_PORT; else fltr->flags |= ICE_TC_FLWR_FIELD_SRC_L4_PORT; - fltr->flags |= ICE_TC_FLWR_FIELD_SRC_L4_PORT; + headers->l4_key.src_port = match.key->src; headers->l4_mask.src_port = match.mask->src; } -- cgit v1.2.3 From fadead80fe4c033b5e514fcbadd20b55c4494112 Mon Sep 17 00:00:00 2001 From: Jacob Keller Date: Mon, 7 Feb 2022 10:23:29 -0800 Subject: ice: fix concurrent reset and removal of VFs Commit c503e63200c6 ("ice: Stop processing VF messages during teardown") introduced a driver state flag, ICE_VF_DEINIT_IN_PROGRESS, which is intended to prevent some issues with concurrently handling messages from VFs while tearing down the VFs. This change was motivated by crashes caused while tearing down and bringing up VFs in rapid succession. It turns out that the fix actually introduces issues with the VF driver caused because the PF no longer responds to any messages sent by the VF during its .remove routine. This results in the VF potentially removing its DMA memory before the PF has shut down the device queues. Additionally, the fix doesn't actually resolve concurrency issues within the ice driver. It is possible for a VF to initiate a reset just prior to the ice driver removing VFs. This can result in the remove task concurrently operating while the VF is being reset. This results in similar memory corruption and panics purportedly fixed by that commit. Fix this concurrency at its root by protecting both the reset and removal flows using the existing VF cfg_lock. This ensures that we cannot remove the VF while any outstanding critical tasks such as a virtchnl message or a reset are occurring. This locking change also fixes the root cause originally fixed by commit c503e63200c6 ("ice: Stop processing VF messages during teardown"), so we can simply revert it. Note that I kept these two changes together because simply reverting the original commit alone would leave the driver vulnerable to worse race conditions. Fixes: c503e63200c6 ("ice: Stop processing VF messages during teardown") Signed-off-by: Jacob Keller Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice.h | 1 - drivers/net/ethernet/intel/ice/ice_main.c | 2 ++ drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c | 42 ++++++++++++++---------- 3 files changed, 27 insertions(+), 18 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h index a9fa701aaa95..473b1f6be9de 100644 --- a/drivers/net/ethernet/intel/ice/ice.h +++ b/drivers/net/ethernet/intel/ice/ice.h @@ -280,7 +280,6 @@ enum ice_pf_state { ICE_VFLR_EVENT_PENDING, ICE_FLTR_OVERFLOW_PROMISC, ICE_VF_DIS, - ICE_VF_DEINIT_IN_PROGRESS, ICE_CFG_BUSY, ICE_SERVICE_SCHED, ICE_SERVICE_DIS, diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index 17a9bb461dc3..f3c346e13b7a 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -1799,7 +1799,9 @@ static void ice_handle_mdd_event(struct ice_pf *pf) * reset, so print the event prior to reset. */ ice_print_vf_rx_mdd_event(vf); + mutex_lock(&pf->vf[i].cfg_lock); ice_reset_vf(&pf->vf[i], false); + mutex_unlock(&pf->vf[i].cfg_lock); } } } diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c index 39b80124d282..408f78e3eb13 100644 --- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c +++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c @@ -500,8 +500,6 @@ void ice_free_vfs(struct ice_pf *pf) struct ice_hw *hw = &pf->hw; unsigned int tmp, i; - set_bit(ICE_VF_DEINIT_IN_PROGRESS, pf->state); - if (!pf->vf) return; @@ -519,22 +517,26 @@ void ice_free_vfs(struct ice_pf *pf) else dev_warn(dev, "VFs are assigned - not disabling SR-IOV\n"); - /* Avoid wait time by stopping all VFs at the same time */ - ice_for_each_vf(pf, i) - ice_dis_vf_qs(&pf->vf[i]); - tmp = pf->num_alloc_vfs; pf->num_qps_per_vf = 0; pf->num_alloc_vfs = 0; for (i = 0; i < tmp; i++) { - if (test_bit(ICE_VF_STATE_INIT, pf->vf[i].vf_states)) { + struct ice_vf *vf = &pf->vf[i]; + + mutex_lock(&vf->cfg_lock); + + ice_dis_vf_qs(vf); + + if (test_bit(ICE_VF_STATE_INIT, vf->vf_states)) { /* disable VF qp mappings and set VF disable state */ - ice_dis_vf_mappings(&pf->vf[i]); - set_bit(ICE_VF_STATE_DIS, pf->vf[i].vf_states); - ice_free_vf_res(&pf->vf[i]); + ice_dis_vf_mappings(vf); + set_bit(ICE_VF_STATE_DIS, vf->vf_states); + ice_free_vf_res(vf); } - mutex_destroy(&pf->vf[i].cfg_lock); + mutex_unlock(&vf->cfg_lock); + + mutex_destroy(&vf->cfg_lock); } if (ice_sriov_free_msix_res(pf)) @@ -570,7 +572,6 @@ void ice_free_vfs(struct ice_pf *pf) i); clear_bit(ICE_VF_DIS, pf->state); - clear_bit(ICE_VF_DEINIT_IN_PROGRESS, pf->state); clear_bit(ICE_FLAG_SRIOV_ENA, pf->flags); } @@ -1498,6 +1499,8 @@ bool ice_reset_all_vfs(struct ice_pf *pf, bool is_vflr) ice_for_each_vf(pf, v) { vf = &pf->vf[v]; + mutex_lock(&vf->cfg_lock); + vf->driver_caps = 0; ice_vc_set_default_allowlist(vf); @@ -1512,6 +1515,8 @@ bool ice_reset_all_vfs(struct ice_pf *pf, bool is_vflr) ice_vf_pre_vsi_rebuild(vf); ice_vf_rebuild_vsi(vf); ice_vf_post_vsi_rebuild(vf); + + mutex_unlock(&vf->cfg_lock); } if (ice_is_eswitch_mode_switchdev(pf)) @@ -1562,6 +1567,8 @@ bool ice_reset_vf(struct ice_vf *vf, bool is_vflr) u32 reg; int i; + lockdep_assert_held(&vf->cfg_lock); + dev = ice_pf_to_dev(pf); if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) { @@ -2061,9 +2068,12 @@ void ice_process_vflr_event(struct ice_pf *pf) bit_idx = (hw->func_caps.vf_base_id + vf_id) % 32; /* read GLGEN_VFLRSTAT register to find out the flr VFs */ reg = rd32(hw, GLGEN_VFLRSTAT(reg_idx)); - if (reg & BIT(bit_idx)) + if (reg & BIT(bit_idx)) { /* GLGEN_VFLRSTAT bit will be cleared in ice_reset_vf */ + mutex_lock(&vf->cfg_lock); ice_reset_vf(vf, true); + mutex_unlock(&vf->cfg_lock); + } } } @@ -2140,7 +2150,9 @@ ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event) if (!vf) return; + mutex_lock(&vf->cfg_lock); ice_vc_reset_vf(vf); + mutex_unlock(&vf->cfg_lock); } /** @@ -4625,10 +4637,6 @@ void ice_vc_process_vf_msg(struct ice_pf *pf, struct ice_rq_event_info *event) struct device *dev; int err = 0; - /* if de-init is underway, don't process messages from VF */ - if (test_bit(ICE_VF_DEINIT_IN_PROGRESS, pf->state)) - return; - dev = ice_pf_to_dev(pf); if (ice_validate_vf_id(pf, vf_id)) { err = -EINVAL; -- cgit v1.2.3 From ed22d9c8d128293fc7b0b086c7d3654bcb99a8dd Mon Sep 17 00:00:00 2001 From: Tom Rix Date: Mon, 14 Feb 2022 06:33:27 -0800 Subject: ice: check the return of ice_ptp_gettimex64 Clang static analysis reports this issue time64.h:69:50: warning: The left operand of '+' is a garbage value set_normalized_timespec64(&ts_delta, lhs.tv_sec + rhs.tv_sec, ~~~~~~~~~~ ^ In ice_ptp_adjtime_nonatomic(), the timespec64 variable 'now' is set by ice_ptp_gettimex64(). This function can fail with -EBUSY, so 'now' can have a gargbage value. So check the return. Fixes: 06c16d89d2cb ("ice: register 1588 PTP clock device object for E810 devices") Signed-off-by: Tom Rix Tested-by: Gurucharan G (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_ptp.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_ptp.c b/drivers/net/ethernet/intel/ice/ice_ptp.c index ae291d442539..000c39d163a2 100644 --- a/drivers/net/ethernet/intel/ice/ice_ptp.c +++ b/drivers/net/ethernet/intel/ice/ice_ptp.c @@ -1533,9 +1533,12 @@ exit: static int ice_ptp_adjtime_nonatomic(struct ptp_clock_info *info, s64 delta) { struct timespec64 now, then; + int ret; then = ns_to_timespec64(delta); - ice_ptp_gettimex64(info, &now, NULL); + ret = ice_ptp_gettimex64(info, &now, NULL); + if (ret) + return ret; now = timespec64_add(now, then); return ice_ptp_settime64(info, (const struct timespec64 *)&now); -- cgit v1.2.3 From 5950bdc88dd1d158f2845fdff8fb1de86476806c Mon Sep 17 00:00:00 2001 From: Tom Rix Date: Mon, 14 Feb 2022 07:40:43 -0800 Subject: ice: initialize local variable 'tlv' Clang static analysis reports this issues ice_common.c:5008:21: warning: The left expression of the compound assignment is an uninitialized value. The computed value will also be garbage ldo->phy_type_low |= ((u64)buf << (i * 16)); ~~~~~~~~~~~~~~~~~ ^ When called from ice_cfg_phy_fec() ldo is the uninitialized local variable tlv. So initialize. Fixes: ea78ce4dab05 ("ice: add link lenient and default override support") Signed-off-by: Tom Rix Tested-by: Gurucharan G (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_common.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c index a6d7d3eff186..e2af99a763ed 100644 --- a/drivers/net/ethernet/intel/ice/ice_common.c +++ b/drivers/net/ethernet/intel/ice/ice_common.c @@ -3340,7 +3340,7 @@ ice_cfg_phy_fec(struct ice_port_info *pi, struct ice_aqc_set_phy_cfg_data *cfg, if (fec == ICE_FEC_AUTO && ice_fw_supports_link_override(hw) && !ice_fw_supports_report_dflt_cfg(hw)) { - struct ice_link_default_override_tlv tlv; + struct ice_link_default_override_tlv tlv = { 0 }; status = ice_get_link_default_override(&tlv, pi); if (status) -- cgit v1.2.3 From fe20371578ef640069e6ae9fa8038f60e7908565 Mon Sep 17 00:00:00 2001 From: Mateusz Palczewski Date: Wed, 23 Feb 2022 09:53:47 -0800 Subject: Revert "i40e: Fix reset bw limit when DCB enabled with 1 TC" Revert of a patch that instead of fixing a AQ error when trying to reset BW limit introduced several regressions related to creation and managing TC. Currently there are errors when creating a TC on both PF and VF. Error log: [17428.783095] i40e 0000:3b:00.1: AQ command Config VSI BW allocation per TC failed = 14 [17428.783107] i40e 0000:3b:00.1: Failed configuring TC map 0 for VSI 391 [17428.783254] i40e 0000:3b:00.1: AQ command Config VSI BW allocation per TC failed = 14 [17428.783259] i40e 0000:3b:00.1: Unable to configure TC map 0 for VSI 391 This reverts commit 3d2504663c41104b4359a15f35670cfa82de1bbf. Fixes: 3d2504663c41 (i40e: Fix reset bw limit when DCB enabled with 1 TC) Signed-off-by: Mateusz Palczewski Signed-off-by: Tony Nguyen Link: https://lore.kernel.org/r/20220223175347.1690692-1-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/intel/i40e/i40e_main.c | 12 +----------- 1 file changed, 1 insertion(+), 11 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index 0c4b7dfb3b35..31b03fe78d3b 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -5372,15 +5372,7 @@ static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc, /* There is no need to reset BW when mqprio mode is on. */ if (pf->flags & I40E_FLAG_TC_MQPRIO) return 0; - - if (!vsi->mqprio_qopt.qopt.hw) { - if (pf->flags & I40E_FLAG_DCB_ENABLED) - goto skip_reset; - - if (IS_ENABLED(CONFIG_I40E_DCB) && - i40e_dcb_hw_get_num_tc(&pf->hw) == 1) - goto skip_reset; - + if (!vsi->mqprio_qopt.qopt.hw && !(pf->flags & I40E_FLAG_DCB_ENABLED)) { ret = i40e_set_bw_limit(vsi, vsi->seid, 0); if (ret) dev_info(&pf->pdev->dev, @@ -5388,8 +5380,6 @@ static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc, vsi->seid); return ret; } - -skip_reset: memset(&bw_data, 0, sizeof(bw_data)); bw_data.tc_valid_bits = enabled_tc; for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) -- cgit v1.2.3 From fc2e6b3b132a907378f6af08356b105a4139c4fb Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:35:49 +0100 Subject: iavf: Rework mutexes for better synchronisation The driver used to crash in multiple spots when put to stress testing of the init, reset and remove paths. The user would experience call traces or hangs when creating, resetting, removing VFs. Depending on the machines, the call traces are happening in random spots, like reset restoring resources racing with driver remove. Make adapter->crit_lock mutex a mandatory lock for guarding the operations performed on all workqueues and functions dealing with resource allocation and disposal. Make __IAVF_REMOVE a final state of the driver respected by workqueues that shall not requeue, when they fail to obtain the crit_lock. Make the IRQ handler not to queue the new work for adminq_task when the __IAVF_REMOVE state is set. Fixes: 5ac49f3c2702 ("iavf: use mutexes for locking of critical sections") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf.h | 1 - drivers/net/ethernet/intel/iavf/iavf_main.c | 66 ++++++++++++++++------------- 2 files changed, 36 insertions(+), 31 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf.h b/drivers/net/ethernet/intel/iavf/iavf.h index 59806d1f7e79..44f83e06486d 100644 --- a/drivers/net/ethernet/intel/iavf/iavf.h +++ b/drivers/net/ethernet/intel/iavf/iavf.h @@ -246,7 +246,6 @@ struct iavf_adapter { struct list_head mac_filter_list; struct mutex crit_lock; struct mutex client_lock; - struct mutex remove_lock; /* Lock to protect accesses to MAC and VLAN lists */ spinlock_t mac_vlan_list_lock; char misc_vector_name[IFNAMSIZ + 9]; diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index 8125b9120615..84ae96e912d7 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -302,8 +302,9 @@ static irqreturn_t iavf_msix_aq(int irq, void *data) rd32(hw, IAVF_VFINT_ICR01); rd32(hw, IAVF_VFINT_ICR0_ENA1); - /* schedule work on the private workqueue */ - queue_work(iavf_wq, &adapter->adminq_task); + if (adapter->state != __IAVF_REMOVE) + /* schedule work on the private workqueue */ + queue_work(iavf_wq, &adapter->adminq_task); return IRQ_HANDLED; } @@ -2374,8 +2375,12 @@ static void iavf_watchdog_task(struct work_struct *work) struct iavf_hw *hw = &adapter->hw; u32 reg_val; - if (!mutex_trylock(&adapter->crit_lock)) + if (!mutex_trylock(&adapter->crit_lock)) { + if (adapter->state == __IAVF_REMOVE) + return; + goto restart_watchdog; + } if (adapter->flags & IAVF_FLAG_PF_COMMS_FAILED) iavf_change_state(adapter, __IAVF_COMM_FAILED); @@ -2601,13 +2606,13 @@ static void iavf_reset_task(struct work_struct *work) /* When device is being removed it doesn't make sense to run the reset * task, just return in such a case. */ - if (mutex_is_locked(&adapter->remove_lock)) - return; + if (!mutex_trylock(&adapter->crit_lock)) { + if (adapter->state != __IAVF_REMOVE) + queue_work(iavf_wq, &adapter->reset_task); - if (iavf_lock_timeout(&adapter->crit_lock, 200)) { - schedule_work(&adapter->reset_task); return; } + while (!mutex_trylock(&adapter->client_lock)) usleep_range(500, 1000); if (CLIENT_ENABLED(adapter)) { @@ -2826,13 +2831,19 @@ static void iavf_adminq_task(struct work_struct *work) if (adapter->flags & IAVF_FLAG_PF_COMMS_FAILED) goto out; + if (!mutex_trylock(&adapter->crit_lock)) { + if (adapter->state == __IAVF_REMOVE) + return; + + queue_work(iavf_wq, &adapter->adminq_task); + goto out; + } + event.buf_len = IAVF_MAX_AQ_BUF_SIZE; event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL); if (!event.msg_buf) goto out; - if (iavf_lock_timeout(&adapter->crit_lock, 200)) - goto freedom; do { ret = iavf_clean_arq_element(hw, &event, &pending); v_op = (enum virtchnl_ops)le32_to_cpu(event.desc.cookie_high); @@ -3800,11 +3811,12 @@ static int iavf_close(struct net_device *netdev) struct iavf_adapter *adapter = netdev_priv(netdev); int status; - if (adapter->state <= __IAVF_DOWN_PENDING) - return 0; + mutex_lock(&adapter->crit_lock); - while (!mutex_trylock(&adapter->crit_lock)) - usleep_range(500, 1000); + if (adapter->state <= __IAVF_DOWN_PENDING) { + mutex_unlock(&adapter->crit_lock); + return 0; + } set_bit(__IAVF_VSI_DOWN, adapter->vsi.state); if (CLIENT_ENABLED(adapter)) @@ -4431,7 +4443,6 @@ static int iavf_probe(struct pci_dev *pdev, const struct pci_device_id *ent) */ mutex_init(&adapter->crit_lock); mutex_init(&adapter->client_lock); - mutex_init(&adapter->remove_lock); mutex_init(&hw->aq.asq_mutex); mutex_init(&hw->aq.arq_mutex); @@ -4556,11 +4567,7 @@ static void iavf_remove(struct pci_dev *pdev) struct iavf_cloud_filter *cf, *cftmp; struct iavf_hw *hw = &adapter->hw; int err; - /* Indicate we are in remove and not to run reset_task */ - mutex_lock(&adapter->remove_lock); - cancel_work_sync(&adapter->reset_task); - cancel_delayed_work_sync(&adapter->watchdog_task); - cancel_delayed_work_sync(&adapter->client_task); + if (adapter->netdev_registered) { unregister_netdev(netdev); adapter->netdev_registered = false; @@ -4572,6 +4579,10 @@ static void iavf_remove(struct pci_dev *pdev) err); } + mutex_lock(&adapter->crit_lock); + dev_info(&adapter->pdev->dev, "Remove device\n"); + iavf_change_state(adapter, __IAVF_REMOVE); + iavf_request_reset(adapter); msleep(50); /* If the FW isn't responding, kick it once, but only once. */ @@ -4579,18 +4590,19 @@ static void iavf_remove(struct pci_dev *pdev) iavf_request_reset(adapter); msleep(50); } - if (iavf_lock_timeout(&adapter->crit_lock, 5000)) - dev_warn(&adapter->pdev->dev, "failed to acquire crit_lock in %s\n", __FUNCTION__); - dev_info(&adapter->pdev->dev, "Removing device\n"); + iavf_misc_irq_disable(adapter); /* Shut down all the garbage mashers on the detention level */ - iavf_change_state(adapter, __IAVF_REMOVE); + cancel_work_sync(&adapter->reset_task); + cancel_delayed_work_sync(&adapter->watchdog_task); + cancel_work_sync(&adapter->adminq_task); + cancel_delayed_work_sync(&adapter->client_task); + adapter->aq_required = 0; adapter->flags &= ~IAVF_FLAG_REINIT_ITR_NEEDED; iavf_free_all_tx_resources(adapter); iavf_free_all_rx_resources(adapter); - iavf_misc_irq_disable(adapter); iavf_free_misc_irq(adapter); /* In case we enter iavf_remove from erroneous state, free traffic irqs @@ -4606,10 +4618,6 @@ static void iavf_remove(struct pci_dev *pdev) iavf_reset_interrupt_capability(adapter); iavf_free_q_vectors(adapter); - cancel_delayed_work_sync(&adapter->watchdog_task); - - cancel_work_sync(&adapter->adminq_task); - iavf_free_rss(adapter); if (hw->aq.asq.count) @@ -4621,8 +4629,6 @@ static void iavf_remove(struct pci_dev *pdev) mutex_destroy(&adapter->client_lock); mutex_unlock(&adapter->crit_lock); mutex_destroy(&adapter->crit_lock); - mutex_unlock(&adapter->remove_lock); - mutex_destroy(&adapter->remove_lock); iounmap(hw->hw_addr); pci_release_regions(pdev); -- cgit v1.2.3 From 974578017fc1fdd06cea8afb9dfa32602e8529ed Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:36:56 +0100 Subject: iavf: Add waiting so the port is initialized in remove There exist races when port is being configured and remove is triggered. unregister_netdev is not and can't be called under crit_lock mutex since it is calling ndo_stop -> iavf_close which requires this lock. Depending on init state the netdev could be still unregistered so unregister_netdev never cleans up, when shortly after that the device could become registered. Make iavf_remove wait until port finishes initialization. All critical state changes are atomic (under crit_lock). Crashes that come from iavf_reset_interrupt_capability and iavf_free_traffic_irqs should now be solved in a graceful manner. Fixes: 605ca7c5c6707 ("iavf: Fix kernel BUG in free_msi_irqs") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_main.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index 84ae96e912d7..5e71b38e9154 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -4558,7 +4558,6 @@ static int __maybe_unused iavf_resume(struct device *dev_d) static void iavf_remove(struct pci_dev *pdev) { struct iavf_adapter *adapter = iavf_pdev_to_adapter(pdev); - enum iavf_state_t prev_state = adapter->last_state; struct net_device *netdev = adapter->netdev; struct iavf_fdir_fltr *fdir, *fdirtmp; struct iavf_vlan_filter *vlf, *vlftmp; @@ -4568,6 +4567,22 @@ static void iavf_remove(struct pci_dev *pdev) struct iavf_hw *hw = &adapter->hw; int err; + /* Wait until port initialization is complete. + * There are flows where register/unregister netdev may race. + */ + while (1) { + mutex_lock(&adapter->crit_lock); + if (adapter->state == __IAVF_RUNNING || + adapter->state == __IAVF_DOWN) { + mutex_unlock(&adapter->crit_lock); + break; + } + + mutex_unlock(&adapter->crit_lock); + usleep_range(500, 1000); + } + cancel_delayed_work_sync(&adapter->watchdog_task); + if (adapter->netdev_registered) { unregister_netdev(netdev); adapter->netdev_registered = false; @@ -4605,16 +4620,6 @@ static void iavf_remove(struct pci_dev *pdev) iavf_free_all_rx_resources(adapter); iavf_free_misc_irq(adapter); - /* In case we enter iavf_remove from erroneous state, free traffic irqs - * here, so as to not cause a kernel crash, when calling - * iavf_reset_interrupt_capability. - */ - if ((adapter->last_state == __IAVF_RESETTING && - prev_state != __IAVF_DOWN) || - (adapter->last_state == __IAVF_RUNNING && - !(netdev->flags & IFF_UP))) - iavf_free_traffic_irqs(adapter); - iavf_reset_interrupt_capability(adapter); iavf_free_q_vectors(adapter); -- cgit v1.2.3 From 3ccd54ef44ebfa0792c5441b6d9c86618f3378d1 Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:37:10 +0100 Subject: iavf: Fix init state closure on remove When init states of the adapter work, the errors like lack of communication with the PF might hop in. If such events occur the driver restores previous states in order to retry initialization in a proper way. When remove task kicks in, this situation could lead to races with unregistering the netdevice as well as resources cleanup. With the commit introducing the waiting in remove for init to complete, this problem turns into an endless waiting if init never recovers from errors. Introduce __IAVF_IN_REMOVE_TASK bit to indicate that the remove thread has started. Make __IAVF_COMM_FAILED adapter state respect the __IAVF_IN_REMOVE_TASK bit and set the __IAVF_INIT_FAILED state and return without any action instead of trying to recover. Make __IAVF_INIT_FAILED adapter state respect the __IAVF_IN_REMOVE_TASK bit and return without any further actions. Make the loop in the remove handler break when adapter has __IAVF_INIT_FAILED state set. Fixes: 898ef1cb1cb2 ("iavf: Combine init and watchdog state machines") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf.h | 4 ++++ drivers/net/ethernet/intel/iavf/iavf_main.c | 24 +++++++++++++++++++++++- 2 files changed, 27 insertions(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf.h b/drivers/net/ethernet/intel/iavf/iavf.h index 44f83e06486d..f259fd517b2c 100644 --- a/drivers/net/ethernet/intel/iavf/iavf.h +++ b/drivers/net/ethernet/intel/iavf/iavf.h @@ -201,6 +201,10 @@ enum iavf_state_t { __IAVF_RUNNING, /* opened, working */ }; +enum iavf_critical_section_t { + __IAVF_IN_REMOVE_TASK, /* device being removed */ +}; + #define IAVF_CLOUD_FIELD_OMAC 0x01 #define IAVF_CLOUD_FIELD_IMAC 0x02 #define IAVF_CLOUD_FIELD_IVLAN 0x04 diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index 5e71b38e9154..be51da978e7c 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -2424,6 +2424,15 @@ static void iavf_watchdog_task(struct work_struct *work) msecs_to_jiffies(1)); return; case __IAVF_INIT_FAILED: + if (test_bit(__IAVF_IN_REMOVE_TASK, + &adapter->crit_section)) { + /* Do not update the state and do not reschedule + * watchdog task, iavf_remove should handle this state + * as it can loop forever + */ + mutex_unlock(&adapter->crit_lock); + return; + } if (++adapter->aq_wait_count > IAVF_AQ_MAX_ERR) { dev_err(&adapter->pdev->dev, "Failed to communicate with PF; waiting before retry\n"); @@ -2440,6 +2449,17 @@ static void iavf_watchdog_task(struct work_struct *work) queue_delayed_work(iavf_wq, &adapter->watchdog_task, HZ); return; case __IAVF_COMM_FAILED: + if (test_bit(__IAVF_IN_REMOVE_TASK, + &adapter->crit_section)) { + /* Set state to __IAVF_INIT_FAILED and perform remove + * steps. Remove IAVF_FLAG_PF_COMMS_FAILED so the task + * doesn't bring the state back to __IAVF_COMM_FAILED. + */ + iavf_change_state(adapter, __IAVF_INIT_FAILED); + adapter->flags &= ~IAVF_FLAG_PF_COMMS_FAILED; + mutex_unlock(&adapter->crit_lock); + return; + } reg_val = rd32(hw, IAVF_VFGEN_RSTAT) & IAVF_VFGEN_RSTAT_VFR_STATE_MASK; if (reg_val == VIRTCHNL_VFR_VFACTIVE || @@ -4567,13 +4587,15 @@ static void iavf_remove(struct pci_dev *pdev) struct iavf_hw *hw = &adapter->hw; int err; + set_bit(__IAVF_IN_REMOVE_TASK, &adapter->crit_section); /* Wait until port initialization is complete. * There are flows where register/unregister netdev may race. */ while (1) { mutex_lock(&adapter->crit_lock); if (adapter->state == __IAVF_RUNNING || - adapter->state == __IAVF_DOWN) { + adapter->state == __IAVF_DOWN || + adapter->state == __IAVF_INIT_FAILED) { mutex_unlock(&adapter->crit_lock); break; } -- cgit v1.2.3 From 0579fafd37fb7efe091f0e6c8ccf968864f40f3e Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:37:50 +0100 Subject: iavf: Fix locking for VIRTCHNL_OP_GET_OFFLOAD_VLAN_V2_CAPS iavf_virtchnl_completion is called under crit_lock but when the code for VIRTCHNL_OP_GET_OFFLOAD_VLAN_V2_CAPS is called, this lock is released in order to obtain rtnl_lock to avoid ABBA deadlock with unregister_netdev. Along with the new way iavf_remove behaves, there exist many risks related to the lock release and attmepts to regrab it. The driver faces crashes related to races between unregister_netdev and netdev_update_features. Yet another risk is that the driver could already obtain the crit_lock in order to destroy it and iavf_virtchnl_completion could crash or block forever. Make iavf_virtchnl_completion never relock crit_lock in it's call paths. Extract rtnl_lock locking logic to the driver for unregister_netdev in order to set the netdev_registered flag inside the lock. Introduce a new flag that will inform adminq_task to perform the code from VIRTCHNL_OP_GET_OFFLOAD_VLAN_V2_CAPS right after it finishes processing messages. Guard this code with remove flags so it's never called when the driver is in remove state. Fixes: 5951a2b9812d ("iavf: Fix VLAN feature flags after VFR") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf.h | 1 + drivers/net/ethernet/intel/iavf/iavf_main.c | 22 +++++++++++++++++++++- drivers/net/ethernet/intel/iavf/iavf_virtchnl.c | 24 +----------------------- 3 files changed, 23 insertions(+), 24 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf.h b/drivers/net/ethernet/intel/iavf/iavf.h index f259fd517b2c..89423947ee65 100644 --- a/drivers/net/ethernet/intel/iavf/iavf.h +++ b/drivers/net/ethernet/intel/iavf/iavf.h @@ -287,6 +287,7 @@ struct iavf_adapter { #define IAVF_FLAG_LEGACY_RX BIT(15) #define IAVF_FLAG_REINIT_ITR_NEEDED BIT(16) #define IAVF_FLAG_QUEUES_DISABLED BIT(17) +#define IAVF_FLAG_SETUP_NETDEV_FEATURES BIT(18) /* duplicates for common code */ #define IAVF_FLAG_DCB_ENABLED 0 /* flags for admin queue service task */ diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index be51da978e7c..67349d24dc90 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -2879,6 +2879,24 @@ static void iavf_adminq_task(struct work_struct *work) } while (pending); mutex_unlock(&adapter->crit_lock); + if ((adapter->flags & IAVF_FLAG_SETUP_NETDEV_FEATURES)) { + if (adapter->netdev_registered || + !test_bit(__IAVF_IN_REMOVE_TASK, &adapter->crit_section)) { + struct net_device *netdev = adapter->netdev; + + rtnl_lock(); + netdev_update_features(netdev); + rtnl_unlock(); + /* Request VLAN offload settings */ + if (VLAN_V2_ALLOWED(adapter)) + iavf_set_vlan_offload_features + (adapter, 0, netdev->features); + + iavf_set_queue_vlan_tag_loc(adapter); + } + + adapter->flags &= ~IAVF_FLAG_SETUP_NETDEV_FEATURES; + } if ((adapter->flags & (IAVF_FLAG_RESET_PENDING | IAVF_FLAG_RESET_NEEDED)) || adapter->state == __IAVF_RESETTING) @@ -4606,8 +4624,10 @@ static void iavf_remove(struct pci_dev *pdev) cancel_delayed_work_sync(&adapter->watchdog_task); if (adapter->netdev_registered) { - unregister_netdev(netdev); + rtnl_lock(); + unregister_netdevice(netdev); adapter->netdev_registered = false; + rtnl_unlock(); } if (CLIENT_ALLOWED(adapter)) { err = iavf_lan_del_device(adapter); diff --git a/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c b/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c index 5ee1d118fd30..88844d68e150 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c +++ b/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c @@ -2146,29 +2146,7 @@ void iavf_virtchnl_completion(struct iavf_adapter *adapter, sizeof(adapter->vlan_v2_caps))); iavf_process_config(adapter); - - /* unlock crit_lock before acquiring rtnl_lock as other - * processes holding rtnl_lock could be waiting for the same - * crit_lock - */ - mutex_unlock(&adapter->crit_lock); - /* VLAN capabilities can change during VFR, so make sure to - * update the netdev features with the new capabilities - */ - rtnl_lock(); - netdev_update_features(netdev); - rtnl_unlock(); - if (iavf_lock_timeout(&adapter->crit_lock, 10000)) - dev_warn(&adapter->pdev->dev, "failed to acquire crit_lock in %s\n", - __FUNCTION__); - - /* Request VLAN offload settings */ - if (VLAN_V2_ALLOWED(adapter)) - iavf_set_vlan_offload_features(adapter, 0, - netdev->features); - - iavf_set_queue_vlan_tag_loc(adapter); - + adapter->flags |= IAVF_FLAG_SETUP_NETDEV_FEATURES; } break; case VIRTCHNL_OP_ENABLE_QUEUES: -- cgit v1.2.3 From a472eb5cbaebb5774672c565e024336c039e9128 Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:38:01 +0100 Subject: iavf: Fix race in init state When iavf_init_version_check sends VIRTCHNL_OP_GET_VF_RESOURCES message, the driver will wait for the response after requeueing the watchdog task in iavf_init_get_resources call stack. The logic is implemented this way that iavf_init_get_resources has to be called in order to allocate adapter->vf_res. It is polling for the AQ response in iavf_get_vf_config function. Expect a call trace from kernel when adminq_task worker handles this message first. adapter->vf_res will be NULL in iavf_virtchnl_completion. Make the watchdog task not queue the adminq_task if the init process is not finished yet. Fixes: 898ef1cb1cb2 ("iavf: Combine init and watchdog state machines") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_main.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index 67349d24dc90..36433d6504b7 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -2532,7 +2532,8 @@ static void iavf_watchdog_task(struct work_struct *work) schedule_delayed_work(&adapter->client_task, msecs_to_jiffies(5)); mutex_unlock(&adapter->crit_lock); restart_watchdog: - queue_work(iavf_wq, &adapter->adminq_task); + if (adapter->state >= __IAVF_DOWN) + queue_work(iavf_wq, &adapter->adminq_task); if (adapter->aq_required) queue_delayed_work(iavf_wq, &adapter->watchdog_task, msecs_to_jiffies(20)); -- cgit v1.2.3 From e85ff9c631e1bf109ce8428848dfc8e8b0041f48 Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:38:31 +0100 Subject: iavf: Fix deadlock in iavf_reset_task There exists a missing mutex_unlock call on crit_lock in iavf_reset_task call path. Unlock the crit_lock before returning from reset task. Fixes: 5ac49f3c2702 ("iavf: use mutexes for locking of critical sections") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_main.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index 36433d6504b7..da50ea3e44c2 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -2688,6 +2688,7 @@ static void iavf_reset_task(struct work_struct *work) reg_val); iavf_disable_vf(adapter); mutex_unlock(&adapter->client_lock); + mutex_unlock(&adapter->crit_lock); return; /* Do not attempt to reinit. It's dead, Jim. */ } -- cgit v1.2.3 From d2c0f45fcceb0995f208c441d9c9a453623f9ccf Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:38:43 +0100 Subject: iavf: Fix missing check for running netdev The driver was queueing reset_task regardless of the netdev state. Do not queue the reset task in iavf_change_mtu if netdev is not running. Fixes: fdd4044ffdc8 ("iavf: Remove timer for work triggering, use delaying work instead") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_main.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index da50ea3e44c2..be97ac251802 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -3905,8 +3905,11 @@ static int iavf_change_mtu(struct net_device *netdev, int new_mtu) iavf_notify_client_l2_params(&adapter->vsi); adapter->flags |= IAVF_FLAG_SERVICE_CLIENT_REQUESTED; } - adapter->flags |= IAVF_FLAG_RESET_NEEDED; - queue_work(iavf_wq, &adapter->reset_task); + + if (netif_running(netdev)) { + adapter->flags |= IAVF_FLAG_RESET_NEEDED; + queue_work(iavf_wq, &adapter->reset_task); + } return 0; } -- cgit v1.2.3 From 14756b2ae265d526b8356e86729090b01778fdf6 Mon Sep 17 00:00:00 2001 From: Slawomir Laba Date: Wed, 23 Feb 2022 13:38:55 +0100 Subject: iavf: Fix __IAVF_RESETTING state usage The setup of __IAVF_RESETTING state in watchdog task had no effect and could lead to slow resets in the driver as the task for __IAVF_RESETTING state only requeues watchdog. Till now the __IAVF_RESETTING was interpreted by reset task as running state which could lead to errors with allocating and resources disposal. Make watchdog_task queue the reset task when it's necessary. Do not update the state to __IAVF_RESETTING so the reset task knows exactly what is the current state of the adapter. Fixes: 898ef1cb1cb2 ("iavf: Combine init and watchdog state machines") Signed-off-by: Slawomir Laba Signed-off-by: Phani Burra Signed-off-by: Jacob Keller Signed-off-by: Mateusz Palczewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_main.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index be97ac251802..dcf24264c7ea 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -1137,8 +1137,7 @@ void iavf_down(struct iavf_adapter *adapter) rss->state = IAVF_ADV_RSS_DEL_REQUEST; spin_unlock_bh(&adapter->adv_rss_lock); - if (!(adapter->flags & IAVF_FLAG_PF_COMMS_FAILED) && - adapter->state != __IAVF_RESETTING) { + if (!(adapter->flags & IAVF_FLAG_PF_COMMS_FAILED)) { /* cancel any current operation */ adapter->current_op = VIRTCHNL_OP_UNKNOWN; /* Schedule operations to close down the HW. Don't wait @@ -2385,11 +2384,12 @@ static void iavf_watchdog_task(struct work_struct *work) if (adapter->flags & IAVF_FLAG_PF_COMMS_FAILED) iavf_change_state(adapter, __IAVF_COMM_FAILED); - if (adapter->flags & IAVF_FLAG_RESET_NEEDED && - adapter->state != __IAVF_RESETTING) { - iavf_change_state(adapter, __IAVF_RESETTING); + if (adapter->flags & IAVF_FLAG_RESET_NEEDED) { adapter->aq_required = 0; adapter->current_op = VIRTCHNL_OP_UNKNOWN; + mutex_unlock(&adapter->crit_lock); + queue_work(iavf_wq, &adapter->reset_task); + return; } switch (adapter->state) { @@ -2697,8 +2697,7 @@ continue_reset: * ndo_open() returning, so we can't assume it means all our open * tasks have finished, since we're not holding the rtnl_lock here. */ - running = ((adapter->state == __IAVF_RUNNING) || - (adapter->state == __IAVF_RESETTING)); + running = adapter->state == __IAVF_RUNNING; if (running) { netdev->flags &= ~IFF_UP; -- cgit v1.2.3 From fda2635466cd26ad237e1bc5d3f6a60f97ad09b6 Mon Sep 17 00:00:00 2001 From: Corinna Vinschen Date: Wed, 16 Feb 2022 14:31:35 +0100 Subject: igc: igc_read_phy_reg_gpy: drop premature return igc_read_phy_reg_gpy checks the return value from igc_read_phy_reg_mdic and if it's not 0, returns immediately. By doing this, it leaves the HW semaphore in the acquired state. Drop this premature return statement, the function returns after releasing the semaphore immediately anyway. Fixes: 5586838fe9ce ("igc: Add code for PHY support") Signed-off-by: Corinna Vinschen Acked-by: Sasha Neftin Tested-by: Naama Meir Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/igc/igc_phy.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/igc/igc_phy.c b/drivers/net/ethernet/intel/igc/igc_phy.c index 5cad31c3c7b0..df91d07ce82a 100644 --- a/drivers/net/ethernet/intel/igc/igc_phy.c +++ b/drivers/net/ethernet/intel/igc/igc_phy.c @@ -779,8 +779,6 @@ s32 igc_read_phy_reg_gpy(struct igc_hw *hw, u32 offset, u16 *data) if (ret_val) return ret_val; ret_val = igc_read_phy_reg_mdic(hw, offset, data); - if (ret_val) - return ret_val; hw->phy.ops.release(hw); } else { ret_val = igc_read_xmdio_reg(hw, (u16)offset, dev_addr, -- cgit v1.2.3 From c4208653a327a09da1e9e7b10299709b6d9b17bf Mon Sep 17 00:00:00 2001 From: Sasha Neftin Date: Sun, 20 Feb 2022 09:29:15 +0200 Subject: igc: igc_write_phy_reg_gpy: drop premature return Similar to "igc_read_phy_reg_gpy: drop premature return" patch. igc_write_phy_reg_gpy checks the return value from igc_write_phy_reg_mdic and if it's not 0, returns immediately. By doing this, it leaves the HW semaphore in the acquired state. Drop this premature return statement, the function returns after releasing the semaphore immediately anyway. Fixes: 5586838fe9ce ("igc: Add code for PHY support") Suggested-by: Dima Ruinskiy Reported-by: Corinna Vinschen Signed-off-by: Sasha Neftin Tested-by: Naama Meir Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/igc/igc_phy.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/igc/igc_phy.c b/drivers/net/ethernet/intel/igc/igc_phy.c index df91d07ce82a..40dbf4b43234 100644 --- a/drivers/net/ethernet/intel/igc/igc_phy.c +++ b/drivers/net/ethernet/intel/igc/igc_phy.c @@ -746,8 +746,6 @@ s32 igc_write_phy_reg_gpy(struct igc_hw *hw, u32 offset, u16 data) if (ret_val) return ret_val; ret_val = igc_write_phy_reg_mdic(hw, offset, data); - if (ret_val) - return ret_val; hw->phy.ops.release(hw); } else { ret_val = igc_write_xmdio_reg(hw, (u16)offset, dev_addr, -- cgit v1.2.3 From 1866aa0d0d6492bc2f8d22d0df49abaccf50cddd Mon Sep 17 00:00:00 2001 From: Sasha Neftin Date: Tue, 25 Jan 2022 19:31:23 +0200 Subject: e1000e: Fix possible HW unit hang after an s0ix exit Disable the OEM bit/Gig Disable/restart AN impact and disable the PHY LAN connected device (LCD) reset during power management flows. This fixes possible HW unit hangs on the s0ix exit on some corporate ADL platforms. Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=214821 Fixes: 3e55d231716e ("e1000e: Add handshake with the CSME to support S0ix") Suggested-by: Dima Ruinskiy Suggested-by: Nir Efrati Signed-off-by: Sasha Neftin Tested-by: Kai-Heng Feng Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/e1000e/hw.h | 1 + drivers/net/ethernet/intel/e1000e/ich8lan.c | 4 ++++ drivers/net/ethernet/intel/e1000e/ich8lan.h | 1 + drivers/net/ethernet/intel/e1000e/netdev.c | 26 ++++++++++++++++++++++++++ 4 files changed, 32 insertions(+) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/e1000e/hw.h b/drivers/net/ethernet/intel/e1000e/hw.h index bcf680e83811..13382df2f2ef 100644 --- a/drivers/net/ethernet/intel/e1000e/hw.h +++ b/drivers/net/ethernet/intel/e1000e/hw.h @@ -630,6 +630,7 @@ struct e1000_phy_info { bool disable_polarity_correction; bool is_mdix; bool polarity_correction; + bool reset_disable; bool speed_downgraded; bool autoneg_wait_to_complete; }; diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c index c908c84b86d2..e298da712758 100644 --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c @@ -2050,6 +2050,10 @@ static s32 e1000_check_reset_block_ich8lan(struct e1000_hw *hw) bool blocked = false; int i = 0; + /* Check the PHY (LCD) reset flag */ + if (hw->phy.reset_disable) + return true; + while ((blocked = !(er32(FWSM) & E1000_ICH_FWSM_RSPCIPHY)) && (i++ < 30)) usleep_range(10000, 11000); diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.h b/drivers/net/ethernet/intel/e1000e/ich8lan.h index 2504b11c3169..638a3ddd7ada 100644 --- a/drivers/net/ethernet/intel/e1000e/ich8lan.h +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.h @@ -271,6 +271,7 @@ #define I217_CGFREG_ENABLE_MTA_RESET 0x0002 #define I217_MEMPWR PHY_REG(772, 26) #define I217_MEMPWR_DISABLE_SMB_RELEASE 0x0010 +#define I217_MEMPWR_MOEM 0x1000 /* Receive Address Initial CRC Calculation */ #define E1000_PCH_RAICC(_n) (0x05F50 + ((_n) * 4)) diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c b/drivers/net/ethernet/intel/e1000e/netdev.c index a42aeb555f34..c5bdef3ffe26 100644 --- a/drivers/net/ethernet/intel/e1000e/netdev.c +++ b/drivers/net/ethernet/intel/e1000e/netdev.c @@ -6987,8 +6987,21 @@ static __maybe_unused int e1000e_pm_suspend(struct device *dev) struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev)); struct e1000_adapter *adapter = netdev_priv(netdev); struct pci_dev *pdev = to_pci_dev(dev); + struct e1000_hw *hw = &adapter->hw; + u16 phy_data; int rc; + if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID && + hw->mac.type >= e1000_pch_adp) { + /* Mask OEM Bits / Gig Disable / Restart AN (772_26[12] = 1) */ + e1e_rphy(hw, I217_MEMPWR, &phy_data); + phy_data |= I217_MEMPWR_MOEM; + e1e_wphy(hw, I217_MEMPWR, phy_data); + + /* Disable LCD reset */ + hw->phy.reset_disable = true; + } + e1000e_flush_lpic(pdev); e1000e_pm_freeze(dev); @@ -7010,6 +7023,8 @@ static __maybe_unused int e1000e_pm_resume(struct device *dev) struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev)); struct e1000_adapter *adapter = netdev_priv(netdev); struct pci_dev *pdev = to_pci_dev(dev); + struct e1000_hw *hw = &adapter->hw; + u16 phy_data; int rc; /* Introduce S0ix implementation */ @@ -7020,6 +7035,17 @@ static __maybe_unused int e1000e_pm_resume(struct device *dev) if (rc) return rc; + if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID && + hw->mac.type >= e1000_pch_adp) { + /* Unmask OEM Bits / Gig Disable / Restart AN 772_26[12] = 0 */ + e1e_rphy(hw, I217_MEMPWR, &phy_data); + phy_data &= ~I217_MEMPWR_MOEM; + e1e_wphy(hw, I217_MEMPWR, phy_data); + + /* Enable LCD reset */ + hw->phy.reset_disable = false; + } + return e1000e_pm_thaw(dev); } -- cgit v1.2.3 From ffd24fa2fcc76ecb2e61e7a4ef8588177bcb42a6 Mon Sep 17 00:00:00 2001 From: Sasha Neftin Date: Thu, 3 Feb 2022 14:21:49 +0200 Subject: e1000e: Correct NVM checksum verification flow Update MAC type check e1000_pch_tgp because for e1000_pch_cnp, NVM checksum update is still possible. Emit a more detailed warning message. Bugzilla: https://bugzilla.opensuse.org/show_bug.cgi?id=1191663 Fixes: 4051f68318ca ("e1000e: Do not take care about recovery NVM checksum") Reported-by: Thomas Bogendoerfer Signed-off-by: Sasha Neftin Tested-by: Naama Meir Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/e1000e/ich8lan.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c index e298da712758..d60e2016d03c 100644 --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c @@ -4140,9 +4140,9 @@ static s32 e1000_validate_nvm_checksum_ich8lan(struct e1000_hw *hw) return ret_val; if (!(data & valid_csum_mask)) { - e_dbg("NVM Checksum Invalid\n"); + e_dbg("NVM Checksum valid bit not set\n"); - if (hw->mac.type < e1000_pch_cnp) { + if (hw->mac.type < e1000_pch_tgp) { data |= valid_csum_mask; ret_val = e1000_write_nvm(hw, word, 1, &data); if (ret_val) -- cgit v1.2.3 From 6c7273a266759d9d36f7c862149f248bcdeddc0f Mon Sep 17 00:00:00 2001 From: Maciej Fijalkowski Date: Wed, 2 Mar 2022 09:59:27 -0800 Subject: ixgbe: xsk: change !netif_carrier_ok() handling in ixgbe_xmit_zc() Commit c685c69fba71 ("ixgbe: don't do any AF_XDP zero-copy transmit if netif is not OK") addressed the ring transient state when MEM_TYPE_XSK_BUFF_POOL was being configured which in turn caused the interface to through down/up. Maurice reported that when carrier is not ok and xsk_pool is present on ring pair, ksoftirqd will consume 100% CPU cycles due to the constant NAPI rescheduling as ixgbe_poll() states that there is still some work to be done. To fix this, do not set work_done to false for a !netif_carrier_ok(). Fixes: c685c69fba71 ("ixgbe: don't do any AF_XDP zero-copy transmit if netif is not OK") Reported-by: Maurice Baijens Tested-by: Maurice Baijens Signed-off-by: Maciej Fijalkowski Tested-by: Sandeep Penigalapati Signed-off-by: Tony Nguyen Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c index b3fd8e5cd85b..6a5e9cf6b5da 100644 --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c @@ -390,12 +390,14 @@ static bool ixgbe_xmit_zc(struct ixgbe_ring *xdp_ring, unsigned int budget) u32 cmd_type; while (budget-- > 0) { - if (unlikely(!ixgbe_desc_unused(xdp_ring)) || - !netif_carrier_ok(xdp_ring->netdev)) { + if (unlikely(!ixgbe_desc_unused(xdp_ring))) { work_done = false; break; } + if (!netif_carrier_ok(xdp_ring->netdev)) + break; + if (!xsk_tx_peek_desc(pool, &desc)) break; -- cgit v1.2.3 From 2cf29e55894886965722e6625f6a03630b4db31d Mon Sep 17 00:00:00 2001 From: Michal Maloszewski Date: Mon, 24 Jan 2022 13:35:43 +0000 Subject: iavf: Fix handling of vlan strip virtual channel messages Modify netdev->features for vlan stripping based on virtual channel messages received from the PF. Change is needed to synchronize vlan strip status between PF sysfs and iavf ethtool. Fixes: 5951a2b9812d ("iavf: Fix VLAN feature flags after VFR") Signed-off-by: Norbert Ciosek Signed-off-by: Michal Maloszewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_virtchnl.c | 40 +++++++++++++++++++++++++ 1 file changed, 40 insertions(+) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c b/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c index 88844d68e150..5263cefe46f5 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c +++ b/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c @@ -1834,6 +1834,22 @@ void iavf_request_reset(struct iavf_adapter *adapter) adapter->current_op = VIRTCHNL_OP_UNKNOWN; } +/** + * iavf_netdev_features_vlan_strip_set - update vlan strip status + * @netdev: ptr to netdev being adjusted + * @enable: enable or disable vlan strip + * + * Helper function to change vlan strip status in netdev->features. + */ +static void iavf_netdev_features_vlan_strip_set(struct net_device *netdev, + const bool enable) +{ + if (enable) + netdev->features |= NETIF_F_HW_VLAN_CTAG_RX; + else + netdev->features &= ~NETIF_F_HW_VLAN_CTAG_RX; +} + /** * iavf_virtchnl_completion * @adapter: adapter structure @@ -2057,8 +2073,18 @@ void iavf_virtchnl_completion(struct iavf_adapter *adapter, } break; case VIRTCHNL_OP_ENABLE_VLAN_STRIPPING: + dev_warn(&adapter->pdev->dev, "Changing VLAN Stripping is not allowed when Port VLAN is configured\n"); + /* Vlan stripping could not be enabled by ethtool. + * Disable it in netdev->features. + */ + iavf_netdev_features_vlan_strip_set(netdev, false); + break; case VIRTCHNL_OP_DISABLE_VLAN_STRIPPING: dev_warn(&adapter->pdev->dev, "Changing VLAN Stripping is not allowed when Port VLAN is configured\n"); + /* Vlan stripping could not be disabled by ethtool. + * Enable it in netdev->features. + */ + iavf_netdev_features_vlan_strip_set(netdev, true); break; default: dev_err(&adapter->pdev->dev, "PF returned error %d (%s) to our request %d\n", @@ -2312,6 +2338,20 @@ void iavf_virtchnl_completion(struct iavf_adapter *adapter, spin_unlock_bh(&adapter->adv_rss_lock); } break; + case VIRTCHNL_OP_ENABLE_VLAN_STRIPPING: + /* PF enabled vlan strip on this VF. + * Update netdev->features if needed to be in sync with ethtool. + */ + if (!v_retval) + iavf_netdev_features_vlan_strip_set(netdev, true); + break; + case VIRTCHNL_OP_DISABLE_VLAN_STRIPPING: + /* PF disabled vlan strip on this VF. + * Update netdev->features if needed to be in sync with ethtool. + */ + if (!v_retval) + iavf_netdev_features_vlan_strip_set(netdev, false); + break; default: if (adapter->current_op && (v_opcode != adapter->current_op)) dev_warn(&adapter->pdev->dev, "Expected response %d from PF, received %d\n", -- cgit v1.2.3 From 57d03f5608c34079f6f15031f4e8b1e2ae95dcb0 Mon Sep 17 00:00:00 2001 From: Michal Maloszewski Date: Wed, 2 Feb 2022 12:44:54 +0000 Subject: iavf: Fix adopting new combined setting In some cases overloaded flag IAVF_FLAG_REINIT_ITR_NEEDED which should indicate that interrupts need to be completely reinitialized during reset leads to RTNL deadlocks using ethtool -C while a reset is in progress. To fix, it was added a new flag IAVF_FLAG_REINIT_MSIX_NEEDED used to trigger MSI-X reinit. New combined setting is fixed adopt after VF reset. This has been implemented by call reinit interrupt scheme during VF reset. Without this fix new combined setting has never been adopted. Fixes: 209f2f9c7181 ("iavf: Add support for VIRTCHNL_VF_OFFLOAD_VLAN_V2 negotiation") Signed-off-by: Grzegorz Szczurek Signed-off-by: Mitch Williams Signed-off-by: Michal Maloszewski Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf.h | 1 + drivers/net/ethernet/intel/iavf/iavf_main.c | 13 +++++++++---- 2 files changed, 10 insertions(+), 4 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/iavf/iavf.h b/drivers/net/ethernet/intel/iavf/iavf.h index 89423947ee65..4babe4705a55 100644 --- a/drivers/net/ethernet/intel/iavf/iavf.h +++ b/drivers/net/ethernet/intel/iavf/iavf.h @@ -288,6 +288,7 @@ struct iavf_adapter { #define IAVF_FLAG_REINIT_ITR_NEEDED BIT(16) #define IAVF_FLAG_QUEUES_DISABLED BIT(17) #define IAVF_FLAG_SETUP_NETDEV_FEATURES BIT(18) +#define IAVF_FLAG_REINIT_MSIX_NEEDED BIT(20) /* duplicates for common code */ #define IAVF_FLAG_DCB_ENABLED 0 /* flags for admin queue service task */ diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index dcf24264c7ea..8e644e9ed8da 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -2120,7 +2120,7 @@ int iavf_parse_vf_resource_msg(struct iavf_adapter *adapter) "Requested %d queues, but PF only gave us %d.\n", num_req_queues, adapter->vsi_res->num_queue_pairs); - adapter->flags |= IAVF_FLAG_REINIT_ITR_NEEDED; + adapter->flags |= IAVF_FLAG_REINIT_MSIX_NEEDED; adapter->num_req_queues = adapter->vsi_res->num_queue_pairs; iavf_schedule_reset(adapter); @@ -2727,7 +2727,8 @@ continue_reset: err); adapter->aq_required = 0; - if (adapter->flags & IAVF_FLAG_REINIT_ITR_NEEDED) { + if ((adapter->flags & IAVF_FLAG_REINIT_MSIX_NEEDED) || + (adapter->flags & IAVF_FLAG_REINIT_ITR_NEEDED)) { err = iavf_reinit_interrupt_scheme(adapter); if (err) goto reset_err; @@ -2799,12 +2800,13 @@ continue_reset: if (err) goto reset_err; - if (adapter->flags & IAVF_FLAG_REINIT_ITR_NEEDED) { + if ((adapter->flags & IAVF_FLAG_REINIT_MSIX_NEEDED) || + (adapter->flags & IAVF_FLAG_REINIT_ITR_NEEDED)) { err = iavf_request_traffic_irqs(adapter, netdev->name); if (err) goto reset_err; - adapter->flags &= ~IAVF_FLAG_REINIT_ITR_NEEDED; + adapter->flags &= ~IAVF_FLAG_REINIT_MSIX_NEEDED; } iavf_configure(adapter); @@ -2819,6 +2821,9 @@ continue_reset: iavf_change_state(adapter, __IAVF_DOWN); wake_up(&adapter->down_waitqueue); } + + adapter->flags &= ~IAVF_FLAG_REINIT_ITR_NEEDED; + mutex_unlock(&adapter->client_lock); mutex_unlock(&adapter->crit_lock); -- cgit v1.2.3 From 5710ab79166504013f7c0ae6a57e7d2fd26e5c43 Mon Sep 17 00:00:00 2001 From: Jacob Keller Date: Wed, 16 Feb 2022 16:51:35 -0800 Subject: i40e: stop disabling VFs due to PF error responses The i40e_vc_send_msg_to_vf_ex (and its wrapper i40e_vc_send_msg_to_vf) function has logic to detect "failure" responses sent to the VF. If a VF is sent more than I40E_DEFAULT_NUM_INVALID_MSGS_ALLOWED, then the VF is marked as disabled. In either case, a dev_info message is printed stating that a VF opcode failed. This logic originates from the early implementation of VF support in commit 5c3c48ac6bf5 ("i40e: implement virtual device interface"). That commit did not go far enough. The "logic" for this behavior seems to be that error responses somehow indicate a malicious VF. This is not really true. The PF might be sending an error for any number of reasons such as lacking resources, an unsupported operation, etc. This does not indicate a malicious VF. We already have a separate robust malicious VF detection which relies on hardware logic to detect and prevent a variety of behaviors. There is no justification for this behavior in the original implementation. In fact, a later commit 18b7af57d9c1 ("i40e: Lower some message levels") reduced the opcode failure message from a dev_err to a dev_info. In addition, recent commit 01cbf50877e6 ("i40e: Fix to not show opcode msg on unsuccessful VF MAC change") changed the logic to allow quieting it for expected failures. That commit prevented this logic from kicking in for specific circumstances. This change did not go far enough. The behavior is not documented nor is it part of any requirement for our products. Other operating systems such as the FreeBSD implementation of our driver do not include this logic. It is clear this check does not make sense, and causes problems which led to ugly workarounds. Fix this by just removing the entire logic and the need for the i40e_vc_send_msg_to_vf_ex function. Fixes: 01cbf50877e6 ("i40e: Fix to not show opcode msg on unsuccessful VF MAC change") Fixes: 5c3c48ac6bf5 ("i40e: implement virtual device interface") Signed-off-by: Jacob Keller Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/i40e/i40e_debugfs.c | 6 +-- drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c | 57 +++------------------- drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h | 5 -- 3 files changed, 9 insertions(+), 59 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/i40e/i40e_debugfs.c b/drivers/net/ethernet/intel/i40e/i40e_debugfs.c index 1e57cc8c47d7..9db5001297c7 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_debugfs.c +++ b/drivers/net/ethernet/intel/i40e/i40e_debugfs.c @@ -742,10 +742,8 @@ static void i40e_dbg_dump_vf(struct i40e_pf *pf, int vf_id) vsi = pf->vsi[vf->lan_vsi_idx]; dev_info(&pf->pdev->dev, "vf %2d: VSI id=%d, seid=%d, qps=%d\n", vf_id, vf->lan_vsi_id, vsi->seid, vf->num_queue_pairs); - dev_info(&pf->pdev->dev, " num MDD=%lld, invalid msg=%lld, valid msg=%lld\n", - vf->num_mdd_events, - vf->num_invalid_msgs, - vf->num_valid_msgs); + dev_info(&pf->pdev->dev, " num MDD=%lld\n", + vf->num_mdd_events); } else { dev_info(&pf->pdev->dev, "invalid VF id %d\n", vf_id); } diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c index dfdb6e786461..2606e8f0f19b 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c +++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c @@ -1917,19 +1917,17 @@ sriov_configure_out: /***********************virtual channel routines******************/ /** - * i40e_vc_send_msg_to_vf_ex + * i40e_vc_send_msg_to_vf * @vf: pointer to the VF info * @v_opcode: virtual channel opcode * @v_retval: virtual channel return value * @msg: pointer to the msg buffer * @msglen: msg length - * @is_quiet: true for not printing unsuccessful return values, false otherwise * * send msg to VF **/ -static int i40e_vc_send_msg_to_vf_ex(struct i40e_vf *vf, u32 v_opcode, - u32 v_retval, u8 *msg, u16 msglen, - bool is_quiet) +static int i40e_vc_send_msg_to_vf(struct i40e_vf *vf, u32 v_opcode, + u32 v_retval, u8 *msg, u16 msglen) { struct i40e_pf *pf; struct i40e_hw *hw; @@ -1944,25 +1942,6 @@ static int i40e_vc_send_msg_to_vf_ex(struct i40e_vf *vf, u32 v_opcode, hw = &pf->hw; abs_vf_id = vf->vf_id + hw->func_caps.vf_base_id; - /* single place to detect unsuccessful return values */ - if (v_retval && !is_quiet) { - vf->num_invalid_msgs++; - dev_info(&pf->pdev->dev, "VF %d failed opcode %d, retval: %d\n", - vf->vf_id, v_opcode, v_retval); - if (vf->num_invalid_msgs > - I40E_DEFAULT_NUM_INVALID_MSGS_ALLOWED) { - dev_err(&pf->pdev->dev, - "Number of invalid messages exceeded for VF %d\n", - vf->vf_id); - dev_err(&pf->pdev->dev, "Use PF Control I/F to enable the VF\n"); - set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states); - } - } else { - vf->num_valid_msgs++; - /* reset the invalid counter, if a valid message is received. */ - vf->num_invalid_msgs = 0; - } - aq_ret = i40e_aq_send_msg_to_vf(hw, abs_vf_id, v_opcode, v_retval, msg, msglen, NULL); if (aq_ret) { @@ -1975,23 +1954,6 @@ static int i40e_vc_send_msg_to_vf_ex(struct i40e_vf *vf, u32 v_opcode, return 0; } -/** - * i40e_vc_send_msg_to_vf - * @vf: pointer to the VF info - * @v_opcode: virtual channel opcode - * @v_retval: virtual channel return value - * @msg: pointer to the msg buffer - * @msglen: msg length - * - * send msg to VF - **/ -static int i40e_vc_send_msg_to_vf(struct i40e_vf *vf, u32 v_opcode, - u32 v_retval, u8 *msg, u16 msglen) -{ - return i40e_vc_send_msg_to_vf_ex(vf, v_opcode, v_retval, - msg, msglen, false); -} - /** * i40e_vc_send_resp_to_vf * @vf: pointer to the VF info @@ -2822,7 +2784,6 @@ error_param: * i40e_check_vf_permission * @vf: pointer to the VF info * @al: MAC address list from virtchnl - * @is_quiet: set true for printing msg without opcode info, false otherwise * * Check that the given list of MAC addresses is allowed. Will return -EPERM * if any address in the list is not valid. Checks the following conditions: @@ -2837,8 +2798,7 @@ error_param: * addresses might not be accurate. **/ static inline int i40e_check_vf_permission(struct i40e_vf *vf, - struct virtchnl_ether_addr_list *al, - bool *is_quiet) + struct virtchnl_ether_addr_list *al) { struct i40e_pf *pf = vf->pf; struct i40e_vsi *vsi = pf->vsi[vf->lan_vsi_idx]; @@ -2846,7 +2806,6 @@ static inline int i40e_check_vf_permission(struct i40e_vf *vf, int mac2add_cnt = 0; int i; - *is_quiet = false; for (i = 0; i < al->num_elements; i++) { struct i40e_mac_filter *f; u8 *addr = al->list[i].addr; @@ -2870,7 +2829,6 @@ static inline int i40e_check_vf_permission(struct i40e_vf *vf, !ether_addr_equal(addr, vf->default_lan_addr.addr)) { dev_err(&pf->pdev->dev, "VF attempting to override administratively set MAC address, bring down and up the VF interface to resume normal operation\n"); - *is_quiet = true; return -EPERM; } @@ -2921,7 +2879,6 @@ static int i40e_vc_add_mac_addr_msg(struct i40e_vf *vf, u8 *msg) (struct virtchnl_ether_addr_list *)msg; struct i40e_pf *pf = vf->pf; struct i40e_vsi *vsi = NULL; - bool is_quiet = false; i40e_status ret = 0; int i; @@ -2938,7 +2895,7 @@ static int i40e_vc_add_mac_addr_msg(struct i40e_vf *vf, u8 *msg) */ spin_lock_bh(&vsi->mac_filter_hash_lock); - ret = i40e_check_vf_permission(vf, al, &is_quiet); + ret = i40e_check_vf_permission(vf, al); if (ret) { spin_unlock_bh(&vsi->mac_filter_hash_lock); goto error_param; @@ -2976,8 +2933,8 @@ static int i40e_vc_add_mac_addr_msg(struct i40e_vf *vf, u8 *msg) error_param: /* send the response to the VF */ - return i40e_vc_send_msg_to_vf_ex(vf, VIRTCHNL_OP_ADD_ETH_ADDR, - ret, NULL, 0, is_quiet); + return i40e_vc_send_msg_to_vf(vf, VIRTCHNL_OP_ADD_ETH_ADDR, + ret, NULL, 0); } /** diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h index 03c42fd0fea1..a554d0a0b09b 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h +++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h @@ -10,8 +10,6 @@ #define I40E_VIRTCHNL_SUPPORTED_QTYPES 2 -#define I40E_DEFAULT_NUM_INVALID_MSGS_ALLOWED 10 - #define I40E_VLAN_PRIORITY_SHIFT 13 #define I40E_VLAN_MASK 0xFFF #define I40E_PRIORITY_MASK 0xE000 @@ -92,9 +90,6 @@ struct i40e_vf { u8 num_queue_pairs; /* num of qps assigned to VF vsis */ u8 num_req_queues; /* num of requested qps */ u64 num_mdd_events; /* num of mdd events detected */ - /* num of continuous malformed or invalid msgs detected */ - u64 num_invalid_msgs; - u64 num_valid_msgs; /* num of valid msgs detected */ unsigned long vf_caps; /* vf's adv. capabilities */ unsigned long vf_states; /* vf's runtime states */ -- cgit v1.2.3 From 79498d5af8e458102242d1667cf44df1f1564e63 Mon Sep 17 00:00:00 2001 From: Jacob Keller Date: Wed, 16 Feb 2022 16:51:36 -0800 Subject: ice: stop disabling VFs due to PF error responses The ice_vc_send_msg_to_vf function has logic to detect "failure" responses being sent to a VF. If a VF is sent more than ICE_DFLT_NUM_INVAL_MSGS_ALLOWED then the VF is marked as disabled. Almost identical logic also existed in the i40e driver. This logic was added to the ice driver in commit 1071a8358a28 ("ice: Implement virtchnl commands for AVF support") which itself copied from the i40e implementation in commit 5c3c48ac6bf5 ("i40e: implement virtual device interface"). Neither commit provides a proper explanation or justification of the check. In fact, later commits to i40e changed the logic to allow bypassing the check in some specific instances. The "logic" for this seems to be that error responses somehow indicate a malicious VF. This is not really true. The PF might be sending an error for any number of reasons such as lack of resources, etc. Additionally, this causes the PF to log an info message for every failed VF response which may confuse users, and can spam the kernel log. This behavior is not documented as part of any requirement for our products and other operating system drivers such as the FreeBSD implementation of our drivers do not include this type of check. In fact, the change from dev_err to dev_info in i40e commit 18b7af57d9c1 ("i40e: Lower some message levels") explains that these messages typically don't actually indicate a real issue. It is quite likely that a user who hits this in practice will be very confused as the VF will be disabled without an obvious way to recover. We already have robust malicious driver detection logic using actual hardware detection mechanisms that detect and prevent invalid device usage. Remove the logic since its not a documented requirement and the behavior is not intuitive. Fixes: 1071a8358a28 ("ice: Implement virtchnl commands for AVF support") Signed-off-by: Jacob Keller Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c | 18 ------------------ drivers/net/ethernet/intel/ice/ice_virtchnl_pf.h | 3 --- 2 files changed, 21 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c index 408f78e3eb13..1be3cd4b2bef 100644 --- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c +++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c @@ -2182,24 +2182,6 @@ ice_vc_send_msg_to_vf(struct ice_vf *vf, u32 v_opcode, dev = ice_pf_to_dev(pf); - /* single place to detect unsuccessful return values */ - if (v_retval) { - vf->num_inval_msgs++; - dev_info(dev, "VF %d failed opcode %d, retval: %d\n", vf->vf_id, - v_opcode, v_retval); - if (vf->num_inval_msgs > ICE_DFLT_NUM_INVAL_MSGS_ALLOWED) { - dev_err(dev, "Number of invalid messages exceeded for VF %d\n", - vf->vf_id); - dev_err(dev, "Use PF Control I/F to enable the VF\n"); - set_bit(ICE_VF_STATE_DIS, vf->vf_states); - return -EIO; - } - } else { - vf->num_valid_msgs++; - /* reset the invalid counter, if a valid message is received. */ - vf->num_inval_msgs = 0; - } - aq_ret = ice_aq_send_msg_to_vf(&pf->hw, vf->vf_id, v_opcode, v_retval, msg, msglen, NULL); if (aq_ret && pf->hw.mailboxq.sq_last_status != ICE_AQ_RC_ENOSYS) { diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.h b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.h index 752487a1bdd6..8f27255cc0cc 100644 --- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.h +++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.h @@ -14,7 +14,6 @@ #define ICE_MAX_MACADDR_PER_VF 18 /* Malicious Driver Detection */ -#define ICE_DFLT_NUM_INVAL_MSGS_ALLOWED 10 #define ICE_MDD_EVENTS_THRESHOLD 30 /* Static VF transaction/status register def */ @@ -134,8 +133,6 @@ struct ice_vf { unsigned int max_tx_rate; /* Maximum Tx bandwidth limit in Mbps */ DECLARE_BITMAP(vf_states, ICE_VF_STATES_NBITS); /* VF runtime states */ - u64 num_inval_msgs; /* number of continuous invalid msgs */ - u64 num_valid_msgs; /* number of valid msgs detected */ unsigned long vf_caps; /* VF's adv. capabilities */ u8 num_req_qs; /* num of queue pairs requested by VF */ u16 num_mac; -- cgit v1.2.3 From 97b0129146b1544bbb0773585327896da3bb4e0a Mon Sep 17 00:00:00 2001 From: Dave Ertman Date: Fri, 18 Feb 2022 12:39:25 -0800 Subject: ice: Fix error with handling of bonding MTU When a bonded interface is destroyed, .ndo_change_mtu can be called during the tear-down process while the RTNL lock is held. This is a problem since the auxiliary driver linked to the LAN driver needs to be notified of the MTU change, and this requires grabbing a device_lock on the auxiliary_device's dev. Currently this is being attempted in the same execution context as the call to .ndo_change_mtu which is causing a dead-lock. Move the notification of the changed MTU to a separate execution context (watchdog service task) and eliminate the "before" notification. Fixes: 348048e724a0e ("ice: Implement iidc operations") Signed-off-by: Dave Ertman Tested-by: Jonathan Toppins Tested-by: Gurucharan G (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 | 29 ++++++++++++++--------------- 2 files changed, 15 insertions(+), 15 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h index 473b1f6be9de..3121f9b04f59 100644 --- a/drivers/net/ethernet/intel/ice/ice.h +++ b/drivers/net/ethernet/intel/ice/ice.h @@ -483,6 +483,7 @@ enum ice_pf_flags { ICE_FLAG_MDD_AUTO_RESET_VF, ICE_FLAG_LINK_LENIENT_MODE_ENA, ICE_FLAG_PLUG_AUX_DEV, + ICE_FLAG_MTU_CHANGED, ICE_PF_FLAGS_NBITS /* must be last */ }; diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index f3c346e13b7a..6fc6514eb007 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -2258,6 +2258,17 @@ static void ice_service_task(struct work_struct *work) if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) ice_plug_aux_dev(pf); + if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) { + struct iidc_event *event; + + event = kzalloc(sizeof(*event), GFP_KERNEL); + if (event) { + set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type); + ice_send_event_to_aux(pf, event); + kfree(event); + } + } + ice_clean_adminq_subtask(pf); ice_check_media_subtask(pf); ice_check_for_hang_subtask(pf); @@ -6822,7 +6833,6 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu) struct ice_netdev_priv *np = netdev_priv(netdev); struct ice_vsi *vsi = np->vsi; struct ice_pf *pf = vsi->back; - struct iidc_event *event; u8 count = 0; int err = 0; @@ -6857,14 +6867,6 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu) return -EBUSY; } - event = kzalloc(sizeof(*event), GFP_KERNEL); - if (!event) - return -ENOMEM; - - set_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type); - ice_send_event_to_aux(pf, event); - clear_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type); - netdev->mtu = (unsigned int)new_mtu; /* if VSI is up, bring it down and then back up */ @@ -6872,21 +6874,18 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu) err = ice_down(vsi); if (err) { netdev_err(netdev, "change MTU if_down err %d\n", err); - goto event_after; + return err; } err = ice_up(vsi); if (err) { netdev_err(netdev, "change MTU if_up err %d\n", err); - goto event_after; + return err; } } netdev_dbg(netdev, "changed MTU to %d\n", new_mtu); -event_after: - set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type); - ice_send_event_to_aux(pf, event); - kfree(event); + set_bit(ICE_FLAG_MTU_CHANGED, pf->flags); return err; } -- cgit v1.2.3 From 3d97f1afd8d831e0c0dc1157418f94b8faa97b54 Mon Sep 17 00:00:00 2001 From: Christophe JAILLET Date: Sun, 16 Jan 2022 19:46:20 +0100 Subject: ice: Don't use GFP_KERNEL in atomic context ice_misc_intr() is an irq handler. It should not sleep. Use GFP_ATOMIC instead of GFP_KERNEL when allocating some memory. Fixes: 348048e724a0 ("ice: Implement iidc operations") Signed-off-by: Christophe JAILLET Tested-by: Leszek Kaliszczuk Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index 6fc6514eb007..83e3e8aae6cf 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -3034,7 +3034,7 @@ static irqreturn_t ice_misc_intr(int __always_unused irq, void *data) struct iidc_event *event; ena_mask &= ~ICE_AUX_CRIT_ERR; - event = kzalloc(sizeof(*event), GFP_KERNEL); + event = kzalloc(sizeof(*event), GFP_ATOMIC); if (event) { set_bit(IIDC_EVENT_CRIT_ERR, event->type); /* report the entire OICR value to AUX driver */ -- cgit v1.2.3 From ad35ffa252af67d4cc7c744b9377a2b577748e3f Mon Sep 17 00:00:00 2001 From: Jedrzej Jagielski Date: Tue, 22 Feb 2022 11:43:04 +0000 Subject: ice: Fix curr_link_speed advertised speed Change curr_link_speed advertised speed, due to link_info.link_speed is not equal phy.curr_user_speed_req. Without this patch it is impossible to set advertised speed to same as link_speed. Testing Hints: Try to set advertised speed to 25G only with 25G default link (use ethtool -s 0x80000000) Fixes: 48cb27f2fd18 ("ice: Implement handlers for ethtool PHY/link operations") Signed-off-by: Grzegorz Siwik Signed-off-by: Jedrzej Jagielski Tested-by: Gurucharan (A Contingent worker at Intel) Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/ice/ice_ethtool.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice_ethtool.c b/drivers/net/ethernet/intel/ice/ice_ethtool.c index e2e3ef7fba7f..a5dc9824e255 100644 --- a/drivers/net/ethernet/intel/ice/ice_ethtool.c +++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c @@ -2298,7 +2298,7 @@ ice_set_link_ksettings(struct net_device *netdev, if (err) goto done; - curr_link_speed = pi->phy.link_info.link_speed; + curr_link_speed = pi->phy.curr_user_speed_req; adv_link_speed = ice_ksettings_find_adv_link_speed(ks); /* If speed didn't get set, set it to what it currently is. -- cgit v1.2.3 From 5cb1ebdbc4342b1c2ce89516e19808d64417bdbc Mon Sep 17 00:00:00 2001 From: Ivan Vecera Date: Thu, 10 Mar 2022 18:16:41 +0100 Subject: ice: Fix race condition during interface enslave Commit 5dbbbd01cbba83 ("ice: Avoid RTNL lock when re-creating auxiliary device") changes a process of re-creation of aux device so ice_plug_aux_dev() is called from ice_service_task() context. This unfortunately opens a race window that can result in dead-lock when interface has left LAG and immediately enters LAG again. Reproducer: ``` #!/bin/sh ip link add lag0 type bond mode 1 miimon 100 ip link set lag0 for n in {1..10}; do echo Cycle: $n ip link set ens7f0 master lag0 sleep 1 ip link set ens7f0 nomaster done ``` This results in: [20976.208697] Workqueue: ice ice_service_task [ice] [20976.213422] Call Trace: [20976.215871] __schedule+0x2d1/0x830 [20976.219364] schedule+0x35/0xa0 [20976.222510] schedule_preempt_disabled+0xa/0x10 [20976.227043] __mutex_lock.isra.7+0x310/0x420 [20976.235071] enum_all_gids_of_dev_cb+0x1c/0x100 [ib_core] [20976.251215] ib_enum_roce_netdev+0xa4/0xe0 [ib_core] [20976.256192] ib_cache_setup_one+0x33/0xa0 [ib_core] [20976.261079] ib_register_device+0x40d/0x580 [ib_core] [20976.266139] irdma_ib_register_device+0x129/0x250 [irdma] [20976.281409] irdma_probe+0x2c1/0x360 [irdma] [20976.285691] auxiliary_bus_probe+0x45/0x70 [20976.289790] really_probe+0x1f2/0x480 [20976.298509] driver_probe_device+0x49/0xc0 [20976.302609] bus_for_each_drv+0x79/0xc0 [20976.306448] __device_attach+0xdc/0x160 [20976.310286] bus_probe_device+0x9d/0xb0 [20976.314128] device_add+0x43c/0x890 [20976.321287] __auxiliary_device_add+0x43/0x60 [20976.325644] ice_plug_aux_dev+0xb2/0x100 [ice] [20976.330109] ice_service_task+0xd0c/0xed0 [ice] [20976.342591] process_one_work+0x1a7/0x360 [20976.350536] worker_thread+0x30/0x390 [20976.358128] kthread+0x10a/0x120 [20976.365547] ret_from_fork+0x1f/0x40 ... [20976.438030] task:ip state:D stack: 0 pid:213658 ppid:213627 flags:0x00004084 [20976.446469] Call Trace: [20976.448921] __schedule+0x2d1/0x830 [20976.452414] schedule+0x35/0xa0 [20976.455559] schedule_preempt_disabled+0xa/0x10 [20976.460090] __mutex_lock.isra.7+0x310/0x420 [20976.464364] device_del+0x36/0x3c0 [20976.467772] ice_unplug_aux_dev+0x1a/0x40 [ice] [20976.472313] ice_lag_event_handler+0x2a2/0x520 [ice] [20976.477288] notifier_call_chain+0x47/0x70 [20976.481386] __netdev_upper_dev_link+0x18b/0x280 [20976.489845] bond_enslave+0xe05/0x1790 [bonding] [20976.494475] do_setlink+0x336/0xf50 [20976.502517] __rtnl_newlink+0x529/0x8b0 [20976.543441] rtnl_newlink+0x43/0x60 [20976.546934] rtnetlink_rcv_msg+0x2b1/0x360 [20976.559238] netlink_rcv_skb+0x4c/0x120 [20976.563079] netlink_unicast+0x196/0x230 [20976.567005] netlink_sendmsg+0x204/0x3d0 [20976.570930] sock_sendmsg+0x4c/0x50 [20976.574423] ____sys_sendmsg+0x1eb/0x250 [20976.586807] ___sys_sendmsg+0x7c/0xc0 [20976.606353] __sys_sendmsg+0x57/0xa0 [20976.609930] do_syscall_64+0x5b/0x1a0 [20976.613598] entry_SYSCALL_64_after_hwframe+0x65/0xca 1. Command 'ip link ... set nomaster' causes that ice_plug_aux_dev() is called from ice_service_task() context, aux device is created and associated device->lock is taken. 2. Command 'ip link ... set master...' calls ice's notifier under RTNL lock and that notifier calls ice_unplug_aux_dev(). That function tries to take aux device->lock but this is already taken by ice_plug_aux_dev() in step 1 3. Later ice_plug_aux_dev() tries to take RTNL lock but this is already taken in step 2 4. Dead-lock The patch fixes this issue by following changes: - Bit ICE_FLAG_PLUG_AUX_DEV is kept to be set during ice_plug_aux_dev() call in ice_service_task() - The bit is checked in ice_clear_rdma_cap() and only if it is not set then ice_unplug_aux_dev() is called. If it is set (in other words plugging of aux device was requested and ice_plug_aux_dev() is potentially running) then the function only clears the bit - Once ice_plug_aux_dev() call (in ice_service_task) is finished the bit ICE_FLAG_PLUG_AUX_DEV is cleared but it is also checked whether it was already cleared by ice_clear_rdma_cap(). If so then aux device is unplugged. Signed-off-by: Ivan Vecera Co-developed-by: Petr Oros Signed-off-by: Petr Oros Reviewed-by: Dave Ertman Link: https://lore.kernel.org/r/20220310171641.3863659-1-ivecera@redhat.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/intel/ice/ice.h | 11 ++++++++++- drivers/net/ethernet/intel/ice/ice_main.c | 12 +++++++++++- 2 files changed, 21 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/intel') diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h index 3121f9b04f59..bea1d1e39fa2 100644 --- a/drivers/net/ethernet/intel/ice/ice.h +++ b/drivers/net/ethernet/intel/ice/ice.h @@ -898,7 +898,16 @@ static inline void ice_set_rdma_cap(struct ice_pf *pf) */ static inline void ice_clear_rdma_cap(struct ice_pf *pf) { - ice_unplug_aux_dev(pf); + /* We can directly unplug aux device here only if the flag bit + * ICE_FLAG_PLUG_AUX_DEV is not set because ice_unplug_aux_dev() + * could race with ice_plug_aux_dev() called from + * ice_service_task(). In this case we only clear that bit now and + * aux device will be unplugged later once ice_plug_aux_device() + * called from ice_service_task() finishes (see ice_service_task()). + */ + if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) + ice_unplug_aux_dev(pf); + clear_bit(ICE_FLAG_RDMA_ENA, pf->flags); clear_bit(ICE_FLAG_AUX_ENA, pf->flags); } diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c index 83e3e8aae6cf..493942e910be 100644 --- a/drivers/net/ethernet/intel/ice/ice_main.c +++ b/drivers/net/ethernet/intel/ice/ice_main.c @@ -2255,9 +2255,19 @@ static void ice_service_task(struct work_struct *work) return; } - if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) + if (test_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) { + /* Plug aux device per request */ ice_plug_aux_dev(pf); + /* Mark plugging as done but check whether unplug was + * requested during ice_plug_aux_dev() call + * (e.g. from ice_clear_rdma_cap()) and if so then + * plug aux device. + */ + if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) + ice_unplug_aux_dev(pf); + } + if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) { struct iidc_event *event; -- cgit v1.2.3