]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/commitdiff
ice: update malicious driver detection event handling
authorPaul Greenwalt <paul.greenwalt@intel.com>
Thu, 13 Feb 2020 21:31:16 +0000 (13:31 -0800)
committerJeff Kirsher <jeffrey.t.kirsher@intel.com>
Wed, 19 Feb 2020 20:56:34 +0000 (12:56 -0800)
Update the PF VFs MDD event message to rate limit once per second and
report the total number Rx|Tx event count. Add support to print pending
MDD events that occur during the rate limit. The use of net_ratelimit did
not allow for per VF Rx|Tx granularity.

Additional PF MDD log messages are guarded by netif_msg_[rx|tx]_err().

Since VF RX MDD events disable the queue, add ethtool private flag
mdd-auto-reset-vf to configure VF reset to re-enable the queue.

Disable anti-spoof detection interrupt to prevent spurious events
during a function reset.

To avoid race condition do not make PF MDD register reads conditional
on global MDD result.

Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
drivers/net/ethernet/intel/ice/ice.h
drivers/net/ethernet/intel/ice/ice_ethtool.c
drivers/net/ethernet/intel/ice/ice_hw_autogen.h
drivers/net/ethernet/intel/ice/ice_main.c
drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
drivers/net/ethernet/intel/ice/ice_virtchnl_pf.h

index cb10abb14e11c7749127ca7e1356fe3e719992d6..2d51ceaa2c8c0846f9cdfcbcae595dde27e876da 100644 (file)
@@ -212,6 +212,7 @@ enum ice_state {
        __ICE_SERVICE_SCHED,
        __ICE_SERVICE_DIS,
        __ICE_OICR_INTR_DIS,            /* Global OICR interrupt disabled */
+       __ICE_MDD_VF_PRINT_PENDING,     /* set when MDD event handle */
        __ICE_STATE_NBITS               /* must be last */
 };
 
@@ -340,6 +341,7 @@ enum ice_pf_flags {
        ICE_FLAG_FW_LLDP_AGENT,
        ICE_FLAG_ETHTOOL_CTXT,          /* set when ethtool holds RTNL lock */
        ICE_FLAG_LEGACY_RX,
+       ICE_FLAG_MDD_AUTO_RESET_VF,
        ICE_PF_FLAGS_NBITS              /* must be last */
 };
 
@@ -363,6 +365,8 @@ struct ice_pf {
        u16 num_vfs_supported;          /* num VFs supported for this PF */
        u16 num_vf_qps;                 /* num queue pairs per VF */
        u16 num_vf_msix;                /* num vectors per VF */
+       /* used to ratelimit the MDD event logging */
+       unsigned long last_printed_mdd_jiffies;
        DECLARE_BITMAP(state, __ICE_STATE_NBITS);
        DECLARE_BITMAP(flags, ICE_PF_FLAGS_NBITS);
        unsigned long *avail_txqs;      /* bitmap to track PF Tx queue usage */
index 583e07fffd5f6e8ebac64b607589c3fd09c7b7aa..63a69ea02d22e9ed08b9a6bff448c4eb180ecfd5 100644 (file)
@@ -157,6 +157,7 @@ struct ice_priv_flag {
 static const struct ice_priv_flag ice_gstrings_priv_flags[] = {
        ICE_PRIV_FLAG("link-down-on-close", ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA),
        ICE_PRIV_FLAG("fw-lldp-agent", ICE_FLAG_FW_LLDP_AGENT),
+       ICE_PRIV_FLAG("mdd-auto-reset-vf", ICE_FLAG_MDD_AUTO_RESET_VF),
        ICE_PRIV_FLAG("legacy-rx", ICE_FLAG_LEGACY_RX),
 };
 
index b99ebfefe06beaf528a04fb8b11335b49421976f..43e4efbccd8e685bd0422db48d277573db933ce9 100644 (file)
 #define VPLAN_TX_QBASE_VFNUMQ_M                        ICE_M(0xFF, 16)
 #define VPLAN_TXQ_MAPENA(_VF)                  (0x00073800 + ((_VF) * 4))
 #define VPLAN_TXQ_MAPENA_TX_ENA_M              BIT(0)
+#define GL_MDCK_TX_TDPU                                0x00049348
+#define GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M BIT(1)
 #define GL_MDET_RX                             0x00294C00
 #define GL_MDET_RX_QNUM_S                      0
 #define GL_MDET_RX_QNUM_M                      ICE_M(0x7FFF, 0)
index 255317e4b1f31538d755c5e2e93f9d2810063638..f11e0934dc03249635d5dd801dca78e0fc8d0c50 100644 (file)
@@ -1187,20 +1187,28 @@ static void ice_service_timer(struct timer_list *t)
  * ice_handle_mdd_event - handle malicious driver detect event
  * @pf: pointer to the PF structure
  *
- * Called from service task. OICR interrupt handler indicates MDD event
+ * Called from service task. OICR interrupt handler indicates MDD event.
+ * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
+ * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
+ * disable the queue, the PF can be configured to reset the VF using ethtool
+ * private flag mdd-auto-reset-vf.
  */
 static void ice_handle_mdd_event(struct ice_pf *pf)
 {
        struct device *dev = ice_pf_to_dev(pf);
        struct ice_hw *hw = &pf->hw;
-       bool mdd_detected = false;
        u32 reg;
        int i;
 
-       if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state))
+       if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state)) {
+               /* Since the VF MDD event logging is rate limited, check if
+                * there are pending MDD events.
+                */
+               ice_print_vfs_mdd_events(pf);
                return;
+       }
 
-       /* find what triggered the MDD event */
+       /* find what triggered an MDD event */
        reg = rd32(hw, GL_MDET_TX_PQM);
        if (reg & GL_MDET_TX_PQM_VALID_M) {
                u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
@@ -1216,7 +1224,6 @@ static void ice_handle_mdd_event(struct ice_pf *pf)
                        dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
                                 event, queue, pf_num, vf_num);
                wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
-               mdd_detected = true;
        }
 
        reg = rd32(hw, GL_MDET_TX_TCLAN);
@@ -1234,7 +1241,6 @@ static void ice_handle_mdd_event(struct ice_pf *pf)
                        dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
                                 event, queue, pf_num, vf_num);
                wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
-               mdd_detected = true;
        }
 
        reg = rd32(hw, GL_MDET_RX);
@@ -1252,85 +1258,85 @@ static void ice_handle_mdd_event(struct ice_pf *pf)
                        dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
                                 event, queue, pf_num, vf_num);
                wr32(hw, GL_MDET_RX, 0xffffffff);
-               mdd_detected = true;
        }
 
-       if (mdd_detected) {
-               bool pf_mdd_detected = false;
-
-               reg = rd32(hw, PF_MDET_TX_PQM);
-               if (reg & PF_MDET_TX_PQM_VALID_M) {
-                       wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
-                       dev_info(dev, "TX driver issue detected, PF reset issued\n");
-                       pf_mdd_detected = true;
-               }
+       /* check to see if this PF caused an MDD event */
+       reg = rd32(hw, PF_MDET_TX_PQM);
+       if (reg & PF_MDET_TX_PQM_VALID_M) {
+               wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
+               if (netif_msg_tx_err(pf))
+                       dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
+       }
 
-               reg = rd32(hw, PF_MDET_TX_TCLAN);
-               if (reg & PF_MDET_TX_TCLAN_VALID_M) {
-                       wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
-                       dev_info(dev, "TX driver issue detected, PF reset issued\n");
-                       pf_mdd_detected = true;
-               }
+       reg = rd32(hw, PF_MDET_TX_TCLAN);
+       if (reg & PF_MDET_TX_TCLAN_VALID_M) {
+               wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
+               if (netif_msg_tx_err(pf))
+                       dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
+       }
 
-               reg = rd32(hw, PF_MDET_RX);
-               if (reg & PF_MDET_RX_VALID_M) {
-                       wr32(hw, PF_MDET_RX, 0xFFFF);
-                       dev_info(dev, "RX driver issue detected, PF reset issued\n");
-                       pf_mdd_detected = true;
-               }
-               /* Queue belongs to the PF initiate a reset */
-               if (pf_mdd_detected) {
-                       set_bit(__ICE_NEEDS_RESTART, pf->state);
-                       ice_service_task_schedule(pf);
-               }
+       reg = rd32(hw, PF_MDET_RX);
+       if (reg & PF_MDET_RX_VALID_M) {
+               wr32(hw, PF_MDET_RX, 0xFFFF);
+               if (netif_msg_rx_err(pf))
+                       dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
        }
 
-       /* check to see if one of the VFs caused the MDD */
+       /* Check to see if one of the VFs caused an MDD event, and then
+        * increment counters and set print pending
+        */
        ice_for_each_vf(pf, i) {
                struct ice_vf *vf = &pf->vf[i];
 
-               bool vf_mdd_detected = false;
-
                reg = rd32(hw, VP_MDET_TX_PQM(i));
                if (reg & VP_MDET_TX_PQM_VALID_M) {
                        wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
-                       vf_mdd_detected = true;
-                       dev_info(dev, "TX driver issue detected on VF %d\n",
-                                i);
+                       vf->mdd_tx_events.count++;
+                       set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
+                       if (netif_msg_tx_err(pf))
+                               dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
+                                        i);
                }
 
                reg = rd32(hw, VP_MDET_TX_TCLAN(i));
                if (reg & VP_MDET_TX_TCLAN_VALID_M) {
                        wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
-                       vf_mdd_detected = true;
-                       dev_info(dev, "TX driver issue detected on VF %d\n",
-                                i);
+                       vf->mdd_tx_events.count++;
+                       set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
+                       if (netif_msg_tx_err(pf))
+                               dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
+                                        i);
                }
 
                reg = rd32(hw, VP_MDET_TX_TDPU(i));
                if (reg & VP_MDET_TX_TDPU_VALID_M) {
                        wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
-                       vf_mdd_detected = true;
-                       dev_info(dev, "TX driver issue detected on VF %d\n",
-                                i);
+                       vf->mdd_tx_events.count++;
+                       set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
+                       if (netif_msg_tx_err(pf))
+                               dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
+                                        i);
                }
 
                reg = rd32(hw, VP_MDET_RX(i));
                if (reg & VP_MDET_RX_VALID_M) {
                        wr32(hw, VP_MDET_RX(i), 0xFFFF);
-                       vf_mdd_detected = true;
-                       dev_info(dev, "RX driver issue detected on VF %d\n",
-                                i);
-               }
-
-               if (vf_mdd_detected) {
-                       vf->num_mdd_events++;
-                       if (vf->num_mdd_events &&
-                           vf->num_mdd_events <= ICE_MDD_EVENTS_THRESHOLD)
-                               dev_info(dev, "VF %d has had %llu MDD events since last boot, Admin might need to reload AVF driver with this number of events\n",
-                                        i, vf->num_mdd_events);
+                       vf->mdd_rx_events.count++;
+                       set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
+                       if (netif_msg_rx_err(pf))
+                               dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
+                                        i);
+
+                       /* Since the queue is disabled on VF Rx MDD events, the
+                        * PF can be configured to reset the VF through ethtool
+                        * private flag mdd-auto-reset-vf.
+                        */
+                       if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags))
+                               ice_reset_vf(&pf->vf[i], false);
                }
        }
+
+       ice_print_vfs_mdd_events(pf);
 }
 
 /**
@@ -1995,6 +2001,14 @@ static void ice_ena_misc_vector(struct ice_pf *pf)
        struct ice_hw *hw = &pf->hw;
        u32 val;
 
+       /* Disable anti-spoof detection interrupt to prevent spurious event
+        * interrupts during a function reset. Anti-spoof functionally is
+        * still supported.
+        */
+       val = rd32(hw, GL_MDCK_TX_TDPU);
+       val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
+       wr32(hw, GL_MDCK_TX_TDPU, val);
+
        /* clear things first */
        wr32(hw, PFINT_OICR_ENA, 0);    /* disable all */
        rd32(hw, PFINT_OICR);           /* read to clear */
index a21f9d2edbbb970f05de605be349f9cd975ad16e..e5c99bb8529e4db46471dafd0c0f27a3d498ef0d 100644 (file)
@@ -171,6 +171,11 @@ static void ice_free_vf_res(struct ice_vf *vf)
        }
 
        last_vector_idx = vf->first_vector_idx + pf->num_vf_msix - 1;
+
+       /* clear VF MDD event information */
+       memset(&vf->mdd_tx_events, 0, sizeof(vf->mdd_tx_events));
+       memset(&vf->mdd_rx_events, 0, sizeof(vf->mdd_rx_events));
+
        /* Disable interrupts so that VF starts in a known state */
        for (i = vf->first_vector_idx; i <= last_vector_idx; i++) {
                wr32(&pf->hw, GLINT_DYN_CTL(i), GLINT_DYN_CTL_CLEARPBA_M);
@@ -1175,7 +1180,7 @@ static bool ice_is_vf_disabled(struct ice_vf *vf)
  *
  * Returns true if the VF is reset, false otherwise.
  */
-static bool ice_reset_vf(struct ice_vf *vf, bool is_vflr)
+bool ice_reset_vf(struct ice_vf *vf, bool is_vflr)
 {
        struct ice_pf *pf = vf->pf;
        struct ice_vsi *vsi;
@@ -3529,3 +3534,52 @@ int ice_get_vf_stats(struct net_device *netdev, int vf_id,
 
        return 0;
 }
+
+/**
+ * ice_print_vfs_mdd_event - print VFs malicious driver detect event
+ * @pf: pointer to the PF structure
+ *
+ * Called from ice_handle_mdd_event to rate limit and print VFs MDD events.
+ */
+void ice_print_vfs_mdd_events(struct ice_pf *pf)
+{
+       struct device *dev = ice_pf_to_dev(pf);
+       struct ice_hw *hw = &pf->hw;
+       int i;
+
+       /* check that there are pending MDD events to print */
+       if (!test_and_clear_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state))
+               return;
+
+       /* VF MDD event logs are rate limited to one second intervals */
+       if (time_is_after_jiffies(pf->last_printed_mdd_jiffies + HZ * 1))
+               return;
+
+       pf->last_printed_mdd_jiffies = jiffies;
+
+       ice_for_each_vf(pf, i) {
+               struct ice_vf *vf = &pf->vf[i];
+
+               /* only print Rx MDD event message if there are new events */
+               if (vf->mdd_rx_events.count != vf->mdd_rx_events.last_printed) {
+                       vf->mdd_rx_events.last_printed =
+                                                       vf->mdd_rx_events.count;
+
+                       dev_info(dev, "%d Rx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n",
+                                vf->mdd_rx_events.count, hw->pf_id, i,
+                                vf->dflt_lan_addr.addr,
+                                test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)
+                                         ? "on" : "off");
+               }
+
+               /* only print Tx MDD event message if there are new events */
+               if (vf->mdd_tx_events.count != vf->mdd_tx_events.last_printed) {
+                       vf->mdd_tx_events.last_printed =
+                                                       vf->mdd_tx_events.count;
+
+                       dev_info(dev, "%d Tx Malicious Driver Detection events detected on PF %d VF %d MAC %pM.\n",
+                                vf->mdd_tx_events.count, hw->pf_id, i,
+                                vf->dflt_lan_addr.addr);
+               }
+       }
+}
index 474b2613f09c8d5ff482a2024dd472ef237053bc..656f1909b38f27a9e306583bb9037a5ffd9100c4 100644 (file)
@@ -55,6 +55,13 @@ enum ice_virtchnl_cap {
        ICE_VIRTCHNL_VF_CAP_PRIVILEGE,
 };
 
+/* VF MDD events print structure */
+struct ice_mdd_vf_events {
+       u16 count;                      /* total count of Rx|Tx events */
+       /* count number of the last printed event */
+       u16 last_printed;
+};
+
 /* VF information structure */
 struct ice_vf {
        struct ice_pf *pf;
@@ -83,13 +90,14 @@ struct ice_vf {
        unsigned int tx_rate;           /* Tx bandwidth limit in Mbps */
        DECLARE_BITMAP(vf_states, ICE_VF_STATES_NBITS); /* VF runtime states */
 
-       u64 num_mdd_events;             /* number of MDD events detected */
        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;
        u16 num_vf_qs;                  /* num of queue configured per VF */
+       struct ice_mdd_vf_events mdd_rx_events;
+       struct ice_mdd_vf_events mdd_tx_events;
 };
 
 #ifdef CONFIG_PCI_IOV
@@ -104,6 +112,7 @@ void ice_vc_process_vf_msg(struct ice_pf *pf, struct ice_rq_event_info *event);
 void ice_vc_notify_link_state(struct ice_pf *pf);
 void ice_vc_notify_reset(struct ice_pf *pf);
 bool ice_reset_all_vfs(struct ice_pf *pf, bool is_vflr);
+bool ice_reset_vf(struct ice_vf *vf, bool is_vflr);
 
 int
 ice_set_vf_port_vlan(struct net_device *netdev, int vf_id, u16 vlan_id, u8 qos,
@@ -123,7 +132,7 @@ ice_get_vf_stats(struct net_device *netdev, int vf_id,
                 struct ifla_vf_stats *vf_stats);
 void
 ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event);
-
+void ice_print_vfs_mdd_events(struct ice_pf *pf);
 #else /* CONFIG_PCI_IOV */
 #define ice_process_vflr_event(pf) do {} while (0)
 #define ice_free_vfs(pf) do {} while (0)
@@ -132,6 +141,7 @@ ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event);
 #define ice_vc_notify_reset(pf) do {} while (0)
 #define ice_set_vf_state_qs_dis(vf) do {} while (0)
 #define ice_vf_lan_overflow_event(pf, event) do {} while (0)
+#define ice_print_vfs_mdd_events(pf) do {} while (0)
 
 static inline bool
 ice_reset_all_vfs(struct ice_pf __always_unused *pf,
@@ -140,6 +150,12 @@ ice_reset_all_vfs(struct ice_pf __always_unused *pf,
        return true;
 }
 
+static inline bool
+ice_reset_vf(struct ice_vf __always_unused *vf, bool __always_unused is_vflr)
+{
+       return true;
+}
+
 static inline int
 ice_sriov_configure(struct pci_dev __always_unused *pdev,
                    int __always_unused num_vfs)