Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
100GbE Intel Wired LAN Driver Updates 2022-03-03

Jacob Keller says:

This series refactors the ice networking driver VF storage from a simple
static array to a hash table. It also introduces krefs and proper locking
and protection to prevent common use-after-free and concurrency issues.

There are two motivations for this work. First is to make the ice driver
more resilient by preventing a whole class of use-after-free bugs that can
occur around concurrent access to VF structures while removing VFs.

The second is to prepare the ice driver for future virtualization work to
support Scalable IOV, an alternative VF implementation compared to Single
Root IOV. The new VF implementation will allow for more dynamic VF creation
and removal, necessitating a more robust implementation for VF storage that
can't rely on the existing mechanisms to prevent concurrent access
violations.

The first few patches are cleanup and preparatory work needed to make the
conversion to the hash table safe. Following this preparatory work is a
patch to migrate the VF structures and variables to a new sub-structure for
code clarity. Next introduce new interface functions to abstract the VF
storage. Finally, the driver is actually converted to the hash table and
kref implementation.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 36 insertions, 28 deletions

diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 95dfd04e8c3a..289e5c99e313 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -505,7 +505,8 @@ ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
 {
 	struct ice_hw *hw = &pf->hw;
 	struct ice_vsi *vsi;
-	unsigned int i;
+	struct ice_vf *vf;
+	unsigned int bkt;
 
 	dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
 
@@ -520,8 +521,10 @@ ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
 		ice_vc_notify_reset(pf);
 
 	/* Disable VFs until reset is completed */
-	ice_for_each_vf(pf, i)
-		ice_set_vf_state_qs_dis(&pf->vf[i]);
+	mutex_lock(&pf->vfs.table_lock);
+	ice_for_each_vf(pf, bkt, vf)
+		ice_set_vf_state_qs_dis(vf);
+	mutex_unlock(&pf->vfs.table_lock);
 
 	if (ice_is_eswitch_mode_switchdev(pf)) {
 		if (reset_type != ICE_RESET_PFR)
@@ -1666,7 +1669,8 @@ static void ice_handle_mdd_event(struct ice_pf *pf)
 {
 	struct device *dev = ice_pf_to_dev(pf);
 	struct ice_hw *hw = &pf->hw;
-	unsigned int i;
+	struct ice_vf *vf;
+	unsigned int bkt;
 	u32 reg;
 
 	if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
@@ -1754,47 +1758,46 @@ static void ice_handle_mdd_event(struct ice_pf *pf)
 	/* 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];
-
-		reg = rd32(hw, VP_MDET_TX_PQM(i));
+	mutex_lock(&pf->vfs.table_lock);
+	ice_for_each_vf(pf, bkt, vf) {
+		reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
 		if (reg & VP_MDET_TX_PQM_VALID_M) {
-			wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
+			wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
 			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);
+					 vf->vf_id);
 		}
 
-		reg = rd32(hw, VP_MDET_TX_TCLAN(i));
+		reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
 		if (reg & VP_MDET_TX_TCLAN_VALID_M) {
-			wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
+			wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
 			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);
+					 vf->vf_id);
 		}
 
-		reg = rd32(hw, VP_MDET_TX_TDPU(i));
+		reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
 		if (reg & VP_MDET_TX_TDPU_VALID_M) {
-			wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
+			wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
 			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);
+					 vf->vf_id);
 		}
 
-		reg = rd32(hw, VP_MDET_RX(i));
+		reg = rd32(hw, VP_MDET_RX(vf->vf_id));
 		if (reg & VP_MDET_RX_VALID_M) {
-			wr32(hw, VP_MDET_RX(i), 0xFFFF);
+			wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
 			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);
+					 vf->vf_id);
 
 			/* Since the queue is disabled on VF Rx MDD events, the
 			 * PF can be configured to reset the VF through ethtool
@@ -1805,12 +1808,13 @@ 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);
+				mutex_lock(&vf->cfg_lock);
+				ice_reset_vf(vf, false);
+				mutex_unlock(&vf->cfg_lock);
 			}
 		}
 	}
+	mutex_unlock(&pf->vfs.table_lock);
 
 	ice_print_vfs_mdd_events(pf);
 }
@@ -2439,7 +2443,7 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 			/* skip this unused q_vector */
 			continue;
 		}
-		if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
+		if (vsi->type == ICE_VSI_CTRL && vsi->vf)
 			err = devm_request_irq(dev, irq_num, vsi->irq_handler,
 					       IRQF_SHARED, q_vector->name,
 					       q_vector);
@@ -3386,14 +3390,14 @@ void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
 static struct ice_vsi *
 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
 {
-	return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID, NULL);
+	return ice_vsi_setup(pf, pi, ICE_VSI_PF, NULL, NULL);
 }
 
 static struct ice_vsi *
 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
 		   struct ice_channel *ch)
 {
-	return ice_vsi_setup(pf, pi, ICE_VSI_CHNL, ICE_INVAL_VFID, ch);
+	return ice_vsi_setup(pf, pi, ICE_VSI_CHNL, NULL, ch);
 }
 
 /**
@@ -3407,7 +3411,7 @@ ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
 static struct ice_vsi *
 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
 {
-	return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID, NULL);
+	return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, NULL, NULL);
 }
 
 /**
@@ -3421,7 +3425,7 @@ ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
 struct ice_vsi *
 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
 {
-	return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID, NULL);
+	return ice_vsi_setup(pf, pi, ICE_VSI_LB, NULL, NULL);
 }
 
 /**
@@ -3680,6 +3684,7 @@ static void ice_deinit_pf(struct ice_pf *pf)
 	mutex_destroy(&pf->sw_mutex);
 	mutex_destroy(&pf->tc_mutex);
 	mutex_destroy(&pf->avail_q_mutex);
+	mutex_destroy(&pf->vfs.table_lock);
 
 	if (pf->avail_txqs) {
 		bitmap_free(pf->avail_txqs);
@@ -3712,7 +3717,7 @@ static void ice_set_pf_caps(struct ice_pf *pf)
 	clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
 	if (func_caps->common_cap.sr_iov_1_1) {
 		set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
-		pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
+		pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
 					      ICE_MAX_VF_COUNT);
 	}
 	clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
@@ -3779,6 +3784,9 @@ static int ice_init_pf(struct ice_pf *pf)
 		return -ENOMEM;
 	}
 
+	mutex_init(&pf->vfs.table_lock);
+	hash_init(pf->vfs.table);
+
 	return 0;
 }