Merge branch 'mlx-next'

Or Gerlitz says:

====================
Mellanox driver update, Oct 14 2015

This series contains two more patches from Eli, patch from Majd
to support PCI error handlers and a fix from Jack to mlx4 VFs
when probed without a provisioned mac address.

The patch set applied on top of net-next commit bbb300eb "Merge branch 'bridge-vlan'"

changes from V0:
  - made the health flag int --> bool to address comment from Dave on patch #1
  - fixed sparse warning noted by the 0-day build tests in patch #2
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/mellanox/mlx4/en_netdev.c

@@ -2816,7 +2816,6 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	struct mlx4_en_priv *priv;
 	int i;
 	int err;
-	u64 mac_u64;
 
 	dev = alloc_etherdev_mqs(sizeof(struct mlx4_en_priv),
 				 MAX_TX_RINGS, MAX_RX_RINGS);
@@ -2908,17 +2907,17 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
 	dev->addr_len = ETH_ALEN;
 	mlx4_en_u64_to_mac(dev->dev_addr, mdev->dev->caps.def_mac[priv->port]);
 	if (!is_valid_ether_addr(dev->dev_addr)) {
-		if (mlx4_is_slave(priv->mdev->dev)) {
-			eth_hw_addr_random(dev);
-			en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
-			mac_u64 = mlx4_mac_to_u64(dev->dev_addr);
-			mdev->dev->caps.def_mac[priv->port] = mac_u64;
-		} else {
-			en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
-			       priv->port, dev->dev_addr);
-			err = -EINVAL;
-			goto out;
-		}
+		en_err(priv, "Port: %d, invalid mac burned: %pM, quiting\n",
+		       priv->port, dev->dev_addr);
+		err = -EINVAL;
+		goto out;
+	} else if (mlx4_is_slave(priv->mdev->dev) &&
+		   (priv->mdev->dev->port_random_macs & 1 << priv->port)) {
+		/* Random MAC was assigned in mlx4_slave_cap
+		 * in mlx4_core module
+		 */
+		dev->addr_assign_type |= NET_ADDR_RANDOM;
+		en_warn(priv, "Assigned random MAC address %pM\n", dev->dev_addr);
 	}
 
 	memcpy(priv->current_mac, dev->dev_addr, sizeof(priv->current_mac));

drivers/net/ethernet/mellanox/mlx4/fw.c

@@ -2840,3 +2840,19 @@ int set_phv_bit(struct mlx4_dev *dev, u8 port, int new_val)
 	return -EOPNOTSUPP;
 }
 EXPORT_SYMBOL(set_phv_bit);
+
+void mlx4_replace_zero_macs(struct mlx4_dev *dev)
+{
+	int i;
+	u8 mac_addr[ETH_ALEN];
+
+	dev->port_random_macs = 0;
+	for (i = 1; i <= dev->caps.num_ports; ++i)
+		if (!dev->caps.def_mac[i] &&
+		    dev->caps.port_type[i] == MLX4_PORT_TYPE_ETH) {
+			eth_random_addr(mac_addr);
+			dev->port_random_macs |= 1 << i;
+			dev->caps.def_mac[i] = mlx4_mac_to_u64(mac_addr);
+		}
+}
+EXPORT_SYMBOL_GPL(mlx4_replace_zero_macs);

drivers/net/ethernet/mellanox/mlx4/main.c

@@ -863,6 +863,8 @@ static int mlx4_slave_cap(struct mlx4_dev *dev)
 		return -ENODEV;
 	}
 
+	mlx4_replace_zero_macs(dev);
+
 	dev->caps.qp0_qkey = kcalloc(dev->caps.num_ports, sizeof(u32), GFP_KERNEL);
 	dev->caps.qp0_tunnel = kcalloc(dev->caps.num_ports, sizeof (u32), GFP_KERNEL);
 	dev->caps.qp0_proxy = kcalloc(dev->caps.num_ports, sizeof (u32), GFP_KERNEL);

drivers/net/ethernet/mellanox/mlx4/mlx4.h

@@ -1378,6 +1378,8 @@ void mlx4_vf_immed_vlan_work_handler(struct work_struct *_work);
 
 void mlx4_init_quotas(struct mlx4_dev *dev);
 
+/* for VFs, replace zero MACs with randomly-generated MACs at driver start */
+void mlx4_replace_zero_macs(struct mlx4_dev *dev);
 int mlx4_get_slave_num_gids(struct mlx4_dev *dev, int slave, int port);
 /* Returns the VF index of slave */
 int mlx4_get_vf_indx(struct mlx4_dev *dev, int slave);

drivers/net/ethernet/mellanox/mlx5/core/cmd.c

@@ -256,8 +256,154 @@ static void dump_buf(void *buf, int size, int data_only, int offset)
 
 enum {
 	MLX5_DRIVER_STATUS_ABORTED = 0xfe,
+	MLX5_DRIVER_SYND = 0xbadd00de,
 };
 
+static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
+				       u32 *synd, u8 *status)
+{
+	*synd = 0;
+	*status = 0;
+
+	switch (op) {
+	case MLX5_CMD_OP_TEARDOWN_HCA:
+	case MLX5_CMD_OP_DISABLE_HCA:
+	case MLX5_CMD_OP_MANAGE_PAGES:
+	case MLX5_CMD_OP_DESTROY_MKEY:
+	case MLX5_CMD_OP_DESTROY_EQ:
+	case MLX5_CMD_OP_DESTROY_CQ:
+	case MLX5_CMD_OP_DESTROY_QP:
+	case MLX5_CMD_OP_DESTROY_PSV:
+	case MLX5_CMD_OP_DESTROY_SRQ:
+	case MLX5_CMD_OP_DESTROY_XRC_SRQ:
+	case MLX5_CMD_OP_DESTROY_DCT:
+	case MLX5_CMD_OP_DEALLOC_Q_COUNTER:
+	case MLX5_CMD_OP_DEALLOC_PD:
+	case MLX5_CMD_OP_DEALLOC_UAR:
+	case MLX5_CMD_OP_DETTACH_FROM_MCG:
+	case MLX5_CMD_OP_DEALLOC_XRCD:
+	case MLX5_CMD_OP_DEALLOC_TRANSPORT_DOMAIN:
+	case MLX5_CMD_OP_DELETE_VXLAN_UDP_DPORT:
+	case MLX5_CMD_OP_DELETE_L2_TABLE_ENTRY:
+	case MLX5_CMD_OP_DESTROY_TIR:
+	case MLX5_CMD_OP_DESTROY_SQ:
+	case MLX5_CMD_OP_DESTROY_RQ:
+	case MLX5_CMD_OP_DESTROY_RMP:
+	case MLX5_CMD_OP_DESTROY_TIS:
+	case MLX5_CMD_OP_DESTROY_RQT:
+	case MLX5_CMD_OP_DESTROY_FLOW_TABLE:
+	case MLX5_CMD_OP_DESTROY_FLOW_GROUP:
+	case MLX5_CMD_OP_DELETE_FLOW_TABLE_ENTRY:
+		return MLX5_CMD_STAT_OK;
+
+	case MLX5_CMD_OP_QUERY_HCA_CAP:
+	case MLX5_CMD_OP_QUERY_ADAPTER:
+	case MLX5_CMD_OP_INIT_HCA:
+	case MLX5_CMD_OP_ENABLE_HCA:
+	case MLX5_CMD_OP_QUERY_PAGES:
+	case MLX5_CMD_OP_SET_HCA_CAP:
+	case MLX5_CMD_OP_QUERY_ISSI:
+	case MLX5_CMD_OP_SET_ISSI:
+	case MLX5_CMD_OP_CREATE_MKEY:
+	case MLX5_CMD_OP_QUERY_MKEY:
+	case MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS:
+	case MLX5_CMD_OP_PAGE_FAULT_RESUME:
+	case MLX5_CMD_OP_CREATE_EQ:
+	case MLX5_CMD_OP_QUERY_EQ:
+	case MLX5_CMD_OP_GEN_EQE:
+	case MLX5_CMD_OP_CREATE_CQ:
+	case MLX5_CMD_OP_QUERY_CQ:
+	case MLX5_CMD_OP_MODIFY_CQ:
+	case MLX5_CMD_OP_CREATE_QP:
+	case MLX5_CMD_OP_RST2INIT_QP:
+	case MLX5_CMD_OP_INIT2RTR_QP:
+	case MLX5_CMD_OP_RTR2RTS_QP:
+	case MLX5_CMD_OP_RTS2RTS_QP:
+	case MLX5_CMD_OP_SQERR2RTS_QP:
+	case MLX5_CMD_OP_2ERR_QP:
+	case MLX5_CMD_OP_2RST_QP:
+	case MLX5_CMD_OP_QUERY_QP:
+	case MLX5_CMD_OP_SQD_RTS_QP:
+	case MLX5_CMD_OP_INIT2INIT_QP:
+	case MLX5_CMD_OP_CREATE_PSV:
+	case MLX5_CMD_OP_CREATE_SRQ:
+	case MLX5_CMD_OP_QUERY_SRQ:
+	case MLX5_CMD_OP_ARM_RQ:
+	case MLX5_CMD_OP_CREATE_XRC_SRQ:
+	case MLX5_CMD_OP_QUERY_XRC_SRQ:
+	case MLX5_CMD_OP_ARM_XRC_SRQ:
+	case MLX5_CMD_OP_CREATE_DCT:
+	case MLX5_CMD_OP_DRAIN_DCT:
+	case MLX5_CMD_OP_QUERY_DCT:
+	case MLX5_CMD_OP_ARM_DCT_FOR_KEY_VIOLATION:
+	case MLX5_CMD_OP_QUERY_VPORT_STATE:
+	case MLX5_CMD_OP_MODIFY_VPORT_STATE:
+	case MLX5_CMD_OP_QUERY_ESW_VPORT_CONTEXT:
+	case MLX5_CMD_OP_MODIFY_ESW_VPORT_CONTEXT:
+	case MLX5_CMD_OP_QUERY_NIC_VPORT_CONTEXT:
+	case MLX5_CMD_OP_MODIFY_NIC_VPORT_CONTEXT:
+	case MLX5_CMD_OP_QUERY_ROCE_ADDRESS:
+	case MLX5_CMD_OP_SET_ROCE_ADDRESS:
+	case MLX5_CMD_OP_QUERY_HCA_VPORT_CONTEXT:
+	case MLX5_CMD_OP_MODIFY_HCA_VPORT_CONTEXT:
+	case MLX5_CMD_OP_QUERY_HCA_VPORT_GID:
+	case MLX5_CMD_OP_QUERY_HCA_VPORT_PKEY:
+	case MLX5_CMD_OP_QUERY_VPORT_COUNTER:
+	case MLX5_CMD_OP_ALLOC_Q_COUNTER:
+	case MLX5_CMD_OP_QUERY_Q_COUNTER:
+	case MLX5_CMD_OP_ALLOC_PD:
+	case MLX5_CMD_OP_ALLOC_UAR:
+	case MLX5_CMD_OP_CONFIG_INT_MODERATION:
+	case MLX5_CMD_OP_ACCESS_REG:
+	case MLX5_CMD_OP_ATTACH_TO_MCG:
+	case MLX5_CMD_OP_GET_DROPPED_PACKET_LOG:
+	case MLX5_CMD_OP_MAD_IFC:
+	case MLX5_CMD_OP_QUERY_MAD_DEMUX:
+	case MLX5_CMD_OP_SET_MAD_DEMUX:
+	case MLX5_CMD_OP_NOP:
+	case MLX5_CMD_OP_ALLOC_XRCD:
+	case MLX5_CMD_OP_ALLOC_TRANSPORT_DOMAIN:
+	case MLX5_CMD_OP_QUERY_CONG_STATUS:
+	case MLX5_CMD_OP_MODIFY_CONG_STATUS:
+	case MLX5_CMD_OP_QUERY_CONG_PARAMS:
+	case MLX5_CMD_OP_MODIFY_CONG_PARAMS:
+	case MLX5_CMD_OP_QUERY_CONG_STATISTICS:
+	case MLX5_CMD_OP_ADD_VXLAN_UDP_DPORT:
+	case MLX5_CMD_OP_SET_L2_TABLE_ENTRY:
+	case MLX5_CMD_OP_QUERY_L2_TABLE_ENTRY:
+	case MLX5_CMD_OP_CREATE_TIR:
+	case MLX5_CMD_OP_MODIFY_TIR:
+	case MLX5_CMD_OP_QUERY_TIR:
+	case MLX5_CMD_OP_CREATE_SQ:
+	case MLX5_CMD_OP_MODIFY_SQ:
+	case MLX5_CMD_OP_QUERY_SQ:
+	case MLX5_CMD_OP_CREATE_RQ:
+	case MLX5_CMD_OP_MODIFY_RQ:
+	case MLX5_CMD_OP_QUERY_RQ:
+	case MLX5_CMD_OP_CREATE_RMP:
+	case MLX5_CMD_OP_MODIFY_RMP:
+	case MLX5_CMD_OP_QUERY_RMP:
+	case MLX5_CMD_OP_CREATE_TIS:
+	case MLX5_CMD_OP_MODIFY_TIS:
+	case MLX5_CMD_OP_QUERY_TIS:
+	case MLX5_CMD_OP_CREATE_RQT:
+	case MLX5_CMD_OP_MODIFY_RQT:
+	case MLX5_CMD_OP_QUERY_RQT:
+	case MLX5_CMD_OP_CREATE_FLOW_TABLE:
+	case MLX5_CMD_OP_QUERY_FLOW_TABLE:
+	case MLX5_CMD_OP_CREATE_FLOW_GROUP:
+	case MLX5_CMD_OP_QUERY_FLOW_GROUP:
+	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
+	case MLX5_CMD_OP_QUERY_FLOW_TABLE_ENTRY:
+		*status = MLX5_DRIVER_STATUS_ABORTED;
+		*synd = MLX5_DRIVER_SYND;
+		return -EIO;
+	default:
+		mlx5_core_err(dev, "Unknown FW command (%d)\n", op);
+		return -EINVAL;
+	}
+}
+
 const char *mlx5_command_str(int command)
 {
 	switch (command) {
@@ -592,6 +738,16 @@ static int wait_func(struct mlx5_core_dev *dev, struct mlx5_cmd_work_ent *ent)
 	return err;
 }
 
+static __be32 *get_synd_ptr(struct mlx5_outbox_hdr *out)
+{
+	return &out->syndrome;
+}
+
+static u8 *get_status_ptr(struct mlx5_outbox_hdr *out)
+{
+	return &out->status;
+}
+
 /*  Notes:
  *    1. Callback functions may not sleep
  *    2. page queue commands do not support asynchrous completion
@@ -1200,6 +1356,11 @@ static struct mlx5_cmd_msg *alloc_msg(struct mlx5_core_dev *dev, int in_size,
 	return msg;
 }
 
+static u16 opcode_from_in(struct mlx5_inbox_hdr *in)
+{
+	return be16_to_cpu(in->opcode);
+}
+
 static int is_manage_pages(struct mlx5_inbox_hdr *in)
 {
 	return be16_to_cpu(in->opcode) == MLX5_CMD_OP_MANAGE_PAGES;
@@ -1214,6 +1375,15 @@ static int cmd_exec(struct mlx5_core_dev *dev, void *in, int in_size, void *out,
 	gfp_t gfp;
 	int err;
 	u8 status = 0;
+	u32 drv_synd;
+
+	if (pci_channel_offline(dev->pdev) ||
+	    dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		err = mlx5_internal_err_ret_value(dev, opcode_from_in(in), &drv_synd, &status);
+		*get_synd_ptr(out) = cpu_to_be32(drv_synd);
+		*get_status_ptr(out) = status;
+		return err;
+	}
 
 	pages_queue = is_manage_pages(in);
 	gfp = callback ? GFP_ATOMIC : GFP_KERNEL;

drivers/net/ethernet/mellanox/mlx5/core/health.c

@@ -34,6 +34,7 @@
 #include <linux/module.h>
 #include <linux/random.h>
 #include <linux/vmalloc.h>
+#include <linux/hardirq.h>
 #include <linux/mlx5/driver.h>
 #include <linux/mlx5/cmd.h>
 #include "mlx5_core.h"
@@ -57,6 +58,91 @@ enum {
 	MLX5_HEALTH_SYNDR_HIGH_TEMP		= 0x10
 };
 
+enum {
+	MLX5_NIC_IFC_FULL		= 0,
+	MLX5_NIC_IFC_DISABLED		= 1,
+	MLX5_NIC_IFC_NO_DRAM_NIC	= 2
+};
+
+static u8 get_nic_interface(struct mlx5_core_dev *dev)
+{
+	return (ioread32be(&dev->iseg->cmdq_addr_l_sz) >> 8) & 3;
+}
+
+static void trigger_cmd_completions(struct mlx5_core_dev *dev)
+{
+	unsigned long flags;
+	u64 vector;
+
+	/* wait for pending handlers to complete */
+	synchronize_irq(dev->priv.msix_arr[MLX5_EQ_VEC_CMD].vector);
+	spin_lock_irqsave(&dev->cmd.alloc_lock, flags);
+	vector = ~dev->cmd.bitmask & ((1ul << (1 << dev->cmd.log_sz)) - 1);
+	if (!vector)
+		goto no_trig;
+
+	vector |= MLX5_TRIGGERED_CMD_COMP;
+	spin_unlock_irqrestore(&dev->cmd.alloc_lock, flags);
+
+	mlx5_core_dbg(dev, "vector 0x%llx\n", vector);
+	mlx5_cmd_comp_handler(dev, vector);
+	return;
+
+no_trig:
+	spin_unlock_irqrestore(&dev->cmd.alloc_lock, flags);
+}
+
+static int in_fatal(struct mlx5_core_dev *dev)
+{
+	struct mlx5_core_health *health = &dev->priv.health;
+	struct health_buffer __iomem *h = health->health;
+
+	if (get_nic_interface(dev) == MLX5_NIC_IFC_DISABLED)
+		return 1;
+
+	if (ioread32be(&h->fw_ver) == 0xffffffff)
+		return 1;
+
+	return 0;
+}
+
+void mlx5_enter_error_state(struct mlx5_core_dev *dev)
+{
+	if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
+		return;
+
+	mlx5_core_err(dev, "start\n");
+	if (pci_channel_offline(dev->pdev) || in_fatal(dev))
+		dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
+
+	mlx5_core_event(dev, MLX5_DEV_EVENT_SYS_ERROR, 0);
+	mlx5_core_err(dev, "end\n");
+}
+
+static void mlx5_handle_bad_state(struct mlx5_core_dev *dev)
+{
+	u8 nic_interface = get_nic_interface(dev);
+
+	switch (nic_interface) {
+	case MLX5_NIC_IFC_FULL:
+		mlx5_core_warn(dev, "Expected to see disabled NIC but it is full driver\n");
+		break;
+
+	case MLX5_NIC_IFC_DISABLED:
+		mlx5_core_warn(dev, "starting teardown\n");
+		break;
+
+	case MLX5_NIC_IFC_NO_DRAM_NIC:
+		mlx5_core_warn(dev, "Expected to see disabled NIC but it is no dram nic\n");
+		break;
+	default:
+		mlx5_core_warn(dev, "Expected to see disabled NIC but it is has invalid value %d\n",
+			       nic_interface);
+	}
+
+	mlx5_disable_device(dev);
+}
+
 static void health_care(struct work_struct *work)
 {
 	struct mlx5_core_health *health;
@@ -67,6 +153,7 @@ static void health_care(struct work_struct *work)
 	priv = container_of(health, struct mlx5_priv, health);
 	dev = container_of(priv, struct mlx5_core_dev, priv);
 	mlx5_core_warn(dev, "handling bad device here\n");
+	mlx5_handle_bad_state(dev);
 }
 
 static const char *hsynd_str(u8 synd)
@@ -122,6 +209,10 @@ static void print_health_info(struct mlx5_core_dev *dev)
 	u32 fw;
 	int i;
 
+	/* If the syndrom is 0, the device is OK and no need to print buffer */
+	if (!ioread8(&h->synd))
+		return;
+
 	for (i = 0; i < ARRAY_SIZE(h->assert_var); i++)
 		dev_err(&dev->pdev->dev, "assert_var[%d] 0x%08x\n", i, ioread32be(h->assert_var + i));
 
@@ -136,13 +227,29 @@ static void print_health_info(struct mlx5_core_dev *dev)
 	dev_err(&dev->pdev->dev, "ext_synd 0x%04x\n", ioread16be(&h->ext_synd));
 }
 
+static unsigned long get_next_poll_jiffies(void)
+{
+	unsigned long next;
+
+	get_random_bytes(&next, sizeof(next));
+	next %= HZ;
+	next += jiffies + MLX5_HEALTH_POLL_INTERVAL;
+
+	return next;
+}
+
 static void poll_health(unsigned long data)
 {
 	struct mlx5_core_dev *dev = (struct mlx5_core_dev *)data;
 	struct mlx5_core_health *health = &dev->priv.health;
-	unsigned long next;
 	u32 count;
 
+	if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+		trigger_cmd_completions(dev);
+		mod_timer(&health->timer, get_next_poll_jiffies());
+		return;
+	}
+
 	count = ioread32be(health->health_counter);
 	if (count == health->prev)
 		++health->miss_counter;
@@ -151,14 +258,16 @@ static void poll_health(unsigned long data)
 
 	health->prev = count;
 	if (health->miss_counter == MAX_MISSES) {
-		mlx5_core_err(dev, "device's health compromised\n");
+		dev_err(&dev->pdev->dev, "device's health compromised - reached miss count\n");
 		print_health_info(dev);
-		queue_work(health->wq, &health->work);
 	} else {
-		get_random_bytes(&next, sizeof(next));
-		next %= HZ;
-		next += jiffies + MLX5_HEALTH_POLL_INTERVAL;
-		mod_timer(&health->timer, next);
+		mod_timer(&health->timer, get_next_poll_jiffies());
+	}
+
+	if (in_fatal(dev) && !health->sick) {
+		health->sick = true;
+		print_health_info(dev);
+		queue_work(health->wq, &health->work);
 	}
 }
 

drivers/net/ethernet/mellanox/mlx5/core/main.c

@@ -39,12 +39,14 @@
 #include <linux/slab.h>
 #include <linux/io-mapping.h>
 #include <linux/interrupt.h>
+#include <linux/delay.h>
 #include <linux/mlx5/driver.h>
 #include <linux/mlx5/cq.h>
 #include <linux/mlx5/qp.h>
 #include <linux/mlx5/srq.h>
 #include <linux/debugfs.h>
 #include <linux/kmod.h>
+#include <linux/delay.h>
 #include <linux/mlx5/mlx5_ifc.h>
 #include "mlx5_core.h"
 
@@ -151,6 +153,25 @@ static struct mlx5_profile profile[] = {
 	},
 };
 
+#define FW_INIT_TIMEOUT_MILI	2000
+#define FW_INIT_WAIT_MS		2
+
+static int wait_fw_init(struct mlx5_core_dev *dev, u32 max_wait_mili)
+{
+	unsigned long end = jiffies + msecs_to_jiffies(max_wait_mili);
+	int err = 0;
+
+	while (fw_initializing(dev)) {
+		if (time_after(jiffies, end)) {
+			err = -EBUSY;
+			break;
+		}
+		msleep(FW_INIT_WAIT_MS);
+	}
+
+	return err;
+}
+
 static int set_dma_caps(struct pci_dev *pdev)
 {
 	int err;
@@ -181,6 +202,34 @@ static int set_dma_caps(struct pci_dev *pdev)
 	return err;
 }
 
+static int mlx5_pci_enable_device(struct mlx5_core_dev *dev)
+{
+	struct pci_dev *pdev = dev->pdev;
+	int err = 0;
+
+	mutex_lock(&dev->pci_status_mutex);
+	if (dev->pci_status == MLX5_PCI_STATUS_DISABLED) {
+		err = pci_enable_device(pdev);
+		if (!err)
+			dev->pci_status = MLX5_PCI_STATUS_ENABLED;
+	}
+	mutex_unlock(&dev->pci_status_mutex);
+
+	return err;
+}
+
+static void mlx5_pci_disable_device(struct mlx5_core_dev *dev)
+{
+	struct pci_dev *pdev = dev->pdev;
+
+	mutex_lock(&dev->pci_status_mutex);
+	if (dev->pci_status == MLX5_PCI_STATUS_ENABLED) {
+		pci_disable_device(pdev);
+		dev->pci_status = MLX5_PCI_STATUS_DISABLED;
+	}
+	mutex_unlock(&dev->pci_status_mutex);
+}
+
 static int request_bar(struct pci_dev *pdev)
 {
 	int err = 0;
@@ -807,7 +856,7 @@ static int mlx5_pci_init(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	if (!priv->dbg_root)
 		return -ENOMEM;
 
-	err = pci_enable_device(pdev);
+	err = mlx5_pci_enable_device(dev);
 	if (err) {
 		dev_err(&pdev->dev, "Cannot enable PCI device, aborting\n");
 		goto err_dbg;
@@ -841,7 +890,7 @@ static int mlx5_pci_init(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	pci_clear_master(dev->pdev);
 	release_bar(dev->pdev);
 err_disable:
-	pci_disable_device(dev->pdev);
+	mlx5_pci_disable_device(dev);
 
 err_dbg:
 	debugfs_remove(priv->dbg_root);
@@ -853,7 +902,7 @@ static void mlx5_pci_close(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	iounmap(dev->iseg);
 	pci_clear_master(dev->pdev);
 	release_bar(dev->pdev);
-	pci_disable_device(dev->pdev);
+	mlx5_pci_disable_device(dev);
 	debugfs_remove(priv->dbg_root);
 }
 
@@ -863,13 +912,32 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	struct pci_dev *pdev = dev->pdev;
 	int err;
 
+	mutex_lock(&dev->intf_state_mutex);
+	if (dev->interface_state == MLX5_INTERFACE_STATE_UP) {
+		dev_warn(&dev->pdev->dev, "%s: interface is up, NOP\n",
+			 __func__);
+		goto out;
+	}
+
 	dev_info(&pdev->dev, "firmware version: %d.%d.%d\n", fw_rev_maj(dev),
 		 fw_rev_min(dev), fw_rev_sub(dev));
 
+	/* on load removing any previous indication of internal error, device is
+	 * up
+	 */
+	dev->state = MLX5_DEVICE_STATE_UP;
+
 	err = mlx5_cmd_init(dev);
 	if (err) {
 		dev_err(&pdev->dev, "Failed initializing command interface, aborting\n");
-		return err;
+		goto out_err;
+	}
+
+	err = wait_fw_init(dev, FW_INIT_TIMEOUT_MILI);
+	if (err) {
+		dev_err(&dev->pdev->dev, "Firmware over %d MS in initializing state, aborting\n",
+			FW_INIT_TIMEOUT_MILI);
+		goto out_err;
 	}
 
 	mlx5_pagealloc_init(dev);
@@ -994,6 +1062,10 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	if (err)
 		pr_info("failed request module on %s\n", MLX5_IB_MOD);
 
+	dev->interface_state = MLX5_INTERFACE_STATE_UP;
+out:
+	mutex_unlock(&dev->intf_state_mutex);
+
 	return 0;
 
 err_reg_dev:
@@ -1024,7 +1096,7 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	mlx5_stop_health_poll(dev);
 	if (mlx5_cmd_teardown_hca(dev)) {
 		dev_err(&dev->pdev->dev, "tear_down_hca failed, skip cleanup\n");
-		return err;
+		goto out_err;
 	}
 
 err_pagealloc_stop:
@@ -1040,13 +1112,23 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	mlx5_pagealloc_cleanup(dev);
 	mlx5_cmd_cleanup(dev);
 
+out_err:
+	dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
+	mutex_unlock(&dev->intf_state_mutex);
+
 	return err;
 }
 
 static int mlx5_unload_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 {
-	int err;
+	int err = 0;
 
+	mutex_lock(&dev->intf_state_mutex);
+	if (dev->interface_state == MLX5_INTERFACE_STATE_DOWN) {
+		dev_warn(&dev->pdev->dev, "%s: interface is down, NOP\n",
+			 __func__);
+		goto out;
+	}
 	mlx5_unregister_device(dev);
 	mlx5_cleanup_mr_table(dev);
 	mlx5_cleanup_srq_table(dev);
@@ -1072,10 +1154,12 @@ static int mlx5_unload_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	mlx5_cmd_cleanup(dev);
 
 out:
+	dev->interface_state = MLX5_INTERFACE_STATE_DOWN;
+	mutex_unlock(&dev->intf_state_mutex);
 	return err;
 }
 
-static void mlx5_core_event(struct mlx5_core_dev *dev, enum mlx5_dev_event event,
+void mlx5_core_event(struct mlx5_core_dev *dev, enum mlx5_dev_event event,
 		     unsigned long param)
 {
 	struct mlx5_priv *priv = &dev->priv;
@@ -1125,6 +1209,8 @@ static int init_one(struct pci_dev *pdev,
 
 	INIT_LIST_HEAD(&priv->ctx_list);
 	spin_lock_init(&priv->ctx_lock);
+	mutex_init(&dev->pci_status_mutex);
+	mutex_init(&dev->intf_state_mutex);
 	err = mlx5_pci_init(dev, priv);
 	if (err) {
 		dev_err(&pdev->dev, "mlx5_pci_init failed with error code %d\n", err);
@@ -1172,6 +1258,112 @@ static void remove_one(struct pci_dev *pdev)
 	kfree(dev);
 }
 
+static pci_ers_result_t mlx5_pci_err_detected(struct pci_dev *pdev,
+					      pci_channel_state_t state)
+{
+	struct mlx5_core_dev *dev = pci_get_drvdata(pdev);
+	struct mlx5_priv *priv = &dev->priv;
+
+	dev_info(&pdev->dev, "%s was called\n", __func__);
+	mlx5_enter_error_state(dev);
+	mlx5_unload_one(dev, priv);
+	mlx5_pci_disable_device(dev);
+	return state == pci_channel_io_perm_failure ?
+		PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_NEED_RESET;
+}
+
+static pci_ers_result_t mlx5_pci_slot_reset(struct pci_dev *pdev)
+{
+	struct mlx5_core_dev *dev = pci_get_drvdata(pdev);
+	int err = 0;
+
+	dev_info(&pdev->dev, "%s was called\n", __func__);
+
+	err = mlx5_pci_enable_device(dev);
+	if (err) {
+		dev_err(&pdev->dev, "%s: mlx5_pci_enable_device failed with error code: %d\n"
+			, __func__, err);
+		return PCI_ERS_RESULT_DISCONNECT;
+	}
+	pci_set_master(pdev);
+	pci_set_power_state(pdev, PCI_D0);
+	pci_restore_state(pdev);
+
+	return err ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED;
+}
+
+void mlx5_disable_device(struct mlx5_core_dev *dev)
+{
+	mlx5_pci_err_detected(dev->pdev, 0);
+}
+
+/* wait for the device to show vital signs. For now we check
+ * that we can read the device ID and that the health buffer
+ * shows a non zero value which is different than 0xffffffff
+ */
+static void wait_vital(struct pci_dev *pdev)
+{
+	struct mlx5_core_dev *dev = pci_get_drvdata(pdev);
+	struct mlx5_core_health *health = &dev->priv.health;
+	const int niter = 100;
+	u32 count;
+	u16 did;
+	int i;
+
+	/* Wait for firmware to be ready after reset */
+	msleep(1000);
+	for (i = 0; i < niter; i++) {
+		if (pci_read_config_word(pdev, 2, &did)) {
+			dev_warn(&pdev->dev, "failed reading config word\n");
+			break;
+		}
+		if (did == pdev->device) {
+			dev_info(&pdev->dev, "device ID correctly read after %d iterations\n", i);
+			break;
+		}
+		msleep(50);
+	}
+	if (i == niter)
+		dev_warn(&pdev->dev, "%s-%d: could not read device ID\n", __func__, __LINE__);
+
+	for (i = 0; i < niter; i++) {
+		count = ioread32be(health->health_counter);
+		if (count && count != 0xffffffff) {
+			dev_info(&pdev->dev, "Counter value 0x%x after %d iterations\n", count, i);
+			break;
+		}
+		msleep(50);
+	}
+
+	if (i == niter)
+		dev_warn(&pdev->dev, "%s-%d: could not read device ID\n", __func__, __LINE__);
+}
+
+static void mlx5_pci_resume(struct pci_dev *pdev)
+{
+	struct mlx5_core_dev *dev = pci_get_drvdata(pdev);
+	struct mlx5_priv *priv = &dev->priv;
+	int err;
+
+	dev_info(&pdev->dev, "%s was called\n", __func__);
+
+	pci_save_state(pdev);
+	wait_vital(pdev);
+
+	err = mlx5_load_one(dev, priv);
+	if (err)
+		dev_err(&pdev->dev, "%s: mlx5_load_one failed with error code: %d\n"
+			, __func__, err);
+	else
+		dev_info(&pdev->dev, "%s: device recovered\n", __func__);
+}
+
+static const struct pci_error_handlers mlx5_err_handler = {
+	.error_detected = mlx5_pci_err_detected,
+	.slot_reset	= mlx5_pci_slot_reset,
+	.resume		= mlx5_pci_resume
+};
+
 static const struct pci_device_id mlx5_core_pci_table[] = {
 	{ PCI_VDEVICE(MELLANOX, 0x1011) }, /* Connect-IB */
 	{ PCI_VDEVICE(MELLANOX, 0x1012) }, /* Connect-IB VF */
@@ -1188,7 +1380,8 @@ static struct pci_driver mlx5_core_driver = {
 	.name           = DRIVER_NAME,
 	.id_table       = mlx5_core_pci_table,
 	.probe          = init_one,
-	.remove         = remove_one
+	.remove         = remove_one,
+	.err_handler	= &mlx5_err_handler
 };
 
 static int __init init(void)

drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h

@@ -86,6 +86,10 @@ int mlx5_query_hca_caps(struct mlx5_core_dev *dev);
 int mlx5_query_board_id(struct mlx5_core_dev *dev);
 int mlx5_cmd_init_hca(struct mlx5_core_dev *dev);
 int mlx5_cmd_teardown_hca(struct mlx5_core_dev *dev);
+void mlx5_core_event(struct mlx5_core_dev *dev, enum mlx5_dev_event event,
+		     unsigned long param);
+void mlx5_enter_error_state(struct mlx5_core_dev *dev);
+void mlx5_disable_device(struct mlx5_core_dev *dev);
 
 void mlx5e_init(void);
 void mlx5e_cleanup(void);

drivers/net/ethernet/mellanox/mlx5/core/pagealloc.c

@@ -493,15 +493,20 @@ int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev)
 	struct fw_page *fwp;
 	struct rb_node *p;
 	int nclaimed = 0;
-	int err;
+	int err = 0;
 
 	do {
 		p = rb_first(&dev->priv.page_root);
 		if (p) {
 			fwp = rb_entry(p, struct fw_page, rb_node);
-			err = reclaim_pages(dev, fwp->func_id,
-					    optimal_reclaimed_pages(),
-					    &nclaimed);
+			if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
+				free_4k(dev, fwp->addr);
+				nclaimed = 1;
+			} else {
+				err = reclaim_pages(dev, fwp->func_id,
+						    optimal_reclaimed_pages(),
+						    &nclaimed);
+			}
 			if (err) {
 				mlx5_core_warn(dev, "failed reclaiming pages (%d)\n",
 					       err);

include/linux/mlx4/device.h

@@ -833,6 +833,7 @@ struct mlx4_dev {
 	struct mlx4_quotas	quotas;
 	struct radix_tree_root	qp_table_tree;
 	u8			rev_id;
+	u8			port_random_macs;
 	char			board_id[MLX4_BOARD_ID_LEN];
 	int			numa_node;
 	int			oper_log_mgm_entry_size;

include/linux/mlx5/device.h

@@ -439,7 +439,8 @@ struct mlx5_init_seg {
 	__be32			cmdq_addr_h;
 	__be32			cmdq_addr_l_sz;
 	__be32			cmd_dbell;
-	__be32			rsvd1[121];
+	__be32			rsvd1[120];
+	__be32			initializing;
 	struct health_buffer	health;
 	__be32			rsvd2[884];
 	__be32			health_counter;

include/linux/mlx5/driver.h

@@ -393,6 +393,7 @@ struct mlx5_core_health {
 	struct timer_list		timer;
 	u32				prev;
 	int				miss_counter;
+	bool				sick;
 	struct workqueue_struct	       *wq;
 	struct work_struct		work;
 };
@@ -486,8 +487,26 @@ struct mlx5_priv {
 	spinlock_t              ctx_lock;
 };
 
+enum mlx5_device_state {
+	MLX5_DEVICE_STATE_UP,
+	MLX5_DEVICE_STATE_INTERNAL_ERROR,
+};
+
+enum mlx5_interface_state {
+	MLX5_INTERFACE_STATE_DOWN,
+	MLX5_INTERFACE_STATE_UP,
+};
+
+enum mlx5_pci_status {
+	MLX5_PCI_STATUS_DISABLED,
+	MLX5_PCI_STATUS_ENABLED,
+};
+
 struct mlx5_core_dev {
 	struct pci_dev	       *pdev;
+	/* sync pci state */
+	struct mutex		pci_status_mutex;
+	enum mlx5_pci_status	pci_status;
 	u8			rev_id;
 	char			board_id[MLX5_BOARD_ID_LEN];
 	struct mlx5_cmd		cmd;
@@ -496,6 +515,10 @@ struct mlx5_core_dev {
 	u32 hca_caps_max[MLX5_CAP_NUM][MLX5_UN_SZ_DW(hca_cap_union)];
 	phys_addr_t		iseg_base;
 	struct mlx5_init_seg __iomem *iseg;
+	enum mlx5_device_state	state;
+	/* sync interface state */
+	struct mutex		intf_state_mutex;
+	enum mlx5_interface_state interface_state;
 	void			(*event) (struct mlx5_core_dev *dev,
 					  enum mlx5_dev_event event,
 					  unsigned long param);
@@ -803,6 +826,11 @@ void mlx5_core_put_rsc(struct mlx5_core_rsc_common *common);
 int mlx5_query_odp_caps(struct mlx5_core_dev *dev,
 			struct mlx5_odp_caps *odp_caps);
 
+static inline int fw_initializing(struct mlx5_core_dev *dev)
+{
+	return ioread32be(&dev->iseg->initializing) >> 31;
+}
+
 static inline u32 mlx5_mkey_to_idx(u32 mkey)
 {
 	return mkey >> 8;