qed: Add infrastructure for error detection and recovery

This patch adds the detection and handling of a parity error ("process kill
event"), including the update of the protocol drivers, and the prevention
of any HW access that will lead to device access towards the host while
recovery is in progress.
It also provides the means for the protocol drivers to trigger a recovery
process on their decision.

Signed-off-by: Tomer Tayar <tomer.tayar@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: Michal Kalderon <michal.kalderon@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

11 files changed, 251 insertions, 16 deletions

diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index b352e313e1f6..3b0955d34716 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -804,6 +804,9 @@ struct qed_dev {
 
 	u32				mcp_nvm_resp;
 
+	/* Recovery */
+	bool recov_in_prog;
+
 	/* Linux specific here */
 	struct  qede_dev		*edev;
 	struct  pci_dev			*pdev;
@@ -943,6 +946,7 @@ void qed_link_update(struct qed_hwfn *hwfn, struct qed_ptt *ptt);
 u32 qed_unzip_data(struct qed_hwfn *p_hwfn,
 		   u32 input_len, u8 *input_buf,
 		   u32 max_size, u8 *unzip_buf);
+void qed_schedule_recovery_handler(struct qed_hwfn *p_hwfn);
 void qed_get_protocol_stats(struct qed_dev *cdev,
 			    enum qed_mcp_protocol_type type,
 			    union qed_mcp_protocol_stats *stats);
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index fa5f07e65672..b17003d9066c 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -2140,6 +2140,11 @@ int qed_hw_init(struct qed_dev *cdev, struct qed_hw_init_params *p_params)
 			   "Load request was sent. Load code: 0x%x\n",
 			   load_code);
 
+		/* Only relevant for recovery:
+		 * Clear the indication after LOAD_REQ is responded by the MFW.
+		 */
+		cdev->recov_in_prog = false;
+
 		qed_mcp_set_capabilities(p_hwfn, p_hwfn->p_main_ptt);
 
 		qed_reset_mb_shadow(p_hwfn, p_hwfn->p_main_ptt);
@@ -2291,6 +2296,9 @@ static void qed_hw_timers_stop(struct qed_dev *cdev,
 	qed_wr(p_hwfn, p_ptt, TM_REG_PF_ENABLE_CONN, 0x0);
 	qed_wr(p_hwfn, p_ptt, TM_REG_PF_ENABLE_TASK, 0x0);
 
+	if (cdev->recov_in_prog)
+		return;
+
 	for (i = 0; i < QED_HW_STOP_RETRY_LIMIT; i++) {
 		if ((!qed_rd(p_hwfn, p_ptt,
 			     TM_REG_PF_SCAN_ACTIVE_CONN)) &&
@@ -2353,12 +2361,14 @@ int qed_hw_stop(struct qed_dev *cdev)
 		p_hwfn->hw_init_done = false;
 
 		/* Send unload command to MCP */
-		rc = qed_mcp_unload_req(p_hwfn, p_ptt);
-		if (rc) {
-			DP_NOTICE(p_hwfn,
-				  "Failed sending a UNLOAD_REQ command. rc = %d.\n",
-				  rc);
-			rc2 = -EINVAL;
+		if (!cdev->recov_in_prog) {
+			rc = qed_mcp_unload_req(p_hwfn, p_ptt);
+			if (rc) {
+				DP_NOTICE(p_hwfn,
+					  "Failed sending a UNLOAD_REQ command. rc = %d.\n",
+					  rc);
+				rc2 = -EINVAL;
+			}
 		}
 
 		qed_slowpath_irq_sync(p_hwfn);
@@ -2400,16 +2410,18 @@ int qed_hw_stop(struct qed_dev *cdev)
 		qed_wr(p_hwfn, p_ptt, DORQ_REG_PF_DB_ENABLE, 0);
 		qed_wr(p_hwfn, p_ptt, QM_REG_PF_EN, 0);
 
-		qed_mcp_unload_done(p_hwfn, p_ptt);
-		if (rc) {
-			DP_NOTICE(p_hwfn,
-				  "Failed sending a UNLOAD_DONE command. rc = %d.\n",
-				  rc);
-			rc2 = -EINVAL;
+		if (!cdev->recov_in_prog) {
+			rc = qed_mcp_unload_done(p_hwfn, p_ptt);
+			if (rc) {
+				DP_NOTICE(p_hwfn,
+					  "Failed sending a UNLOAD_DONE command. rc = %d.\n",
+					  rc);
+				rc2 = -EINVAL;
+			}
 		}
 	}
 
-	if (IS_PF(cdev)) {
+	if (IS_PF(cdev) && !cdev->recov_in_prog) {
 		p_hwfn = QED_LEADING_HWFN(cdev);
 		p_ptt = QED_LEADING_HWFN(cdev)->p_main_ptt;
 
@@ -3459,6 +3471,7 @@ static int qed_hw_prepare_single(struct qed_hwfn *p_hwfn,
 				 void __iomem *p_doorbells,
 				 enum qed_pci_personality personality)
 {
+	struct qed_dev *cdev = p_hwfn->cdev;
 	int rc = 0;
 
 	/* Split PCI bars evenly between hwfns */
@@ -3511,7 +3524,7 @@ static int qed_hw_prepare_single(struct qed_hwfn *p_hwfn,
 	/* Sending a mailbox to the MFW should be done after qed_get_hw_info()
 	 * is called as it sets the ports number in an engine.
 	 */
-	if (IS_LEAD_HWFN(p_hwfn)) {
+	if (IS_LEAD_HWFN(p_hwfn) && !cdev->recov_in_prog) {
 		rc = qed_mcp_initiate_pf_flr(p_hwfn, p_hwfn->p_main_ptt);
 		if (rc)
 			DP_NOTICE(p_hwfn, "Failed to initiate PF FLR\n");
diff --git a/drivers/net/ethernet/qlogic/qed/qed_hsi.h b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
index b13cfb449d8f..417121e74ee9 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_hsi.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
@@ -12827,7 +12827,7 @@ enum MFW_DRV_MSG_TYPE {
 	MFW_DRV_MSG_LLDP_DATA_UPDATED,
 	MFW_DRV_MSG_DCBX_REMOTE_MIB_UPDATED,
 	MFW_DRV_MSG_DCBX_OPERATIONAL_MIB_UPDATED,
-	MFW_DRV_MSG_RESERVED4,
+	MFW_DRV_MSG_ERROR_RECOVERY,
 	MFW_DRV_MSG_BW_UPDATE,
 	MFW_DRV_MSG_S_TAG_UPDATE,
 	MFW_DRV_MSG_GET_LAN_STATS,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_hw.c b/drivers/net/ethernet/qlogic/qed/qed_hw.c
index 70504dcf4087..72ec1c6bdf70 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_hw.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_hw.c
@@ -703,6 +703,17 @@ static int qed_dmae_execute_command(struct qed_hwfn *p_hwfn,
 	int qed_status = 0;
 	u32 offset = 0;
 
+	if (p_hwfn->cdev->recov_in_prog) {
+		DP_VERBOSE(p_hwfn,
+			   NETIF_MSG_HW,
+			   "Recovery is in progress. Avoid DMAE transaction [{src: addr 0x%llx, type %d}, {dst: addr 0x%llx, type %d}, size %d].\n",
+			   src_addr, src_type, dst_addr, dst_type,
+			   size_in_dwords);
+
+		/* Let the flow complete w/o any error handling */
+		return 0;
+	}
+
 	qed_dmae_opcode(p_hwfn,
 			(src_type == QED_DMAE_ADDRESS_GRC),
 			(dst_type == QED_DMAE_ADDRESS_GRC),
diff --git a/drivers/net/ethernet/qlogic/qed/qed_main.c b/drivers/net/ethernet/qlogic/qed/qed_main.c
index 6adf5bda9811..b47352643fb5 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_main.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_main.c
@@ -359,6 +359,8 @@ static struct qed_dev *qed_probe(struct pci_dev *pdev,
 
 	qed_init_dp(cdev, params->dp_module, params->dp_level);
 
+	cdev->recov_in_prog = params->recov_in_prog;
+
 	rc = qed_init_pci(cdev, pdev);
 	if (rc) {
 		DP_ERR(cdev, "init pci failed\n");
@@ -2203,6 +2205,15 @@ static int qed_nvm_get_image(struct qed_dev *cdev, enum qed_nvm_images type,
 	return qed_mcp_get_nvm_image(hwfn, type, buf, len);
 }
 
+void qed_schedule_recovery_handler(struct qed_hwfn *p_hwfn)
+{
+	struct qed_common_cb_ops *ops = p_hwfn->cdev->protocol_ops.common;
+	void *cookie = p_hwfn->cdev->ops_cookie;
+
+	if (ops && ops->schedule_recovery_handler)
+		ops->schedule_recovery_handler(cookie);
+}
+
 static int qed_set_coalesce(struct qed_dev *cdev, u16 rx_coal, u16 tx_coal,
 			    void *handle)
 {
@@ -2226,6 +2237,23 @@ static int qed_set_led(struct qed_dev *cdev, enum qed_led_mode mode)
 	return status;
 }
 
+static int qed_recovery_process(struct qed_dev *cdev)
+{
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
+	struct qed_ptt *p_ptt;
+	int rc = 0;
+
+	p_ptt = qed_ptt_acquire(p_hwfn);
+	if (!p_ptt)
+		return -EAGAIN;
+
+	rc = qed_start_recovery_process(p_hwfn, p_ptt);
+
+	qed_ptt_release(p_hwfn, p_ptt);
+
+	return rc;
+}
+
 static int qed_update_wol(struct qed_dev *cdev, bool enabled)
 {
 	struct qed_hwfn *hwfn = QED_LEADING_HWFN(cdev);
@@ -2380,6 +2408,8 @@ const struct qed_common_ops qed_common_ops_pass = {
 	.nvm_get_image = &qed_nvm_get_image,
 	.set_coalesce = &qed_set_coalesce,
 	.set_led = &qed_set_led,
+	.recovery_process = &qed_recovery_process,
+	.recovery_prolog = &qed_recovery_prolog,
 	.update_drv_state = &qed_update_drv_state,
 	.update_mac = &qed_update_mac,
 	.update_mtu = &qed_update_mtu,
diff --git a/drivers/net/ethernet/qlogic/qed/qed_mcp.c b/drivers/net/ethernet/qlogic/qed/qed_mcp.c
index 1024484d7dd8..bb8541847aa5 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_mcp.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_mcp.c
@@ -1549,6 +1549,60 @@ int qed_mcp_set_link(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, bool b_up)
 	return 0;
 }
 
+u32 qed_get_process_kill_counter(struct qed_hwfn *p_hwfn,
+				 struct qed_ptt *p_ptt)
+{
+	u32 path_offsize_addr, path_offsize, path_addr, proc_kill_cnt;
+
+	if (IS_VF(p_hwfn->cdev))
+		return -EINVAL;
+
+	path_offsize_addr = SECTION_OFFSIZE_ADDR(p_hwfn->mcp_info->public_base,
+						 PUBLIC_PATH);
+	path_offsize = qed_rd(p_hwfn, p_ptt, path_offsize_addr);
+	path_addr = SECTION_ADDR(path_offsize, QED_PATH_ID(p_hwfn));
+
+	proc_kill_cnt = qed_rd(p_hwfn, p_ptt,
+			       path_addr +
+			       offsetof(struct public_path, process_kill)) &
+			PROCESS_KILL_COUNTER_MASK;
+
+	return proc_kill_cnt;
+}
+
+static void qed_mcp_handle_process_kill(struct qed_hwfn *p_hwfn,
+					struct qed_ptt *p_ptt)
+{
+	struct qed_dev *cdev = p_hwfn->cdev;
+	u32 proc_kill_cnt;
+
+	/* Prevent possible attentions/interrupts during the recovery handling
+	 * and till its load phase, during which they will be re-enabled.
+	 */
+	qed_int_igu_disable_int(p_hwfn, p_ptt);
+
+	DP_NOTICE(p_hwfn, "Received a process kill indication\n");
+
+	/* The following operations should be done once, and thus in CMT mode
+	 * are carried out by only the first HW function.
+	 */
+	if (p_hwfn != QED_LEADING_HWFN(cdev))
+		return;
+
+	if (cdev->recov_in_prog) {
+		DP_NOTICE(p_hwfn,
+			  "Ignoring the indication since a recovery process is already in progress\n");
+		return;
+	}
+
+	cdev->recov_in_prog = true;
+
+	proc_kill_cnt = qed_get_process_kill_counter(p_hwfn, p_ptt);
+	DP_NOTICE(p_hwfn, "Process kill counter: %d\n", proc_kill_cnt);
+
+	qed_schedule_recovery_handler(p_hwfn);
+}
+
 static void qed_mcp_send_protocol_stats(struct qed_hwfn *p_hwfn,
 					struct qed_ptt *p_ptt,
 					enum MFW_DRV_MSG_TYPE type)
@@ -1779,6 +1833,9 @@ int qed_mcp_handle_events(struct qed_hwfn *p_hwfn,
 		case MFW_DRV_MSG_TRANSCEIVER_STATE_CHANGE:
 			qed_mcp_handle_transceiver_change(p_hwfn, p_ptt);
 			break;
+		case MFW_DRV_MSG_ERROR_RECOVERY:
+			qed_mcp_handle_process_kill(p_hwfn, p_ptt);
+			break;
 		case MFW_DRV_MSG_GET_LAN_STATS:
 		case MFW_DRV_MSG_GET_FCOE_STATS:
 		case MFW_DRV_MSG_GET_ISCSI_STATS:
@@ -2324,6 +2381,43 @@ int qed_mcp_get_flash_size(struct qed_hwfn *p_hwfn,
 	return 0;
 }
 
+int qed_start_recovery_process(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	struct qed_dev *cdev = p_hwfn->cdev;
+
+	if (cdev->recov_in_prog) {
+		DP_NOTICE(p_hwfn,
+			  "Avoid triggering a recovery since such a process is already in progress\n");
+		return -EAGAIN;
+	}
+
+	DP_NOTICE(p_hwfn, "Triggering a recovery process\n");
+	qed_wr(p_hwfn, p_ptt, MISC_REG_AEU_GENERAL_ATTN_35, 0x1);
+
+	return 0;
+}
+
+#define QED_RECOVERY_PROLOG_SLEEP_MS    100
+
+int qed_recovery_prolog(struct qed_dev *cdev)
+{
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
+	struct qed_ptt *p_ptt = p_hwfn->p_main_ptt;
+	int rc;
+
+	/* Allow ongoing PCIe transactions to complete */
+	msleep(QED_RECOVERY_PROLOG_SLEEP_MS);
+
+	/* Clear the PF's internal FID_enable in the PXP */
+	rc = qed_pglueb_set_pfid_enable(p_hwfn, p_ptt, false);
+	if (rc)
+		DP_NOTICE(p_hwfn,
+			  "qed_pglueb_set_pfid_enable() failed. rc = %d.\n",
+			  rc);
+
+	return rc;
+}
+
 static int
 qed_mcp_config_vf_msix_bb(struct qed_hwfn *p_hwfn,
 			  struct qed_ptt *p_ptt, u8 vf_id, u8 num)
diff --git a/drivers/net/ethernet/qlogic/qed/qed_mcp.h b/drivers/net/ethernet/qlogic/qed/qed_mcp.h
index 387c5e649136..6e1d72a669ae 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_mcp.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_mcp.h
@@ -441,6 +441,38 @@ qed_mcp_send_drv_version(struct qed_hwfn *p_hwfn,
 			 struct qed_mcp_drv_version *p_ver);
 
 /**
+ * @brief Read the MFW process kill counter
+ *
+ * @param p_hwfn
+ * @param p_ptt
+ *
+ * @return u32
+ */
+u32 qed_get_process_kill_counter(struct qed_hwfn *p_hwfn,
+				 struct qed_ptt *p_ptt);
+
+/**
+ * @brief Trigger a recovery process
+ *
+ *  @param p_hwfn
+ *  @param p_ptt
+ *
+ * @return int
+ */
+int qed_start_recovery_process(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
+
+/**
+ * @brief A recovery handler must call this function as its first step.
+ *        It is assumed that the handler is not run from an interrupt context.
+ *
+ *  @param cdev
+ *  @param p_ptt
+ *
+ * @return int
+ */
+int qed_recovery_prolog(struct qed_dev *cdev);
+
+/**
  * @brief Notify MFW about the change in base device properties
  *
  *  @param p_hwfn
diff --git a/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h b/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h
index 8939ed6e08b7..5ce825ca5f24 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h
@@ -518,6 +518,8 @@
 	0x180824UL
 #define  MISC_REG_AEU_GENERAL_ATTN_0 \
 	0x008400UL
+#define MISC_REG_AEU_GENERAL_ATTN_35 \
+	0x00848cUL
 #define  CAU_REG_SB_ADDR_MEMORY \
 	0x1c8000UL
 #define  CAU_REG_SB_VAR_MEMORY \
diff --git a/drivers/net/ethernet/qlogic/qed/qed_spq.c b/drivers/net/ethernet/qlogic/qed/qed_spq.c
index eb88bbc6b193..3e0f7c46bb1b 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_spq.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_spq.c
@@ -790,6 +790,17 @@ static int qed_spq_pend_post(struct qed_hwfn *p_hwfn)
 				 SPQ_HIGH_PRI_RESERVE_DEFAULT);
 }
 
+static void qed_spq_recov_set_ret_code(struct qed_spq_entry *p_ent,
+				       u8 *fw_return_code)
+{
+	if (!fw_return_code)
+		return;
+
+	if (p_ent->elem.hdr.protocol_id == PROTOCOLID_ROCE ||
+	    p_ent->elem.hdr.protocol_id == PROTOCOLID_IWARP)
+		*fw_return_code = RDMA_RETURN_OK;
+}
+
 /* Avoid overriding of SPQ entries when getting out-of-order completions, by
  * marking the completions in a bitmap and increasing the chain consumer only
  * for the first successive completed entries.
@@ -825,6 +836,17 @@ int qed_spq_post(struct qed_hwfn *p_hwfn,
 		return -EINVAL;
 	}
 
+	if (p_hwfn->cdev->recov_in_prog) {
+		DP_VERBOSE(p_hwfn,
+			   QED_MSG_SPQ,
+			   "Recovery is in progress. Skip spq post [cmd %02x protocol %02x]\n",
+			   p_ent->elem.hdr.cmd_id, p_ent->elem.hdr.protocol_id);
+
+		/* Let the flow complete w/o any error handling */
+		qed_spq_recov_set_ret_code(p_ent, fw_return_code);
+		return 0;
+	}
+
 	/* Complete the entry */
 	rc = qed_spq_fill_entry(p_hwfn, p_ent);
 
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sriov.c b/drivers/net/ethernet/qlogic/qed/qed_sriov.c
index ca6290fa0f30..71e28be58102 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sriov.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_sriov.c
@@ -4447,6 +4447,13 @@ int qed_sriov_disable(struct qed_dev *cdev, bool pci_enabled)
 	if (cdev->p_iov_info && cdev->p_iov_info->num_vfs && pci_enabled)
 		pci_disable_sriov(cdev->pdev);
 
+	if (cdev->recov_in_prog) {
+		DP_VERBOSE(cdev,
+			   QED_MSG_IOV,
+			   "Skip SRIOV disable operations in the device since a recovery is in progress\n");
+		goto out;
+	}
+
 	for_each_hwfn(cdev, i) {
 		struct qed_hwfn *hwfn = &cdev->hwfns[i];
 		struct qed_ptt *ptt = qed_ptt_acquire(hwfn);
@@ -4486,7 +4493,7 @@ int qed_sriov_disable(struct qed_dev *cdev, bool pci_enabled)
 
 		qed_ptt_release(hwfn, ptt);
 	}
-
+out:
 	qed_iov_set_vfs_to_disable(cdev, false);
 
 	return 0;
diff --git a/include/linux/qed/qed_if.h b/include/linux/qed/qed_if.h
index 5f818fda96bd..35170f74ed80 100644
--- a/include/linux/qed/qed_if.h
+++ b/include/linux/qed/qed_if.h
@@ -763,6 +763,7 @@ struct qed_probe_params {
 	u32 dp_module;
 	u8 dp_level;
 	bool is_vf;
+	bool recov_in_prog;
 };
 
 #define QED_DRV_VER_STR_SIZE 12
@@ -809,6 +810,7 @@ struct qed_common_cb_ops {
 	void (*arfs_filter_op)(void *dev, void *fltr, u8 fw_rc);
 	void	(*link_update)(void			*dev,
 			       struct qed_link_output	*link);
+	void (*schedule_recovery_handler)(void *dev);
 	void	(*dcbx_aen)(void *dev, struct qed_dcbx_get *get, u32 mib_type);
 	void (*get_generic_tlv_data)(void *dev, struct qed_generic_tlvs *data);
 	void (*get_protocol_tlv_data)(void *dev, void *data);
@@ -1057,6 +1059,24 @@ struct qed_common_ops {
 			       void __iomem *db_addr, void *db_data);
 
 /**
+ * @brief recovery_process - Trigger a recovery process
+ *
+ * @param cdev
+ *
+ * @return 0 on success, error otherwise.
+ */
+	int (*recovery_process)(struct qed_dev *cdev);
+
+/**
+ * @brief recovery_prolog - Execute the prolog operations of a recovery process
+ *
+ * @param cdev
+ *
+ * @return 0 on success, error otherwise.
+ */
+	int (*recovery_prolog)(struct qed_dev *cdev);
+
+/**
  * @brief update_drv_state - API to inform the change in the driver state.
  *
  * @param cdev