[SCSI] qla2xxx: Extend base EEH support in qla2xxx.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

1 files changed, 75 insertions, 8 deletions

diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index 2f873d237325..1ab358210c6a 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -475,11 +475,11 @@ qla2xxx_queuecommand(struct scsi_cmnd *cmd, void (*done)(struct scsi_cmnd *))
 	srb_t *sp;
 	int rval;
 
-	if (unlikely(pci_channel_offline(ha->pdev))) {
-		if (ha->pdev->error_state == pci_channel_io_frozen)
-			cmd->result = DID_REQUEUE << 16;
-		else
+	if (ha->flags.eeh_busy) {
+		if (ha->flags.pci_channel_io_perm_failure)
 			cmd->result = DID_NO_CONNECT << 16;
+		else
+			cmd->result = DID_REQUEUE << 16;
 		goto qc24_fail_command;
 	}
 
@@ -552,8 +552,15 @@ qla2x00_eh_wait_on_command(struct scsi_cmnd *cmd)
 #define ABORT_POLLING_PERIOD	1000
 #define ABORT_WAIT_ITER		((10 * 1000) / (ABORT_POLLING_PERIOD))
 	unsigned long wait_iter = ABORT_WAIT_ITER;
+	scsi_qla_host_t *vha = shost_priv(cmd->device->host);
+	struct qla_hw_data *ha = vha->hw;
 	int ret = QLA_SUCCESS;
 
+	if (unlikely(pci_channel_offline(ha->pdev)) || ha->flags.eeh_busy) {
+		DEBUG17(qla_printk(KERN_WARNING, ha, "return:eh_wait\n"));
+		return ret;
+	}
+
 	while (CMD_SP(cmd) && wait_iter--) {
 		msleep(ABORT_POLLING_PERIOD);
 	}
@@ -2174,6 +2181,24 @@ qla2x00_free_device(scsi_qla_host_t *vha)
 {
 	struct qla_hw_data *ha = vha->hw;
 
+	qla2x00_abort_all_cmds(vha, DID_NO_CONNECT << 16);
+
+	/* Disable timer */
+	if (vha->timer_active)
+		qla2x00_stop_timer(vha);
+
+	/* Kill the kernel thread for this host */
+	if (ha->dpc_thread) {
+		struct task_struct *t = ha->dpc_thread;
+
+		/*
+		 * qla2xxx_wake_dpc checks for ->dpc_thread
+		 * so we need to zero it out.
+		 */
+		ha->dpc_thread = NULL;
+		kthread_stop(t);
+	}
+
 	qla25xx_delete_queues(vha);
 
 	if (ha->flags.fce_enabled)
@@ -2185,6 +2210,8 @@ qla2x00_free_device(scsi_qla_host_t *vha)
 	/* Stop currently executing firmware. */
 	qla2x00_try_to_stop_firmware(vha);
 
+	vha->flags.online = 0;
+
 	/* turn-off interrupts on the card */
 	if (ha->interrupts_on)
 		ha->isp_ops->disable_intrs(ha);
@@ -2859,6 +2886,13 @@ qla2x00_do_dpc(void *data)
 		if (!base_vha->flags.init_done)
 			continue;
 
+		if (ha->flags.eeh_busy) {
+			DEBUG17(qla_printk(KERN_WARNING, ha,
+			    "qla2x00_do_dpc: dpc_flags: %lx\n",
+			    base_vha->dpc_flags));
+			continue;
+		}
+
 		DEBUG3(printk("scsi(%ld): DPC handler\n", base_vha->host_no));
 
 		ha->dpc_active = 1;
@@ -3049,8 +3083,13 @@ qla2x00_timer(scsi_qla_host_t *vha)
 	int		index;
 	srb_t		*sp;
 	int		t;
+	uint16_t        w;
 	struct qla_hw_data *ha = vha->hw;
 	struct req_que *req;
+
+	/* Hardware read to raise pending EEH errors during mailbox waits. */
+	if (!pci_channel_offline(ha->pdev))
+		pci_read_config_word(ha->pdev, PCI_VENDOR_ID, &w);
 	/*
 	 * Ports - Port down timer.
 	 *
@@ -3252,16 +3291,23 @@ qla2x00_release_firmware(void)
 static pci_ers_result_t
 qla2xxx_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state)
 {
-	scsi_qla_host_t *base_vha = pci_get_drvdata(pdev);
+	scsi_qla_host_t *vha = pci_get_drvdata(pdev);
+	struct qla_hw_data *ha = vha->hw;
+
+	DEBUG2(qla_printk(KERN_WARNING, ha, "error_detected:state %x\n",
+	    state));
 
 	switch (state) {
 	case pci_channel_io_normal:
+		ha->flags.eeh_busy = 0;
 		return PCI_ERS_RESULT_CAN_RECOVER;
 	case pci_channel_io_frozen:
+		ha->flags.eeh_busy = 1;
 		pci_disable_device(pdev);
 		return PCI_ERS_RESULT_NEED_RESET;
 	case pci_channel_io_perm_failure:
-		qla2x00_abort_all_cmds(base_vha, DID_NO_CONNECT << 16);
+		ha->flags.pci_channel_io_perm_failure = 1;
+		qla2x00_abort_all_cmds(vha, DID_NO_CONNECT << 16);
 		return PCI_ERS_RESULT_DISCONNECT;
 	}
 	return PCI_ERS_RESULT_NEED_RESET;
@@ -3312,6 +3358,8 @@ qla2xxx_pci_slot_reset(struct pci_dev *pdev)
 	struct qla_hw_data *ha = base_vha->hw;
 	int rc;
 
+	DEBUG17(qla_printk(KERN_WARNING, ha, "slot_reset\n"));
+
 	if (ha->mem_only)
 		rc = pci_enable_device_mem(pdev);
 	else
@@ -3320,19 +3368,33 @@ qla2xxx_pci_slot_reset(struct pci_dev *pdev)
 	if (rc) {
 		qla_printk(KERN_WARNING, ha,
 		    "Can't re-enable PCI device after reset.\n");
-
 		return ret;
 	}
-	pci_set_master(pdev);
 
 	if (ha->isp_ops->pci_config(base_vha))
 		return ret;
 
+#ifdef QL_DEBUG_LEVEL_17
+	{
+		uint8_t b;
+		uint32_t i;
+
+		printk("slot_reset_1: ");
+		for (i = 0; i < 256; i++) {
+			pci_read_config_byte(ha->pdev, i, &b);
+			printk("%s%02x", (i%16) ? " " : "\n", b);
+		}
+		printk("\n");
+	}
+#endif
 	set_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags);
 	if (qla2x00_abort_isp(base_vha) == QLA_SUCCESS)
 		ret =  PCI_ERS_RESULT_RECOVERED;
 	clear_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags);
 
+	DEBUG17(qla_printk(KERN_WARNING, ha,
+	    "slot_reset-return:ret=%x\n", ret));
+
 	return ret;
 }
 
@@ -3343,12 +3405,17 @@ qla2xxx_pci_resume(struct pci_dev *pdev)
 	struct qla_hw_data *ha = base_vha->hw;
 	int ret;
 
+	DEBUG17(qla_printk(KERN_WARNING, ha, "pci_resume\n"));
+
 	ret = qla2x00_wait_for_hba_online(base_vha);
 	if (ret != QLA_SUCCESS) {
 		qla_printk(KERN_ERR, ha,
 		    "the device failed to resume I/O "
 		    "from slot/link_reset");
 	}
+
+	ha->flags.eeh_busy = 0;
+
 	pci_cleanup_aer_uncorrect_error_status(pdev);
 }