scsi: qla2xxx: Flush mailbox commands on chip reset

Flush pending mailbox commands on chip reset.  Wake up command that's
waiting for an interrupt and wait for mailbox counters to go to zero.

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

4 files changed, 74 insertions, 5 deletions

diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index ee4d1f4fdf95..40bcf938cf4f 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -3598,6 +3598,7 @@ struct qla_hw_data {
                 uint32_t        detected_lr_sfp:1;
                 uint32_t        using_lr_setting:1;
                 uint32_t        rida_fmt2:1;
+                uint32_t        purge_mbox:1;
         } flags;
 
         uint16_t max_exchg;
@@ -3843,6 +3844,9 @@ struct qla_hw_data {
         int             port_down_retry_count;
         uint8_t         mbx_count;
         uint8_t         aen_mbx_count;
+        atomic_t        num_pend_mbx_stage1;
+        atomic_t        num_pend_mbx_stage2;
+        atomic_t        num_pend_mbx_stage3;
 
         uint32_t        login_retry_count;
         /* SNS command interfaces. */
@@ -4156,6 +4160,7 @@ struct qla_hw_data {
         struct work_struct board_disable;
 
         struct mr_data_fx00 mr;
+        uint32_t chip_reset;
 
         struct qlt_hw_data tgt;
         int     allow_cna_fw_dump;
diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c
index bb565f556c2f..a10a8bb895e9 100644
--- a/drivers/scsi/qla2xxx/qla_init.c
+++ b/drivers/scsi/qla2xxx/qla_init.c
@@ -6283,6 +6283,7 @@ qla2x00_abort_isp_cleanup(scsi_qla_host_t *vha)
         ql_log(ql_log_info, vha, 0x00af,
             "Performing ISP error recovery - ha=%p.\n", ha);
 
+        ha->flags.purge_mbox = 1;
         /* For ISP82XX, reset_chip is just disabling interrupts.
          * Driver waits for the completion of the commands.
          * the interrupts need to be enabled.
@@ -6297,13 +6298,31 @@ qla2x00_abort_isp_cleanup(scsi_qla_host_t *vha)
         ha->current_topology = 0;
         ha->flags.fw_started = 0;
         ha->flags.fw_init_done = 0;
-        ha->base_qpair->chip_reset++;
+        ha->chip_reset++;
+        ha->base_qpair->chip_reset = ha->chip_reset;
         for (i = 0; i < ha->max_qpairs; i++) {
                 if (ha->queue_pair_map[i])
                         ha->queue_pair_map[i]->chip_reset =
                                 ha->base_qpair->chip_reset;
         }
 
+        /* purge MBox commands */
+        if (atomic_read(&ha->num_pend_mbx_stage3)) {
+                clear_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags);
+                complete(&ha->mbx_intr_comp);
+        }
+
+        i = 0;
+        while (atomic_read(&ha->num_pend_mbx_stage3) ||
+            atomic_read(&ha->num_pend_mbx_stage2) ||
+            atomic_read(&ha->num_pend_mbx_stage1)) {
+                msleep(20);
+                i++;
+                if (i > 50)
+                        break;
+        }
+        ha->flags.purge_mbox = 0;
+
         atomic_set(&vha->loop_down_timer, LOOP_DOWN_TIME);
         if (atomic_read(&vha->loop_state) != LOOP_DOWN) {
                 atomic_set(&vha->loop_state, LOOP_DOWN);
diff --git a/drivers/scsi/qla2xxx/qla_mbx.c b/drivers/scsi/qla2xxx/qla_mbx.c
index 17537f0b3b54..10847cdca093 100644
--- a/drivers/scsi/qla2xxx/qla_mbx.c
+++ b/drivers/scsi/qla2xxx/qla_mbx.c
@@ -110,6 +110,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
         unsigned long   wait_time;
         struct qla_hw_data *ha = vha->hw;
         scsi_qla_host_t *base_vha = pci_get_drvdata(ha->pdev);
+        u32 chip_reset;
 
 
         ql_dbg(ql_dbg_mbx, vha, 0x1000, "Entered %s.\n", __func__);
@@ -140,7 +141,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
 
         rval = QLA_SUCCESS;
         abort_active = test_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags);
-
+        chip_reset = ha->chip_reset;
 
         if (ha->flags.pci_channel_io_perm_failure) {
                 ql_log(ql_log_warn, vha, 0x1003,
@@ -167,6 +168,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                 return QLA_FUNCTION_TIMEOUT;
         }
 
+        atomic_inc(&ha->num_pend_mbx_stage1);
         /*
          * Wait for active mailbox commands to finish by waiting at most tov
          * seconds. This is to serialize actual issuing of mailbox cmds during
@@ -177,8 +179,14 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                 ql_log(ql_log_warn, vha, 0xd035,
                     "Cmd access timeout, cmd=0x%x, Exiting.\n",
                     mcp->mb[0]);
+                atomic_dec(&ha->num_pend_mbx_stage1);
                 return QLA_FUNCTION_TIMEOUT;
         }
+        atomic_dec(&ha->num_pend_mbx_stage1);
+        if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) {
+                rval = QLA_ABORTED;
+                goto premature_exit;
+        }
 
         ha->flags.mbox_busy = 1;
         /* Save mailbox command for debug */
@@ -189,6 +197,13 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
 
         spin_lock_irqsave(&ha->hardware_lock, flags);
 
+        if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) {
+                rval = QLA_ABORTED;
+                ha->flags.mbox_busy = 0;
+                spin_unlock_irqrestore(&ha->hardware_lock, flags);
+                goto premature_exit;
+        }
+
         /* Load mailbox registers. */
         if (IS_P3P_TYPE(ha))
                 optr = (uint16_t __iomem *)&reg->isp82.mailbox_in[0];
@@ -231,7 +246,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
             "jiffies=%lx.\n", jiffies);
 
         /* Wait for mbx cmd completion until timeout */
-
+        atomic_inc(&ha->num_pend_mbx_stage2);
         if ((!abort_active && io_lock_on) || IS_NOPOLLING_TYPE(ha)) {
                 set_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags);
 
@@ -241,6 +256,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                                 spin_unlock_irqrestore(&ha->hardware_lock,
                                         flags);
                                 ha->flags.mbox_busy = 0;
+                                atomic_dec(&ha->num_pend_mbx_stage2);
                                 ql_dbg(ql_dbg_mbx, vha, 0x1010,
                                     "Pending mailbox timeout, exiting.\n");
                                 rval = QLA_FUNCTION_TIMEOUT;
@@ -254,6 +270,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                 spin_unlock_irqrestore(&ha->hardware_lock, flags);
 
                 wait_time = jiffies;
+                atomic_inc(&ha->num_pend_mbx_stage3);
                 if (!wait_for_completion_timeout(&ha->mbx_intr_comp,
                     mcp->tov * HZ)) {
                         ql_dbg(ql_dbg_mbx, vha, 0x117a,
@@ -261,7 +278,17 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                         spin_lock_irqsave(&ha->hardware_lock, flags);
                         clear_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags);
                         spin_unlock_irqrestore(&ha->hardware_lock, flags);
+
+                } else if (ha->flags.purge_mbox ||
+                    chip_reset != ha->chip_reset) {
+                        ha->flags.mbox_busy = 0;
+                        atomic_dec(&ha->num_pend_mbx_stage2);
+                        atomic_dec(&ha->num_pend_mbx_stage3);
+                        rval = QLA_ABORTED;
+                        goto premature_exit;
                 }
+                atomic_dec(&ha->num_pend_mbx_stage3);
+
                 if (time_after(jiffies, wait_time + 5 * HZ))
                         ql_log(ql_log_warn, vha, 0x1015, "cmd=0x%x, waited %d msecs\n",
                             command, jiffies_to_msecs(jiffies - wait_time));
@@ -275,6 +302,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                                 spin_unlock_irqrestore(&ha->hardware_lock,
                                         flags);
                                 ha->flags.mbox_busy = 0;
+                                atomic_dec(&ha->num_pend_mbx_stage2);
                                 ql_dbg(ql_dbg_mbx, vha, 0x1012,
                                     "Pending mailbox timeout, exiting.\n");
                                 rval = QLA_FUNCTION_TIMEOUT;
@@ -289,6 +317,14 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
 
                 wait_time = jiffies + mcp->tov * HZ; /* wait at most tov secs */
                 while (!ha->flags.mbox_int) {
+                        if (ha->flags.purge_mbox ||
+                            chip_reset != ha->chip_reset) {
+                                ha->flags.mbox_busy = 0;
+                                atomic_dec(&ha->num_pend_mbx_stage2);
+                                rval = QLA_ABORTED;
+                                goto premature_exit;
+                        }
+
                         if (time_after(jiffies, wait_time))
                                 break;
 
@@ -312,6 +348,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                     "Waited %d sec.\n",
                     (uint)((jiffies - (wait_time - (mcp->tov * HZ)))/HZ));
         }
+        atomic_dec(&ha->num_pend_mbx_stage2);
 
         /* Check whether we timed out */
         if (ha->flags.mbox_int) {
@@ -390,7 +427,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp)
                 /* Capture FW dump only, if PCI device active */
                 if (!pci_channel_offline(vha->hw->pdev)) {
                         pci_read_config_word(ha->pdev, PCI_VENDOR_ID, &w);
-                        if (w == 0xffff || ictrl == 0xffffffff) {
+                        if (w == 0xffff || ictrl == 0xffffffff ||
+                            (chip_reset != ha->chip_reset)) {
                                 /* This is special case if there is unload
                                  * of driver happening and if PCI device go
                                  * into bad state due to PCI error condition
@@ -497,7 +535,11 @@ premature_exit:
         complete(&ha->mbx_cmd_comp);
 
 mbx_done:
-        if (rval) {
+        if (rval == QLA_ABORTED) {
+                ql_log(ql_log_info, vha, 0xd035,
+                    "Chip Reset in progress. Purging Mbox cmd=0x%x.\n",
+                    mcp->mb[0]);
+        } else if (rval) {
                 if (ql2xextended_error_logging & (ql_dbg_disc|ql_dbg_mbx)) {
                         pr_warn("%s [%s]-%04x:%ld: **** Failed", QL_MSGHDR,
                             dev_name(&ha->pdev->dev), 0x1020+0x800,
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index e218d68de532..04e0c7f51e68 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -2815,6 +2815,9 @@ qla2x00_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
         ha->link_data_rate = PORT_SPEED_UNKNOWN;
         ha->optrom_size = OPTROM_SIZE_2300;
         ha->max_exchg = FW_MAX_EXCHANGES_CNT;
+        atomic_set(&ha->num_pend_mbx_stage1, 0);
+        atomic_set(&ha->num_pend_mbx_stage2, 0);
+        atomic_set(&ha->num_pend_mbx_stage3, 0);
 
         /* Assign ISP specific operations. */
         if (IS_QLA2100(ha)) {