scsi: lpfc: Correct irq handling via locks when taking adapter offline

When taking the board offline while performing i/o, unsafe locking errors
occurred and irq level isn't properly managed.

In lpfc_sli_hba_down, spin_lock_irqsave(&phba->hbalock, flags) does not
disable softirqs raised from timer expiry.  It is possible that a softirq is
raised from the lpfc_els_retry_delay routine and recursively requests the same
phba->hbalock spinlock causing deadlock.

Address the deadlocks by creating a new port_list lock. The softirq behavior
can then be managed a level deeper into the calling sequences.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

1 files changed, 24 insertions, 1 deletions

diff --git a/drivers/scsi/lpfc/lpfc_sli.c b/drivers/scsi/lpfc/lpfc_sli.c
index a95c823cd1a4..495de99ed82d 100644
--- a/drivers/scsi/lpfc/lpfc_sli.c
+++ b/drivers/scsi/lpfc/lpfc_sli.c
@@ -10273,8 +10273,12 @@ lpfc_sli_mbox_sys_flush(struct lpfc_hba *phba)
 	LPFC_MBOXQ_t *pmb;
 	unsigned long iflag;
 
+	/* Disable softirqs, including timers from obtaining phba->hbalock */
+	local_bh_disable();
+
 	/* Flush all the mailbox commands in the mbox system */
 	spin_lock_irqsave(&phba->hbalock, iflag);
+
 	/* The pending mailbox command queue */
 	list_splice_init(&phba->sli.mboxq, &completions);
 	/* The outstanding active mailbox command */
@@ -10287,6 +10291,9 @@ lpfc_sli_mbox_sys_flush(struct lpfc_hba *phba)
 	list_splice_init(&phba->sli.mboxq_cmpl, &completions);
 	spin_unlock_irqrestore(&phba->hbalock, iflag);
 
+	/* Enable softirqs again, done with phba->hbalock */
+	local_bh_enable();
+
 	/* Return all flushed mailbox commands with MBX_NOT_FINISHED status */
 	while (!list_empty(&completions)) {
 		list_remove_head(&completions, pmb, LPFC_MBOXQ_t, list);
@@ -10426,6 +10433,9 @@ lpfc_sli_hba_down(struct lpfc_hba *phba)
 
 	lpfc_hba_down_prep(phba);
 
+	/* Disable softirqs, including timers from obtaining phba->hbalock */
+	local_bh_disable();
+
 	lpfc_fabric_abort_hba(phba);
 
 	spin_lock_irqsave(&phba->hbalock, flags);
@@ -10479,6 +10489,9 @@ lpfc_sli_hba_down(struct lpfc_hba *phba)
 		kfree(buf_ptr);
 	}
 
+	/* Enable softirqs again, done with phba->hbalock */
+	local_bh_enable();
+
 	/* Return any active mbox cmds */
 	del_timer_sync(&psli->mbox_tmo);
 
@@ -11782,6 +11795,9 @@ lpfc_sli_mbox_sys_shutdown(struct lpfc_hba *phba, int mbx_action)
 	}
 	timeout = msecs_to_jiffies(LPFC_MBOX_TMO * 1000) + jiffies;
 
+	/* Disable softirqs, including timers from obtaining phba->hbalock */
+	local_bh_disable();
+
 	spin_lock_irq(&phba->hbalock);
 	psli->sli_flag |= LPFC_SLI_ASYNC_MBX_BLK;
 
@@ -11795,6 +11811,9 @@ lpfc_sli_mbox_sys_shutdown(struct lpfc_hba *phba, int mbx_action)
 						1000) + jiffies;
 		spin_unlock_irq(&phba->hbalock);
 
+		/* Enable softirqs again, done with phba->hbalock */
+		local_bh_enable();
+
 		while (phba->sli.mbox_active) {
 			/* Check active mailbox complete status every 2ms */
 			msleep(2);
@@ -11804,9 +11823,13 @@ lpfc_sli_mbox_sys_shutdown(struct lpfc_hba *phba, int mbx_action)
 				 */
 				break;
 		}
-	} else
+	} else {
 		spin_unlock_irq(&phba->hbalock);
 
+		/* Enable softirqs again, done with phba->hbalock */
+		local_bh_enable();
+	}
+
 	lpfc_sli_mbox_sys_flush(phba);
 }