[3/8] lpfc: Fix list corruption on the completion queue.

Commit Message

James Smart Aug. 1, 2018, 12:23 a.m. UTC

Enabling list_debug showed the drivers txcmplq was suffering list
corruption. The systems will eventually crash because the
iocb free list gets crossed linked with the prings txcmplq.
Most systems will run for a while after the corruption, but will
eventually crash when a scsi eh reset occurs and the txcmplq is
attempted to be flushed. The flush gets stuck in an endless loop.

The problem is the abort handler does not hold the sli4 ring lock while
validating the IO so the IO could complete while the driver is still
preping the abort.  The erroneously generated, abort when it completes,
has pointers to the original io that has already completed, and the
io manipulation (for the second time) corrupts the list.

Correct by taking the ring lock early in the abort handler so the
erroneous abort won't be sent if the io has/is completing.

Signed-off-by: Dick Kennedy <[e-mail]>
Signed-off-by: James Smart <[e-mail]>
---
 drivers/scsi/lpfc/lpfc_scsi.c | 44 +++++++++++++++++++++++++++++++++++--------
 1 file changed, 36 insertions(+), 8 deletions(-)

Patch

diff --git a/drivers/scsi/lpfc/lpfc_scsi.c b/drivers/scsi/lpfc/lpfc_scsi.c
index 25ef2b6ccc34..51ce244cf10a 100644
--- a/drivers/scsi/lpfc/lpfc_scsi.c
+++ b/drivers/scsi/lpfc/lpfc_scsi.c
@@ -4727,7 +4727,7 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 	struct lpfc_scsi_buf *lpfc_cmd;
 	IOCB_t *cmd, *icmd;
 	int ret = SUCCESS, status = 0;
-	struct lpfc_sli_ring *pring_s4;
+	struct lpfc_sli_ring *pring_s4 = NULL;
 	int ret_val;
 	unsigned long flags;
 	DECLARE_WAIT_QUEUE_HEAD_ONSTACK(waitq);
@@ -4757,8 +4757,25 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 	}
 
 	iocb = &lpfc_cmd->cur_iocbq;
+	if (phba->sli_rev == LPFC_SLI_REV4) {
+		if (!(phba->cfg_fof) ||
+		    (!(iocb->iocb_flag & LPFC_IO_FOF))) {
+			pring_s4 =
+				phba->sli4_hba.fcp_wq[iocb->hba_wqidx]->pring;
+		} else {
+			iocb->hba_wqidx = 0;
+			pring_s4 = phba->sli4_hba.oas_wq->pring;
+		}
+		if (!pring_s4) {
+			ret = FAILED;
+			goto out_unlock;
+		}
+		spin_lock(&pring_s4->ring_lock);
+	}
 	/* the command is in process of being cancelled */
 	if (!(iocb->iocb_flag & LPFC_IO_ON_TXCMPLQ)) {
+		if (phba->sli_rev == LPFC_SLI_REV4)
+			spin_unlock(&pring_s4->ring_lock);
 		spin_unlock_irqrestore(&phba->hbalock, flags);
 		lpfc_printf_vlog(vport, KERN_WARNING, LOG_FCP,
 			"3169 SCSI Layer abort requested I/O has been "
@@ -4772,6 +4789,8 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 	 * see the completion before the eh fired. Just return SUCCESS.
 	 */
 	if (lpfc_cmd->pCmd != cmnd) {
+		if (phba->sli_rev == LPFC_SLI_REV4)
+			spin_unlock(&pring_s4->ring_lock);
 		lpfc_printf_vlog(vport, KERN_WARNING, LOG_FCP,
 			"3170 SCSI Layer abort requested I/O has been "
 			"completed by LLD.\n");
@@ -4784,6 +4803,8 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 	if (iocb->iocb_flag & LPFC_DRIVER_ABORTED) {
 		lpfc_printf_vlog(vport, KERN_WARNING, LOG_FCP,
 			 "3389 SCSI Layer I/O Abort Request is pending\n");
+		if (phba->sli_rev == LPFC_SLI_REV4)
+			spin_unlock(&pring_s4->ring_lock);
 		spin_unlock_irqrestore(&phba->hbalock, flags);
 		goto wait_for_cmpl;
 	}
@@ -4791,6 +4812,8 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 	abtsiocb = __lpfc_sli_get_iocbq(phba);
 	if (abtsiocb == NULL) {
 		ret = FAILED;
+		if (phba->sli_rev == LPFC_SLI_REV4)
+			spin_unlock(&pring_s4->ring_lock);
 		goto out_unlock;
 	}
 
@@ -4828,14 +4851,9 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 
 	abtsiocb->iocb_cmpl = lpfc_sli_abort_fcp_cmpl;
 	abtsiocb->vport = vport;
+	lpfc_cmd->waitq = &waitq;
 	if (phba->sli_rev == LPFC_SLI_REV4) {
-		pring_s4 = lpfc_sli4_calc_ring(phba, abtsiocb);
-		if (pring_s4 == NULL) {
-			ret = FAILED;
-			goto out_unlock;
-		}
 		/* Note: both hbalock and ring_lock must be set here */
-		spin_lock(&pring_s4->ring_lock);
 		ret_val = __lpfc_sli_issue_iocb(phba, pring_s4->ringno,
 						abtsiocb, 0);
 		spin_unlock(&pring_s4->ring_lock);
@@ -4848,6 +4866,17 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 
 
 	if (ret_val == IOCB_ERROR) {
+		if (phba->sli_rev == LPFC_SLI_REV4)
+			spin_lock_irqsave(&pring_s4->ring_lock, flags);
+		else
+			spin_lock_irqsave(&phba->hbalock, flags);
+		/* Indicate the IO is not being aborted by the driver. */
+		iocb->iocb_flag &= ~LPFC_DRIVER_ABORTED;
+		lpfc_cmd->waitq = NULL;
+		if (phba->sli_rev == LPFC_SLI_REV4)
+			spin_unlock_irqrestore(&pring_s4->ring_lock, flags);
+		else
+			spin_unlock_irqrestore(&phba->hbalock, flags);
 		lpfc_sli_release_iocbq(phba, abtsiocb);
 		ret = FAILED;
 		goto out;
@@ -4858,7 +4887,6 @@  lpfc_abort_handler(struct scsi_cmnd *cmnd)
 			&phba->sli.sli3_ring[LPFC_FCP_RING], HA_R0RE_REQ);
 
 wait_for_cmpl:
-	lpfc_cmd->waitq = &waitq;
 	/* Wait for abort to complete */
 	wait_event_timeout(waitq,
 			  (lpfc_cmd->pCmd != cmnd),