diff mbox series

[05/42] lpfc: Fix failure to clear non-zero eq_delay after io rate reduction

Message ID 20190814235712.4487-6-jsmart2021@gmail.com (mailing list archive)
State Accepted
Headers show
Series lpfc: Update lpfc to revision 12.4.0.0 | expand

Commit Message

James Smart Aug. 14, 2019, 11:56 p.m. UTC
Unusually high IO latency can be observed with little IO in
progress. The latency may remain high regardless of amount of IO and
can only be cleared by forcing lpfc_fcp_imax values to non-zero and
then back to zero.

The driver's eq_delay mechanism that scales the interrupt coalescing
based on io completion load failed to reduce or turn off coalescing
when load decreased. Specifically, if no io completed on a cpu within
an eq_delay polling window, the eq delay processing was skipped and no
change was made to the coalescing values. This left the coalescing
values set when they were no longer applicable.

Fix by always clearing the percpu counters for each time period and
always run the eq_delay calculations if an eq has a non-zero
coalescing value.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
---
 drivers/scsi/lpfc/lpfc_init.c | 32 +++++++++++++++++++++-----------
 1 file changed, 21 insertions(+), 11 deletions(-)
diff mbox series

Patch

diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c
index 50d641f65af9..72353c9c0fa9 100644
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/drivers/scsi/lpfc/lpfc_init.c
@@ -1262,6 +1262,7 @@  lpfc_hb_eq_delay_work(struct work_struct *work)
 	unsigned char *eqcnt = NULL;
 	uint32_t usdelay;
 	int i;
+	bool update = false;
 
 	if (!phba->cfg_auto_imax || phba->pport->load_flag & FC_UNLOADING)
 		return;
@@ -1275,20 +1276,29 @@  lpfc_hb_eq_delay_work(struct work_struct *work)
 	if (!eqcnt)
 		goto requeue;
 
-	/* Loop thru all IRQ vectors */
-	for (i = 0; i < phba->cfg_irq_chann; i++) {
-		/* Get the EQ corresponding to the IRQ vector */
-		eq = phba->sli4_hba.hba_eq_hdl[i].eq;
-		if (eq && eqcnt[eq->last_cpu] < 2)
-			eqcnt[eq->last_cpu]++;
-		continue;
-	}
+	if (phba->cfg_irq_chann > 1) {
+		/* Loop thru all IRQ vectors */
+		for (i = 0; i < phba->cfg_irq_chann; i++) {
+			/* Get the EQ corresponding to the IRQ vector */
+			eq = phba->sli4_hba.hba_eq_hdl[i].eq;
+			if (!eq)
+				continue;
+			if (eq->q_mode) {
+				update = true;
+				break;
+			}
+			if (eqcnt[eq->last_cpu] < 2)
+				eqcnt[eq->last_cpu]++;
+		}
+	} else
+		update = true;
 
 	for_each_present_cpu(i) {
-		if (phba->cfg_irq_chann > 1 && eqcnt[i] < 2)
-			continue;
-
 		eqi = per_cpu_ptr(phba->sli4_hba.eq_info, i);
+		if (!update && eqcnt[i] < 2) {
+			eqi->icnt = 0;
+			continue;
+		}
 
 		usdelay = (eqi->icnt / LPFC_IMAX_THRESHOLD) *
 			   LPFC_EQ_DELAY_STEP;