[SCSI] lpfc 8.1.1 : Add polled-mode support
- Add functionality to run in polled mode only. Includes run time
attribute to enable mode.
- Enable runtime writable hba settings for coallescing and delay parameters
Customers have requested a mode in the driver to run strictly polled.
This is generally to support an environment where the server is extremely
loaded and is looking to reclaim some cpu cycles from adapter interrupt
handling.
This patch adds a new "poll" attribute, and the following behavior:
if value is 0 (default):
The driver uses the normal method for i/o completion. It uses the
firmware feature of interrupt coalesing. The firmware allows a
minimum number of i/o completions before an interrupt, or a maximum
time delay between interrupts. By default, the driver sets these
to no delay (disabled) or 1 i/o - meaning coalescing is disabled.
Attributes were provided to change the coalescing values, but it was
a module-load time only and global across all adapters.
This patch allows them to be writable on a per-adapter basis.
if value is 1 :
Interrupts are left enabled, expecting that the user has tuned the
interrupt coalescing values. When this setting is enabled, the driver
will attempt to service completed i/o whenever new i/o is submitted
to the adapter. If the coalescing values are large, and the i/o
generation rate steady, an interrupt will be avoided by servicing
completed i/o prior to the coalescing thresholds kicking in. However,
if the i/o completion load is high enough or i/o generation slow, the
coalescion values will ensure that completed i/o is serviced in a timely
fashion.
if value is 3 :
Turns off FCP i/o interrupts altogether. The coalescing values now have
no effect. A new attribute "poll_tmo" (default 10ms) exists to set
the polling interval for i/o completion. When this setting is enabled,
the driver will attempt to service completed i/o and restart the
interval timer whenever new i/o is submitted. This behavior allows for
servicing of completed i/o sooner than the interval timer, but ensures
that if no i/o is being issued, then the interval timer will kick in
to service the outstanding i/o.
Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
diff --git a/drivers/scsi/lpfc/lpfc_sli.c b/drivers/scsi/lpfc/lpfc_sli.c
index e2c08c5..7b785ad 100644
--- a/drivers/scsi/lpfc/lpfc_sli.c
+++ b/drivers/scsi/lpfc/lpfc_sli.c
@@ -886,6 +886,182 @@
return rc;
}
+static void lpfc_sli_rsp_pointers_error(struct lpfc_hba * phba,
+ struct lpfc_sli_ring * pring)
+{
+ struct lpfc_pgp *pgp = &phba->slim2p->mbx.us.s2.port[pring->ringno];
+ /*
+ * Ring <ringno> handler: portRspPut <portRspPut> is bigger then
+ * rsp ring <portRspMax>
+ */
+ lpfc_printf_log(phba, KERN_ERR, LOG_SLI,
+ "%d:0312 Ring %d handler: portRspPut %d "
+ "is bigger then rsp ring %d\n",
+ phba->brd_no, pring->ringno,
+ le32_to_cpu(pgp->rspPutInx),
+ pring->numRiocb);
+
+ phba->hba_state = LPFC_HBA_ERROR;
+
+ /*
+ * All error attention handlers are posted to
+ * worker thread
+ */
+ phba->work_ha |= HA_ERATT;
+ phba->work_hs = HS_FFER3;
+ if (phba->work_wait)
+ wake_up(phba->work_wait);
+
+ return;
+}
+
+void lpfc_sli_poll_fcp_ring(struct lpfc_hba * phba)
+{
+ struct lpfc_sli * psli = &phba->sli;
+ struct lpfc_sli_ring * pring = &psli->ring[LPFC_FCP_RING];
+ IOCB_t *irsp = NULL;
+ IOCB_t *entry = NULL;
+ struct lpfc_iocbq *cmdiocbq = NULL;
+ struct lpfc_iocbq rspiocbq;
+ struct lpfc_pgp *pgp;
+ uint32_t status;
+ uint32_t portRspPut, portRspMax;
+ int type;
+ uint32_t rsp_cmpl = 0;
+ void __iomem *to_slim;
+ uint32_t ha_copy;
+
+ pring->stats.iocb_event++;
+
+ /* The driver assumes SLI-2 mode */
+ pgp = &phba->slim2p->mbx.us.s2.port[pring->ringno];
+
+ /*
+ * The next available response entry should never exceed the maximum
+ * entries. If it does, treat it as an adapter hardware error.
+ */
+ portRspMax = pring->numRiocb;
+ portRspPut = le32_to_cpu(pgp->rspPutInx);
+ if (unlikely(portRspPut >= portRspMax)) {
+ lpfc_sli_rsp_pointers_error(phba, pring);
+ return;
+ }
+
+ rmb();
+ while (pring->rspidx != portRspPut) {
+
+ entry = IOCB_ENTRY(pring->rspringaddr, pring->rspidx);
+
+ if (++pring->rspidx >= portRspMax)
+ pring->rspidx = 0;
+
+ lpfc_sli_pcimem_bcopy((uint32_t *) entry,
+ (uint32_t *) &rspiocbq.iocb,
+ sizeof (IOCB_t));
+ irsp = &rspiocbq.iocb;
+ type = lpfc_sli_iocb_cmd_type(irsp->ulpCommand & CMD_IOCB_MASK);
+ pring->stats.iocb_rsp++;
+ rsp_cmpl++;
+
+ if (unlikely(irsp->ulpStatus)) {
+ /* Rsp ring <ringno> error: IOCB */
+ lpfc_printf_log(phba, KERN_WARNING, LOG_SLI,
+ "%d:0326 Rsp Ring %d error: IOCB Data: "
+ "x%x x%x x%x x%x x%x x%x x%x x%x\n",
+ phba->brd_no, pring->ringno,
+ irsp->un.ulpWord[0],
+ irsp->un.ulpWord[1],
+ irsp->un.ulpWord[2],
+ irsp->un.ulpWord[3],
+ irsp->un.ulpWord[4],
+ irsp->un.ulpWord[5],
+ *(((uint32_t *) irsp) + 6),
+ *(((uint32_t *) irsp) + 7));
+ }
+
+ switch (type) {
+ case LPFC_ABORT_IOCB:
+ case LPFC_SOL_IOCB:
+ /*
+ * Idle exchange closed via ABTS from port. No iocb
+ * resources need to be recovered.
+ */
+ if (unlikely(irsp->ulpCommand == CMD_XRI_ABORTED_CX)) {
+ printk(KERN_INFO "%s: IOCB cmd 0x%x processed."
+ " Skipping completion\n", __FUNCTION__,
+ irsp->ulpCommand);
+ break;
+ }
+
+ cmdiocbq = lpfc_sli_iocbq_lookup(phba, pring,
+ &rspiocbq);
+ if ((cmdiocbq) && (cmdiocbq->iocb_cmpl)) {
+ (cmdiocbq->iocb_cmpl)(phba, cmdiocbq,
+ &rspiocbq);
+ }
+ break;
+ default:
+ if (irsp->ulpCommand == CMD_ADAPTER_MSG) {
+ char adaptermsg[LPFC_MAX_ADPTMSG];
+ memset(adaptermsg, 0, LPFC_MAX_ADPTMSG);
+ memcpy(&adaptermsg[0], (uint8_t *) irsp,
+ MAX_MSG_DATA);
+ dev_warn(&((phba->pcidev)->dev), "lpfc%d: %s",
+ phba->brd_no, adaptermsg);
+ } else {
+ /* Unknown IOCB command */
+ lpfc_printf_log(phba, KERN_ERR, LOG_SLI,
+ "%d:0321 Unknown IOCB command "
+ "Data: x%x, x%x x%x x%x x%x\n",
+ phba->brd_no, type,
+ irsp->ulpCommand,
+ irsp->ulpStatus,
+ irsp->ulpIoTag,
+ irsp->ulpContext);
+ }
+ break;
+ }
+
+ /*
+ * The response IOCB has been processed. Update the ring
+ * pointer in SLIM. If the port response put pointer has not
+ * been updated, sync the pgp->rspPutInx and fetch the new port
+ * response put pointer.
+ */
+ to_slim = phba->MBslimaddr +
+ (SLIMOFF + (pring->ringno * 2) + 1) * 4;
+ writeb(pring->rspidx, to_slim);
+
+ if (pring->rspidx == portRspPut)
+ portRspPut = le32_to_cpu(pgp->rspPutInx);
+ }
+
+ ha_copy = readl(phba->HAregaddr);
+ ha_copy >>= (LPFC_FCP_RING * 4);
+
+ if ((rsp_cmpl > 0) && (ha_copy & HA_R0RE_REQ)) {
+ pring->stats.iocb_rsp_full++;
+ status = ((CA_R0ATT | CA_R0RE_RSP) << (LPFC_FCP_RING * 4));
+ writel(status, phba->CAregaddr);
+ readl(phba->CAregaddr);
+ }
+ if ((ha_copy & HA_R0CE_RSP) &&
+ (pring->flag & LPFC_CALL_RING_AVAILABLE)) {
+ pring->flag &= ~LPFC_CALL_RING_AVAILABLE;
+ pring->stats.iocb_cmd_empty++;
+
+ /* Force update of the local copy of cmdGetInx */
+ pring->local_getidx = le32_to_cpu(pgp->cmdGetInx);
+ lpfc_sli_resume_iocb(phba, pring);
+
+ if ((pring->lpfc_sli_cmd_available))
+ (pring->lpfc_sli_cmd_available) (phba, pring);
+
+ }
+
+ return;
+}
+
/*
* This routine presumes LPFC_FCP_RING handling and doesn't bother
* to check it explicitly.
@@ -917,24 +1093,7 @@
portRspMax = pring->numRiocb;
portRspPut = le32_to_cpu(pgp->rspPutInx);
if (unlikely(portRspPut >= portRspMax)) {
- /*
- * Ring <ringno> handler: portRspPut <portRspPut> is bigger then
- * rsp ring <portRspMax>
- */
- lpfc_printf_log(phba, KERN_ERR, LOG_SLI,
- "%d:0312 Ring %d handler: portRspPut %d "
- "is bigger then rsp ring %d\n",
- phba->brd_no, pring->ringno, portRspPut,
- portRspMax);
-
- phba->hba_state = LPFC_HBA_ERROR;
-
- /* All error attention handlers are posted to worker thread */
- phba->work_ha |= HA_ERATT;
- phba->work_hs = HS_FFER3;
- if (phba->work_wait)
- wake_up(phba->work_wait);
-
+ lpfc_sli_rsp_pointers_error(phba, pring);
spin_unlock_irqrestore(phba->host->host_lock, iflag);
return 1;
}
@@ -947,6 +1106,10 @@
* network byte order and pci byte orders are different.
*/
entry = IOCB_ENTRY(pring->rspringaddr, pring->rspidx);
+
+ if (++pring->rspidx >= portRspMax)
+ pring->rspidx = 0;
+
lpfc_sli_pcimem_bcopy((uint32_t *) entry,
(uint32_t *) &rspiocbq.iocb,
sizeof (IOCB_t));
@@ -1020,9 +1183,6 @@
* been updated, sync the pgp->rspPutInx and fetch the new port
* response put pointer.
*/
- if (++pring->rspidx >= portRspMax)
- pring->rspidx = 0;
-
to_slim = phba->MBslimaddr +
(SLIMOFF + (pring->ringno * 2) + 1) * 4;
writel(pring->rspidx, to_slim);
@@ -2615,6 +2775,7 @@
DECLARE_WAIT_QUEUE_HEAD(done_q);
long timeleft, timeout_req = 0;
int retval = IOCB_SUCCESS;
+ uint32_t creg_val;
/*
* If the caller has provided a response iocbq buffer, then context2
@@ -2630,6 +2791,13 @@
piocb->context_un.wait_queue = &done_q;
piocb->iocb_flag &= ~LPFC_IO_WAKE;
+ if (phba->cfg_poll & DISABLE_FCP_RING_INT) {
+ creg_val = readl(phba->HCregaddr);
+ creg_val |= (HC_R0INT_ENA << LPFC_FCP_RING);
+ writel(creg_val, phba->HCregaddr);
+ readl(phba->HCregaddr); /* flush */
+ }
+
retval = lpfc_sli_issue_iocb(phba, pring, piocb, 0);
if (retval == IOCB_SUCCESS) {
timeout_req = timeout * HZ;
@@ -2663,6 +2831,13 @@
retval = IOCB_ERROR;
}
+ if (phba->cfg_poll & DISABLE_FCP_RING_INT) {
+ creg_val = readl(phba->HCregaddr);
+ creg_val &= ~(HC_R0INT_ENA << LPFC_FCP_RING);
+ writel(creg_val, phba->HCregaddr);
+ readl(phba->HCregaddr); /* flush */
+ }
+
if (prspiocbq)
piocb->context2 = NULL;