xiaomi-sm8450/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

scsi: lpfc: Merge per-protocol WQ/CQ pairs into single per-cpu pair

Browse Source 
Currently, each hardware queue, typically allocated per-cpu, consists of a
WQ/CQ pair per protocol. Meaning if both SCSI and NVMe are supported 2
WQ/CQ pairs will exist for the hardware queue. Separate queues are
unnecessary. The current implementation wastes memory backing the 2nd set
of queues, and the use of double the SLI-4 WQ/CQ's means less hardware
queues can be supported which means there may not always be enough to have
a pair per cpu. If there is only 1 pair per cpu, more cpu's may get their
own WQ/CQ.

Rework the implementation to use a single WQ/CQ pair by both protocols.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This commit is contained in:
James Smart
2019-08-14 16:57:11 -07:00
committed by Martin K. Petersen
parent 0d8af09643
commit c00f62e6c5
10 changed files with 216 additions and 555 deletions
Download Patch File Download Diff File Expand all files Collapse all files

145
drivers/scsi/lpfc/lpfc_sli.c
View File

@@ -3978,7 +3978,7 @@ lpfc_sli_abort_fcp_rings(struct lpfc_hba *phba)
/* Look on all the FCP Rings for the iotag */
if (phba->sli_rev >= LPFC_SLI_REV4) {
for (i = 0; i < phba->cfg_hdw_queue; i++) {
pring = phba->sli4_hba.hdwq[i].fcp_wq->pring;
pring = phba->sli4_hba.hdwq[i].io_wq->pring;
lpfc_sli_abort_iocb_ring(phba, pring);
}
} else {
@@ -3988,17 +3988,17 @@ lpfc_sli_abort_fcp_rings(struct lpfc_hba *phba)
}
/**
* lpfc_sli_flush_fcp_rings - flush all iocbs in the fcp ring
* lpfc_sli_flush_io_rings - flush all iocbs in the IO ring
* @phba: Pointer to HBA context object.
*
* This function flushes all iocbs in the fcp ring and frees all the iocb
* This function flushes all iocbs in the IO ring and frees all the iocb
* objects in txq and txcmplq. This function will not issue abort iocbs
* for all the iocb commands in txcmplq, they will just be returned with
* IOERR_SLI_DOWN. This function is invoked with EEH when device's PCI
* slot has been permanently disabled.
**/
void
lpfc_sli_flush_fcp_rings(struct lpfc_hba *phba)
lpfc_sli_flush_io_rings(struct lpfc_hba *phba)
{
LIST_HEAD(txq);
LIST_HEAD(txcmplq);
@@ -4009,13 +4009,13 @@ lpfc_sli_flush_fcp_rings(struct lpfc_hba *phba)
spin_lock_irq(&phba->hbalock);
/* Indicate the I/O queues are flushed */
phba->hba_flag |= HBA_FCP_IOQ_FLUSH;
phba->hba_flag |= HBA_IOQ_FLUSH;
spin_unlock_irq(&phba->hbalock);
/* Look on all the FCP Rings for the iotag */
if (phba->sli_rev >= LPFC_SLI_REV4) {
for (i = 0; i < phba->cfg_hdw_queue; i++) {
pring = phba->sli4_hba.hdwq[i].fcp_wq->pring;
pring = phba->sli4_hba.hdwq[i].io_wq->pring;
spin_lock_irq(&pring->ring_lock);
/* Retrieve everything on txq */
@@ -4062,56 +4062,6 @@ lpfc_sli_flush_fcp_rings(struct lpfc_hba *phba)
}
}
/**
* lpfc_sli_flush_nvme_rings - flush all wqes in the nvme rings
* @phba: Pointer to HBA context object.
*
* This function flushes all wqes in the nvme rings and frees all resources
* in the txcmplq. This function does not issue abort wqes for the IO
* commands in txcmplq, they will just be returned with
* IOERR_SLI_DOWN. This function is invoked with EEH when device's PCI
* slot has been permanently disabled.
**/
void
lpfc_sli_flush_nvme_rings(struct lpfc_hba *phba)
{
LIST_HEAD(txcmplq);
struct lpfc_sli_ring *pring;
uint32_t i;
struct lpfc_iocbq *piocb, *next_iocb;
if ((phba->sli_rev < LPFC_SLI_REV4) ||
!(phba->cfg_enable_fc4_type & LPFC_ENABLE_NVME))
return;
/* Hint to other driver operations that a flush is in progress. */
spin_lock_irq(&phba->hbalock);
phba->hba_flag |= HBA_NVME_IOQ_FLUSH;
spin_unlock_irq(&phba->hbalock);
/* Cycle through all NVME rings and complete each IO with
* a local driver reason code. This is a flush so no
* abort exchange to FW.
*/
for (i = 0; i < phba->cfg_hdw_queue; i++) {
pring = phba->sli4_hba.hdwq[i].nvme_wq->pring;
spin_lock_irq(&pring->ring_lock);
list_for_each_entry_safe(piocb, next_iocb,
&pring->txcmplq, list)
piocb->iocb_flag &= ~LPFC_IO_ON_TXCMPLQ;
/* Retrieve everything on the txcmplq */
list_splice_init(&pring->txcmplq, &txcmplq);
pring->txcmplq_cnt = 0;
spin_unlock_irq(&pring->ring_lock);
/* Flush the txcmpq &&&PAE */
lpfc_sli_cancel_iocbs(phba, &txcmplq,
IOSTAT_LOCAL_REJECT,
IOERR_SLI_DOWN);
}
}
/**
* lpfc_sli_brdready_s3 - Check for sli3 host ready status
* @phba: Pointer to HBA context object.
@@ -5603,10 +5553,8 @@ lpfc_sli4_arm_cqeq_intr(struct lpfc_hba *phba)
for (qidx = 0; qidx < phba->cfg_hdw_queue; qidx++) {
qp = &sli4_hba->hdwq[qidx];
/* ARM the corresponding CQ */
sli4_hba->sli4_write_cq_db(phba, qp->fcp_cq, 0,
LPFC_QUEUE_REARM);
sli4_hba->sli4_write_cq_db(phba, qp->nvme_cq, 0,
LPFC_QUEUE_REARM);
sli4_hba->sli4_write_cq_db(phba, qp[qidx].io_cq, 0,
LPFC_QUEUE_REARM);
}
/* Loop thru all IRQ vectors */
@@ -7262,7 +7210,7 @@ lpfc_sli4_hba_setup(struct lpfc_hba *phba)
else
phba->hba_flag &= ~HBA_FIP_SUPPORT;
phba->hba_flag &= ~HBA_FCP_IOQ_FLUSH;
phba->hba_flag &= ~HBA_IOQ_FLUSH;
if (phba->sli_rev != LPFC_SLI_REV4) {
lpfc_printf_log(phba, KERN_ERR, LOG_MBOX | LOG_SLI,
@@ -9930,7 +9878,7 @@ __lpfc_sli_issue_iocb_s4(struct lpfc_hba *phba, uint32_t ring_number,
/* Get the WQ */
if ((piocb->iocb_flag & LPFC_IO_FCP) ||
(piocb->iocb_flag & LPFC_USE_FCPWQIDX)) {
wq = phba->sli4_hba.hdwq[piocb->hba_wqidx].fcp_wq;
wq = phba->sli4_hba.hdwq[piocb->hba_wqidx].io_wq;
} else {
wq = phba->sli4_hba.els_wq;
}
@@ -10077,7 +10025,7 @@ lpfc_sli4_calc_ring(struct lpfc_hba *phba, struct lpfc_iocbq *piocb)
lpfc_cmd = (struct lpfc_io_buf *)piocb->context1;
piocb->hba_wqidx = lpfc_cmd->hdwq_no;
}
return phba->sli4_hba.hdwq[piocb->hba_wqidx].fcp_wq->pring;
return phba->sli4_hba.hdwq[piocb->hba_wqidx].io_wq->pring;
} else {
if (unlikely(!phba->sli4_hba.els_wq))
return NULL;
@@ -10530,7 +10478,7 @@ lpfc_sli4_queue_init(struct lpfc_hba *phba)
INIT_LIST_HEAD(&psli->mboxq_cmpl);
/* Initialize list headers for txq and txcmplq as double linked lists */
for (i = 0; i < phba->cfg_hdw_queue; i++) {
pring = phba->sli4_hba.hdwq[i].fcp_wq->pring;
pring = phba->sli4_hba.hdwq[i].io_wq->pring;
pring->flag = 0;
pring->ringno = LPFC_FCP_RING;
pring->txcmplq_cnt = 0;
@@ -10549,16 +10497,6 @@ lpfc_sli4_queue_init(struct lpfc_hba *phba)
spin_lock_init(&pring->ring_lock);
if (phba->cfg_enable_fc4_type & LPFC_ENABLE_NVME) {
for (i = 0; i < phba->cfg_hdw_queue; i++) {
pring = phba->sli4_hba.hdwq[i].nvme_wq->pring;
pring->flag = 0;
pring->ringno = LPFC_FCP_RING;
pring->txcmplq_cnt = 0;
INIT_LIST_HEAD(&pring->txq);
INIT_LIST_HEAD(&pring->txcmplq);
INIT_LIST_HEAD(&pring->iocb_continueq);
spin_lock_init(&pring->ring_lock);
}
pring = phba->sli4_hba.nvmels_wq->pring;
pring->flag = 0;
pring->ringno = LPFC_ELS_RING;
@@ -11522,7 +11460,7 @@ lpfc_sli_abort_iocb(struct lpfc_vport *vport, struct lpfc_sli_ring *pring,
int i;
/* all I/Os are in process of being flushed */
if (phba->hba_flag & HBA_FCP_IOQ_FLUSH)
if (phba->hba_flag & HBA_IOQ_FLUSH)
return errcnt;
for (i = 1; i <= phba->sli.last_iotag; i++) {
@@ -11632,7 +11570,7 @@ lpfc_sli_abort_taskmgmt(struct lpfc_vport *vport, struct lpfc_sli_ring *pring,
spin_lock_irqsave(&phba->hbalock, iflags);
/* all I/Os are in process of being flushed */
if (phba->hba_flag & HBA_FCP_IOQ_FLUSH) {
if (phba->hba_flag & HBA_IOQ_FLUSH) {
spin_unlock_irqrestore(&phba->hbalock, iflags);
return 0;
}
@@ -11656,7 +11594,7 @@ lpfc_sli_abort_taskmgmt(struct lpfc_vport *vport, struct lpfc_sli_ring *pring,
if (phba->sli_rev == LPFC_SLI_REV4) {
pring_s4 =
phba->sli4_hba.hdwq[iocbq->hba_wqidx].fcp_wq->pring;
phba->sli4_hba.hdwq[iocbq->hba_wqidx].io_wq->pring;
if (!pring_s4) {
spin_unlock(&lpfc_cmd->buf_lock);
continue;
@@ -13365,8 +13303,13 @@ lpfc_sli4_sp_handle_abort_xri_wcqe(struct lpfc_hba *phba,
unsigned long iflags;
switch (cq->subtype) {
case LPFC_FCP:
lpfc_sli4_fcp_xri_aborted(phba, wcqe, cq->hdwq);
case LPFC_IO:
lpfc_sli4_io_xri_aborted(phba, wcqe, cq->hdwq);
if (phba->cfg_enable_fc4_type & LPFC_ENABLE_NVME) {
/* Notify aborted XRI for NVME work queue */
if (phba->nvmet_support)
lpfc_sli4_nvmet_xri_aborted(phba, wcqe);
}
workposted = false;
break;
case LPFC_NVME_LS: /* NVME LS uses ELS resources */
@@ -13384,15 +13327,6 @@ lpfc_sli4_sp_handle_abort_xri_wcqe(struct lpfc_hba *phba,
spin_unlock_irqrestore(&phba->hbalock, iflags);
workposted = true;
break;
case LPFC_NVME:
/* Notify aborted XRI for NVME work queue */
if (phba->nvmet_support)
lpfc_sli4_nvmet_xri_aborted(phba, wcqe);
else
lpfc_sli4_nvme_xri_aborted(phba, wcqe, cq->hdwq);
workposted = false;
break;
default:
lpfc_printf_log(phba, KERN_ERR, LOG_SLI,
"0603 Invalid CQ subtype %d: "
@@ -13720,7 +13654,7 @@ __lpfc_sli4_sp_process_cq(struct lpfc_queue *cq)
&delay);
break;
case LPFC_WCQ:
if (cq->subtype == LPFC_FCP || cq->subtype == LPFC_NVME)
if (cq->subtype == LPFC_IO)
workposted |= __lpfc_sli4_process_cq(phba, cq,
lpfc_sli4_fp_handle_cqe,
&delay);
@@ -14037,10 +13971,7 @@ lpfc_sli4_fp_handle_cqe(struct lpfc_hba *phba, struct lpfc_queue *cq,
cq->CQ_wq++;
/* Process the WQ complete event */
phba->last_completion_time = jiffies;
if ((cq->subtype == LPFC_FCP) || (cq->subtype == LPFC_NVME))
lpfc_sli4_fp_handle_fcp_wcqe(phba, cq,
(struct lpfc_wcqe_complete *)&wcqe);
if (cq->subtype == LPFC_NVME_LS)
if (cq->subtype == LPFC_IO || cq->subtype == LPFC_NVME_LS)
lpfc_sli4_fp_handle_fcp_wcqe(phba, cq,
(struct lpfc_wcqe_complete *)&wcqe);
break;
@@ -19506,7 +19437,7 @@ lpfc_drain_txq(struct lpfc_hba *phba)
if (phba->link_flag & LS_MDS_LOOPBACK) {
/* MDS WQE are posted only to first WQ*/
wq = phba->sli4_hba.hdwq[0].fcp_wq;
wq = phba->sli4_hba.hdwq[0].io_wq;
if (unlikely(!wq))
return 0;
pring = wq->pring;
@@ -19757,10 +19688,10 @@ lpfc_sli4_issue_wqe(struct lpfc_hba *phba, struct lpfc_sli4_hdw_queue *qp,
/* NVME_FCREQ and NVME_ABTS requests */
if (pwqe->iocb_flag & LPFC_IO_NVME) {
/* Get the IO distribution (hba_wqidx) for WQ assignment. */
wq = qp->nvme_wq;
wq = qp->io_wq;
pring = wq->pring;
bf_set(wqe_cqid, &wqe->generic.wqe_com, qp->nvme_cq_map);
bf_set(wqe_cqid, &wqe->generic.wqe_com, qp->io_cq_map);
lpfc_qp_spin_lock_irqsave(&pring->ring_lock, iflags,
qp, wq_access);
@@ -19777,7 +19708,7 @@ lpfc_sli4_issue_wqe(struct lpfc_hba *phba, struct lpfc_sli4_hdw_queue *qp,
/* NVMET requests */
if (pwqe->iocb_flag & LPFC_IO_NVMET) {
/* Get the IO distribution (hba_wqidx) for WQ assignment. */
wq = qp->nvme_wq;
wq = qp->io_wq;
pring = wq->pring;
ctxp = pwqe->context2;
@@ -19788,7 +19719,7 @@ lpfc_sli4_issue_wqe(struct lpfc_hba *phba, struct lpfc_sli4_hdw_queue *qp,
}
bf_set(wqe_xri_tag, &pwqe->wqe.xmit_bls_rsp.wqe_com,
pwqe->sli4_xritag);
bf_set(wqe_cqid, &wqe->generic.wqe_com, qp->nvme_cq_map);
bf_set(wqe_cqid, &wqe->generic.wqe_com, qp->io_cq_map);
lpfc_qp_spin_lock_irqsave(&pring->ring_lock, iflags,
qp, wq_access);
@@ -19835,9 +19766,7 @@ void lpfc_snapshot_mxp(struct lpfc_hba *phba, u32 hwqid)
if (multixri_pool->stat_snapshot_taken == LPFC_MXP_SNAPSHOT_TAKEN) {
pvt_pool = &qp->p_multixri_pool->pvt_pool;
pbl_pool = &qp->p_multixri_pool->pbl_pool;
txcmplq_cnt = qp->fcp_wq->pring->txcmplq_cnt;
if (qp->nvme_wq)
txcmplq_cnt += qp->nvme_wq->pring->txcmplq_cnt;
txcmplq_cnt = qp->io_wq->pring->txcmplq_cnt;
multixri_pool->stat_pbl_count = pbl_pool->count;
multixri_pool->stat_pvt_count = pvt_pool->count;
@@ -19907,12 +19836,9 @@ void lpfc_adjust_high_watermark(struct lpfc_hba *phba, u32 hwqid)
watermark_max = xri_limit;
watermark_min = xri_limit / 2;
txcmplq_cnt = qp->fcp_wq->pring->txcmplq_cnt;
txcmplq_cnt = qp->io_wq->pring->txcmplq_cnt;
abts_io_bufs = qp->abts_scsi_io_bufs;
if (qp->nvme_wq) {
txcmplq_cnt += qp->nvme_wq->pring->txcmplq_cnt;
abts_io_bufs += qp->abts_nvme_io_bufs;
}
abts_io_bufs += qp->abts_nvme_io_bufs;
new_watermark = txcmplq_cnt + abts_io_bufs;
new_watermark = min(watermark_max, new_watermark);
@@ -20187,12 +20113,9 @@ void lpfc_release_io_buf(struct lpfc_hba *phba, struct lpfc_io_buf *lpfc_ncmd,
pbl_pool = &qp->p_multixri_pool->pbl_pool;
pvt_pool = &qp->p_multixri_pool->pvt_pool;
txcmplq_cnt = qp->fcp_wq->pring->txcmplq_cnt;
txcmplq_cnt = qp->io_wq->pring->txcmplq_cnt;
abts_io_bufs = qp->abts_scsi_io_bufs;
if (qp->nvme_wq) {
txcmplq_cnt += qp->nvme_wq->pring->txcmplq_cnt;
abts_io_bufs += qp->abts_nvme_io_bufs;
}
abts_io_bufs += qp->abts_nvme_io_bufs;
xri_owned = pvt_pool->count + txcmplq_cnt + abts_io_bufs;
xri_limit = qp->p_multixri_pool->xri_limit;