scsi: lpfc: Remove lock contention target write path

Lower IOps performance with write operations. Perf tool shows lock
contention in dma_pool_alloc and dma_pool_free related to the
txrdy_payload_pool.

The allocations are for dma buffers for XFER_RDY's, which actually are not
needed for the FCP_TRECEIVE command as the command contents are used by the
adapter to generate the IU.

Remove the allocations and the associated buffer pool.  Rather than leaving
NULLs in buffer pointer locations, set command and sgl to indicate skipped
SGLE indexes.

Link: https://lore.kernel.org/r/20191018211832.7917-10-jsmart2021@gmail.com
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>

5 files changed, 14 insertions, 54 deletions

diff --git a/drivers/scsi/lpfc/lpfc.h b/drivers/scsi/lpfc/lpfc.h
index 291335e16806..dabb5eac5fed 100644
--- a/drivers/scsi/lpfc/lpfc.h
+++ b/drivers/scsi/lpfc/lpfc.h
@@ -989,7 +989,6 @@ struct lpfc_hba {
 	struct dma_pool *lpfc_drb_pool; /* data receive buffer pool */
 	struct dma_pool *lpfc_nvmet_drb_pool; /* data receive buffer pool */
 	struct dma_pool *lpfc_hbq_pool;	/* SLI3 hbq buffer pool */
-	struct dma_pool *txrdy_payload_pool;
 	struct dma_pool *lpfc_cmd_rsp_buf_pool;
 	struct lpfc_dma_pool lpfc_mbuf_safety_pool;
 
diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c
index 1d14aa22f973..8292b66e4b07 100644
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/drivers/scsi/lpfc/lpfc_init.c
@@ -7556,18 +7556,10 @@ lpfc_create_shost(struct lpfc_hba *phba)
 
 	if (phba->nvmet_support) {
 		/* Only 1 vport (pport) will support NVME target */
-		if (phba->txrdy_payload_pool == NULL) {
-			phba->txrdy_payload_pool = dma_pool_create(
-				"txrdy_pool", &phba->pcidev->dev,
-				TXRDY_PAYLOAD_LEN, 16, 0);
-			if (phba->txrdy_payload_pool) {
-				phba->targetport = NULL;
-				phba->cfg_enable_fc4_type = LPFC_ENABLE_NVME;
-				lpfc_printf_log(phba, KERN_INFO,
-						LOG_INIT | LOG_NVME_DISC,
-						"6076 NVME Target Found\n");
-			}
-		}
+		phba->targetport = NULL;
+		phba->cfg_enable_fc4_type = LPFC_ENABLE_NVME;
+		lpfc_printf_log(phba, KERN_INFO, LOG_INIT | LOG_NVME_DISC,
+				"6076 NVME Target Found\n");
 	}
 
 	lpfc_debugfs_initialize(vport);
diff --git a/drivers/scsi/lpfc/lpfc_mem.c b/drivers/scsi/lpfc/lpfc_mem.c
index ae09bb863497..7082279e4c01 100644
--- a/drivers/scsi/lpfc/lpfc_mem.c
+++ b/drivers/scsi/lpfc/lpfc_mem.c
@@ -230,9 +230,6 @@ lpfc_mem_free(struct lpfc_hba *phba)
 	dma_pool_destroy(phba->lpfc_hrb_pool);
 	phba->lpfc_hrb_pool = NULL;
 
-	dma_pool_destroy(phba->txrdy_payload_pool);
-	phba->txrdy_payload_pool = NULL;
-
 	dma_pool_destroy(phba->lpfc_hbq_pool);
 	phba->lpfc_hbq_pool = NULL;
 
diff --git a/drivers/scsi/lpfc/lpfc_nvmet.c b/drivers/scsi/lpfc/lpfc_nvmet.c
index c9481a618b85..e3ac9059d33d 100644
--- a/drivers/scsi/lpfc/lpfc_nvmet.c
+++ b/drivers/scsi/lpfc/lpfc_nvmet.c
@@ -378,13 +378,6 @@ lpfc_nvmet_ctxbuf_post(struct lpfc_hba *phba, struct lpfc_nvmet_ctxbuf *ctx_buf)
 	int cpu;
 	unsigned long iflag;
 
-	if (ctxp->txrdy) {
-		dma_pool_free(phba->txrdy_payload_pool, ctxp->txrdy,
-			      ctxp->txrdy_phys);
-		ctxp->txrdy = NULL;
-		ctxp->txrdy_phys = 0;
-	}
-
 	if (ctxp->state == LPFC_NVMET_STE_FREE) {
 		lpfc_printf_log(phba, KERN_ERR, LOG_NVME_IOERR,
 				"6411 NVMET free, already free IO x%x: %d %d\n",
@@ -430,7 +423,6 @@ lpfc_nvmet_ctxbuf_post(struct lpfc_hba *phba, struct lpfc_nvmet_ctxbuf *ctx_buf)
 
 		ctxp = (struct lpfc_nvmet_rcv_ctx *)ctx_buf->context;
 		ctxp->wqeq = NULL;
-		ctxp->txrdy = NULL;
 		ctxp->offset = 0;
 		ctxp->phba = phba;
 		ctxp->size = size;
@@ -2327,7 +2319,6 @@ lpfc_nvmet_unsol_fcp_buffer(struct lpfc_hba *phba,
 				ctxp->state, ctxp->entry_cnt, ctxp->oxid);
 	}
 	ctxp->wqeq = NULL;
-	ctxp->txrdy = NULL;
 	ctxp->offset = 0;
 	ctxp->phba = phba;
 	ctxp->size = size;
@@ -2606,7 +2597,6 @@ lpfc_nvmet_prep_fcp_wqe(struct lpfc_hba *phba,
 	struct scatterlist *sgel;
 	union lpfc_wqe128 *wqe;
 	struct ulp_bde64 *bde;
-	uint32_t *txrdy;
 	dma_addr_t physaddr;
 	int i, cnt;
 	int do_pbde;
@@ -2768,23 +2758,11 @@ lpfc_nvmet_prep_fcp_wqe(struct lpfc_hba *phba,
 		       &lpfc_treceive_cmd_template.words[3],
 		       sizeof(uint32_t) * 9);
 
-		/* Words 0 - 2 : The first sg segment */
-		txrdy = dma_pool_alloc(phba->txrdy_payload_pool,
-				       GFP_KERNEL, &physaddr);
-		if (!txrdy) {
-			lpfc_printf_log(phba, KERN_ERR, LOG_NVME_IOERR,
-					"6041 Bad txrdy buffer: oxid x%x\n",
-					ctxp->oxid);
-			return NULL;
-		}
-		ctxp->txrdy = txrdy;
-		ctxp->txrdy_phys = physaddr;
-		wqe->fcp_treceive.bde.tus.f.bdeFlags = BUFF_TYPE_BDE_64;
-		wqe->fcp_treceive.bde.tus.f.bdeSize = TXRDY_PAYLOAD_LEN;
-		wqe->fcp_treceive.bde.addrLow =
-			cpu_to_le32(putPaddrLow(physaddr));
-		wqe->fcp_treceive.bde.addrHigh =
-			cpu_to_le32(putPaddrHigh(physaddr));
+		/* Words 0 - 2 : First SGE is skipped, set invalid BDE type */
+		wqe->fcp_treceive.bde.tus.f.bdeFlags = LPFC_SGE_TYPE_SKIP;
+		wqe->fcp_treceive.bde.tus.f.bdeSize = 0;
+		wqe->fcp_treceive.bde.addrLow = 0;
+		wqe->fcp_treceive.bde.addrHigh = 0;
 
 		/* Word 4 */
 		wqe->fcp_treceive.relative_offset = ctxp->offset;
@@ -2819,17 +2797,13 @@ lpfc_nvmet_prep_fcp_wqe(struct lpfc_hba *phba,
 		/* Word 12 */
 		wqe->fcp_tsend.fcp_data_len = rsp->transfer_length;
 
-		/* Setup 1 TXRDY and 1 SKIP SGE */
-		txrdy[0] = 0;
-		txrdy[1] = cpu_to_be32(rsp->transfer_length);
-		txrdy[2] = 0;
-
-		sgl->addr_hi = putPaddrHigh(physaddr);
-		sgl->addr_lo = putPaddrLow(physaddr);
+		/* Setup 2 SKIP SGEs */
+		sgl->addr_hi = 0;
+		sgl->addr_lo = 0;
 		sgl->word2 = 0;
-		bf_set(lpfc_sli4_sge_type, sgl, LPFC_SGE_TYPE_DATA);
+		bf_set(lpfc_sli4_sge_type, sgl, LPFC_SGE_TYPE_SKIP);
 		sgl->word2 = cpu_to_le32(sgl->word2);
-		sgl->sge_len = cpu_to_le32(TXRDY_PAYLOAD_LEN);
+		sgl->sge_len = 0;
 		sgl++;
 		sgl->addr_hi = 0;
 		sgl->addr_lo = 0;
diff --git a/drivers/scsi/lpfc/lpfc_nvmet.h b/drivers/scsi/lpfc/lpfc_nvmet.h
index 8ff67deac10a..b80b1639b9a7 100644
--- a/drivers/scsi/lpfc/lpfc_nvmet.h
+++ b/drivers/scsi/lpfc/lpfc_nvmet.h
@@ -112,9 +112,7 @@ struct lpfc_nvmet_rcv_ctx {
 	struct lpfc_hba *phba;
 	struct lpfc_iocbq *wqeq;
 	struct lpfc_iocbq *abort_wqeq;
-	dma_addr_t txrdy_phys;
 	spinlock_t ctxlock; /* protect flag access */
-	uint32_t *txrdy;
 	uint32_t sid;
 	uint32_t offset;
 	uint16_t oxid;