scsi: lpfc: raise sg count for nvme to use available sg resources

The driver allocates a sg list per io struture based on a fixed maximum
size. When it registers with the protocol transports and indicates the max sg
list size it supports, the driver manipulates the fixed value to report a
lesser amount so that it has reserved space for sg elements that are used for
DIF.

The driver initialization path sets the cfg_sg_seg_cnt field to the
manipulated value for scsi. NVME initialization ran afterward and capped it's
maximum by the manipulated value for SCSI. This erroneously made NVME report
the SCSI-reduce-for-DIF value that reduced the max io size for nvme and wasted
sg elements.

Rework the driver so that cfg_sg_seg_cnt becomes the overall maximum size and
allow the max size to be tunable.  A separate (new) scsi sg count is then
setup with the scsi-modified reduced value. NVME then initializes based off
the overall maximum.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/lpfc/lpfc.h

@@ -52,7 +52,7 @@ struct lpfc_sli2_slim;
 		downloads using bsg */
 
 #define LPFC_MIN_SG_SLI4_BUF_SZ	0x800	/* based on LPFC_DEFAULT_SG_SEG_CNT */
-#define LPFC_MAX_SG_SLI4_SEG_CNT_DIF 128 /* sg element count per scsi cmnd */
+#define LPFC_MAX_BG_SLI4_SEG_CNT_DIF 128 /* sg element count for BlockGuard */
 #define LPFC_MAX_SG_SEG_CNT_DIF 512	/* sg element count per scsi cmnd  */
 #define LPFC_MAX_SG_SEG_CNT	4096	/* sg element count per scsi cmnd */
 #define LPFC_MIN_SG_SEG_CNT	32	/* sg element count per scsi cmnd */
@@ -790,6 +790,7 @@ struct lpfc_hba {
 	uint32_t cfg_total_seg_cnt;
 	uint32_t cfg_sg_seg_cnt;
 	uint32_t cfg_nvme_seg_cnt;
+	uint32_t cfg_scsi_seg_cnt;
 	uint32_t cfg_sg_dma_buf_size;
 	uint64_t cfg_soft_wwnn;
 	uint64_t cfg_soft_wwpn;

drivers/scsi/lpfc/lpfc_attr.c

@@ -5353,15 +5353,74 @@ LPFC_ATTR(delay_discovery, 0, 0, 1,
 
 /*
  * lpfc_sg_seg_cnt - Initial Maximum DMA Segment Count
- * This value can be set to values between 64 and 4096. The default value is
+ * This value can be set to values between 64 and 4096. The default value
- * 64, but may be increased to allow for larger Max I/O sizes. The scsi layer
+ * is 64, but may be increased to allow for larger Max I/O sizes. The scsi
- * will be allowed to request I/Os of sizes up to (MAX_SEG_COUNT * SEG_SIZE).
+ * and nvme layers will allow I/O sizes up to (MAX_SEG_COUNT * SEG_SIZE).
  * Because of the additional overhead involved in setting up T10-DIF,
  * this parameter will be limited to 128 if BlockGuard is enabled under SLI4
  * and will be limited to 512 if BlockGuard is enabled under SLI3.
  */
-LPFC_ATTR_R(sg_seg_cnt, LPFC_DEFAULT_SG_SEG_CNT, LPFC_MIN_SG_SEG_CNT,
+static uint lpfc_sg_seg_cnt = LPFC_DEFAULT_SG_SEG_CNT;
-	    LPFC_MAX_SG_SEG_CNT, "Max Scatter Gather Segment Count");
+module_param(lpfc_sg_seg_cnt, uint, 0444);
+MODULE_PARM_DESC(lpfc_sg_seg_cnt, "Max Scatter Gather Segment Count");
+
+/**
+ * lpfc_sg_seg_cnt_show - Display the scatter/gather list sizes
+ *    configured for the adapter
+ * @dev: class converted to a Scsi_host structure.
+ * @attr: device attribute, not used.
+ * @buf: on return contains a string with the list sizes
+ *
+ * Returns: size of formatted string.
+ **/
+static ssize_t
+lpfc_sg_seg_cnt_show(struct device *dev, struct device_attribute *attr,
+		     char *buf)
+{
+	struct Scsi_Host  *shost = class_to_shost(dev);
+	struct lpfc_vport *vport = (struct lpfc_vport *)shost->hostdata;
+	struct lpfc_hba   *phba = vport->phba;
+	int len;
+
+	len = snprintf(buf, PAGE_SIZE, "SGL sz: %d  total SGEs: %d\n",
+		       phba->cfg_sg_dma_buf_size, phba->cfg_total_seg_cnt);
+
+	len += snprintf(buf + len, PAGE_SIZE, "Cfg: %d  SCSI: %d  NVME: %d\n",
+			phba->cfg_sg_seg_cnt, phba->cfg_scsi_seg_cnt,
+			phba->cfg_nvme_seg_cnt);
+	return len;
+}
+
+static DEVICE_ATTR_RO(lpfc_sg_seg_cnt);
+
+/**
+ * lpfc_sg_seg_cnt_init - Set the hba sg_seg_cnt initial value
+ * @phba: lpfc_hba pointer.
+ * @val: contains the initial value
+ *
+ * Description:
+ * Validates the initial value is within range and assigns it to the
+ * adapter. If not in range, an error message is posted and the
+ * default value is assigned.
+ *
+ * Returns:
+ * zero if value is in range and is set
+ * -EINVAL if value was out of range
+ **/
+static int
+lpfc_sg_seg_cnt_init(struct lpfc_hba *phba, int val)
+{
+	if (val >= LPFC_MIN_SG_SEG_CNT && val <= LPFC_MAX_SG_SEG_CNT) {
+		phba->cfg_sg_seg_cnt = val;
+		return 0;
+	}
+	lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+			"0409 "LPFC_DRIVER_NAME"_sg_seg_cnt attribute cannot "
+			"be set to %d, allowed range is [%d, %d]\n",
+			val, LPFC_MIN_SG_SEG_CNT, LPFC_MAX_SG_SEG_CNT);
+	phba->cfg_sg_seg_cnt = LPFC_DEFAULT_SG_SEG_CNT;
+	return -EINVAL;
+}
 
 /*
  * lpfc_enable_mds_diags: Enable MDS Diagnostics

drivers/scsi/lpfc/lpfc_init.c

@@ -3956,7 +3956,7 @@ lpfc_create_port(struct lpfc_hba *phba, int instance, struct device *dev)
 	if (phba->sli_rev == LPFC_SLI_REV4) {
 		shost->dma_boundary =
 			phba->sli4_hba.pc_sli4_params.sge_supp_len-1;
-		shost->sg_tablesize = phba->cfg_sg_seg_cnt;
+		shost->sg_tablesize = phba->cfg_scsi_seg_cnt;
 	}
 
 	/*
@@ -5919,8 +5919,6 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 	 * There are going to be 2 reserved SGEs: 1 FCP cmnd + 1 FCP rsp
 	 */
 	max_buf_size = (2 * SLI4_PAGE_SIZE);
-	if (phba->cfg_sg_seg_cnt > LPFC_MAX_SGL_SEG_CNT - extra)
-		phba->cfg_sg_seg_cnt = LPFC_MAX_SGL_SEG_CNT - extra;
 
 	/*
 	 * Since lpfc_sg_seg_cnt is module param, the sg_dma_buf_size
@@ -5942,9 +5940,16 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 		/* Total SGEs for scsi_sg_list and scsi_sg_prot_list */
 		phba->cfg_total_seg_cnt = LPFC_MAX_SGL_SEG_CNT;
 
-		if (phba->cfg_sg_seg_cnt > LPFC_MAX_SG_SLI4_SEG_CNT_DIF)
+		/*
-			phba->cfg_sg_seg_cnt =
+		 * If supporting DIF, reduce the seg count for scsi to
-				LPFC_MAX_SG_SLI4_SEG_CNT_DIF;
+		 * allow room for the DIF sges.
+		 */
+		if (phba->cfg_enable_bg &&
+		    phba->cfg_sg_seg_cnt > LPFC_MAX_BG_SLI4_SEG_CNT_DIF)
+			phba->cfg_scsi_seg_cnt = LPFC_MAX_BG_SLI4_SEG_CNT_DIF;
+		else
+			phba->cfg_scsi_seg_cnt = phba->cfg_sg_seg_cnt;
+
 	} else {
 		/*
 		 * The scsi_buf for a regular I/O holds the FCP cmnd,
@@ -5958,6 +5963,7 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 
 		/* Total SGEs for scsi_sg_list */
 		phba->cfg_total_seg_cnt = phba->cfg_sg_seg_cnt + extra;
+		phba->cfg_scsi_seg_cnt = phba->cfg_sg_seg_cnt;
 
 		/*
 		 * NOTE: if (phba->cfg_sg_seg_cnt + extra) <= 256 we only
@@ -5965,10 +5971,22 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 		 */
 	}
 
+	/* Limit to LPFC_MAX_NVME_SEG_CNT for NVME. */
+	if (phba->cfg_enable_fc4_type & LPFC_ENABLE_NVME) {
+		if (phba->cfg_sg_seg_cnt > LPFC_MAX_NVME_SEG_CNT) {
+			lpfc_printf_log(phba, KERN_INFO, LOG_NVME | LOG_INIT,
+					"6300 Reducing NVME sg segment "
+					"cnt to %d\n",
+					LPFC_MAX_NVME_SEG_CNT);
+			phba->cfg_nvme_seg_cnt = LPFC_MAX_NVME_SEG_CNT;
+		} else
+			phba->cfg_nvme_seg_cnt = phba->cfg_sg_seg_cnt;
+	}
+
 	/* Initialize the host templates with the updated values. */
-	lpfc_vport_template.sg_tablesize = phba->cfg_sg_seg_cnt;
+	lpfc_vport_template.sg_tablesize = phba->cfg_scsi_seg_cnt;
-	lpfc_template.sg_tablesize = phba->cfg_sg_seg_cnt;
+	lpfc_template.sg_tablesize = phba->cfg_scsi_seg_cnt;
-	lpfc_template_no_hr.sg_tablesize = phba->cfg_sg_seg_cnt;
+	lpfc_template_no_hr.sg_tablesize = phba->cfg_scsi_seg_cnt;
 
 	if (phba->cfg_sg_dma_buf_size  <= LPFC_MIN_SG_SLI4_BUF_SZ)
 		phba->cfg_sg_dma_buf_size = LPFC_MIN_SG_SLI4_BUF_SZ;
@@ -5977,9 +5995,11 @@ lpfc_sli4_driver_resource_setup(struct lpfc_hba *phba)
 			SLI4_PAGE_ALIGN(phba->cfg_sg_dma_buf_size);
 
 	lpfc_printf_log(phba, KERN_INFO, LOG_INIT | LOG_FCP,
-			"9087 sg_tablesize:%d dmabuf_size:%d total_sge:%d\n",
+			"9087 sg_seg_cnt:%d dmabuf_size:%d "
+			"total:%d scsi:%d nvme:%d\n",
 			phba->cfg_sg_seg_cnt, phba->cfg_sg_dma_buf_size,
-			phba->cfg_total_seg_cnt);
+			phba->cfg_total_seg_cnt,  phba->cfg_scsi_seg_cnt,
+			phba->cfg_nvme_seg_cnt);
 
 	/* Initialize buffer queue management fields */
 	INIT_LIST_HEAD(&phba->hbqs[LPFC_ELS_HBQ].hbq_buffer_list);

drivers/scsi/lpfc/lpfc_nvme.c

@@ -2462,17 +2462,10 @@ lpfc_nvme_create_localport(struct lpfc_vport *vport)
 	nfcp_info.node_name = wwn_to_u64(vport->fc_nodename.u.wwn);
 	nfcp_info.port_name = wwn_to_u64(vport->fc_portname.u.wwn);
 
-	/* Limit to LPFC_MAX_NVME_SEG_CNT.
+	/* We need to tell the transport layer + 1 because it takes page
-	 * For now need + 1 to get around NVME transport logic.
+	 * alignment into account. When space for the SGL is allocated we
+	 * allocate + 3, one for cmd, one for rsp and one for this alignment
 	 */
-	if (phba->cfg_sg_seg_cnt > LPFC_MAX_NVME_SEG_CNT) {
-		lpfc_printf_vlog(vport, KERN_INFO, LOG_NVME | LOG_INIT,
-				 "6300 Reducing sg segment cnt to %d\n",
-				 LPFC_MAX_NVME_SEG_CNT);
-		phba->cfg_nvme_seg_cnt = LPFC_MAX_NVME_SEG_CNT;
-	} else {
-		phba->cfg_nvme_seg_cnt = phba->cfg_sg_seg_cnt;
-	}
 	lpfc_nvme_template.max_sgl_segments = phba->cfg_nvme_seg_cnt + 1;
 	lpfc_nvme_template.max_hw_queues = phba->cfg_nvme_io_channel;
 

drivers/scsi/lpfc/lpfc_nvmet.c

@@ -1373,17 +1373,10 @@ lpfc_nvmet_create_targetport(struct lpfc_hba *phba)
 	pinfo.port_name = wwn_to_u64(vport->fc_portname.u.wwn);
 	pinfo.port_id = vport->fc_myDID;
 
-	/* Limit to LPFC_MAX_NVME_SEG_CNT.
+	/* We need to tell the transport layer + 1 because it takes page
-	 * For now need + 1 to get around NVME transport logic.
+	 * alignment into account. When space for the SGL is allocated we
+	 * allocate + 3, one for cmd, one for rsp and one for this alignment
 	 */
-	if (phba->cfg_sg_seg_cnt > LPFC_MAX_NVME_SEG_CNT) {
-		lpfc_printf_log(phba, KERN_INFO, LOG_NVME | LOG_INIT,
-				"6400 Reducing sg segment cnt to %d\n",
-				LPFC_MAX_NVME_SEG_CNT);
-		phba->cfg_nvme_seg_cnt = LPFC_MAX_NVME_SEG_CNT;
-	} else {
-		phba->cfg_nvme_seg_cnt = phba->cfg_sg_seg_cnt;
-	}
 	lpfc_tgttemplate.max_sgl_segments = phba->cfg_nvme_seg_cnt + 1;
 	lpfc_tgttemplate.max_hw_queues = phba->cfg_nvme_io_channel;
 	lpfc_tgttemplate.target_features = NVMET_FCTGTFEAT_READDATA_RSP;