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@xxxxxxxxxxxx>
Signed-off-by: James Smart <james.smart@xxxxxxxxxxxx>
---
drivers/scsi/lpfc/lpfc.h | 3 +-
drivers/scsi/lpfc/lpfc_attr.c | 69 +++++++++++++++++++++++++++++++++++++++---
drivers/scsi/lpfc/lpfc_init.c | 42 ++++++++++++++++++-------
drivers/scsi/lpfc/lpfc_nvme.c | 13 ++------
drivers/scsi/lpfc/lpfc_nvmet.c | 13 ++------
5 files changed, 103 insertions(+), 37 deletions(-)
diff --git a/drivers/scsi/lpfc/lpfc.h b/drivers/scsi/lpfc/lpfc.h
index 43732e8d1347..322b24f826a4 100644
--- a/drivers/scsi/lpfc/lpfc.h
+++ b/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;
diff --git a/drivers/scsi/lpfc/lpfc_attr.c b/drivers/scsi/lpfc/lpfc_attr.c
index 057a60abe664..7d81f44630ee 100644
--- a/drivers/scsi/lpfc/lpfc_attr.c
+++ b/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
- * 64, but may be increased to allow for larger Max I/O sizes. The scsi layer
- * will be allowed to request I/Os of sizes up to (MAX_SEG_COUNT * SEG_SIZE).
+ * This value can be set to values between 64 and 4096. The default value
+ * is 64, but may be increased to allow for larger Max I/O sizes. The scsi
+ * 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,
- LPFC_MAX_SG_SEG_CNT, "Max Scatter Gather Segment Count");
+static uint lpfc_sg_seg_cnt = LPFC_DEFAULT_SG_SEG_CNT;
+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
diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c
index f3cae733ae2d..90fb83f88179 100644
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/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 =
- LPFC_MAX_SG_SLI4_SEG_CNT_DIF;
+ /*
+ * If supporting DIF, reduce the seg count for scsi to
+ * 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_template.sg_tablesize = phba->cfg_sg_seg_cnt;
- lpfc_template_no_hr.sg_tablesize = phba->cfg_sg_seg_cnt;
+ lpfc_vport_template.sg_tablesize = phba->cfg_scsi_seg_cnt;
+ lpfc_template.sg_tablesize = phba->cfg_scsi_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);
diff --git a/drivers/scsi/lpfc/lpfc_nvme.c b/drivers/scsi/lpfc/lpfc_nvme.c
index 028462e5994d..a84299c36c9f 100644
--- a/drivers/scsi/lpfc/lpfc_nvme.c
+++ b/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.
- * For now need + 1 to get around NVME transport logic.
+ /* We need to tell the transport layer + 1 because it takes page
+ * 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;
diff --git a/drivers/scsi/lpfc/lpfc_nvmet.c b/drivers/scsi/lpfc/lpfc_nvmet.c
index b766afe10d3d..4926ca6f0b8c 100644
--- a/drivers/scsi/lpfc/lpfc_nvmet.c
+++ b/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.
- * For now need + 1 to get around NVME transport logic.
+ /* We need to tell the transport layer + 1 because it takes page
+ * 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;
--
2.13.1
