RDMA/i40iw: Fix SD calculation for initial HMC creation

Correct SD calculation by using base address returned from commit FPM.
This alleviates any assumptions on resource ordering and alignment
requirement. Also consolidate SD estimation code into i40iw_est_sd().

Signed-off-by: Mustafa Ismail <mustafa.ismail@intel.com>
Signed-off-by: Faisal Latif <faisal.latif@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

drivers/infiniband/hw/i40iw/i40iw_ctrl.c

@@ -114,16 +114,21 @@ static enum i40iw_status_code i40iw_cqp_poll_registers(
  * i40iw_sc_parse_fpm_commit_buf - parse fpm commit buffer
  * @buf: ptr to fpm commit buffer
  * @info: ptr to i40iw_hmc_obj_info struct
+ * @sd: number of SDs for HMC objects
  *
  * parses fpm commit info and copy base value
  * of hmc objects in hmc_info
  */
 static enum i40iw_status_code i40iw_sc_parse_fpm_commit_buf(
 				u64 *buf,
-				struct i40iw_hmc_obj_info *info)
+				struct i40iw_hmc_obj_info *info,
+				u32 *sd)
 {
 	u64 temp;
+	u64 size;
+	u64 base = 0;
 	u32 i, j;
+	u32 k = 0;
 	u32 low;
 
 	/* copy base values in obj_info */
@@ -131,10 +136,20 @@ static enum i40iw_status_code i40iw_sc_parse_fpm_commit_buf(
 			i <= I40IW_HMC_IW_PBLE; i++, j += 8) {
 		get_64bit_val(buf, j, &temp);
 		info[i].base = RS_64_1(temp, 32) * 512;
+		if (info[i].base > base) {
+			base = info[i].base;
+			k = i;
+		}
 		low = (u32)(temp);
 		if (low)
 			info[i].cnt = low;
 	}
+	size = info[k].cnt * info[k].size + info[k].base;
+	if (size & 0x1FFFFF)
+		*sd = (u32)((size >> 21) + 1); /* add 1 for remainder */
+	else
+		*sd = (u32)(size >> 21);
+
 	return 0;
 }
 
@@ -3206,7 +3221,7 @@ enum i40iw_status_code i40iw_sc_init_iw_hmc(struct i40iw_sc_dev *dev, u8 hmc_fn_
 		i40iw_cqp_commit_fpm_values_cmd(dev, &query_fpm_mem, hmc_fn_id);
 
 		/* parse the fpm_commit_buf and fill hmc obj info */
-		i40iw_sc_parse_fpm_commit_buf((u64 *)query_fpm_mem.va, hmc_info->hmc_obj);
+		i40iw_sc_parse_fpm_commit_buf((u64 *)query_fpm_mem.va, hmc_info->hmc_obj, &hmc_info->sd_table.sd_cnt);
 		mem_size = sizeof(struct i40iw_hmc_sd_entry) *
 			   (hmc_info->sd_table.sd_cnt + hmc_info->first_sd_index);
 		ret_code = i40iw_allocate_virt_mem(dev->hw, &virt_mem, mem_size);
@@ -3280,7 +3295,9 @@ static enum i40iw_status_code i40iw_sc_configure_iw_fpm(struct i40iw_sc_dev *dev
 
 	/* parse the fpm_commit_buf and fill hmc obj info */
 	if (!ret_code)
-		ret_code = i40iw_sc_parse_fpm_commit_buf(dev->fpm_commit_buf, hmc_info->hmc_obj);
+		ret_code = i40iw_sc_parse_fpm_commit_buf(dev->fpm_commit_buf,
+							 hmc_info->hmc_obj,
+							 &hmc_info->sd_table.sd_cnt);
 
 	i40iw_debug_buf(dev, I40IW_DEBUG_HMC, "COMMIT FPM BUFFER",
 			commit_fpm_mem.va, I40IW_COMMIT_FPM_BUF_SIZE);
@@ -3527,6 +3544,40 @@ static bool i40iw_ring_full(struct i40iw_sc_cqp *cqp)
 	return I40IW_RING_FULL_ERR(cqp->sq_ring);
 }
 
+/**
+ * i40iw_est_sd - returns approximate number of SDs for HMC
+ * @dev: sc device struct
+ * @hmc_info: hmc structure, size and count for HMC objects
+ */
+static u64 i40iw_est_sd(struct i40iw_sc_dev *dev, struct i40iw_hmc_info *hmc_info)
+{
+	int i;
+	u64 size = 0;
+	u64 sd;
+
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_PBLE; i++)
+		size += hmc_info->hmc_obj[i].cnt * hmc_info->hmc_obj[i].size;
+
+	if (dev->is_pf)
+		size += hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt * hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].size;
+
+	if (size & 0x1FFFFF)
+		sd = (size >> 21) + 1; /* add 1 for remainder */
+	else
+		sd = size >> 21;
+
+	if (!dev->is_pf) {
+		/* 2MB alignment for VF PBLE HMC */
+		size = hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt * hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].size;
+		if (size & 0x1FFFFF)
+			sd += (size >> 21) + 1; /* add 1 for remainder */
+		else
+			sd += size >> 21;
+	}
+
+	return sd;
+}
+
 /**
  * i40iw_config_fpm_values - configure HMC objects
  * @dev: sc device struct
@@ -3538,7 +3589,7 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 	u32 i, mem_size;
 	u32 qpwantedoriginal, qpwanted, mrwanted, pblewanted;
 	u32 powerof2;
-	u64 sd_needed, bytes_needed;
+	u64 sd_needed;
 	u32 loop_count = 0;
 
 	struct i40iw_hmc_info *hmc_info;
@@ -3556,23 +3607,15 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 		return ret_code;
 	}
 
-	bytes_needed = 0;
-	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++) {
+	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++)
 		hmc_info->hmc_obj[i].cnt = hmc_info->hmc_obj[i].max_cnt;
-		bytes_needed +=
-		    (hmc_info->hmc_obj[i].max_cnt) * (hmc_info->hmc_obj[i].size);
-		i40iw_debug(dev, I40IW_DEBUG_HMC,
-			    "%s i[%04d] max_cnt[0x%04X] size[0x%04llx]\n",
-			    __func__, i, hmc_info->hmc_obj[i].max_cnt,
-			    hmc_info->hmc_obj[i].size);
-	}
-	sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1; /* round up */
+	sd_needed = i40iw_est_sd(dev, hmc_info);
 	i40iw_debug(dev, I40IW_DEBUG_HMC,
 		    "%s: FW initial max sd_count[%08lld] first_sd_index[%04d]\n",
 		    __func__, sd_needed, hmc_info->first_sd_index);
 	i40iw_debug(dev, I40IW_DEBUG_HMC,
-		    "%s: bytes_needed=0x%llx sd count %d where max sd is %d\n",
-		    __func__, bytes_needed, hmc_info->sd_table.sd_cnt,
+		    "%s: sd count %d where max sd is %d\n",
+		    __func__, hmc_info->sd_table.sd_cnt,
 		    hmc_fpm_misc->max_sds);
 
 	qpwanted = min(qp_count, hmc_info->hmc_obj[I40IW_HMC_IW_QP].max_cnt);
@@ -3614,11 +3657,7 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 		hmc_info->hmc_obj[I40IW_HMC_IW_PBLE].cnt = pblewanted;
 
 		/* How much memory is needed for all the objects. */
-		bytes_needed = 0;
-		for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++)
-			bytes_needed +=
-			    (hmc_info->hmc_obj[i].cnt) * (hmc_info->hmc_obj[i].size);
-		sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1;
+		sd_needed = i40iw_est_sd(dev, hmc_info);
 		if ((loop_count > 1000) ||
 		    ((!(loop_count % 10)) &&
 		    (qpwanted > qpwantedoriginal * 2 / 3))) {
@@ -3639,15 +3678,7 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 			pblewanted -= FPM_MULTIPLIER * 1000;
 	} while (sd_needed > hmc_fpm_misc->max_sds && loop_count < 2000);
 
-	bytes_needed = 0;
-	for (i = I40IW_HMC_IW_QP; i < I40IW_HMC_IW_MAX; i++) {
-		bytes_needed += (hmc_info->hmc_obj[i].cnt) * (hmc_info->hmc_obj[i].size);
-		i40iw_debug(dev, I40IW_DEBUG_HMC,
-			    "%s i[%04d] cnt[0x%04x] size[0x%04llx]\n",
-			    __func__, i, hmc_info->hmc_obj[i].cnt,
-			    hmc_info->hmc_obj[i].size);
-	}
-	sd_needed = (bytes_needed / I40IW_HMC_DIRECT_BP_SIZE) + 1;    /* round up not truncate. */
+	sd_needed = i40iw_est_sd(dev, hmc_info);
 
 	i40iw_debug(dev, I40IW_DEBUG_HMC,
 		    "loop_cnt=%d, sd_needed=%lld, qpcnt = %d, cqcnt=%d, mrcnt=%d, pblecnt=%d\n",
@@ -3665,8 +3696,6 @@ enum i40iw_status_code i40iw_config_fpm_values(struct i40iw_sc_dev *dev, u32 qp_
 		return ret_code;
 	}
 
-	hmc_info->sd_table.sd_cnt = (u32)sd_needed;
-
 	mem_size = sizeof(struct i40iw_hmc_sd_entry) *
 		   (hmc_info->sd_table.sd_cnt + hmc_info->first_sd_index + 1);
 	ret_code = i40iw_allocate_virt_mem(dev->hw, &virt_mem, mem_size);

drivers/infiniband/hw/i40iw/i40iw_type.h

@@ -1112,7 +1112,7 @@ struct i40iw_hmc_ops {
 	enum i40iw_status_code (*parse_fpm_query_buf)(u64 *, struct i40iw_hmc_info *,
 						      struct i40iw_hmc_fpm_misc *);
 	enum i40iw_status_code (*configure_iw_fpm)(struct i40iw_sc_dev *, u8);
-	enum i40iw_status_code (*parse_fpm_commit_buf)(u64 *, struct i40iw_hmc_obj_info *);
+	enum i40iw_status_code (*parse_fpm_commit_buf)(u64 *, struct i40iw_hmc_obj_info *, u32 *sd);
 	enum i40iw_status_code (*create_hmc_object)(struct i40iw_sc_dev *dev,
 						    struct i40iw_hmc_create_obj_info *);
 	enum i40iw_status_code (*del_hmc_object)(struct i40iw_sc_dev *dev,