From: Shiraz Saleem <shiraz.saleem@intel.com>
Date: Mon, 16 Oct 2017 15:46:02 -0500
Subject: i40iw: Refactor queue depth calculation
Patch-mainline: v4.15-rc1
Git-commit: 1196923838cf1d2b7abcb9a1007eede09ad93acd
References: bsc#1114687 FATE#326314

Queue depth calculations use a mix of work requests
and actual number of bytes. Consolidate all calculations
using minimum WQE size to avoid confusion.

Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 drivers/infiniband/hw/i40iw/i40iw_d.h     |    9 +++-
 drivers/infiniband/hw/i40iw/i40iw_uk.c    |   63 +++++++++++++++++++++++-------
 drivers/infiniband/hw/i40iw/i40iw_user.h  |    4 +
 drivers/infiniband/hw/i40iw/i40iw_verbs.c |   34 +++-------------
 4 files changed, 67 insertions(+), 43 deletions(-)

--- a/drivers/infiniband/hw/i40iw/i40iw_d.h
+++ b/drivers/infiniband/hw/i40iw/i40iw_d.h
@@ -1301,8 +1301,13 @@
 	(0xffffffffULL << I40IWQPC_LOCAL_IPADDR0_SHIFT)
 
 /* wqe size considering 32 bytes per wqe*/
-#define I40IWQP_SW_MIN_WQSIZE 4		/* 128 bytes */
-#define I40IWQP_SW_MAX_WQSIZE 2048	/* 2048 bytes */
+#define I40IW_QP_SW_MIN_WQSIZE 4		/*in WRs*/
+#define I40IW_SQ_RSVD 2
+#define I40IW_RQ_RSVD 1
+#define I40IW_MAX_QUANTAS_PER_WR 2
+#define I40IW_QP_SW_MAX_SQ_QUANTAS 2048
+#define I40IW_QP_SW_MAX_RQ_QUANTAS 16384
+#define I40IW_MAX_QP_WRS ((I40IW_QP_SW_MAX_SQ_QUANTAS / I40IW_MAX_QUANTAS_PER_WR) - 1)
 
 #define I40IWQP_OP_RDMA_WRITE 0
 #define I40IWQP_OP_RDMA_READ 1
--- a/drivers/infiniband/hw/i40iw/i40iw_uk.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_uk.c
@@ -894,8 +894,21 @@ exit:
 }
 
 /**
+ * i40iw_qp_roundup - return round up QP WQ depth
+ * @wqdepth: WQ depth in quantas to round up
+ */
+static int i40iw_qp_round_up(u32 wqdepth)
+{
+	int scount = 1;
+
+	for (wqdepth--; scount <= 16; scount *= 2)
+		wqdepth |= wqdepth >> scount;
+
+	return ++wqdepth;
+}
+
+/**
  * i40iw_get_wqe_shift - get shift count for maximum wqe size
- * @wqdepth: depth of wq required.
  * @sge: Maximum Scatter Gather Elements wqe
  * @inline_data: Maximum inline data size
  * @shift: Returns the shift needed based on sge
@@ -905,22 +918,48 @@ exit:
  * For 2 or 3 SGEs or inline data <= 48, shift = 1 (wqe size of 64 bytes).
  * Shift of 2 otherwise (wqe size of 128 bytes).
  */
-enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u32 sge, u32 inline_data, u8 *shift)
+void i40iw_get_wqe_shift(u32 sge, u32 inline_data, u8 *shift)
 {
-	u32 size;
-
 	*shift = 0;
 	if (sge > 1 || inline_data > 16)
 		*shift = (sge < 4 && inline_data <= 48) ? 1 : 2;
+}
 
-	/* check if wqdepth is multiple of 2 or not */
+/*
+ * i40iw_get_sqdepth - get SQ depth (quantas)
+ * @sq_size: SQ size
+ * @shift: shift which determines size of WQE
+ * @sqdepth: depth of SQ
+ *
+ */
+enum i40iw_status_code i40iw_get_sqdepth(u32 sq_size, u8 shift, u32 *sqdepth)
+{
+	*sqdepth = i40iw_qp_round_up((sq_size << shift) + I40IW_SQ_RSVD);
 
-	if ((wqdepth < I40IWQP_SW_MIN_WQSIZE) || (wqdepth & (wqdepth - 1)))
+	if (*sqdepth < (I40IW_QP_SW_MIN_WQSIZE << shift))
+		*sqdepth = I40IW_QP_SW_MIN_WQSIZE << shift;
+	else if (*sqdepth > I40IW_QP_SW_MAX_SQ_QUANTAS)
 		return I40IW_ERR_INVALID_SIZE;
 
-	size = wqdepth << *shift;	/* multiple of 32 bytes count */
-	if (size > I40IWQP_SW_MAX_WQSIZE)
+	return 0;
+}
+
+/*
+ * i40iw_get_rq_depth - get RQ depth (quantas)
+ * @rq_size: RQ size
+ * @shift: shift which determines size of WQE
+ * @rqdepth: depth of RQ
+ *
+ */
+enum i40iw_status_code i40iw_get_rqdepth(u32 rq_size, u8 shift, u32 *rqdepth)
+{
+	*rqdepth = i40iw_qp_round_up((rq_size << shift) + I40IW_RQ_RSVD);
+
+	if (*rqdepth < (I40IW_QP_SW_MIN_WQSIZE << shift))
+		*rqdepth = I40IW_QP_SW_MIN_WQSIZE << shift;
+	else if (*rqdepth > I40IW_QP_SW_MAX_RQ_QUANTAS)
 		return I40IW_ERR_INVALID_SIZE;
+
 	return 0;
 }
 
@@ -974,9 +1013,7 @@ enum i40iw_status_code i40iw_qp_uk_init(
 
 	if (info->max_rq_frag_cnt > I40IW_MAX_WQ_FRAGMENT_COUNT)
 		return I40IW_ERR_INVALID_FRAG_COUNT;
-	ret_code = i40iw_get_wqe_shift(info->sq_size, info->max_sq_frag_cnt, info->max_inline_data, &sqshift);
-	if (ret_code)
-		return ret_code;
+	i40iw_get_wqe_shift(info->max_sq_frag_cnt, info->max_inline_data, &sqshift);
 
 	qp->sq_base = info->sq;
 	qp->rq_base = info->rq;
@@ -1010,9 +1047,7 @@ enum i40iw_status_code i40iw_qp_uk_init(
 		I40IW_RING_INIT(qp->rq_ring, qp->rq_size);
 		switch (info->abi_ver) {
 		case 4:
-			ret_code = i40iw_get_wqe_shift(info->rq_size, info->max_rq_frag_cnt, 0, &rqshift);
-			if (ret_code)
-				return ret_code;
+			i40iw_get_wqe_shift(info->max_rq_frag_cnt, 0, &rqshift);
 			break;
 		case 5: /* fallthrough until next ABI version */
 		default:
--- a/drivers/infiniband/hw/i40iw/i40iw_user.h
+++ b/drivers/infiniband/hw/i40iw/i40iw_user.h
@@ -425,5 +425,7 @@ enum i40iw_status_code i40iw_fragcnt_to_
 enum i40iw_status_code i40iw_fragcnt_to_wqesize_rq(u32 frag_cnt, u8 *wqe_size);
 enum i40iw_status_code i40iw_inline_data_size_to_wqesize(u32 data_size,
 							 u8 *wqe_size);
-enum i40iw_status_code i40iw_get_wqe_shift(u32 wqdepth, u32 sge, u32 inline_data, u8 *shift);
+void i40iw_get_wqe_shift(u32 sge, u32 inline_data, u8 *shift);
+enum i40iw_status_code i40iw_get_sqdepth(u32 sq_size, u8 shift, u32 *sqdepth);
+enum i40iw_status_code i40iw_get_rqdepth(u32 rq_size, u8 shift, u32 *rqdepth);
 #endif
--- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
@@ -69,7 +69,7 @@ static int i40iw_query_device(struct ib_
 	props->hw_ver = (u32)iwdev->sc_dev.hw_rev;
 	props->max_mr_size = I40IW_MAX_OUTBOUND_MESSAGE_SIZE;
 	props->max_qp = iwdev->max_qp - iwdev->used_qps;
-	props->max_qp_wr = (I40IW_MAX_WQ_ENTRIES >> 2) - 1;
+	props->max_qp_wr = I40IW_MAX_QP_WRS;
 	props->max_sge = I40IW_MAX_WQ_FRAGMENT_COUNT;
 	props->max_cq = iwdev->max_cq - iwdev->used_cqs;
 	props->max_cqe = iwdev->max_cqe;
@@ -381,22 +381,6 @@ static int i40iw_dealloc_pd(struct ib_pd
 }
 
 /**
- * i40iw_qp_roundup - return round up qp ring size
- * @wr_ring_size: ring size to round up
- */
-static int i40iw_qp_roundup(u32 wr_ring_size)
-{
-	int scount = 1;
-
-	if (wr_ring_size < I40IWQP_SW_MIN_WQSIZE)
-		wr_ring_size = I40IWQP_SW_MIN_WQSIZE;
-
-	for (wr_ring_size--; scount <= 16; scount *= 2)
-		wr_ring_size |= wr_ring_size >> scount;
-	return ++wr_ring_size;
-}
-
-/**
  * i40iw_get_pbl - Retrieve pbl from a list given a virtual
  * address
  * @va: user virtual address
@@ -515,21 +499,19 @@ static int i40iw_setup_kmode_qp(struct i
 {
 	struct i40iw_dma_mem *mem = &iwqp->kqp.dma_mem;
 	u32 sqdepth, rqdepth;
-	u32 sq_size, rq_size;
 	u8 sqshift;
 	u32 size;
 	enum i40iw_status_code status;
 	struct i40iw_qp_uk_init_info *ukinfo = &info->qp_uk_init_info;
 
-	sq_size = i40iw_qp_roundup(ukinfo->sq_size + 1);
-	rq_size = i40iw_qp_roundup(ukinfo->rq_size + 1);
-
-	status = i40iw_get_wqe_shift(sq_size, ukinfo->max_sq_frag_cnt, ukinfo->max_inline_data, &sqshift);
+	i40iw_get_wqe_shift(ukinfo->max_sq_frag_cnt, ukinfo->max_inline_data, &sqshift);
+	status = i40iw_get_sqdepth(ukinfo->sq_size, sqshift, &sqdepth);
 	if (status)
 		return -ENOMEM;
 
-	sqdepth = sq_size << sqshift;
-	rqdepth = rq_size << I40IW_MAX_RQ_WQE_SHIFT;
+	status = i40iw_get_rqdepth(ukinfo->rq_size, I40IW_MAX_RQ_WQE_SHIFT, &rqdepth);
+	if (status)
+		return -ENOMEM;
 
 	size = sqdepth * sizeof(struct i40iw_sq_uk_wr_trk_info) + (rqdepth << 3);
 	iwqp->kqp.wrid_mem = kzalloc(size, GFP_KERNEL);
@@ -559,8 +541,8 @@ static int i40iw_setup_kmode_qp(struct i
 	ukinfo->shadow_area = ukinfo->rq[rqdepth].elem;
 	info->shadow_area_pa = info->rq_pa + (rqdepth * I40IW_QP_WQE_MIN_SIZE);
 
-	ukinfo->sq_size = sq_size;
-	ukinfo->rq_size = rq_size;
+	ukinfo->sq_size = sqdepth >> sqshift;
+	ukinfo->rq_size = rqdepth >> I40IW_MAX_RQ_WQE_SHIFT;
 	ukinfo->qp_id = iwqp->ibqp.qp_num;
 	return 0;
 }