From: Tariq Toukan <tariqt@mellanox.com>
Date: Wed, 12 Sep 2018 15:36:41 +0300
Subject: net/mlx5: Refactor fragmented buffer struct fields and init flow
Patch-mainline: v4.20-rc1
Git-commit: 4972e6fa3a04032830bc3d6bb343d08ab3546773
References: bsc#1103990 FATE#326006
Take struct mlx5_frag_buf out of mlx5_frag_buf_ctrl, as it is not
needed to manage and control the datapath of the fragmented buffers API.
struct mlx5_frag_buf contains control info to manage the allocation
and de-allocation of the fragmented buffer.
Its fields are not relevant for datapath, so here I take them out of the
struct mlx5_frag_buf_ctrl, except for the fragments array itself.
In addition, modified mlx5_fill_fbc to initialise the frags pointers
as well. This implies that the buffer must be allocated before the
function is called.
A set of type-specific *_get_byte_size() functions are replaced by
a generic one.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
drivers/infiniband/hw/mlx5/cq.c | 31 ++----
drivers/infiniband/hw/mlx5/mlx5_ib.h | 1
drivers/net/ethernet/mellanox/mlx5/core/wq.c | 122 ++++++++++-----------------
include/linux/mlx5/driver.h | 22 +---
4 files changed, 70 insertions(+), 106 deletions(-)
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -393,7 +393,7 @@ static void handle_atomics(struct mlx5_i
static void free_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf)
{
- mlx5_frag_buf_free(dev->mdev, &buf->fbc.frag_buf);
+ mlx5_frag_buf_free(dev->mdev, &buf->frag_buf);
}
static void get_sig_err_item(struct mlx5_sig_err_cqe *cqe,
@@ -728,16 +728,11 @@ static int alloc_cq_frag_buf(struct mlx5
int nent,
int cqe_size)
{
- struct mlx5_frag_buf_ctrl *c = &buf->fbc;
- struct mlx5_frag_buf *frag_buf = &c->frag_buf;
- u32 cqc_buff[MLX5_ST_SZ_DW(cqc)] = {0};
+ struct mlx5_frag_buf *frag_buf = &buf->frag_buf;
+ u8 log_wq_stride = 6 + (cqe_size == 128 ? 1 : 0);
+ u8 log_wq_sz = ilog2(cqe_size);
int err;
- MLX5_SET(cqc, cqc_buff, log_cq_size, ilog2(cqe_size));
- MLX5_SET(cqc, cqc_buff, cqe_sz, (cqe_size == 128) ? 1 : 0);
-
- mlx5_core_init_cq_frag_buf(&buf->fbc, cqc_buff);
-
err = mlx5_frag_buf_alloc_node(dev->mdev,
nent * cqe_size,
frag_buf,
@@ -745,6 +740,8 @@ static int alloc_cq_frag_buf(struct mlx5
if (err)
return err;
+ mlx5_init_fbc(frag_buf->frags, log_wq_stride, log_wq_sz, &buf->fbc);
+
buf->cqe_size = cqe_size;
buf->nent = nent;
@@ -934,7 +931,7 @@ static int create_cq_kernel(struct mlx5_
*inlen = MLX5_ST_SZ_BYTES(create_cq_in) +
MLX5_FLD_SZ_BYTES(create_cq_in, pas[0]) *
- cq->buf.fbc.frag_buf.npages;
+ cq->buf.frag_buf.npages;
*cqb = kvzalloc(*inlen, GFP_KERNEL);
if (!*cqb) {
err = -ENOMEM;
@@ -942,11 +939,11 @@ static int create_cq_kernel(struct mlx5_
}
pas = (__be64 *)MLX5_ADDR_OF(create_cq_in, *cqb, pas);
- mlx5_fill_page_frag_array(&cq->buf.fbc.frag_buf, pas);
+ mlx5_fill_page_frag_array(&cq->buf.frag_buf, pas);
cqc = MLX5_ADDR_OF(create_cq_in, *cqb, cq_context);
MLX5_SET(cqc, cqc, log_page_size,
- cq->buf.fbc.frag_buf.page_shift -
+ cq->buf.frag_buf.page_shift -
MLX5_ADAPTER_PAGE_SHIFT);
*index = dev->mdev->priv.uar->index;
@@ -1365,11 +1362,10 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq
cqe_size = 64;
err = resize_kernel(dev, cq, entries, cqe_size);
if (!err) {
- struct mlx5_frag_buf_ctrl *c;
+ struct mlx5_frag_buf *frag_buf = &cq->resize_buf->frag_buf;
- c = &cq->resize_buf->fbc;
- npas = c->frag_buf.npages;
- page_shift = c->frag_buf.page_shift;
+ npas = frag_buf->npages;
+ page_shift = frag_buf->page_shift;
}
}
@@ -1390,8 +1386,7 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq
mlx5_ib_populate_pas(dev, cq->resize_umem, page_shift,
pas, 0);
else
- mlx5_fill_page_frag_array(&cq->resize_buf->fbc.frag_buf,
- pas);
+ mlx5_fill_page_frag_array(&cq->resize_buf->frag_buf, pas);
MLX5_SET(modify_cq_in, in,
modify_field_select_resize_field_select.resize_field_select.resize_field_select,
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -435,6 +435,7 @@ struct mlx5_ib_qp {
struct mlx5_ib_cq_buf {
struct mlx5_frag_buf_ctrl fbc;
+ struct mlx5_frag_buf frag_buf;
struct ib_umem *umem;
int cqe_size;
int nent;
--- a/drivers/net/ethernet/mellanox/mlx5/core/wq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/wq.c
@@ -49,54 +49,37 @@ u32 mlx5_wq_ll_get_size(struct mlx5_wq_l
return (u32)wq->fbc.sz_m1 + 1;
}
-static u32 mlx5_wq_cyc_get_byte_size(struct mlx5_wq_cyc *wq)
+static u32 wq_get_byte_sz(u8 log_sz, u8 log_stride)
{
- return mlx5_wq_cyc_get_size(wq) << wq->fbc.log_stride;
-}
-
-static u32 mlx5_wq_qp_get_byte_size(struct mlx5_wq_qp *wq)
-{
- return mlx5_wq_cyc_get_byte_size(&wq->rq) +
- mlx5_wq_cyc_get_byte_size(&wq->sq);
-}
-
-static u32 mlx5_cqwq_get_byte_size(struct mlx5_cqwq *wq)
-{
- return mlx5_cqwq_get_size(wq) << wq->fbc.log_stride;
-}
-
-static u32 mlx5_wq_ll_get_byte_size(struct mlx5_wq_ll *wq)
-{
- return mlx5_wq_ll_get_size(wq) << wq->fbc.log_stride;
+ return ((u32)1 << log_sz) << log_stride;
}
int mlx5_wq_cyc_create(struct mlx5_core_dev *mdev, struct mlx5_wq_param *param,
void *wqc, struct mlx5_wq_cyc *wq,
struct mlx5_wq_ctrl *wq_ctrl)
{
+ u8 log_wq_stride = MLX5_GET(wq, wqc, log_wq_stride);
+ u8 log_wq_sz = MLX5_GET(wq, wqc, log_wq_sz);
struct mlx5_frag_buf_ctrl *fbc = &wq->fbc;
int err;
- mlx5_fill_fbc(MLX5_GET(wq, wqc, log_wq_stride),
- MLX5_GET(wq, wqc, log_wq_sz),
- fbc);
- wq->sz = wq->fbc.sz_m1 + 1;
-
err = mlx5_db_alloc_node(mdev, &wq_ctrl->db, param->db_numa_node);
if (err) {
mlx5_core_warn(mdev, "mlx5_db_alloc_node() failed, %d\n", err);
return err;
}
- err = mlx5_frag_buf_alloc_node(mdev, mlx5_wq_cyc_get_byte_size(wq),
+ wq->db = wq_ctrl->db.db;
+
+ err = mlx5_frag_buf_alloc_node(mdev, wq_get_byte_sz(log_wq_sz, log_wq_stride),
&wq_ctrl->buf, param->buf_numa_node);
if (err) {
mlx5_core_warn(mdev, "mlx5_frag_buf_alloc_node() failed, %d\n", err);
goto err_db_free;
}
- fbc->frag_buf = wq_ctrl->buf;
- wq->db = wq_ctrl->db.db;
+ mlx5_init_fbc(wq_ctrl->buf.frags, log_wq_stride, log_wq_sz, fbc);
+ wq->sz = mlx5_wq_cyc_get_size(wq);
wq_ctrl->mdev = mdev;
@@ -108,46 +91,19 @@ err_db_free:
return err;
}
-static void mlx5_qp_set_frag_buf(struct mlx5_frag_buf *buf,
- struct mlx5_wq_qp *qp)
-{
- struct mlx5_frag_buf_ctrl *sq_fbc;
- struct mlx5_frag_buf *rqb, *sqb;
-
- rqb = &qp->rq.fbc.frag_buf;
- *rqb = *buf;
- rqb->size = mlx5_wq_cyc_get_byte_size(&qp->rq);
- rqb->npages = DIV_ROUND_UP(rqb->size, PAGE_SIZE);
-
- sq_fbc = &qp->sq.fbc;
- sqb = &sq_fbc->frag_buf;
- *sqb = *buf;
- sqb->size = mlx5_wq_cyc_get_byte_size(&qp->sq);
- sqb->npages = DIV_ROUND_UP(sqb->size, PAGE_SIZE);
- sqb->frags += rqb->npages; /* first part is for the rq */
- if (sq_fbc->strides_offset)
- sqb->frags--;
-}
-
int mlx5_wq_qp_create(struct mlx5_core_dev *mdev, struct mlx5_wq_param *param,
void *qpc, struct mlx5_wq_qp *wq,
struct mlx5_wq_ctrl *wq_ctrl)
{
- u16 sq_strides_offset;
- u32 rq_pg_remainder;
+ u8 log_rq_stride = MLX5_GET(qpc, qpc, log_rq_stride) + 4;
+ u8 log_rq_sz = MLX5_GET(qpc, qpc, log_rq_size);
+ u8 log_sq_stride = ilog2(MLX5_SEND_WQE_BB);
+ u8 log_sq_sz = MLX5_GET(qpc, qpc, log_sq_size);
+
+ u32 rq_byte_size;
int err;
- mlx5_fill_fbc(MLX5_GET(qpc, qpc, log_rq_stride) + 4,
- MLX5_GET(qpc, qpc, log_rq_size),
- &wq->rq.fbc);
-
- rq_pg_remainder = mlx5_wq_cyc_get_byte_size(&wq->rq) % PAGE_SIZE;
- sq_strides_offset = rq_pg_remainder / MLX5_SEND_WQE_BB;
-
- mlx5_fill_fbc_offset(ilog2(MLX5_SEND_WQE_BB),
- MLX5_GET(qpc, qpc, log_sq_size),
- sq_strides_offset,
- &wq->sq.fbc);
+
err = mlx5_db_alloc_node(mdev, &wq_ctrl->db, param->db_numa_node);
if (err) {
@@ -155,14 +111,32 @@ int mlx5_wq_qp_create(struct mlx5_core_d
return err;
}
- err = mlx5_frag_buf_alloc_node(mdev, mlx5_wq_qp_get_byte_size(wq),
+ err = mlx5_frag_buf_alloc_node(mdev,
+ wq_get_byte_sz(log_rq_sz, log_rq_stride) +
+ wq_get_byte_sz(log_sq_sz, log_sq_stride),
&wq_ctrl->buf, param->buf_numa_node);
if (err) {
mlx5_core_warn(mdev, "mlx5_frag_buf_alloc_node() failed, %d\n", err);
goto err_db_free;
}
- mlx5_qp_set_frag_buf(&wq_ctrl->buf, wq);
+ mlx5_init_fbc(wq_ctrl->buf.frags, log_rq_stride, log_rq_sz, &wq->rq.fbc);
+
+ rq_byte_size = wq_get_byte_sz(log_rq_sz, log_rq_stride);
+
+ if (rq_byte_size < PAGE_SIZE) {
+ /* SQ starts within the same page of the RQ */
+ u16 sq_strides_offset = rq_byte_size / MLX5_SEND_WQE_BB;
+
+ mlx5_init_fbc_offset(wq_ctrl->buf.frags,
+ log_sq_stride, log_sq_sz, sq_strides_offset,
+ &wq->sq.fbc);
+ } else {
+ u16 rq_npages = rq_byte_size >> PAGE_SHIFT;
+
+ mlx5_init_fbc(wq_ctrl->buf.frags + rq_npages,
+ log_sq_stride, log_sq_sz, &wq->sq.fbc);
+ }
wq->rq.db = &wq_ctrl->db.db[MLX5_RCV_DBR];
wq->sq.db = &wq_ctrl->db.db[MLX5_SND_DBR];
@@ -181,17 +155,19 @@ int mlx5_cqwq_create(struct mlx5_core_de
void *cqc, struct mlx5_cqwq *wq,
struct mlx5_wq_ctrl *wq_ctrl)
{
+ u8 log_wq_stride = MLX5_GET(cqc, cqc, cqe_sz) + 6;
+ u8 log_wq_sz = MLX5_GET(cqc, cqc, log_cq_size);
int err;
- mlx5_core_init_cq_frag_buf(&wq->fbc, cqc);
-
err = mlx5_db_alloc_node(mdev, &wq_ctrl->db, param->db_numa_node);
if (err) {
mlx5_core_warn(mdev, "mlx5_db_alloc_node() failed, %d\n", err);
return err;
}
- err = mlx5_frag_buf_alloc_node(mdev, mlx5_cqwq_get_byte_size(wq),
+ wq->db = wq_ctrl->db.db;
+
+ err = mlx5_frag_buf_alloc_node(mdev, wq_get_byte_sz(log_wq_sz, log_wq_stride),
&wq_ctrl->buf,
param->buf_numa_node);
if (err) {
@@ -200,8 +176,7 @@ int mlx5_cqwq_create(struct mlx5_core_de
goto err_db_free;
}
- wq->fbc.frag_buf = wq_ctrl->buf;
- wq->db = wq_ctrl->db.db;
+ mlx5_init_fbc(wq_ctrl->buf.frags, log_wq_stride, log_wq_sz, &wq->fbc);
wq_ctrl->mdev = mdev;
@@ -217,30 +192,29 @@ int mlx5_wq_ll_create(struct mlx5_core_d
void *wqc, struct mlx5_wq_ll *wq,
struct mlx5_wq_ctrl *wq_ctrl)
{
+ u8 log_wq_stride = MLX5_GET(wq, wqc, log_wq_stride);
+ u8 log_wq_sz = MLX5_GET(wq, wqc, log_wq_sz);
struct mlx5_frag_buf_ctrl *fbc = &wq->fbc;
struct mlx5_wqe_srq_next_seg *next_seg;
int err;
int i;
- mlx5_fill_fbc(MLX5_GET(wq, wqc, log_wq_stride),
- MLX5_GET(wq, wqc, log_wq_sz),
- fbc);
-
err = mlx5_db_alloc_node(mdev, &wq_ctrl->db, param->db_numa_node);
if (err) {
mlx5_core_warn(mdev, "mlx5_db_alloc_node() failed, %d\n", err);
return err;
}
- err = mlx5_frag_buf_alloc_node(mdev, mlx5_wq_ll_get_byte_size(wq),
+ wq->db = wq_ctrl->db.db;
+
+ err = mlx5_frag_buf_alloc_node(mdev, wq_get_byte_sz(log_wq_sz, log_wq_stride),
&wq_ctrl->buf, param->buf_numa_node);
if (err) {
mlx5_core_warn(mdev, "mlx5_frag_buf_alloc_node() failed, %d\n", err);
goto err_db_free;
}
- wq->fbc.frag_buf = wq_ctrl->buf;
- wq->db = wq_ctrl->db.db;
+ mlx5_init_fbc(wq_ctrl->buf.frags, log_wq_stride, log_wq_sz, fbc);
for (i = 0; i < fbc->sz_m1; i++) {
next_seg = mlx5_wq_ll_get_wqe(wq, i);
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -357,7 +357,7 @@ struct mlx5_frag_buf {
};
struct mlx5_frag_buf_ctrl {
- struct mlx5_frag_buf frag_buf;
+ struct mlx5_buf_list *frags;
u32 sz_m1;
u16 frag_sz_m1;
u16 strides_offset;
@@ -994,10 +994,12 @@ static inline u32 mlx5_base_mkey(const u
return key & 0xffffff00u;
}
-static inline void mlx5_fill_fbc_offset(u8 log_stride, u8 log_sz,
+static inline void mlx5_init_fbc_offset(struct mlx5_buf_list *frags,
+ u8 log_stride, u8 log_sz,
u16 strides_offset,
struct mlx5_frag_buf_ctrl *fbc)
{
+ fbc->frags = frags;
fbc->log_stride = log_stride;
fbc->log_sz = log_sz;
fbc->sz_m1 = (1 << fbc->log_sz) - 1;
@@ -1006,18 +1008,11 @@ static inline void mlx5_fill_fbc_offset(
fbc->strides_offset = strides_offset;
}
-static inline void mlx5_fill_fbc(u8 log_stride, u8 log_sz,
+static inline void mlx5_init_fbc(struct mlx5_buf_list *frags,
+ u8 log_stride, u8 log_sz,
struct mlx5_frag_buf_ctrl *fbc)
{
- mlx5_fill_fbc_offset(log_stride, log_sz, 0, fbc);
-}
-
-static inline void mlx5_core_init_cq_frag_buf(struct mlx5_frag_buf_ctrl *fbc,
- void *cqc)
-{
- mlx5_fill_fbc(6 + MLX5_GET(cqc, cqc, cqe_sz),
- MLX5_GET(cqc, cqc, log_cq_size),
- fbc);
+ mlx5_init_fbc_offset(frags, log_stride, log_sz, 0, fbc);
}
static inline void *mlx5_frag_buf_get_wqe(struct mlx5_frag_buf_ctrl *fbc,
@@ -1028,8 +1023,7 @@ static inline void *mlx5_frag_buf_get_wq
ix += fbc->strides_offset;
frag = ix >> fbc->log_frag_strides;
- return fbc->frag_buf.frags[frag].buf +
- ((fbc->frag_sz_m1 & ix) << fbc->log_stride);
+ return fbc->frags[frag].buf + ((fbc->frag_sz_m1 & ix) << fbc->log_stride);
}
static inline u32