aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/mlx5
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/infiniband/hw/mlx5')
-rw-r--r--drivers/infiniband/hw/mlx5/Kconfig1
-rw-r--r--drivers/infiniband/hw/mlx5/Makefile1
-rw-r--r--drivers/infiniband/hw/mlx5/cmd.c9
-rw-r--r--drivers/infiniband/hw/mlx5/cq.c56
-rw-r--r--drivers/infiniband/hw/mlx5/ib_rep.c13
-rw-r--r--drivers/infiniband/hw/mlx5/ib_rep.h12
-rw-r--r--drivers/infiniband/hw/mlx5/mad.c60
-rw-r--r--drivers/infiniband/hw/mlx5/main.c61
-rw-r--r--drivers/infiniband/hw/mlx5/mem.c20
-rw-r--r--drivers/infiniband/hw/mlx5/mlx5_ib.h29
-rw-r--r--drivers/infiniband/hw/mlx5/mr.c554
-rw-r--r--drivers/infiniband/hw/mlx5/odp.c23
-rw-r--r--drivers/infiniband/hw/mlx5/qp.c236
13 files changed, 775 insertions, 300 deletions
diff --git a/drivers/infiniband/hw/mlx5/Kconfig b/drivers/infiniband/hw/mlx5/Kconfig
index 8d651c05de62..ea248def4556 100644
--- a/drivers/infiniband/hw/mlx5/Kconfig
+++ b/drivers/infiniband/hw/mlx5/Kconfig
@@ -1,3 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0-only
config MLX5_INFINIBAND
tristate "Mellanox 5th generation network adapters (ConnectX series) support"
depends on NETDEVICES && ETHERNET && PCI && MLX5_CORE
diff --git a/drivers/infiniband/hw/mlx5/Makefile b/drivers/infiniband/hw/mlx5/Makefile
index 33f5adb14e4e..9924be8384d8 100644
--- a/drivers/infiniband/hw/mlx5/Makefile
+++ b/drivers/infiniband/hw/mlx5/Makefile
@@ -1,3 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0-only
obj-$(CONFIG_MLX5_INFINIBAND) += mlx5_ib.o
mlx5_ib-y := main.o cq.o doorbell.o qp.o mem.o srq_cmd.o \
diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c
index e3ec79b8f7f5..6c8645033102 100644
--- a/drivers/infiniband/hw/mlx5/cmd.c
+++ b/drivers/infiniband/hw/mlx5/cmd.c
@@ -190,12 +190,12 @@ int mlx5_cmd_alloc_sw_icm(struct mlx5_dm *dm, int type, u64 length,
u16 uid, phys_addr_t *addr, u32 *obj_id)
{
struct mlx5_core_dev *dev = dm->dev;
- u32 num_blocks = DIV_ROUND_UP(length, MLX5_SW_ICM_BLOCK_SIZE(dev));
u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {};
u32 in[MLX5_ST_SZ_DW(create_sw_icm_in)] = {};
unsigned long *block_map;
u64 icm_start_addr;
u32 log_icm_size;
+ u32 num_blocks;
u32 max_blocks;
u64 block_idx;
void *sw_icm;
@@ -224,6 +224,8 @@ int mlx5_cmd_alloc_sw_icm(struct mlx5_dm *dm, int type, u64 length,
return -EINVAL;
}
+ num_blocks = (length + MLX5_SW_ICM_BLOCK_SIZE(dev) - 1) >>
+ MLX5_LOG_SW_ICM_BLOCK_SIZE(dev);
max_blocks = BIT(log_icm_size - MLX5_LOG_SW_ICM_BLOCK_SIZE(dev));
spin_lock(&dm->lock);
block_idx = bitmap_find_next_zero_area(block_map,
@@ -266,13 +268,16 @@ int mlx5_cmd_dealloc_sw_icm(struct mlx5_dm *dm, int type, u64 length,
u16 uid, phys_addr_t addr, u32 obj_id)
{
struct mlx5_core_dev *dev = dm->dev;
- u32 num_blocks = DIV_ROUND_UP(length, MLX5_SW_ICM_BLOCK_SIZE(dev));
u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)] = {};
u32 in[MLX5_ST_SZ_DW(general_obj_in_cmd_hdr)] = {};
unsigned long *block_map;
+ u32 num_blocks;
u64 start_idx;
int err;
+ num_blocks = (length + MLX5_SW_ICM_BLOCK_SIZE(dev) - 1) >>
+ MLX5_LOG_SW_ICM_BLOCK_SIZE(dev);
+
switch (type) {
case MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM:
start_idx =
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index 4efbbd2fce0c..45f48cde6b9d 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -884,15 +884,15 @@ static void notify_soft_wc_handler(struct work_struct *work)
cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
}
-struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
- const struct ib_cq_init_attr *attr,
- struct ib_udata *udata)
+int mlx5_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
+ struct ib_udata *udata)
{
+ struct ib_device *ibdev = ibcq->device;
int entries = attr->cqe;
int vector = attr->comp_vector;
struct mlx5_ib_dev *dev = to_mdev(ibdev);
+ struct mlx5_ib_cq *cq = to_mcq(ibcq);
u32 out[MLX5_ST_SZ_DW(create_cq_out)];
- struct mlx5_ib_cq *cq;
int uninitialized_var(index);
int uninitialized_var(inlen);
u32 *cqb = NULL;
@@ -904,18 +904,14 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
if (entries < 0 ||
(entries > (1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz))))
- return ERR_PTR(-EINVAL);
+ return -EINVAL;
if (check_cq_create_flags(attr->flags))
- return ERR_PTR(-EOPNOTSUPP);
+ return -EOPNOTSUPP;
entries = roundup_pow_of_two(entries + 1);
if (entries > (1 << MLX5_CAP_GEN(dev->mdev, log_max_cq_sz)))
- return ERR_PTR(-EINVAL);
-
- cq = kzalloc(sizeof(*cq), GFP_KERNEL);
- if (!cq)
- return ERR_PTR(-ENOMEM);
+ return -EINVAL;
cq->ibcq.cqe = entries - 1;
mutex_init(&cq->resize_mutex);
@@ -930,13 +926,13 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
err = create_cq_user(dev, udata, cq, entries, &cqb, &cqe_size,
&index, &inlen);
if (err)
- goto err_create;
+ return err;
} else {
cqe_size = cache_line_size() == 128 ? 128 : 64;
err = create_cq_kernel(dev, cq, entries, cqe_size, &cqb,
&index, &inlen);
if (err)
- goto err_create;
+ return err;
INIT_WORK(&cq->notify_work, notify_soft_wc_handler);
}
@@ -981,7 +977,7 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
kvfree(cqb);
- return &cq->ibcq;
+ return 0;
err_cmd:
mlx5_core_destroy_cq(dev->mdev, &cq->mcq);
@@ -992,14 +988,10 @@ err_cqb:
destroy_cq_user(cq, udata);
else
destroy_cq_kernel(dev, cq);
-
-err_create:
- kfree(cq);
-
- return ERR_PTR(err);
+ return err;
}
-int mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
+void mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
{
struct mlx5_ib_dev *dev = to_mdev(cq->device);
struct mlx5_ib_cq *mcq = to_mcq(cq);
@@ -1009,10 +1001,6 @@ int mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
destroy_cq_user(mcq, udata);
else
destroy_cq_kernel(dev, mcq);
-
- kfree(mcq);
-
- return 0;
}
static int is_equal_rsn(struct mlx5_cqe64 *cqe64, u32 rsn)
@@ -1138,11 +1126,6 @@ static int resize_user(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
return 0;
}
-static void un_resize_user(struct mlx5_ib_cq *cq)
-{
- ib_umem_release(cq->resize_umem);
-}
-
static int resize_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
int entries, int cqe_size)
{
@@ -1165,12 +1148,6 @@ ex:
return err;
}
-static void un_resize_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq)
-{
- free_cq_buf(dev, cq->resize_buf);
- cq->resize_buf = NULL;
-}
-
static int copy_resize_cqes(struct mlx5_ib_cq *cq)
{
struct mlx5_ib_dev *dev = to_mdev(cq->ibcq.device);
@@ -1351,10 +1328,11 @@ ex_alloc:
kvfree(in);
ex_resize:
- if (udata)
- un_resize_user(cq);
- else
- un_resize_kernel(dev, cq);
+ ib_umem_release(cq->resize_umem);
+ if (!udata) {
+ free_cq_buf(dev, cq->resize_buf);
+ cq->resize_buf = NULL;
+ }
ex:
mutex_unlock(&cq->resize_mutex);
return err;
diff --git a/drivers/infiniband/hw/mlx5/ib_rep.c b/drivers/infiniband/hw/mlx5/ib_rep.c
index 3065c5d0ee96..f30e0d881368 100644
--- a/drivers/infiniband/hw/mlx5/ib_rep.c
+++ b/drivers/infiniband/hw/mlx5/ib_rep.c
@@ -114,15 +114,15 @@ u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw)
}
struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
- int vport_index)
+ u16 vport_num)
{
- return mlx5_eswitch_get_proto_dev(esw, vport_index, REP_IB);
+ return mlx5_eswitch_get_proto_dev(esw, vport_num, REP_IB);
}
struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw,
- int vport_index)
+ u16 vport_num)
{
- return mlx5_eswitch_get_proto_dev(esw, vport_index, REP_ETH);
+ return mlx5_eswitch_get_proto_dev(esw, vport_num, REP_ETH);
}
struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw)
@@ -130,9 +130,10 @@ struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw)
return mlx5_eswitch_uplink_get_proto_dev(esw, REP_IB);
}
-struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw, int vport)
+struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw,
+ u16 vport_num)
{
- return mlx5_eswitch_vport_rep(esw, vport);
+ return mlx5_eswitch_vport_rep(esw, vport_num);
}
struct mlx5_flow_handle *create_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
diff --git a/drivers/infiniband/hw/mlx5/ib_rep.h b/drivers/infiniband/hw/mlx5/ib_rep.h
index 478503ce20df..de43b423bafc 100644
--- a/drivers/infiniband/hw/mlx5/ib_rep.h
+++ b/drivers/infiniband/hw/mlx5/ib_rep.h
@@ -14,17 +14,17 @@ extern const struct mlx5_ib_profile uplink_rep_profile;
u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw);
struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
- int vport_index);
+ u16 vport_num);
struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw);
struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw,
- int vport_index);
+ u16 vport_num);
void mlx5_ib_register_vport_reps(struct mlx5_core_dev *mdev);
void mlx5_ib_unregister_vport_reps(struct mlx5_core_dev *mdev);
struct mlx5_flow_handle *create_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
struct mlx5_ib_sq *sq,
u16 port);
struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw,
- int vport_index);
+ u16 vport_num);
#else /* CONFIG_MLX5_ESWITCH */
static inline u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw)
{
@@ -33,7 +33,7 @@ static inline u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw)
static inline
struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
- int vport_index)
+ u16 vport_num)
{
return NULL;
}
@@ -46,7 +46,7 @@ struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw)
static inline
struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw,
- int vport_index)
+ u16 vport_num)
{
return NULL;
}
@@ -63,7 +63,7 @@ struct mlx5_flow_handle *create_flow_rule_vport_sq(struct mlx5_ib_dev *dev,
static inline
struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw,
- int vport_index)
+ u16 vport_num)
{
return NULL;
}
diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c
index 6c529e6f3a01..348c1df69cdc 100644
--- a/drivers/infiniband/hw/mlx5/mad.c
+++ b/drivers/infiniband/hw/mlx5/mad.c
@@ -200,19 +200,33 @@ static void pma_cnt_assign(struct ib_pma_portcounters *pma_cnt,
vl_15_dropped);
}
-static int process_pma_cmd(struct mlx5_core_dev *mdev, u8 port_num,
+static int process_pma_cmd(struct mlx5_ib_dev *dev, u8 port_num,
const struct ib_mad *in_mad, struct ib_mad *out_mad)
{
- int err;
+ struct mlx5_core_dev *mdev;
+ bool native_port = true;
+ u8 mdev_port_num;
void *out_cnt;
+ int err;
+ mdev = mlx5_ib_get_native_port_mdev(dev, port_num, &mdev_port_num);
+ if (!mdev) {
+ /* Fail to get the native port, likely due to 2nd port is still
+ * unaffiliated. In such case default to 1st port and attached
+ * PF device.
+ */
+ native_port = false;
+ mdev = dev->mdev;
+ mdev_port_num = 1;
+ }
/* Declaring support of extended counters */
if (in_mad->mad_hdr.attr_id == IB_PMA_CLASS_PORT_INFO) {
struct ib_class_port_info cpi = {};
cpi.capability_mask = IB_PMA_CLASS_CAP_EXT_WIDTH;
memcpy((out_mad->data + 40), &cpi, sizeof(cpi));
- return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+ err = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+ goto done;
}
if (in_mad->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS_EXT) {
@@ -221,11 +235,13 @@ static int process_pma_cmd(struct mlx5_core_dev *mdev, u8 port_num,
int sz = MLX5_ST_SZ_BYTES(query_vport_counter_out);
out_cnt = kvzalloc(sz, GFP_KERNEL);
- if (!out_cnt)
- return IB_MAD_RESULT_FAILURE;
+ if (!out_cnt) {
+ err = IB_MAD_RESULT_FAILURE;
+ goto done;
+ }
err = mlx5_core_query_vport_counter(mdev, 0, 0,
- port_num, out_cnt, sz);
+ mdev_port_num, out_cnt, sz);
if (!err)
pma_cnt_ext_assign(pma_cnt_ext, out_cnt);
} else {
@@ -234,20 +250,23 @@ static int process_pma_cmd(struct mlx5_core_dev *mdev, u8 port_num,
int sz = MLX5_ST_SZ_BYTES(ppcnt_reg);
out_cnt = kvzalloc(sz, GFP_KERNEL);
- if (!out_cnt)
- return IB_MAD_RESULT_FAILURE;
+ if (!out_cnt) {
+ err = IB_MAD_RESULT_FAILURE;
+ goto done;
+ }
- err = mlx5_core_query_ib_ppcnt(mdev, port_num,
+ err = mlx5_core_query_ib_ppcnt(mdev, mdev_port_num,
out_cnt, sz);
if (!err)
pma_cnt_assign(pma_cnt, out_cnt);
- }
-
+ }
kvfree(out_cnt);
- if (err)
- return IB_MAD_RESULT_FAILURE;
-
- return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+ err = err ? IB_MAD_RESULT_FAILURE :
+ IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+done:
+ if (native_port)
+ mlx5_ib_put_native_port_mdev(dev, port_num);
+ return err;
}
int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
@@ -259,8 +278,6 @@ int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
struct mlx5_ib_dev *dev = to_mdev(ibdev);
const struct ib_mad *in_mad = (const struct ib_mad *)in;
struct ib_mad *out_mad = (struct ib_mad *)out;
- struct mlx5_core_dev *mdev;
- u8 mdev_port_num;
int ret;
if (WARN_ON_ONCE(in_mad_size != sizeof(*in_mad) ||
@@ -269,19 +286,14 @@ int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
memset(out_mad->data, 0, sizeof(out_mad->data));
- mdev = mlx5_ib_get_native_port_mdev(dev, port_num, &mdev_port_num);
- if (!mdev)
- return IB_MAD_RESULT_FAILURE;
-
- if (MLX5_CAP_GEN(mdev, vport_counters) &&
+ if (MLX5_CAP_GEN(dev->mdev, vport_counters) &&
in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT &&
in_mad->mad_hdr.method == IB_MGMT_METHOD_GET) {
- ret = process_pma_cmd(mdev, mdev_port_num, in_mad, out_mad);
+ ret = process_pma_cmd(dev, port_num, in_mad, out_mad);
} else {
ret = process_mad(ibdev, mad_flags, port_num, in_wc, in_grh,
in_mad, out_mad);
}
- mlx5_ib_put_native_port_mdev(dev, port_num);
return ret;
}
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 26b1ce2359ba..9db8c06aa01e 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -52,6 +52,7 @@
#include <linux/mlx5/port.h>
#include <linux/mlx5/vport.h>
#include <linux/mlx5/fs.h>
+#include <linux/mlx5/eswitch.h>
#include <linux/list.h>
#include <rdma/ib_smi.h>
#include <rdma/ib_umem.h>
@@ -888,7 +889,7 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
}
props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
if (MLX5_CAP_GEN(mdev, sho)) {
- props->device_cap_flags |= IB_DEVICE_SIGNATURE_HANDOVER;
+ props->device_cap_flags |= IB_DEVICE_INTEGRITY_HANDOVER;
/* At this stage no support for signature handover */
props->sig_prot_cap = IB_PROT_T10DIF_TYPE_1 |
IB_PROT_T10DIF_TYPE_2 |
@@ -1008,6 +1009,8 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
props->max_srq_sge = max_rq_sg - 1;
props->max_fast_reg_page_list_len =
1 << MLX5_CAP_GEN(mdev, log_max_klm_list_size);
+ props->max_pi_fast_reg_page_list_len =
+ props->max_fast_reg_page_list_len / 2;
get_atomic_caps_qp(dev, props);
props->masked_atomic_cap = IB_ATOMIC_NONE;
props->max_mcast_grp = 1 << MLX5_CAP_GEN(mdev, log_max_mcg);
@@ -2344,7 +2347,7 @@ static int handle_alloc_dm_sw_icm(struct ib_ucontext *ctx,
/* Allocation size must a multiple of the basic block size
* and a power of 2.
*/
- act_size = roundup(attr->length, MLX5_SW_ICM_BLOCK_SIZE(dm_db->dev));
+ act_size = round_up(attr->length, MLX5_SW_ICM_BLOCK_SIZE(dm_db->dev));
act_size = roundup_pow_of_two(act_size);
dm->size = act_size;
@@ -3257,11 +3260,14 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
int max_table_size;
int num_entries;
int num_groups;
+ bool esw_encap;
u32 flags = 0;
int priority;
max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
log_max_ft_size));
+ esw_encap = mlx5_eswitch_get_encap_mode(dev->mdev) !=
+ DEVLINK_ESWITCH_ENCAP_MODE_NONE;
if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
enum mlx5_flow_namespace_type fn_type;
@@ -3274,10 +3280,10 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
if (ft_type == MLX5_IB_FT_RX) {
fn_type = MLX5_FLOW_NAMESPACE_BYPASS;
prio = &dev->flow_db->prios[priority];
- if (!dev->is_rep &&
+ if (!dev->is_rep && !esw_encap &&
MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, decap))
flags |= MLX5_FLOW_TABLE_TUNNEL_EN_DECAP;
- if (!dev->is_rep &&
+ if (!dev->is_rep && !esw_encap &&
MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
reformat_l3_tunnel_to_l2))
flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
@@ -3287,7 +3293,7 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
log_max_ft_size));
fn_type = MLX5_FLOW_NAMESPACE_EGRESS;
prio = &dev->flow_db->egress_prios[priority];
- if (!dev->is_rep &&
+ if (!dev->is_rep && !esw_encap &&
MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, reformat))
flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
}
@@ -3923,6 +3929,7 @@ _get_flow_table(struct mlx5_ib_dev *dev,
struct mlx5_flow_namespace *ns = NULL;
struct mlx5_ib_flow_prio *prio = NULL;
int max_table_size = 0;
+ bool esw_encap;
u32 flags = 0;
int priority;
@@ -3931,22 +3938,30 @@ _get_flow_table(struct mlx5_ib_dev *dev,
else
priority = ib_prio_to_core_prio(fs_matcher->priority, false);
+ esw_encap = mlx5_eswitch_get_encap_mode(dev->mdev) !=
+ DEVLINK_ESWITCH_ENCAP_MODE_NONE;
if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_BYPASS) {
max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
log_max_ft_size));
- if (MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, decap))
+ if (MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, decap) && !esw_encap)
flags |= MLX5_FLOW_TABLE_TUNNEL_EN_DECAP;
if (MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
- reformat_l3_tunnel_to_l2))
+ reformat_l3_tunnel_to_l2) &&
+ !esw_encap)
flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
} else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_EGRESS) {
max_table_size = BIT(
MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, log_max_ft_size));
- if (MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, reformat))
+ if (MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, reformat) && !esw_encap)
flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
} else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_FDB) {
max_table_size = BIT(
MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev, log_max_ft_size));
+ if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev, decap) && esw_encap)
+ flags |= MLX5_FLOW_TABLE_TUNNEL_EN_DECAP;
+ if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev, reformat_l3_tunnel_to_l2) &&
+ esw_encap)
+ flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
priority = FDB_BYPASS_PATH;
}
@@ -4926,18 +4941,19 @@ static int create_dev_resources(struct mlx5_ib_resources *devr)
if (ret)
goto error0;
- devr->c0 = mlx5_ib_create_cq(&dev->ib_dev, &cq_attr, NULL);
- if (IS_ERR(devr->c0)) {
- ret = PTR_ERR(devr->c0);
+ devr->c0 = rdma_zalloc_drv_obj(ibdev, ib_cq);
+ if (!devr->c0) {
+ ret = -ENOMEM;
goto error1;
}
- devr->c0->device = &dev->ib_dev;
- devr->c0->uobject = NULL;
- devr->c0->comp_handler = NULL;
- devr->c0->event_handler = NULL;
- devr->c0->cq_context = NULL;
+
+ devr->c0->device = &dev->ib_dev;
atomic_set(&devr->c0->usecnt, 0);
+ ret = mlx5_ib_create_cq(devr->c0, &cq_attr, NULL);
+ if (ret)
+ goto err_create_cq;
+
devr->x0 = mlx5_ib_alloc_xrcd(&dev->ib_dev, NULL);
if (IS_ERR(devr->x0)) {
ret = PTR_ERR(devr->x0);
@@ -5029,6 +5045,8 @@ error3:
mlx5_ib_dealloc_xrcd(devr->x0, NULL);
error2:
mlx5_ib_destroy_cq(devr->c0, NULL);
+err_create_cq:
+ kfree(devr->c0);
error1:
mlx5_ib_dealloc_pd(devr->p0, NULL);
error0:
@@ -5047,6 +5065,7 @@ static void destroy_dev_resources(struct mlx5_ib_resources *devr)
mlx5_ib_dealloc_xrcd(devr->x0, NULL);
mlx5_ib_dealloc_xrcd(devr->x1, NULL);
mlx5_ib_destroy_cq(devr->c0, NULL);
+ kfree(devr->c0);
mlx5_ib_dealloc_pd(devr->p0, NULL);
kfree(devr->p0);
@@ -6079,7 +6098,6 @@ static int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev)
if (mlx5_use_mad_ifc(dev))
get_ext_port_caps(dev);
- dev->ib_dev.owner = THIS_MODULE;
dev->ib_dev.node_type = RDMA_NODE_IB_CA;
dev->ib_dev.local_dma_lkey = 0 /* not supported for now */;
dev->ib_dev.phys_port_cnt = dev->num_ports;
@@ -6159,8 +6177,13 @@ static void mlx5_ib_stage_flow_db_cleanup(struct mlx5_ib_dev *dev)
}
static const struct ib_device_ops mlx5_ib_dev_ops = {
+ .owner = THIS_MODULE,
+ .driver_id = RDMA_DRIVER_MLX5,
+ .uverbs_abi_ver = MLX5_IB_UVERBS_ABI_VERSION,
+
.add_gid = mlx5_ib_add_gid,
.alloc_mr = mlx5_ib_alloc_mr,
+ .alloc_mr_integrity = mlx5_ib_alloc_mr_integrity,
.alloc_pd = mlx5_ib_alloc_pd,
.alloc_ucontext = mlx5_ib_alloc_ucontext,
.attach_mcast = mlx5_ib_mcg_attach,
@@ -6190,6 +6213,7 @@ static const struct ib_device_ops mlx5_ib_dev_ops = {
.get_dma_mr = mlx5_ib_get_dma_mr,
.get_link_layer = mlx5_ib_port_link_layer,
.map_mr_sg = mlx5_ib_map_mr_sg,
+ .map_mr_sg_pi = mlx5_ib_map_mr_sg_pi,
.mmap = mlx5_ib_mmap,
.modify_cq = mlx5_ib_modify_cq,
.modify_device = mlx5_ib_modify_device,
@@ -6214,6 +6238,7 @@ static const struct ib_device_ops mlx5_ib_dev_ops = {
.resize_cq = mlx5_ib_resize_cq,
INIT_RDMA_OBJ_SIZE(ib_ah, mlx5_ib_ah, ibah),
+ INIT_RDMA_OBJ_SIZE(ib_cq, mlx5_ib_cq, ibcq),
INIT_RDMA_OBJ_SIZE(ib_pd, mlx5_ib_pd, ibpd),
INIT_RDMA_OBJ_SIZE(ib_srq, mlx5_ib_srq, ibsrq),
INIT_RDMA_OBJ_SIZE(ib_ucontext, mlx5_ib_ucontext, ibucontext),
@@ -6256,7 +6281,6 @@ static int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
struct mlx5_core_dev *mdev = dev->mdev;
int err;
- dev->ib_dev.uverbs_abi_ver = MLX5_IB_UVERBS_ABI_VERSION;
dev->ib_dev.uverbs_cmd_mask =
(1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE) |
@@ -6325,7 +6349,6 @@ static int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
if (mlx5_accel_ipsec_device_caps(dev->mdev) &
MLX5_ACCEL_IPSEC_CAP_DEVICE)
ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_flow_ipsec_ops);
- dev->ib_dev.driver_id = RDMA_DRIVER_MLX5;
ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_ops);
if (IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS))
diff --git a/drivers/infiniband/hw/mlx5/mem.c b/drivers/infiniband/hw/mlx5/mem.c
index 9f90be296ee0..fe1a76d8531c 100644
--- a/drivers/infiniband/hw/mlx5/mem.c
+++ b/drivers/infiniband/hw/mlx5/mem.c
@@ -55,9 +55,10 @@ void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
int i = 0;
struct scatterlist *sg;
int entry;
- unsigned long page_shift = umem->page_shift;
if (umem->is_odp) {
+ unsigned int page_shift = to_ib_umem_odp(umem)->page_shift;
+
*ncont = ib_umem_page_count(umem);
*count = *ncont << (page_shift - PAGE_SHIFT);
*shift = page_shift;
@@ -67,15 +68,15 @@ void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
return;
}
- addr = addr >> page_shift;
+ addr = addr >> PAGE_SHIFT;
tmp = (unsigned long)addr;
m = find_first_bit(&tmp, BITS_PER_LONG);
if (max_page_shift)
- m = min_t(unsigned long, max_page_shift - page_shift, m);
+ m = min_t(unsigned long, max_page_shift - PAGE_SHIFT, m);
for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
- len = sg_dma_len(sg) >> page_shift;
- pfn = sg_dma_address(sg) >> page_shift;
+ len = sg_dma_len(sg) >> PAGE_SHIFT;
+ pfn = sg_dma_address(sg) >> PAGE_SHIFT;
if (base + p != pfn) {
/* If either the offset or the new
* base are unaligned update m
@@ -107,7 +108,7 @@ void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
*ncont = 0;
}
- *shift = page_shift + m;
+ *shift = PAGE_SHIFT + m;
*count = i;
}
@@ -140,8 +141,7 @@ void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
int page_shift, size_t offset, size_t num_pages,
__be64 *pas, int access_flags)
{
- unsigned long umem_page_shift = umem->page_shift;
- int shift = page_shift - umem_page_shift;
+ int shift = page_shift - PAGE_SHIFT;
int mask = (1 << shift) - 1;
int i, k, idx;
u64 cur = 0;
@@ -165,7 +165,7 @@ void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
i = 0;
for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
- len = sg_dma_len(sg) >> umem_page_shift;
+ len = sg_dma_len(sg) >> PAGE_SHIFT;
base = sg_dma_address(sg);
/* Skip elements below offset */
@@ -184,7 +184,7 @@ void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
for (; k < len; k++) {
if (!(i & mask)) {
- cur = base + (k << umem_page_shift);
+ cur = base + (k << PAGE_SHIFT);
cur |= access_flags;
idx = (i >> shift) - offset;
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index ee73dc122d28..305d26cdf7f3 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -431,9 +431,6 @@ struct mlx5_ib_qp {
int create_type;
- /* Store signature errors */
- bool signature_en;
-
struct list_head qps_list;
struct list_head cq_recv_list;
struct list_head cq_send_list;
@@ -587,6 +584,9 @@ struct mlx5_ib_mr {
void *descs;
dma_addr_t desc_map;
int ndescs;
+ int data_length;
+ int meta_ndescs;
+ int meta_length;
int max_descs;
int desc_size;
int access_mode;
@@ -605,6 +605,13 @@ struct mlx5_ib_mr {
int access_flags; /* Needed for rereg MR */
struct mlx5_ib_mr *parent;
+ /* Needed for IB_MR_TYPE_INTEGRITY */
+ struct mlx5_ib_mr *pi_mr;
+ struct mlx5_ib_mr *klm_mr;
+ struct mlx5_ib_mr *mtt_mr;
+ u64 data_iova;
+ u64 pi_iova;
+
atomic_t num_leaf_free;
wait_queue_head_t q_leaf_free;
struct mlx5_async_work cb_work;
@@ -1115,10 +1122,9 @@ int mlx5_ib_read_user_wqe_rq(struct mlx5_ib_qp *qp, int wqe_index, void *buffer,
int buflen, size_t *bc);
int mlx5_ib_read_user_wqe_srq(struct mlx5_ib_srq *srq, int wqe_index,
void *buffer, int buflen, size_t *bc);
-struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
- const struct ib_cq_init_attr *attr,
- struct ib_udata *udata);
-int mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
+int mlx5_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
+ struct ib_udata *udata);
+void mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
int mlx5_ib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc);
int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
int mlx5_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
@@ -1148,8 +1154,15 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
int mlx5_ib_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata);
struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
u32 max_num_sg, struct ib_udata *udata);
+struct ib_mr *mlx5_ib_alloc_mr_integrity(struct ib_pd *pd,
+ u32 max_num_sg,
+ u32 max_num_meta_sg);
int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
unsigned int *sg_offset);
+int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
+ int data_sg_nents, unsigned int *data_sg_offset,
+ struct scatterlist *meta_sg, int meta_sg_nents,
+ unsigned int *meta_sg_offset);
int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
const struct ib_wc *in_wc, const struct ib_grh *in_grh,
const struct ib_mad_hdr *in, size_t in_mad_size,
@@ -1201,7 +1214,7 @@ int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
struct ib_wq *mlx5_ib_create_wq(struct ib_pd *pd,
struct ib_wq_init_attr *init_attr,
struct ib_udata *udata);
-int mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata);
+void mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata);
int mlx5_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
u32 wq_attr_mask, struct ib_udata *udata);
struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 83b452d977d4..20ece6e0b2fc 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -1507,10 +1507,9 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
return 0;
err:
- if (mr->umem) {
- ib_umem_release(mr->umem);
- mr->umem = NULL;
- }
+ ib_umem_release(mr->umem);
+ mr->umem = NULL;
+
clean_mr(dev, mr);
return err;
}
@@ -1606,8 +1605,9 @@ static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
synchronize_srcu(&dev->mr_srcu);
/* Destroy all page mappings */
if (umem_odp->page_list)
- mlx5_ib_invalidate_range(umem_odp, ib_umem_start(umem),
- ib_umem_end(umem));
+ mlx5_ib_invalidate_range(umem_odp,
+ ib_umem_start(umem_odp),
+ ib_umem_end(umem_odp));
else
mlx5_ib_free_implicit_mr(mr);
/*
@@ -1629,28 +1629,85 @@ static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
* remove the DMA mapping.
*/
mlx5_mr_cache_free(dev, mr);
- if (umem) {
- ib_umem_release(umem);
+ ib_umem_release(umem);
+ if (umem)
atomic_sub(npages, &dev->mdev->priv.reg_pages);
- }
+
if (!mr->allocated_from_cache)
kfree(mr);
}
int mlx5_ib_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata)
{
- dereg_mr(to_mdev(ibmr->device), to_mmr(ibmr));
+ struct mlx5_ib_mr *mmr = to_mmr(ibmr);
+
+ if (ibmr->type == IB_MR_TYPE_INTEGRITY) {
+ dereg_mr(to_mdev(mmr->mtt_mr->ibmr.device), mmr->mtt_mr);
+ dereg_mr(to_mdev(mmr->klm_mr->ibmr.device), mmr->klm_mr);
+ }
+
+ dereg_mr(to_mdev(ibmr->device), mmr);
+
return 0;
}
-struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
- u32 max_num_sg, struct ib_udata *udata)
+static void mlx5_set_umr_free_mkey(struct ib_pd *pd, u32 *in, int ndescs,
+ int access_mode, int page_shift)
+{
+ void *mkc;
+
+ mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+ MLX5_SET(mkc, mkc, free, 1);
+ MLX5_SET(mkc, mkc, qpn, 0xffffff);
+ MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+ MLX5_SET(mkc, mkc, translations_octword_size, ndescs);
+ MLX5_SET(mkc, mkc, access_mode_1_0, access_mode & 0x3);
+ MLX5_SET(mkc, mkc, access_mode_4_2, (access_mode >> 2) & 0x7);
+ MLX5_SET(mkc, mkc, umr_en, 1);
+ MLX5_SET(mkc, mkc, log_page_size, page_shift);
+}
+
+static int _mlx5_alloc_mkey_descs(struct ib_pd *pd, struct mlx5_ib_mr *mr,
+ int ndescs, int desc_size, int page_shift,
+ int access_mode, u32 *in, int inlen)
{
struct mlx5_ib_dev *dev = to_mdev(pd->device);
+ int err;
+
+ mr->access_mode = access_mode;
+ mr->desc_size = desc_size;
+ mr->max_descs = ndescs;
+
+ err = mlx5_alloc_priv_descs(pd->device, mr, ndescs, desc_size);
+ if (err)
+ return err;
+
+ mlx5_set_umr_free_mkey(pd, in, ndescs, access_mode, page_shift);
+
+ err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, inlen);
+ if (err)
+ goto err_free_descs;
+
+ mr->mmkey.type = MLX5_MKEY_MR;
+ mr->ibmr.lkey = mr->mmkey.key;
+ mr->ibmr.rkey = mr->mmkey.key;
+
+ return 0;
+
+err_free_descs:
+ mlx5_free_priv_descs(mr);
+ return err;
+}
+
+static struct mlx5_ib_mr *mlx5_ib_alloc_pi_mr(struct ib_pd *pd,
+ u32 max_num_sg, u32 max_num_meta_sg,
+ int desc_size, int access_mode)
+{
int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
- int ndescs = ALIGN(max_num_sg, 4);
+ int ndescs = ALIGN(max_num_sg + max_num_meta_sg, 4);
+ int page_shift = 0;
struct mlx5_ib_mr *mr;
- void *mkc;
u32 *in;
int err;
@@ -1658,99 +1715,168 @@ struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
if (!mr)
return ERR_PTR(-ENOMEM);
+ mr->ibmr.pd = pd;
+ mr->ibmr.device = pd->device;
+
in = kzalloc(inlen, GFP_KERNEL);
if (!in) {
err = -ENOMEM;
goto err_free;
}
+ if (access_mode == MLX5_MKC_ACCESS_MODE_MTT)
+ page_shift = PAGE_SHIFT;
+
+ err = _mlx5_alloc_mkey_descs(pd, mr, ndescs, desc_size, page_shift,
+ access_mode, in, inlen);
+ if (err)
+ goto err_free_in;
+
+ mr->umem = NULL;
+ kfree(in);
+
+ return mr;
+
+err_free_in:
+ kfree(in);
+err_free:
+ kfree(mr);
+ return ERR_PTR(err);
+}
+
+static int mlx5_alloc_mem_reg_descs(struct ib_pd *pd, struct mlx5_ib_mr *mr,
+ int ndescs, u32 *in, int inlen)
+{
+ return _mlx5_alloc_mkey_descs(pd, mr, ndescs, sizeof(struct mlx5_mtt),
+ PAGE_SHIFT, MLX5_MKC_ACCESS_MODE_MTT, in,
+ inlen);
+}
+
+static int mlx5_alloc_sg_gaps_descs(struct ib_pd *pd, struct mlx5_ib_mr *mr,
+ int ndescs, u32 *in, int inlen)
+{
+ return _mlx5_alloc_mkey_descs(pd, mr, ndescs, sizeof(struct mlx5_klm),
+ 0, MLX5_MKC_ACCESS_MODE_KLMS, in, inlen);
+}
+
+static int mlx5_alloc_integrity_descs(struct ib_pd *pd, struct mlx5_ib_mr *mr,
+ int max_num_sg, int max_num_meta_sg,
+ u32 *in, int inlen)
+{
+ struct mlx5_ib_dev *dev = to_mdev(pd->device);
+ u32 psv_index[2];
+ void *mkc;
+ int err;
+
+ mr->sig = kzalloc(sizeof(*mr->sig), GFP_KERNEL);
+ if (!mr->sig)
+ return -ENOMEM;
+
+ /* create mem & wire PSVs */
+ err = mlx5_core_create_psv(dev->mdev, to_mpd(pd)->pdn, 2, psv_index);
+ if (err)
+ goto err_free_sig;
+
+ mr->sig->psv_memory.psv_idx = psv_index[0];
+ mr->sig->psv_wire.psv_idx = psv_index[1];
+
+ mr->sig->sig_status_checked = true;
+ mr->sig->sig_err_exists = false;
+ /* Next UMR, Arm SIGERR */
+ ++mr->sig->sigerr_count;
+ mr->klm_mr = mlx5_ib_alloc_pi_mr(pd, max_num_sg, max_num_meta_sg,
+ sizeof(struct mlx5_klm),
+ MLX5_MKC_ACCESS_MODE_KLMS);
+ if (IS_ERR(mr->klm_mr)) {
+ err = PTR_ERR(mr->klm_mr);
+ goto err_destroy_psv;
+ }
+ mr->mtt_mr = mlx5_ib_alloc_pi_mr(pd, max_num_sg, max_num_meta_sg,
+ sizeof(struct mlx5_mtt),
+ MLX5_MKC_ACCESS_MODE_MTT);
+ if (IS_ERR(mr->mtt_mr)) {
+ err = PTR_ERR(mr->mtt_mr);
+ goto err_free_klm_mr;
+ }
+
+ /* Set bsf descriptors for mkey */
mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
- MLX5_SET(mkc, mkc, free, 1);
- MLX5_SET(mkc, mkc, translations_octword_size, ndescs);
- MLX5_SET(mkc, mkc, qpn, 0xffffff);
- MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+ MLX5_SET(mkc, mkc, bsf_en, 1);
+ MLX5_SET(mkc, mkc, bsf_octword_size, MLX5_MKEY_BSF_OCTO_SIZE);
- if (mr_type == IB_MR_TYPE_MEM_REG) {
- mr->access_mode = MLX5_MKC_ACCESS_MODE_MTT;
- MLX5_SET(mkc, mkc, log_page_size, PAGE_SHIFT);
- err = mlx5_alloc_priv_descs(pd->device, mr,
- ndescs, sizeof(struct mlx5_mtt));
- if (err)
- goto err_free_in;
+ err = _mlx5_alloc_mkey_descs(pd, mr, 4, sizeof(struct mlx5_klm), 0,
+ MLX5_MKC_ACCESS_MODE_KLMS, in, inlen);
+ if (err)
+ goto err_free_mtt_mr;
- mr->desc_size = sizeof(struct mlx5_mtt);
- mr->max_descs = ndescs;
- } else if (mr_type == IB_MR_TYPE_SG_GAPS) {
- mr->access_mode = MLX5_MKC_ACCESS_MODE_KLMS;
+ return 0;
- err = mlx5_alloc_priv_descs(pd->device, mr,
- ndescs, sizeof(struct mlx5_klm));
- if (err)
- goto err_free_in;
- mr->desc_size = sizeof(struct mlx5_klm);
- mr->max_descs = ndescs;
- } else if (mr_type == IB_MR_TYPE_SIGNATURE) {
- u32 psv_index[2];
-
- MLX5_SET(mkc, mkc, bsf_en, 1);
- MLX5_SET(mkc, mkc, bsf_octword_size, MLX5_MKEY_BSF_OCTO_SIZE);
- mr->sig = kzalloc(sizeof(*mr->sig), GFP_KERNEL);
- if (!mr->sig) {
- err = -ENOMEM;
- goto err_free_in;
- }
+err_free_mtt_mr:
+ dereg_mr(to_mdev(mr->mtt_mr->ibmr.device), mr->mtt_mr);
+ mr->mtt_mr = NULL;
+err_free_klm_mr:
+ dereg_mr(to_mdev(mr->klm_mr->ibmr.device), mr->klm_mr);
+ mr->klm_mr = NULL;
+err_destroy_psv:
+ if (mlx5_core_destroy_psv(dev->mdev, mr->sig->psv_memory.psv_idx))
+ mlx5_ib_warn(dev, "failed to destroy mem psv %d\n",
+ mr->sig->psv_memory.psv_idx);
+ if (mlx5_core_destroy_psv(dev->mdev, mr->sig->psv_wire.psv_idx))
+ mlx5_ib_warn(dev, "failed to destroy wire psv %d\n",
+ mr->sig->psv_wire.psv_idx);
+err_free_sig:
+ kfree(mr->sig);
- /* create mem & wire PSVs */
- err = mlx5_core_create_psv(dev->mdev, to_mpd(pd)->pdn,
- 2, psv_index);
- if (err)
- goto err_free_sig;
+ return err;
+}
+
+static struct ib_mr *__mlx5_ib_alloc_mr(struct ib_pd *pd,
+ enum ib_mr_type mr_type, u32 max_num_sg,
+ u32 max_num_meta_sg)
+{
+ struct mlx5_ib_dev *dev = to_mdev(pd->device);
+ int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+ int ndescs = ALIGN(max_num_sg, 4);
+ struct mlx5_ib_mr *mr;
+ u32 *in;
+ int err;
- mr->access_mode = MLX5_MKC_ACCESS_MODE_KLMS;
- mr->sig->psv_memory.psv_idx = psv_index[0];
- mr->sig->psv_wire.psv_idx = psv_index[1];
+ mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+ if (!mr)
+ return ERR_PTR(-ENOMEM);
- mr->sig->sig_status_checked = true;
- mr->sig->sig_err_exists = false;
- /* Next UMR, Arm SIGERR */
- ++mr->sig->sigerr_count;
- } else {
+ in = kzalloc(inlen, GFP_KERNEL);
+ if (!in) {
+ err = -ENOMEM;
+ goto err_free;
+ }
+
+ mr->ibmr.device = pd->device;
+ mr->umem = NULL;
+
+ switch (mr_type) {
+ case IB_MR_TYPE_MEM_REG:
+ err = mlx5_alloc_mem_reg_descs(pd, mr, ndescs, in, inlen);
+ break;
+ case IB_MR_TYPE_SG_GAPS:
+ err = mlx5_alloc_sg_gaps_descs(pd, mr, ndescs, in, inlen);
+ break;
+ case IB_MR_TYPE_INTEGRITY:
+ err = mlx5_alloc_integrity_descs(pd, mr, max_num_sg,
+ max_num_meta_sg, in, inlen);
+ break;
+ default:
mlx5_ib_warn(dev, "Invalid mr type %d\n", mr_type);
err = -EINVAL;
- goto err_free_in;
}
- MLX5_SET(mkc, mkc, access_mode_1_0, mr->access_mode & 0x3);
- MLX5_SET(mkc, mkc, access_mode_4_2, (mr->access_mode >> 2) & 0x7);
- MLX5_SET(mkc, mkc, umr_en, 1);
-
- mr->ibmr.device = pd->device;
- err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, inlen);
if (err)
- goto err_destroy_psv;
+ goto err_free_in;
- mr->mmkey.type = MLX5_MKEY_MR;
- mr->ibmr.lkey = mr->mmkey.key;
- mr->ibmr.rkey = mr->mmkey.key;
- mr->umem = NULL;
kfree(in);
return &mr->ibmr;
-err_destroy_psv:
- if (mr->sig) {
- if (mlx5_core_destroy_psv(dev->mdev,
- mr->sig->psv_memory.psv_idx))
- mlx5_ib_warn(dev, "failed to destroy mem psv %d\n",
- mr->sig->psv_memory.psv_idx);
- if (mlx5_core_destroy_psv(dev->mdev,
- mr->sig->psv_wire.psv_idx))
- mlx5_ib_warn(dev, "failed to destroy wire psv %d\n",
- mr->sig->psv_wire.psv_idx);
- }
- mlx5_free_priv_descs(mr);
-err_free_sig:
- kfree(mr->sig);
err_free_in:
kfree(in);
err_free:
@@ -1758,6 +1884,19 @@ err_free:
return ERR_PTR(err);
}
+struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
+ u32 max_num_sg, struct ib_udata *udata)
+{
+ return __mlx5_ib_alloc_mr(pd, mr_type, max_num_sg, 0);
+}
+
+struct ib_mr *mlx5_ib_alloc_mr_integrity(struct ib_pd *pd,
+ u32 max_num_sg, u32 max_num_meta_sg)
+{
+ return __mlx5_ib_alloc_mr(pd, IB_MR_TYPE_INTEGRITY, max_num_sg,
+ max_num_meta_sg);
+}
+
struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
struct ib_udata *udata)
{
@@ -1887,16 +2026,53 @@ done:
}
static int
+mlx5_ib_map_pa_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
+ int data_sg_nents, unsigned int *data_sg_offset,
+ struct scatterlist *meta_sg, int meta_sg_nents,
+ unsigned int *meta_sg_offset)
+{
+ struct mlx5_ib_mr *mr = to_mmr(ibmr);
+ unsigned int sg_offset = 0;
+ int n = 0;
+
+ mr->meta_length = 0;
+ if (data_sg_nents == 1) {
+ n++;
+ mr->ndescs = 1;
+ if (data_sg_offset)
+ sg_offset = *data_sg_offset;
+ mr->data_length = sg_dma_len(data_sg) - sg_offset;
+ mr->data_iova = sg_dma_address(data_sg) + sg_offset;
+ if (meta_sg_nents == 1) {
+ n++;
+ mr->meta_ndescs = 1;
+ if (meta_sg_offset)
+ sg_offset = *meta_sg_offset;
+ else
+ sg_offset = 0;
+ mr->meta_length = sg_dma_len(meta_sg) - sg_offset;
+ mr->pi_iova = sg_dma_address(meta_sg) + sg_offset;
+ }
+ ibmr->length = mr->data_length + mr->meta_length;
+ }
+
+ return n;
+}
+
+static int
mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
struct scatterlist *sgl,
unsigned short sg_nents,
- unsigned int *sg_offset_p)
+ unsigned int *sg_offset_p,
+ struct scatterlist *meta_sgl,
+ unsigned short meta_sg_nents,
+ unsigned int *meta_sg_offset_p)
{
struct scatterlist *sg = sgl;
struct mlx5_klm *klms = mr->descs;
unsigned int sg_offset = sg_offset_p ? *sg_offset_p : 0;
u32 lkey = mr->ibmr.pd->local_dma_lkey;
- int i;
+ int i, j = 0;
mr->ibmr.iova = sg_dma_address(sg) + sg_offset;
mr->ibmr.length = 0;
@@ -1911,12 +2087,36 @@ mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
sg_offset = 0;
}
- mr->ndescs = i;
if (sg_offset_p)
*sg_offset_p = sg_offset;
- return i;
+ mr->ndescs = i;
+ mr->data_length = mr->ibmr.length;
+
+ if (meta_sg_nents) {
+ sg = meta_sgl;
+ sg_offset = meta_sg_offset_p ? *meta_sg_offset_p : 0;
+ for_each_sg(meta_sgl, sg, meta_sg_nents, j) {
+ if (unlikely(i + j >= mr->max_descs))
+ break;
+ klms[i + j].va = cpu_to_be64(sg_dma_address(sg) +
+ sg_offset);
+ klms[i + j].bcount = cpu_to_be32(sg_dma_len(sg) -
+ sg_offset);
+ klms[i + j].key = cpu_to_be32(lkey);
+ mr->ibmr.length += sg_dma_len(sg) - sg_offset;
+
+ sg_offset = 0;
+ }
+ if (meta_sg_offset_p)
+ *meta_sg_offset_p = sg_offset;
+
+ mr->meta_ndescs = j;
+ mr->meta_length = mr->ibmr.length - mr->data_length;
+ }
+
+ return i + j;
}
static int mlx5_set_page(struct ib_mr *ibmr, u64 addr)
@@ -1933,6 +2133,181 @@ static int mlx5_set_page(struct ib_mr *ibmr, u64 addr)
return 0;
}
+static int mlx5_set_page_pi(struct ib_mr *ibmr, u64 addr)
+{
+ struct mlx5_ib_mr *mr = to_mmr(ibmr);
+ __be64 *descs;
+
+ if (unlikely(mr->ndescs + mr->meta_ndescs == mr->max_descs))
+ return -ENOMEM;
+
+ descs = mr->descs;
+ descs[mr->ndescs + mr->meta_ndescs++] =
+ cpu_to_be64(addr | MLX5_EN_RD | MLX5_EN_WR);
+
+ return 0;
+}
+
+static int
+mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
+ int data_sg_nents, unsigned int *data_sg_offset,
+ struct scatterlist *meta_sg, int meta_sg_nents,
+ unsigned int *meta_sg_offset)
+{
+ struct mlx5_ib_mr *mr = to_mmr(ibmr);
+ struct mlx5_ib_mr *pi_mr = mr->mtt_mr;
+ int n;
+
+ pi_mr->ndescs = 0;
+ pi_mr->meta_ndescs = 0;
+ pi_mr->meta_length = 0;
+
+ ib_dma_sync_single_for_cpu(ibmr->device, pi_mr->desc_map,
+ pi_mr->desc_size * pi_mr->max_descs,
+ DMA_TO_DEVICE);
+
+ pi_mr->ibmr.page_size = ibmr->page_size;
+ n = ib_sg_to_pages(&pi_mr->ibmr, data_sg, data_sg_nents, data_sg_offset,
+ mlx5_set_page);
+ if (n != data_sg_nents)
+ return n;
+
+ pi_mr->data_iova = pi_mr->ibmr.iova;
+ pi_mr->data_length = pi_mr->ibmr.length;
+ pi_mr->ibmr.length = pi_mr->data_length;
+ ibmr->length = pi_mr->data_length;
+
+ if (meta_sg_nents) {
+ u64 page_mask = ~((u64)ibmr->page_size - 1);
+ u64 iova = pi_mr->data_iova;
+
+ n += ib_sg_to_pages(&pi_mr->ibmr, meta_sg, meta_sg_nents,
+ meta_sg_offset, mlx5_set_page_pi);
+
+ pi_mr->meta_length = pi_mr->ibmr.length;
+ /*
+ * PI address for the HW is the offset of the metadata address
+ * relative to the first data page address.
+ * It equals to first data page address + size of data pages +
+ * metadata offset at the first metadata page
+ */
+ pi_mr->pi_iova = (iova & page_mask) +
+ pi_mr->ndescs * ibmr->page_size +
+ (pi_mr->ibmr.iova & ~page_mask);
+ /*
+ * In order to use one MTT MR for data and metadata, we register
+ * also the gaps between the end of the data and the start of
+ * the metadata (the sig MR will verify that the HW will access
+ * to right addresses). This mapping is safe because we use
+ * internal mkey for the registration.
+ */
+ pi_mr->ibmr.length = pi_mr->pi_iova + pi_mr->meta_length - iova;
+ pi_mr->ibmr.iova = iova;
+ ibmr->length += pi_mr->meta_length;
+ }
+
+ ib_dma_sync_single_for_device(ibmr->device, pi_mr->desc_map,
+ pi_mr->desc_size * pi_mr->max_descs,
+ DMA_TO_DEVICE);
+
+ return n;
+}
+
+static int
+mlx5_ib_map_klm_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
+ int data_sg_nents, unsigned int *data_sg_offset,
+ struct scatterlist *meta_sg, int meta_sg_nents,
+ unsigned int *meta_sg_offset)
+{
+ struct mlx5_ib_mr *mr = to_mmr(ibmr);
+ struct mlx5_ib_mr *pi_mr = mr->klm_mr;
+ int n;
+
+ pi_mr->ndescs = 0;
+ pi_mr->meta_ndescs = 0;
+ pi_mr->meta_length = 0;
+
+ ib_dma_sync_single_for_cpu(ibmr->device, pi_mr->desc_map,
+ pi_mr->desc_size * pi_mr->max_descs,
+ DMA_TO_DEVICE);
+
+ n = mlx5_ib_sg_to_klms(pi_mr, data_sg, data_sg_nents, data_sg_offset,
+ meta_sg, meta_sg_nents, meta_sg_offset);
+
+ ib_dma_sync_single_for_device(ibmr->device, pi_mr->desc_map,
+ pi_mr->desc_size * pi_mr->max_descs,
+ DMA_TO_DEVICE);
+
+ /* This is zero-based memory region */
+ pi_mr->data_iova = 0;
+ pi_mr->ibmr.iova = 0;
+ pi_mr->pi_iova = pi_mr->data_length;
+ ibmr->length = pi_mr->ibmr.length;
+
+ return n;
+}
+
+int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
+ int data_sg_nents, unsigned int *data_sg_offset,
+ struct scatterlist *meta_sg, int meta_sg_nents,
+ unsigned int *meta_sg_offset)
+{
+ struct mlx5_ib_mr *mr = to_mmr(ibmr);
+ struct mlx5_ib_mr *pi_mr = NULL;
+ int n;
+
+ WARN_ON(ibmr->type != IB_MR_TYPE_INTEGRITY);
+
+ mr->ndescs = 0;
+ mr->data_length = 0;
+ mr->data_iova = 0;
+ mr->meta_ndescs = 0;
+ mr->pi_iova = 0;
+ /*
+ * As a performance optimization, if possible, there is no need to
+ * perform UMR operation to register the data/metadata buffers.
+ * First try to map the sg lists to PA descriptors with local_dma_lkey.
+ * Fallback to UMR only in case of a failure.
+ */
+ n = mlx5_ib_map_pa_mr_sg_pi(ibmr, data_sg, data_sg_nents,
+ data_sg_offset, meta_sg, meta_sg_nents,
+ meta_sg_offset);
+ if (n == data_sg_nents + meta_sg_nents)
+ goto out;
+ /*
+ * As a performance optimization, if possible, there is no need to map
+ * the sg lists to KLM descriptors. First try to map the sg lists to MTT
+ * descriptors and fallback to KLM only in case of a failure.
+ * It's more efficient for the HW to work with MTT descriptors
+ * (especially in high load).
+ * Use KLM (indirect access) only if it's mandatory.
+ */
+ pi_mr = mr->mtt_mr;
+ n = mlx5_ib_map_mtt_mr_sg_pi(ibmr, data_sg, data_sg_nents,
+ data_sg_offset, meta_sg, meta_sg_nents,
+ meta_sg_offset);
+ if (n == data_sg_nents + meta_sg_nents)
+ goto out;
+
+ pi_mr = mr->klm_mr;
+ n = mlx5_ib_map_klm_mr_sg_pi(ibmr, data_sg, data_sg_nents,
+ data_sg_offset, meta_sg, meta_sg_nents,
+ meta_sg_offset);
+ if (unlikely(n != data_sg_nents + meta_sg_nents))
+ return -ENOMEM;
+
+out:
+ /* This is zero-based memory region */
+ ibmr->iova = 0;
+ mr->pi_mr = pi_mr;
+ if (pi_mr)
+ ibmr->sig_attrs->meta_length = pi_mr->meta_length;
+ else
+ ibmr->sig_attrs->meta_length = mr->meta_length;
+
+ return 0;
+}
+
int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
unsigned int *sg_offset)
{
@@ -1946,7 +2321,8 @@ int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
DMA_TO_DEVICE);
if (mr->access_mode == MLX5_MKC_ACCESS_MODE_KLMS)
- n = mlx5_ib_sg_to_klms(mr, sg, sg_nents, sg_offset);
+ n = mlx5_ib_sg_to_klms(mr, sg, sg_nents, sg_offset, NULL, 0,
+ NULL);
else
n = ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset,
mlx5_set_page);
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 831c450b271a..5b642d81e617 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -150,7 +150,7 @@ static struct ib_umem_odp *odp_lookup(u64 start, u64 length,
if (!rb)
goto not_found;
odp = rb_entry(rb, struct ib_umem_odp, interval_tree.rb);
- if (ib_umem_start(&odp->umem) > start + length)
+ if (ib_umem_start(odp) > start + length)
goto not_found;
}
not_found:
@@ -200,7 +200,7 @@ void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset,
static void mr_leaf_free_action(struct work_struct *work)
{
struct ib_umem_odp *odp = container_of(work, struct ib_umem_odp, work);
- int idx = ib_umem_start(&odp->umem) >> MLX5_IMR_MTT_SHIFT;
+ int idx = ib_umem_start(odp) >> MLX5_IMR_MTT_SHIFT;
struct mlx5_ib_mr *mr = odp->private, *imr = mr->parent;
mr->parent = NULL;
@@ -224,7 +224,6 @@ void mlx5_ib_invalidate_range(struct ib_umem_odp *umem_odp, unsigned long start,
const u64 umr_block_mask = (MLX5_UMR_MTT_ALIGNMENT /
sizeof(struct mlx5_mtt)) - 1;
u64 idx = 0, blk_start_idx = 0;
- struct ib_umem *umem;
int in_block = 0;
u64 addr;
@@ -232,15 +231,14 @@ void mlx5_ib_invalidate_range(struct ib_umem_odp *umem_odp, unsigned long start,
pr_err("invalidation called on NULL umem or non-ODP umem\n");
return;
}
- umem = &umem_odp->umem;
mr = umem_odp->private;
if (!mr || !mr->ibmr.pd)
return;
- start = max_t(u64, ib_umem_start(umem), start);
- end = min_t(u64, ib_umem_end(umem), end);
+ start = max_t(u64, ib_umem_start(umem_odp), start);
+ end = min_t(u64, ib_umem_end(umem_odp), end);
/*
* Iteration one - zap the HW's MTTs. The notifiers_count ensures that
@@ -249,8 +247,8 @@ void mlx5_ib_invalidate_range(struct ib_umem_odp *umem_odp, unsigned long start,
* but they will write 0s as well, so no difference in the end result.
*/
- for (addr = start; addr < end; addr += BIT(umem->page_shift)) {
- idx = (addr - ib_umem_start(umem)) >> umem->page_shift;
+ for (addr = start; addr < end; addr += BIT(umem_odp->page_shift)) {
+ idx = (addr - ib_umem_start(umem_odp)) >> umem_odp->page_shift;
/*
* Strive to write the MTTs in chunks, but avoid overwriting
* non-existing MTTs. The huristic here can be improved to
@@ -544,13 +542,12 @@ static int mr_leaf_free(struct ib_umem_odp *umem_odp, u64 start, u64 end,
void *cookie)
{
struct mlx5_ib_mr *mr = umem_odp->private, *imr = cookie;
- struct ib_umem *umem = &umem_odp->umem;
if (mr->parent != imr)
return 0;
- ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem),
- ib_umem_end(umem));
+ ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem_odp),
+ ib_umem_end(umem_odp));
if (umem_odp->dying)
return 0;
@@ -602,9 +599,9 @@ static int pagefault_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
}
next_mr:
- size = min_t(size_t, bcnt, ib_umem_end(&odp->umem) - io_virt);
+ size = min_t(size_t, bcnt, ib_umem_end(odp) - io_virt);
- page_shift = mr->umem->page_shift;
+ page_shift = odp->page_shift;
page_mask = ~(BIT(page_shift) - 1);
start_idx = (io_virt - (mr->mmkey.iova & page_mask)) >> page_shift;
access_mask = ODP_READ_ALLOWED_BIT;
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 768c7e81f688..8b7a60ada92c 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -442,9 +442,9 @@ static int calc_send_wqe(struct ib_qp_init_attr *attr)
}
size += attr->cap.max_send_sge * sizeof(struct mlx5_wqe_data_seg);
- if (attr->create_flags & IB_QP_CREATE_SIGNATURE_EN &&
+ if (attr->create_flags & IB_QP_CREATE_INTEGRITY_EN &&
ALIGN(max_t(int, inl_size, size), MLX5_SEND_WQE_BB) < MLX5_SIG_WQE_SIZE)
- return MLX5_SIG_WQE_SIZE;
+ return MLX5_SIG_WQE_SIZE;
else
return ALIGN(max_t(int, inl_size, size), MLX5_SEND_WQE_BB);
}
@@ -496,9 +496,6 @@ static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
sizeof(struct mlx5_wqe_inline_seg);
attr->cap.max_inline_data = qp->max_inline_data;
- if (attr->create_flags & IB_QP_CREATE_SIGNATURE_EN)
- qp->signature_en = true;
-
wq_size = roundup_pow_of_two(attr->cap.max_send_wr * wqe_size);
qp->sq.wqe_cnt = wq_size / MLX5_SEND_WQE_BB;
if (qp->sq.wqe_cnt > (1 << MLX5_CAP_GEN(dev->mdev, log_max_qp_sz))) {
@@ -790,8 +787,7 @@ static void destroy_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd,
atomic_dec(&dev->delay_drop.rqs_cnt);
mlx5_ib_db_unmap_user(context, &rwq->db);
- if (rwq->umem)
- ib_umem_release(rwq->umem);
+ ib_umem_release(rwq->umem);
}
static int create_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd,
@@ -977,8 +973,7 @@ err_free:
kvfree(*in);
err_umem:
- if (ubuffer->umem)
- ib_umem_release(ubuffer->umem);
+ ib_umem_release(ubuffer->umem);
err_bfreg:
if (bfregn != MLX5_IB_INVALID_BFREG)
@@ -997,8 +992,7 @@ static void destroy_qp_user(struct mlx5_ib_dev *dev, struct ib_pd *pd,
ibucontext);
mlx5_ib_db_unmap_user(context, &qp->db);
- if (base->ubuffer.umem)
- ib_umem_release(base->ubuffer.umem);
+ ib_umem_release(base->ubuffer.umem);
/*
* Free only the BFREGs which are handled by the kernel.
@@ -1042,7 +1036,7 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
void *qpc;
int err;
- if (init_attr->create_flags & ~(IB_QP_CREATE_SIGNATURE_EN |
+ if (init_attr->create_flags & ~(IB_QP_CREATE_INTEGRITY_EN |
IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK |
IB_QP_CREATE_IPOIB_UD_LSO |
IB_QP_CREATE_NETIF_QP |
@@ -4170,15 +4164,13 @@ static __be64 sig_mkey_mask(void)
}
static void set_reg_umr_seg(struct mlx5_wqe_umr_ctrl_seg *umr,
- struct mlx5_ib_mr *mr, bool umr_inline)
+ struct mlx5_ib_mr *mr, u8 flags)
{
- int size = mr->ndescs * mr->desc_size;
+ int size = (mr->ndescs + mr->meta_ndescs) * mr->desc_size;
memset(umr, 0, sizeof(*umr));
- umr->flags = MLX5_UMR_CHECK_NOT_FREE;
- if (umr_inline)
- umr->flags |= MLX5_UMR_INLINE;
+ umr->flags = flags;
umr->xlt_octowords = cpu_to_be16(get_xlt_octo(size));
umr->mkey_mask = frwr_mkey_mask();
}
@@ -4305,7 +4297,7 @@ static void set_reg_mkey_seg(struct mlx5_mkey_seg *seg,
struct mlx5_ib_mr *mr,
u32 key, int access)
{
- int ndescs = ALIGN(mr->ndescs, 8) >> 1;
+ int ndescs = ALIGN(mr->ndescs + mr->meta_ndescs, 8) >> 1;
memset(seg, 0, sizeof(*seg));
@@ -4356,7 +4348,7 @@ static void set_reg_data_seg(struct mlx5_wqe_data_seg *dseg,
struct mlx5_ib_mr *mr,
struct mlx5_ib_pd *pd)
{
- int bcount = mr->desc_size * mr->ndescs;
+ int bcount = mr->desc_size * (mr->ndescs + mr->meta_ndescs);
dseg->addr = cpu_to_be64(mr->desc_map);
dseg->byte_count = cpu_to_be32(ALIGN(bcount, 64));
@@ -4549,23 +4541,37 @@ static int mlx5_set_bsf(struct ib_mr *sig_mr,
return 0;
}
-static int set_sig_data_segment(const struct ib_sig_handover_wr *wr,
- struct mlx5_ib_qp *qp, void **seg,
- int *size, void **cur_edge)
+static int set_sig_data_segment(const struct ib_send_wr *send_wr,
+ struct ib_mr *sig_mr,
+ struct ib_sig_attrs *sig_attrs,
+ struct mlx5_ib_qp *qp, void **seg, int *size,
+ void **cur_edge)
{
- struct ib_sig_attrs *sig_attrs = wr->sig_attrs;
- struct ib_mr *sig_mr = wr->sig_mr;
struct mlx5_bsf *bsf;
- u32 data_len = wr->wr.sg_list->length;
- u32 data_key = wr->wr.sg_list->lkey;
- u64 data_va = wr->wr.sg_list->addr;
+ u32 data_len;
+ u32 data_key;
+ u64 data_va;
+ u32 prot_len = 0;
+ u32 prot_key = 0;
+ u64 prot_va = 0;
+ bool prot = false;
int ret;
int wqe_size;
+ struct mlx5_ib_mr *mr = to_mmr(sig_mr);
+ struct mlx5_ib_mr *pi_mr = mr->pi_mr;
+
+ data_len = pi_mr->data_length;
+ data_key = pi_mr->ibmr.lkey;
+ data_va = pi_mr->data_iova;
+ if (pi_mr->meta_ndescs) {
+ prot_len = pi_mr->meta_length;
+ prot_key = pi_mr->ibmr.lkey;
+ prot_va = pi_mr->pi_iova;
+ prot = true;
+ }
- if (!wr->prot ||
- (data_key == wr->prot->lkey &&
- data_va == wr->prot->addr &&
- data_len == wr->prot->length)) {
+ if (!prot || (data_key == prot_key && data_va == prot_va &&
+ data_len == prot_len)) {
/**
* Source domain doesn't contain signature information
* or data and protection are interleaved in memory.
@@ -4599,8 +4605,6 @@ static int set_sig_data_segment(const struct ib_sig_handover_wr *wr,
struct mlx5_stride_block_ctrl_seg *sblock_ctrl;
struct mlx5_stride_block_entry *data_sentry;
struct mlx5_stride_block_entry *prot_sentry;
- u32 prot_key = wr->prot->lkey;
- u64 prot_va = wr->prot->addr;
u16 block_size = sig_attrs->mem.sig.dif.pi_interval;
int prot_size;
@@ -4650,17 +4654,15 @@ static int set_sig_data_segment(const struct ib_sig_handover_wr *wr,
}
static void set_sig_mkey_segment(struct mlx5_mkey_seg *seg,
- const struct ib_sig_handover_wr *wr, u32 size,
- u32 length, u32 pdn)
+ struct ib_mr *sig_mr, int access_flags,
+ u32 size, u32 length, u32 pdn)
{
- struct ib_mr *sig_mr = wr->sig_mr;
u32 sig_key = sig_mr->rkey;
u8 sigerr = to_mmr(sig_mr)->sig->sigerr_count & 1;
memset(seg, 0, sizeof(*seg));
- seg->flags = get_umr_flags(wr->access_flags) |
- MLX5_MKC_ACCESS_MODE_KLMS;
+ seg->flags = get_umr_flags(access_flags) | MLX5_MKC_ACCESS_MODE_KLMS;
seg->qpn_mkey7_0 = cpu_to_be32((sig_key & 0xff) | 0xffffff00);
seg->flags_pd = cpu_to_be32(MLX5_MKEY_REMOTE_INVAL | sigerr << 26 |
MLX5_MKEY_BSF_EN | pdn);
@@ -4680,49 +4682,50 @@ static void set_sig_umr_segment(struct mlx5_wqe_umr_ctrl_seg *umr,
umr->mkey_mask = sig_mkey_mask();
}
-
-static int set_sig_umr_wr(const struct ib_send_wr *send_wr,
- struct mlx5_ib_qp *qp, void **seg, int *size,
- void **cur_edge)
+static int set_pi_umr_wr(const struct ib_send_wr *send_wr,
+ struct mlx5_ib_qp *qp, void **seg, int *size,
+ void **cur_edge)
{
- const struct ib_sig_handover_wr *wr = sig_handover_wr(send_wr);
- struct mlx5_ib_mr *sig_mr = to_mmr(wr->sig_mr);
+ const struct ib_reg_wr *wr = reg_wr(send_wr);
+ struct mlx5_ib_mr *sig_mr = to_mmr(wr->mr);
+ struct mlx5_ib_mr *pi_mr = sig_mr->pi_mr;
+ struct ib_sig_attrs *sig_attrs = sig_mr->ibmr.sig_attrs;
u32 pdn = get_pd(qp)->pdn;
u32 xlt_size;
int region_len, ret;
- if (unlikely(wr->wr.num_sge != 1) ||
- unlikely(wr->access_flags & IB_ACCESS_REMOTE_ATOMIC) ||
- unlikely(!sig_mr->sig) || unlikely(!qp->signature_en) ||
+ if (unlikely(send_wr->num_sge != 0) ||
+ unlikely(wr->access & IB_ACCESS_REMOTE_ATOMIC) ||
+ unlikely(!sig_mr->sig) || unlikely(!qp->ibqp.integrity_en) ||
unlikely(!sig_mr->sig->sig_status_checked))
return -EINVAL;
/* length of the protected region, data + protection */
- region_len = wr->wr.sg_list->length;
- if (wr->prot &&
- (wr->prot->lkey != wr->wr.sg_list->lkey ||
- wr->prot->addr != wr->wr.sg_list->addr ||
- wr->prot->length != wr->wr.sg_list->length))
- region_len += wr->prot->length;
+ region_len = pi_mr->ibmr.length;
/**
* KLM octoword size - if protection was provided
* then we use strided block format (3 octowords),
* else we use single KLM (1 octoword)
**/
- xlt_size = wr->prot ? 0x30 : sizeof(struct mlx5_klm);
+ if (sig_attrs->mem.sig_type != IB_SIG_TYPE_NONE)
+ xlt_size = 0x30;
+ else
+ xlt_size = sizeof(struct mlx5_klm);
set_sig_umr_segment(*seg, xlt_size);
*seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
*size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
handle_post_send_edge(&qp->sq, seg, *size, cur_edge);
- set_sig_mkey_segment(*seg, wr, xlt_size, region_len, pdn);
+ set_sig_mkey_segment(*seg, wr->mr, wr->access, xlt_size, region_len,
+ pdn);
*seg += sizeof(struct mlx5_mkey_seg);
*size += sizeof(struct mlx5_mkey_seg) / 16;
handle_post_send_edge(&qp->sq, seg, *size, cur_edge);
- ret = set_sig_data_segment(wr, qp, seg, size, cur_edge);
+ ret = set_sig_data_segment(send_wr, wr->mr, sig_attrs, qp, seg, size,
+ cur_edge);
if (ret)
return ret;
@@ -4759,12 +4762,14 @@ static int set_psv_wr(struct ib_sig_domain *domain,
static int set_reg_wr(struct mlx5_ib_qp *qp,
const struct ib_reg_wr *wr,
- void **seg, int *size, void **cur_edge)
+ void **seg, int *size, void **cur_edge,
+ bool check_not_free)
{
struct mlx5_ib_mr *mr = to_mmr(wr->mr);
struct mlx5_ib_pd *pd = to_mpd(qp->ibqp.pd);
- size_t mr_list_size = mr->ndescs * mr->desc_size;
+ int mr_list_size = (mr->ndescs + mr->meta_ndescs) * mr->desc_size;
bool umr_inline = mr_list_size <= MLX5_IB_SQ_UMR_INLINE_THRESHOLD;
+ u8 flags = 0;
if (unlikely(wr->wr.send_flags & IB_SEND_INLINE)) {
mlx5_ib_warn(to_mdev(qp->ibqp.device),
@@ -4772,7 +4777,12 @@ static int set_reg_wr(struct mlx5_ib_qp *qp,
return -EINVAL;
}
- set_reg_umr_seg(*seg, mr, umr_inline);
+ if (check_not_free)
+ flags |= MLX5_UMR_CHECK_NOT_FREE;
+ if (umr_inline)
+ flags |= MLX5_UMR_INLINE;
+
+ set_reg_umr_seg(*seg, mr, flags);
*seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
*size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
handle_post_send_edge(&qp->sq, seg, *size, cur_edge);
@@ -4898,8 +4908,12 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
struct mlx5_wqe_ctrl_seg *ctrl = NULL; /* compiler warning */
struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
struct mlx5_core_dev *mdev = dev->mdev;
+ struct ib_reg_wr reg_pi_wr;
struct mlx5_ib_qp *qp;
struct mlx5_ib_mr *mr;
+ struct mlx5_ib_mr *pi_mr;
+ struct mlx5_ib_mr pa_pi_mr;
+ struct ib_sig_attrs *sig_attrs;
struct mlx5_wqe_xrc_seg *xrc;
struct mlx5_bf *bf;
void *cur_edge;
@@ -4953,7 +4967,8 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
goto out;
}
- if (wr->opcode == IB_WR_REG_MR) {
+ if (wr->opcode == IB_WR_REG_MR ||
+ wr->opcode == IB_WR_REG_MR_INTEGRITY) {
fence = dev->umr_fence;
next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
} else {
@@ -5003,7 +5018,7 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
qp->sq.wr_data[idx] = IB_WR_REG_MR;
ctrl->imm = cpu_to_be32(reg_wr(wr)->key);
err = set_reg_wr(qp, reg_wr(wr), &seg, &size,
- &cur_edge);
+ &cur_edge, true);
if (err) {
*bad_wr = wr;
goto out;
@@ -5011,26 +5026,82 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
num_sge = 0;
break;
- case IB_WR_REG_SIG_MR:
- qp->sq.wr_data[idx] = IB_WR_REG_SIG_MR;
- mr = to_mmr(sig_handover_wr(wr)->sig_mr);
-
+ case IB_WR_REG_MR_INTEGRITY:
+ qp->sq.wr_data[idx] = IB_WR_REG_MR_INTEGRITY;
+
+ mr = to_mmr(reg_wr(wr)->mr);
+ pi_mr = mr->pi_mr;
+
+ if (pi_mr) {
+ memset(&reg_pi_wr, 0,
+ sizeof(struct ib_reg_wr));
+
+ reg_pi_wr.mr = &pi_mr->ibmr;
+ reg_pi_wr.access = reg_wr(wr)->access;
+ reg_pi_wr.key = pi_mr->ibmr.rkey;
+
+ ctrl->imm = cpu_to_be32(reg_pi_wr.key);
+ /* UMR for data + prot registration */
+ err = set_reg_wr(qp, &reg_pi_wr, &seg,
+ &size, &cur_edge,
+ false);
+ if (err) {
+ *bad_wr = wr;
+ goto out;
+ }
+ finish_wqe(qp, ctrl, seg, size,
+ cur_edge, idx, wr->wr_id,
+ nreq, fence,
+ MLX5_OPCODE_UMR);
+
+ err = begin_wqe(qp, &seg, &ctrl, wr,
+ &idx, &size, &cur_edge,
+ nreq);
+ if (err) {
+ mlx5_ib_warn(dev, "\n");
+ err = -ENOMEM;
+ *bad_wr = wr;
+ goto out;
+ }
+ } else {
+ memset(&pa_pi_mr, 0,
+ sizeof(struct mlx5_ib_mr));
+ /* No UMR, use local_dma_lkey */
+ pa_pi_mr.ibmr.lkey =
+ mr->ibmr.pd->local_dma_lkey;
+
+ pa_pi_mr.ndescs = mr->ndescs;
+ pa_pi_mr.data_length = mr->data_length;
+ pa_pi_mr.data_iova = mr->data_iova;
+ if (mr->meta_ndescs) {
+ pa_pi_mr.meta_ndescs =
+ mr->meta_ndescs;
+ pa_pi_mr.meta_length =
+ mr->meta_length;
+ pa_pi_mr.pi_iova = mr->pi_iova;
+ }
+
+ pa_pi_mr.ibmr.length = mr->ibmr.length;
+ mr->pi_mr = &pa_pi_mr;
+ }
ctrl->imm = cpu_to_be32(mr->ibmr.rkey);
- err = set_sig_umr_wr(wr, qp, &seg, &size,
- &cur_edge);
+ /* UMR for sig MR */
+ err = set_pi_umr_wr(wr, qp, &seg, &size,
+ &cur_edge);
if (err) {
mlx5_ib_warn(dev, "\n");
*bad_wr = wr;
goto out;
}
-
finish_wqe(qp, ctrl, seg, size, cur_edge, idx,
wr->wr_id, nreq, fence,
MLX5_OPCODE_UMR);
+
/*
* SET_PSV WQEs are not signaled and solicited
* on error
*/
+ sig_attrs = mr->ibmr.sig_attrs;
err = __begin_wqe(qp, &seg, &ctrl, wr, &idx,
&size, &cur_edge, nreq, false,
true);
@@ -5040,19 +5111,18 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
*bad_wr = wr;
goto out;
}
-
- err = set_psv_wr(&sig_handover_wr(wr)->sig_attrs->mem,
- mr->sig->psv_memory.psv_idx, &seg,
- &size);
+ err = set_psv_wr(&sig_attrs->mem,
+ mr->sig->psv_memory.psv_idx,
+ &seg, &size);
if (err) {
mlx5_ib_warn(dev, "\n");
*bad_wr = wr;
goto out;
}
-
finish_wqe(qp, ctrl, seg, size, cur_edge, idx,
- wr->wr_id, nreq, fence,
+ wr->wr_id, nreq, next_fence,
MLX5_OPCODE_SET_PSV);
+
err = __begin_wqe(qp, &seg, &ctrl, wr, &idx,
&size, &cur_edge, nreq, false,
true);
@@ -5062,20 +5132,20 @@ static int _mlx5_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
*bad_wr = wr;
goto out;
}
-
- err = set_psv_wr(&sig_handover_wr(wr)->sig_attrs->wire,
- mr->sig->psv_wire.psv_idx, &seg,
- &size);
+ err = set_psv_wr(&sig_attrs->wire,
+ mr->sig->psv_wire.psv_idx,
+ &seg, &size);
if (err) {
mlx5_ib_warn(dev, "\n");
*bad_wr = wr;
goto out;
}
-
finish_wqe(qp, ctrl, seg, size, cur_edge, idx,
- wr->wr_id, nreq, fence,
+ wr->wr_id, nreq, next_fence,
MLX5_OPCODE_SET_PSV);
- qp->next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
+
+ qp->next_fence =
+ MLX5_FENCE_MODE_INITIATOR_SMALL;
num_sge = 0;
goto skip_psv;
@@ -6047,7 +6117,7 @@ err:
return ERR_PTR(err);
}
-int mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata)
+void mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata)
{
struct mlx5_ib_dev *dev = to_mdev(wq->device);
struct mlx5_ib_rwq *rwq = to_mrwq(wq);
@@ -6055,8 +6125,6 @@ int mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata)
mlx5_core_destroy_rq_tracked(dev->mdev, &rwq->core_qp);
destroy_user_rq(dev, wq->pd, rwq, udata);
kfree(rwq);
-
- return 0;
}
struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
Powered by cgit, Gruvboxed by Blaster4385.