Skip to content

Commit

Permalink
net/mlx5: Allocate individual capability
Browse files Browse the repository at this point in the history 
Currently mlx5_core_dev contains array of capabilities. It contains 19
valid capabilities of the device, 2 reserved entries and 12 holes.
Due to this for 14 unused entries, mlx5_core_dev allocates 14 * 8K = 112K
bytes of memory which is never used. Due to this mlx5_core_dev structure
size is 270Kbytes odd. This allocation further aligns to next power of 2
to 512Kbytes.

By skipping non-existent entries,
(a) 112Kbyte is saved,
(b) mlx5_core_dev reduces to 8KB with alignment
(c) 350KB saved in alignment

In future individual capability allocation can be used to skip its
allocation when such capability is disabled at the device level. This
patch prepares mlx5_core_dev to hold capability using a pointer instead
of inline array.

Signed-off-by: Parav Pandit <[email protected]>
Reviewed-by: Leon Romanovsky <[email protected]>
Reviewed-by: Shay Drory <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
  • Loading branch information
@paravmellanox
paravmellanox authored and Saeed Mahameed committed Aug 11, 2021
1 parent 5958a6f commit 48f02ee
Show file tree
Hide file tree
Showing 4 changed files with 104 additions and 40 deletions.
2 changes: 1 addition & 1 deletion drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2343,7 +2343,7 @@ static int create_leaf_prios(struct mlx5_flow_namespace *ns, int prio,

#define FLOW_TABLE_BIT_SZ 1
#define GET_FLOW_TABLE_CAP(dev, offset) \
((be32_to_cpu(*((__be32 *)(dev->caps.hca[MLX5_CAP_FLOW_TABLE].cur) + \
((be32_to_cpu(*((__be32 *)(dev->caps.hca[MLX5_CAP_FLOW_TABLE]->cur) + \
offset / 32)) >> \
(32 - FLOW_TABLE_BIT_SZ - (offset & 0x1f))) & FLOW_TABLE_BIT_SZ)
static bool has_required_caps(struct mlx5_core_dev *dev, struct node_caps *caps)
Expand Down
71 changes: 66 additions & 5 deletions drivers/net/ethernet/mellanox/mlx5/core/main.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -389,11 +389,11 @@ static int mlx5_core_get_caps_mode(struct mlx5_core_dev *dev,

switch (cap_mode) {
case HCA_CAP_OPMOD_GET_MAX:
memcpy(dev->caps.hca[cap_type].max, hca_caps,
memcpy(dev->caps.hca[cap_type]->max, hca_caps,
MLX5_UN_SZ_BYTES(hca_cap_union));
break;
case HCA_CAP_OPMOD_GET_CUR:
memcpy(dev->caps.hca[cap_type].cur, hca_caps,
memcpy(dev->caps.hca[cap_type]->cur, hca_caps,
MLX5_UN_SZ_BYTES(hca_cap_union));
break;
default:
Expand Down Expand Up @@ -469,7 +469,7 @@ static int handle_hca_cap_odp(struct mlx5_core_dev *dev, void *set_ctx)
return err;

set_hca_cap = MLX5_ADDR_OF(set_hca_cap_in, set_ctx, capability);
memcpy(set_hca_cap, dev->caps.hca[MLX5_CAP_ODP].cur,
memcpy(set_hca_cap, dev->caps.hca[MLX5_CAP_ODP]->cur,
MLX5_ST_SZ_BYTES(odp_cap));

#define ODP_CAP_SET_MAX(dev, field) \
Expand Down Expand Up @@ -514,7 +514,7 @@ static int handle_hca_cap(struct mlx5_core_dev *dev, void *set_ctx)

set_hca_cap = MLX5_ADDR_OF(set_hca_cap_in, set_ctx,
capability);
memcpy(set_hca_cap, dev->caps.hca[MLX5_CAP_GENERAL].cur,
memcpy(set_hca_cap, dev->caps.hca[MLX5_CAP_GENERAL]->cur,
MLX5_ST_SZ_BYTES(cmd_hca_cap));

mlx5_core_dbg(dev, "Current Pkey table size %d Setting new size %d\n",
Expand Down Expand Up @@ -596,7 +596,7 @@ static int handle_hca_cap_roce(struct mlx5_core_dev *dev, void *set_ctx)
return 0;

set_hca_cap = MLX5_ADDR_OF(set_hca_cap_in, set_ctx, capability);
memcpy(set_hca_cap, dev->caps.hca[MLX5_CAP_ROCE].cur,
memcpy(set_hca_cap, dev->caps.hca[MLX5_CAP_ROCE]->cur,
MLX5_ST_SZ_BYTES(roce_cap));
MLX5_SET(roce_cap, set_hca_cap, sw_r_roce_src_udp_port, 1);

Expand Down Expand Up @@ -1375,6 +1375,60 @@ void mlx5_unload_one(struct mlx5_core_dev *dev)
mutex_unlock(&dev->intf_state_mutex);
}

static const int types[] = {
MLX5_CAP_GENERAL,
MLX5_CAP_GENERAL_2,
MLX5_CAP_ETHERNET_OFFLOADS,
MLX5_CAP_IPOIB_ENHANCED_OFFLOADS,
MLX5_CAP_ODP,
MLX5_CAP_ATOMIC,
MLX5_CAP_ROCE,
MLX5_CAP_IPOIB_OFFLOADS,
MLX5_CAP_FLOW_TABLE,
MLX5_CAP_ESWITCH_FLOW_TABLE,
MLX5_CAP_ESWITCH,
MLX5_CAP_VECTOR_CALC,
MLX5_CAP_QOS,
MLX5_CAP_DEBUG,
MLX5_CAP_DEV_MEM,
MLX5_CAP_DEV_EVENT,
MLX5_CAP_TLS,
MLX5_CAP_VDPA_EMULATION,
MLX5_CAP_IPSEC,
};

static void mlx5_hca_caps_free(struct mlx5_core_dev *dev)
{
int type;
int i;

for (i = 0; i < ARRAY_SIZE(types); i++) {
type = types[i];
kfree(dev->caps.hca[type]);
}
}

static int mlx5_hca_caps_alloc(struct mlx5_core_dev *dev)
{
struct mlx5_hca_cap *cap;
int type;
int i;

for (i = 0; i < ARRAY_SIZE(types); i++) {
cap = kzalloc(sizeof(*cap), GFP_KERNEL);
if (!cap)
goto err;
type = types[i];
dev->caps.hca[type] = cap;
}

return 0;

err:
mlx5_hca_caps_free(dev);
return -ENOMEM;
}

int mlx5_mdev_init(struct mlx5_core_dev *dev, int profile_idx)
{
struct mlx5_priv *priv = &dev->priv;
Expand Down Expand Up @@ -1410,8 +1464,14 @@ int mlx5_mdev_init(struct mlx5_core_dev *dev, int profile_idx)
if (err)
goto err_adev_init;

err = mlx5_hca_caps_alloc(dev);
if (err)
goto err_hca_caps;

return 0;

err_hca_caps:
mlx5_adev_cleanup(dev);
err_adev_init:
mlx5_pagealloc_cleanup(dev);
err_pagealloc_init:
Expand All @@ -1430,6 +1490,7 @@ void mlx5_mdev_uninit(struct mlx5_core_dev *dev)
{
struct mlx5_priv *priv = &dev->priv;

mlx5_hca_caps_free(dev);
mlx5_adev_cleanup(dev);
mlx5_pagealloc_cleanup(dev);
mlx5_health_cleanup(dev);
Expand Down
69 changes: 36 additions & 33 deletions include/linux/mlx5/device.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1157,6 +1157,9 @@ enum mlx5_cap_mode {
HCA_CAP_OPMOD_GET_CUR = 1,
};

/* Any new cap addition must update mlx5_hca_caps_alloc() to allocate
* capability memory.
*/
enum mlx5_cap_type {
MLX5_CAP_GENERAL = 0,
MLX5_CAP_ETHERNET_OFFLOADS,
Expand Down Expand Up @@ -1213,55 +1216,55 @@ enum mlx5_qcam_feature_groups {

/* GET Dev Caps macros */
#define MLX5_CAP_GEN(mdev, cap) \
MLX5_GET(cmd_hca_cap, mdev->caps.hca[MLX5_CAP_GENERAL].cur, cap)
MLX5_GET(cmd_hca_cap, mdev->caps.hca[MLX5_CAP_GENERAL]->cur, cap)

#define MLX5_CAP_GEN_64(mdev, cap) \
MLX5_GET64(cmd_hca_cap, mdev->caps.hca[MLX5_CAP_GENERAL].cur, cap)
MLX5_GET64(cmd_hca_cap, mdev->caps.hca[MLX5_CAP_GENERAL]->cur, cap)

#define MLX5_CAP_GEN_MAX(mdev, cap) \
MLX5_GET(cmd_hca_cap, mdev->caps.hca[MLX5_CAP_GENERAL].max, cap)
MLX5_GET(cmd_hca_cap, mdev->caps.hca[MLX5_CAP_GENERAL]->max, cap)

#define MLX5_CAP_GEN_2(mdev, cap) \
MLX5_GET(cmd_hca_cap_2, mdev->caps.hca[MLX5_CAP_GENERAL_2].cur, cap)
MLX5_GET(cmd_hca_cap_2, mdev->caps.hca[MLX5_CAP_GENERAL_2]->cur, cap)

#define MLX5_CAP_GEN_2_64(mdev, cap) \
MLX5_GET64(cmd_hca_cap_2, mdev->caps.hca[MLX5_CAP_GENERAL_2].cur, cap)
MLX5_GET64(cmd_hca_cap_2, mdev->caps.hca[MLX5_CAP_GENERAL_2]->cur, cap)

#define MLX5_CAP_GEN_2_MAX(mdev, cap) \
MLX5_GET(cmd_hca_cap_2, mdev->caps.hca[MLX5_CAP_GENERAL_2].max, cap)
MLX5_GET(cmd_hca_cap_2, mdev->caps.hca[MLX5_CAP_GENERAL_2]->max, cap)

#define MLX5_CAP_ETH(mdev, cap) \
MLX5_GET(per_protocol_networking_offload_caps,\
mdev->caps.hca[MLX5_CAP_ETHERNET_OFFLOADS].cur, cap)
mdev->caps.hca[MLX5_CAP_ETHERNET_OFFLOADS]->cur, cap)

#define MLX5_CAP_ETH_MAX(mdev, cap) \
MLX5_GET(per_protocol_networking_offload_caps,\
mdev->caps.hca[MLX5_CAP_ETHERNET_OFFLOADS].max, cap)
mdev->caps.hca[MLX5_CAP_ETHERNET_OFFLOADS]->max, cap)

#define MLX5_CAP_IPOIB_ENHANCED(mdev, cap) \
MLX5_GET(per_protocol_networking_offload_caps,\
mdev->caps.hca[MLX5_CAP_IPOIB_ENHANCED_OFFLOADS].cur, cap)
mdev->caps.hca[MLX5_CAP_IPOIB_ENHANCED_OFFLOADS]->cur, cap)

#define MLX5_CAP_ROCE(mdev, cap) \
MLX5_GET(roce_cap, mdev->caps.hca[MLX5_CAP_ROCE].cur, cap)
MLX5_GET(roce_cap, mdev->caps.hca[MLX5_CAP_ROCE]->cur, cap)

#define MLX5_CAP_ROCE_MAX(mdev, cap) \
MLX5_GET(roce_cap, mdev->caps.hca[MLX5_CAP_ROCE].max, cap)
MLX5_GET(roce_cap, mdev->caps.hca[MLX5_CAP_ROCE]->max, cap)

#define MLX5_CAP_ATOMIC(mdev, cap) \
MLX5_GET(atomic_caps, mdev->caps.hca[MLX5_CAP_ATOMIC].cur, cap)
MLX5_GET(atomic_caps, mdev->caps.hca[MLX5_CAP_ATOMIC]->cur, cap)

#define MLX5_CAP_ATOMIC_MAX(mdev, cap) \
MLX5_GET(atomic_caps, mdev->caps.hca[MLX5_CAP_ATOMIC].max, cap)
MLX5_GET(atomic_caps, mdev->caps.hca[MLX5_CAP_ATOMIC]->max, cap)

#define MLX5_CAP_FLOWTABLE(mdev, cap) \
MLX5_GET(flow_table_nic_cap, mdev->caps.hca[MLX5_CAP_FLOW_TABLE].cur, cap)
MLX5_GET(flow_table_nic_cap, mdev->caps.hca[MLX5_CAP_FLOW_TABLE]->cur, cap)

#define MLX5_CAP64_FLOWTABLE(mdev, cap) \
MLX5_GET64(flow_table_nic_cap, (mdev)->caps.hca[MLX5_CAP_FLOW_TABLE].cur, cap)
MLX5_GET64(flow_table_nic_cap, (mdev)->caps.hca[MLX5_CAP_FLOW_TABLE]->cur, cap)

#define MLX5_CAP_FLOWTABLE_MAX(mdev, cap) \
MLX5_GET(flow_table_nic_cap, mdev->caps.hca[MLX5_CAP_FLOW_TABLE].max, cap)
MLX5_GET(flow_table_nic_cap, mdev->caps.hca[MLX5_CAP_FLOW_TABLE]->max, cap)

#define MLX5_CAP_FLOWTABLE_NIC_RX(mdev, cap) \
MLX5_CAP_FLOWTABLE(mdev, flow_table_properties_nic_receive.cap)
Expand Down Expand Up @@ -1301,11 +1304,11 @@ enum mlx5_qcam_feature_groups {

#define MLX5_CAP_ESW_FLOWTABLE(mdev, cap) \
MLX5_GET(flow_table_eswitch_cap, \
mdev->caps.hca[MLX5_CAP_ESWITCH_FLOW_TABLE].cur, cap)
mdev->caps.hca[MLX5_CAP_ESWITCH_FLOW_TABLE]->cur, cap)

#define MLX5_CAP_ESW_FLOWTABLE_MAX(mdev, cap) \
MLX5_GET(flow_table_eswitch_cap, \
mdev->caps.hca[MLX5_CAP_ESWITCH_FLOW_TABLE].max, cap)
mdev->caps.hca[MLX5_CAP_ESWITCH_FLOW_TABLE]->max, cap)

#define MLX5_CAP_ESW_FLOWTABLE_FDB(mdev, cap) \
MLX5_CAP_ESW_FLOWTABLE(mdev, flow_table_properties_nic_esw_fdb.cap)
Expand All @@ -1327,31 +1330,31 @@ enum mlx5_qcam_feature_groups {

#define MLX5_CAP_ESW(mdev, cap) \
MLX5_GET(e_switch_cap, \
mdev->caps.hca[MLX5_CAP_ESWITCH].cur, cap)
mdev->caps.hca[MLX5_CAP_ESWITCH]->cur, cap)

#define MLX5_CAP64_ESW_FLOWTABLE(mdev, cap) \
MLX5_GET64(flow_table_eswitch_cap, \
(mdev)->caps.hca[MLX5_CAP_ESWITCH_FLOW_TABLE].cur, cap)
(mdev)->caps.hca[MLX5_CAP_ESWITCH_FLOW_TABLE]->cur, cap)

#define MLX5_CAP_ESW_MAX(mdev, cap) \
MLX5_GET(e_switch_cap, \
mdev->caps.hca[MLX5_CAP_ESWITCH].max, cap)
mdev->caps.hca[MLX5_CAP_ESWITCH]->max, cap)

#define MLX5_CAP_ODP(mdev, cap)\
MLX5_GET(odp_cap, mdev->caps.hca[MLX5_CAP_ODP].cur, cap)
MLX5_GET(odp_cap, mdev->caps.hca[MLX5_CAP_ODP]->cur, cap)

#define MLX5_CAP_ODP_MAX(mdev, cap)\
MLX5_GET(odp_cap, mdev->caps.hca[MLX5_CAP_ODP].max, cap)
MLX5_GET(odp_cap, mdev->caps.hca[MLX5_CAP_ODP]->max, cap)

#define MLX5_CAP_VECTOR_CALC(mdev, cap) \
MLX5_GET(vector_calc_cap, \
mdev->caps.hca[MLX5_CAP_VECTOR_CALC].cur, cap)
mdev->caps.hca[MLX5_CAP_VECTOR_CALC]->cur, cap)

#define MLX5_CAP_QOS(mdev, cap)\
MLX5_GET(qos_cap, mdev->caps.hca[MLX5_CAP_QOS].cur, cap)
MLX5_GET(qos_cap, mdev->caps.hca[MLX5_CAP_QOS]->cur, cap)

#define MLX5_CAP_DEBUG(mdev, cap)\
MLX5_GET(debug_cap, mdev->caps.hca[MLX5_CAP_DEBUG].cur, cap)
MLX5_GET(debug_cap, mdev->caps.hca[MLX5_CAP_DEBUG]->cur, cap)

#define MLX5_CAP_PCAM_FEATURE(mdev, fld) \
MLX5_GET(pcam_reg, (mdev)->caps.pcam, feature_cap_mask.enhanced_features.fld)
Expand Down Expand Up @@ -1387,27 +1390,27 @@ enum mlx5_qcam_feature_groups {
MLX5_GET64(fpga_cap, (mdev)->caps.fpga, cap)

#define MLX5_CAP_DEV_MEM(mdev, cap)\
MLX5_GET(device_mem_cap, mdev->caps.hca[MLX5_CAP_DEV_MEM].cur, cap)
MLX5_GET(device_mem_cap, mdev->caps.hca[MLX5_CAP_DEV_MEM]->cur, cap)

#define MLX5_CAP64_DEV_MEM(mdev, cap)\
MLX5_GET64(device_mem_cap, mdev->caps.hca[MLX5_CAP_DEV_MEM].cur, cap)
MLX5_GET64(device_mem_cap, mdev->caps.hca[MLX5_CAP_DEV_MEM]->cur, cap)

#define MLX5_CAP_TLS(mdev, cap) \
MLX5_GET(tls_cap, (mdev)->caps.hca[MLX5_CAP_TLS].cur, cap)
MLX5_GET(tls_cap, (mdev)->caps.hca[MLX5_CAP_TLS]->cur, cap)

#define MLX5_CAP_DEV_EVENT(mdev, cap)\
MLX5_ADDR_OF(device_event_cap, (mdev)->caps.hca[MLX5_CAP_DEV_EVENT].cur, cap)
MLX5_ADDR_OF(device_event_cap, (mdev)->caps.hca[MLX5_CAP_DEV_EVENT]->cur, cap)

#define MLX5_CAP_DEV_VDPA_EMULATION(mdev, cap)\
MLX5_GET(virtio_emulation_cap, \
(mdev)->caps.hca[MLX5_CAP_VDPA_EMULATION].cur, cap)
(mdev)->caps.hca[MLX5_CAP_VDPA_EMULATION]->cur, cap)

#define MLX5_CAP64_DEV_VDPA_EMULATION(mdev, cap)\
MLX5_GET64(virtio_emulation_cap, \
(mdev)->caps.hca[MLX5_CAP_VDPA_EMULATION].cur, cap)
(mdev)->caps.hca[MLX5_CAP_VDPA_EMULATION]->cur, cap)

#define MLX5_CAP_IPSEC(mdev, cap)\
MLX5_GET(ipsec_cap, (mdev)->caps.hca[MLX5_CAP_IPSEC].cur, cap)
MLX5_GET(ipsec_cap, (mdev)->caps.hca[MLX5_CAP_IPSEC]->cur, cap)

enum {
MLX5_CMD_STAT_OK = 0x0,
Expand Down
2 changes: 1 addition & 1 deletion include/linux/mlx5/driver.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -745,7 +745,7 @@ struct mlx5_core_dev {
char board_id[MLX5_BOARD_ID_LEN];
struct mlx5_cmd cmd;
struct {
struct mlx5_hca_cap hca[MLX5_CAP_NUM];
struct mlx5_hca_cap *hca[MLX5_CAP_NUM];
u32 pcam[MLX5_ST_SZ_DW(pcam_reg)];
u32 mcam[MLX5_MCAM_REGS_NUM][MLX5_ST_SZ_DW(mcam_reg)];
u32 fpga[MLX5_ST_SZ_DW(fpga_cap)];
Expand Down

0 comments on commit 48f02ee

Please sign in to comment.