[6/9] net/mlx5: add transfer attribute to matcher
Checks
Commit Message
In current implementation the DV steering supported only NIC steering.
This commit adds the transfer attribute in order to create a matcher
on the FDB tabels.
Signed-off-by: Ori Kam <orika@mellanox.com>
---
drivers/net/mlx5/mlx5_flow.c | 1 +
drivers/net/mlx5/mlx5_flow.h | 2 ++
drivers/net/mlx5/mlx5_flow_dv.c | 22 ++++++++++++++++++----
3 files changed, 21 insertions(+), 4 deletions(-)
Comments
On Sun, Apr 14, 2019 at 09:12:34PM +0000, Ori Kam wrote:
> In current implementation the DV steering supported only NIC steering.
> This commit adds the transfer attribute in order to create a matcher
> on the FDB tabels.
>
> Signed-off-by: Ori Kam <orika@mellanox.com>
> ---
> drivers/net/mlx5/mlx5_flow.c | 1 +
> drivers/net/mlx5/mlx5_flow.h | 2 ++
> drivers/net/mlx5/mlx5_flow_dv.c | 22 ++++++++++++++++++----
> 3 files changed, 21 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
> index 83abc14..9fd5096 100644
> --- a/drivers/net/mlx5/mlx5_flow.c
> +++ b/drivers/net/mlx5/mlx5_flow.c
> @@ -2095,6 +2095,7 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev *dev, int32_t priority,
> flow = rte_calloc(__func__, 1, flow_size, 0);
> flow->drv_type = flow_get_drv_type(dev, attr);
> flow->ingress = attr->ingress;
> + flow->transfer = attr->transfer;
> assert(flow->drv_type > MLX5_FLOW_TYPE_MIN &&
> flow->drv_type < MLX5_FLOW_TYPE_MAX);
> flow->queue = (void *)(flow + 1);
> diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
> index 85954c2..9d72024 100644
> --- a/drivers/net/mlx5/mlx5_flow.h
> +++ b/drivers/net/mlx5/mlx5_flow.h
> @@ -210,6 +210,7 @@ struct mlx5_flow_dv_matcher {
> uint16_t crc; /**< CRC of key. */
> uint16_t priority; /**< Priority of matcher. */
> uint8_t egress; /**< Egress matcher. */
> + uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
egress and transfer can be bit fields? Those come from rte_flow_attr but I don't
understand why it needs to be uint8_t individually.
> uint32_t group; /**< The matcher group. */
> struct mlx5_flow_dv_match_params mask; /**< Matcher mask. */
> };
> @@ -382,6 +383,7 @@ struct rte_flow {
> struct mlx5_fdir *fdir; /**< Pointer to associated FDIR if any. */
> uint8_t ingress; /**< 1 if the flow is ingress. */
> uint32_t group; /**< The group index. */
> + uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
Bit-field?
Just out of curiosity, flow->ingress vs matcher->egress, why?
rte_flow_attr has both ingress and egress because a flow can be applied for both
directions. But, in mlx5 PMD, it looks exclusive - !ingress == egress vice
versa. Then, I don't understand why one has ingress and the other one has
egress even wasting bits.
> };
>
> typedef int (*mlx5_flow_validate_t)(struct rte_eth_dev *dev,
> diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
> index e66ee34..b4ca9ca 100644
> --- a/drivers/net/mlx5/mlx5_flow_dv.c
> +++ b/drivers/net/mlx5/mlx5_flow_dv.c
> @@ -3203,6 +3203,8 @@ struct field_modify_info modify_tcp[] = {
> * Table id to use.
> * @param[in] egress
> * Direction of the table.
> + * @param[in] transfer
> + * E-Switch or Nic flow..
Redundant periods (..)
Nic -> NIC
Thanks,
Yongseok
> * @param[out] error
> * pointer to error structure.
> *
> @@ -3212,6 +3214,7 @@ struct field_modify_info modify_tcp[] = {
> static struct mlx5_flow_tbl_resource *
> flow_dv_tbl_resource_get(struct rte_eth_dev *dev,
> uint32_t table_id, uint8_t egress,
> + uint8_t transfer,
> struct rte_flow_error *error)
> {
> struct mlx5_priv *priv = dev->data->dev_private;
> @@ -3219,7 +3222,12 @@ struct field_modify_info modify_tcp[] = {
> struct mlx5_flow_tbl_resource *tbl;
>
> #ifdef HAVE_MLX5DV_DR
> - if (egress) {
> + if (transfer) {
> + tbl = &sh->fdb_tbl[table_id];
> + if (!tbl->obj)
> + tbl->obj = mlx5_glue->dr_create_flow_tbl
> + (sh->fdb_ns, table_id);
> + } else if (egress) {
> tbl = &sh->tx_tbl[table_id];
> if (!tbl->obj)
> tbl->obj = mlx5_glue->dr_create_flow_tbl
> @@ -3241,7 +3249,9 @@ struct field_modify_info modify_tcp[] = {
> #else
> (void)error;
> (void)tbl;
> - if (egress)
> + if (transfer)
> + return &sh->fdb_tbl[table_id];
> + else if (egress)
> return &sh->tx_tbl[table_id];
> else
> return &sh->rx_tbl[table_id];
> @@ -3306,6 +3316,7 @@ struct field_modify_info modify_tcp[] = {
> matcher->priority == cache_matcher->priority &&
> matcher->egress == cache_matcher->egress &&
> matcher->group == cache_matcher->group &&
> + matcher->transfer == cache_matcher->transfer &&
> !memcmp((const void *)matcher->mask.buf,
> (const void *)cache_matcher->mask.buf,
> cache_matcher->mask.size)) {
> @@ -3327,7 +3338,8 @@ struct field_modify_info modify_tcp[] = {
> RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
> "cannot allocate matcher memory");
> tbl = flow_dv_tbl_resource_get(dev, matcher->group * MLX5_GROUP_FACTOR,
> - matcher->egress, error);
> + matcher->egress, matcher->transfer,
> + error);
> if (!tbl) {
> rte_free(cache_matcher);
> return rte_flow_error_set(error, ENOMEM,
> @@ -3654,7 +3666,8 @@ struct field_modify_info modify_tcp[] = {
> jump_data = action->conf;
> tbl = flow_dv_tbl_resource_get(dev, jump_data->group *
> MLX5_GROUP_FACTOR,
> - attr->egress, error);
> + attr->egress,
> + attr->transfer, error);
> if (!tbl)
> return rte_flow_error_set
> (error, errno,
> @@ -3882,6 +3895,7 @@ struct field_modify_info modify_tcp[] = {
> matcher.priority);
> matcher.egress = attr->egress;
> matcher.group = attr->group;
> + matcher.transfer = attr->transfer;
> if (flow_dv_matcher_register(dev, &matcher, dev_flow, error))
> return -rte_errno;
> return 0;
> --
> 1.8.3.1
>
Hi Yongseok,
PSB
> -----Original Message-----
> From: Yongseok Koh
> Sent: Thursday, April 18, 2019 3:39 AM
> To: Ori Kam <orika@mellanox.com>
> Cc: Shahaf Shuler <shahafs@mellanox.com>; Matan Azrad
> <matan@mellanox.com>; Slava Ovsiienko <viacheslavo@mellanox.com>; Moti
> Haimovsky <motih@mellanox.com>; dev@dpdk.org
> Subject: Re: [PATCH 6/9] net/mlx5: add transfer attribute to matcher
>
> On Sun, Apr 14, 2019 at 09:12:34PM +0000, Ori Kam wrote:
> > In current implementation the DV steering supported only NIC steering.
> > This commit adds the transfer attribute in order to create a matcher
> > on the FDB tabels.
> >
> > Signed-off-by: Ori Kam <orika@mellanox.com>
> > ---
> > drivers/net/mlx5/mlx5_flow.c | 1 +
> > drivers/net/mlx5/mlx5_flow.h | 2 ++
> > drivers/net/mlx5/mlx5_flow_dv.c | 22 ++++++++++++++++++----
> > 3 files changed, 21 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
> > index 83abc14..9fd5096 100644
> > --- a/drivers/net/mlx5/mlx5_flow.c
> > +++ b/drivers/net/mlx5/mlx5_flow.c
> > @@ -2095,6 +2095,7 @@ uint32_t mlx5_flow_adjust_priority(struct
> rte_eth_dev *dev, int32_t priority,
> > flow = rte_calloc(__func__, 1, flow_size, 0);
> > flow->drv_type = flow_get_drv_type(dev, attr);
> > flow->ingress = attr->ingress;
> > + flow->transfer = attr->transfer;
> > assert(flow->drv_type > MLX5_FLOW_TYPE_MIN &&
> > flow->drv_type < MLX5_FLOW_TYPE_MAX);
> > flow->queue = (void *)(flow + 1);
> > diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
> > index 85954c2..9d72024 100644
> > --- a/drivers/net/mlx5/mlx5_flow.h
> > +++ b/drivers/net/mlx5/mlx5_flow.h
> > @@ -210,6 +210,7 @@ struct mlx5_flow_dv_matcher {
> > uint16_t crc; /**< CRC of key. */
> > uint16_t priority; /**< Priority of matcher. */
> > uint8_t egress; /**< Egress matcher. */
> > + uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
>
> egress and transfer can be bit fields? Those come from rte_flow_attr but I don't
> understand why it needs to be uint8_t individually.
>
You are correct but since the egress was already defined like this I didn't want to add my code differently.
I don't care to change the new code to bit mask and in later patch (after this release) modify also the egress;
> > uint32_t group; /**< The matcher group. */
> > struct mlx5_flow_dv_match_params mask; /**< Matcher mask. */
> > };
> > @@ -382,6 +383,7 @@ struct rte_flow {
> > struct mlx5_fdir *fdir; /**< Pointer to associated FDIR if any. */
> > uint8_t ingress; /**< 1 if the flow is ingress. */
> > uint32_t group; /**< The group index. */
> > + uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
>
> Bit-field?
>
Same as above comment to keep with the ingress variable.
> Just out of curiosity, flow->ingress vs matcher->egress, why?
> rte_flow_attr has both ingress and egress because a flow can be applied for
> both
> directions. But, in mlx5 PMD, it looks exclusive - !ingress == egress vice
> versa. Then, I don't understand why one has ingress and the other one has
> egress even wasting bits.
>
First I agree it is confusing. But it doesn't waste bits since in any case we must store a bit in each structure.
In any case I think I also should put it in my future commites.
> > };
> >
> > typedef int (*mlx5_flow_validate_t)(struct rte_eth_dev *dev,
> > diff --git a/drivers/net/mlx5/mlx5_flow_dv.c
> b/drivers/net/mlx5/mlx5_flow_dv.c
> > index e66ee34..b4ca9ca 100644
> > --- a/drivers/net/mlx5/mlx5_flow_dv.c
> > +++ b/drivers/net/mlx5/mlx5_flow_dv.c
> > @@ -3203,6 +3203,8 @@ struct field_modify_info modify_tcp[] = {
> > * Table id to use.
> > * @param[in] egress
> > * Direction of the table.
> > + * @param[in] transfer
> > + * E-Switch or Nic flow..
>
> Redundant periods (..)
> Nic -> NIC
>
Will fix.
>
> Thanks,
> Yongseok
>
> > * @param[out] error
> > * pointer to error structure.
> > *
> > @@ -3212,6 +3214,7 @@ struct field_modify_info modify_tcp[] = {
> > static struct mlx5_flow_tbl_resource *
> > flow_dv_tbl_resource_get(struct rte_eth_dev *dev,
> > uint32_t table_id, uint8_t egress,
> > + uint8_t transfer,
> > struct rte_flow_error *error)
> > {
> > struct mlx5_priv *priv = dev->data->dev_private;
> > @@ -3219,7 +3222,12 @@ struct field_modify_info modify_tcp[] = {
> > struct mlx5_flow_tbl_resource *tbl;
> >
> > #ifdef HAVE_MLX5DV_DR
> > - if (egress) {
> > + if (transfer) {
> > + tbl = &sh->fdb_tbl[table_id];
> > + if (!tbl->obj)
> > + tbl->obj = mlx5_glue->dr_create_flow_tbl
> > + (sh->fdb_ns, table_id);
> > + } else if (egress) {
> > tbl = &sh->tx_tbl[table_id];
> > if (!tbl->obj)
> > tbl->obj = mlx5_glue->dr_create_flow_tbl
> > @@ -3241,7 +3249,9 @@ struct field_modify_info modify_tcp[] = {
> > #else
> > (void)error;
> > (void)tbl;
> > - if (egress)
> > + if (transfer)
> > + return &sh->fdb_tbl[table_id];
> > + else if (egress)
> > return &sh->tx_tbl[table_id];
> > else
> > return &sh->rx_tbl[table_id];
> > @@ -3306,6 +3316,7 @@ struct field_modify_info modify_tcp[] = {
> > matcher->priority == cache_matcher->priority &&
> > matcher->egress == cache_matcher->egress &&
> > matcher->group == cache_matcher->group &&
> > + matcher->transfer == cache_matcher->transfer &&
> > !memcmp((const void *)matcher->mask.buf,
> > (const void *)cache_matcher->mask.buf,
> > cache_matcher->mask.size)) {
> > @@ -3327,7 +3338,8 @@ struct field_modify_info modify_tcp[] = {
> >
> RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
> > "cannot allocate matcher memory");
> > tbl = flow_dv_tbl_resource_get(dev, matcher->group *
> MLX5_GROUP_FACTOR,
> > - matcher->egress, error);
> > + matcher->egress, matcher->transfer,
> > + error);
> > if (!tbl) {
> > rte_free(cache_matcher);
> > return rte_flow_error_set(error, ENOMEM,
> > @@ -3654,7 +3666,8 @@ struct field_modify_info modify_tcp[] = {
> > jump_data = action->conf;
> > tbl = flow_dv_tbl_resource_get(dev, jump_data->group
> *
> > MLX5_GROUP_FACTOR,
> > - attr->egress, error);
> > + attr->egress,
> > + attr->transfer, error);
> > if (!tbl)
> > return rte_flow_error_set
> > (error, errno,
> > @@ -3882,6 +3895,7 @@ struct field_modify_info modify_tcp[] = {
> > matcher.priority);
> > matcher.egress = attr->egress;
> > matcher.group = attr->group;
> > + matcher.transfer = attr->transfer;
> > if (flow_dv_matcher_register(dev, &matcher, dev_flow, error))
> > return -rte_errno;
> > return 0;
> > --
> > 1.8.3.1
> >
@@ -2095,6 +2095,7 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev *dev, int32_t priority,
flow = rte_calloc(__func__, 1, flow_size, 0);
flow->drv_type = flow_get_drv_type(dev, attr);
flow->ingress = attr->ingress;
+ flow->transfer = attr->transfer;
assert(flow->drv_type > MLX5_FLOW_TYPE_MIN &&
flow->drv_type < MLX5_FLOW_TYPE_MAX);
flow->queue = (void *)(flow + 1);
@@ -210,6 +210,7 @@ struct mlx5_flow_dv_matcher {
uint16_t crc; /**< CRC of key. */
uint16_t priority; /**< Priority of matcher. */
uint8_t egress; /**< Egress matcher. */
+ uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
uint32_t group; /**< The matcher group. */
struct mlx5_flow_dv_match_params mask; /**< Matcher mask. */
};
@@ -382,6 +383,7 @@ struct rte_flow {
struct mlx5_fdir *fdir; /**< Pointer to associated FDIR if any. */
uint8_t ingress; /**< 1 if the flow is ingress. */
uint32_t group; /**< The group index. */
+ uint8_t transfer; /**< 1 if the flow is E-Switch flow. */
};
typedef int (*mlx5_flow_validate_t)(struct rte_eth_dev *dev,
@@ -3203,6 +3203,8 @@ struct field_modify_info modify_tcp[] = {
* Table id to use.
* @param[in] egress
* Direction of the table.
+ * @param[in] transfer
+ * E-Switch or Nic flow..
* @param[out] error
* pointer to error structure.
*
@@ -3212,6 +3214,7 @@ struct field_modify_info modify_tcp[] = {
static struct mlx5_flow_tbl_resource *
flow_dv_tbl_resource_get(struct rte_eth_dev *dev,
uint32_t table_id, uint8_t egress,
+ uint8_t transfer,
struct rte_flow_error *error)
{
struct mlx5_priv *priv = dev->data->dev_private;
@@ -3219,7 +3222,12 @@ struct field_modify_info modify_tcp[] = {
struct mlx5_flow_tbl_resource *tbl;
#ifdef HAVE_MLX5DV_DR
- if (egress) {
+ if (transfer) {
+ tbl = &sh->fdb_tbl[table_id];
+ if (!tbl->obj)
+ tbl->obj = mlx5_glue->dr_create_flow_tbl
+ (sh->fdb_ns, table_id);
+ } else if (egress) {
tbl = &sh->tx_tbl[table_id];
if (!tbl->obj)
tbl->obj = mlx5_glue->dr_create_flow_tbl
@@ -3241,7 +3249,9 @@ struct field_modify_info modify_tcp[] = {
#else
(void)error;
(void)tbl;
- if (egress)
+ if (transfer)
+ return &sh->fdb_tbl[table_id];
+ else if (egress)
return &sh->tx_tbl[table_id];
else
return &sh->rx_tbl[table_id];
@@ -3306,6 +3316,7 @@ struct field_modify_info modify_tcp[] = {
matcher->priority == cache_matcher->priority &&
matcher->egress == cache_matcher->egress &&
matcher->group == cache_matcher->group &&
+ matcher->transfer == cache_matcher->transfer &&
!memcmp((const void *)matcher->mask.buf,
(const void *)cache_matcher->mask.buf,
cache_matcher->mask.size)) {
@@ -3327,7 +3338,8 @@ struct field_modify_info modify_tcp[] = {
RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
"cannot allocate matcher memory");
tbl = flow_dv_tbl_resource_get(dev, matcher->group * MLX5_GROUP_FACTOR,
- matcher->egress, error);
+ matcher->egress, matcher->transfer,
+ error);
if (!tbl) {
rte_free(cache_matcher);
return rte_flow_error_set(error, ENOMEM,
@@ -3654,7 +3666,8 @@ struct field_modify_info modify_tcp[] = {
jump_data = action->conf;
tbl = flow_dv_tbl_resource_get(dev, jump_data->group *
MLX5_GROUP_FACTOR,
- attr->egress, error);
+ attr->egress,
+ attr->transfer, error);
if (!tbl)
return rte_flow_error_set
(error, errno,
@@ -3882,6 +3895,7 @@ struct field_modify_info modify_tcp[] = {
matcher.priority);
matcher.egress = attr->egress;
matcher.group = attr->group;
+ matcher.transfer = attr->transfer;
if (flow_dv_matcher_register(dev, &matcher, dev_flow, error))
return -rte_errno;
return 0;