[net-next,v2] net: mvpp2: cls: Add Classification offload support
diff mbox series

Message ID 20190423075031.26074-1-maxime.chevallier@bootlin.com
State New, archived
Headers show
Series
  • [net-next,v2] net: mvpp2: cls: Add Classification offload support
Related show

Commit Message

Maxime Chevallier April 23, 2019, 7:50 a.m. UTC
This commit introduces basic classification offloading support for the
PPv2 controller.

The PPv2 classifier has many classification engines, for now we only use
the C2 TCAM match engine.

This engine allows to perform ternary lookups on 64 bits keys (called
Header Extracted Key), that are built by extracting fields from the packet
header and concatenating them. At most 4 fields can be extracted for a
single lookup.

This basic implementation allows to build the HEK from the following
fields :
 - L4 source and destination ports (for UDP and TCP)

More fields are to be added in the future.

Classification flows are added through the ethtool interface, using the
newly introduced flow_rule infrastructure as an internal rule
representation, allowing to more easily implement tc flower rules if
need be.

Flows are separated based on their flow-type (supported flow-types are
tcp4, udp4, tcp6, udp6, ip4, ip6 and ether), each type can contain up to
7 classification rules. Note that the most specific flow-type is always
used, meaning that a rule in the tcp4 flow will superseed a rule in the
ip4 flow matching the same key)

When inserting a rule in a given flow, the location given is relative to
the flow :

ethtool -N eth0 flow-type udp4 dst-port 1234 action 2 loc 0

ethtool -N eth0 flow-type tcp4 dst-port 1234 action 3 loc 0

However when removing a rule, the global location is to be used. This
location can be retrieved by using ethtool -n <interface>.

This commit only introduces support for the "steer to rxq" action.

Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
---
V2: - Re-arranged struct mvpp2_rfs_rule to minimize gaps on 64 bits systems,
    as suggested by David.

    - Fixed a smatch error indicating a possible out-of-bound array access
    when accessing the internal list of rules.

    - Other smatch warnings saying that some indices can't be negatived were
      not fixed on purpose, even though no path leads to these indicies being
      negative right now, future work could lead to cases where indices aren't
      valid.

 drivers/net/ethernet/marvell/mvpp2/mvpp2.h    |  50 ++
 .../net/ethernet/marvell/mvpp2/mvpp2_cls.c    | 458 +++++++++++++++++-
 .../net/ethernet/marvell/mvpp2/mvpp2_cls.h    |  56 ++-
 .../net/ethernet/marvell/mvpp2/mvpp2_main.c   |  24 +-
 4 files changed, 573 insertions(+), 15 deletions(-)

Comments

Saeed Mahameed April 23, 2019, 5:58 p.m. UTC | #1
On Tue, 2019-04-23 at 09:50 +0200, Maxime Chevallier wrote:
> This commit introduces basic classification offloading support for
> the
> PPv2 controller.
> 
> The PPv2 classifier has many classification engines, for now we only
> use
> the C2 TCAM match engine.
> 
> This engine allows to perform ternary lookups on 64 bits keys (called
> Header Extracted Key), that are built by extracting fields from the
> packet
> header and concatenating them. At most 4 fields can be extracted for
> a
> single lookup.
> 
> This basic implementation allows to build the HEK from the following
> fields :
>  - L4 source and destination ports (for UDP and TCP)
> 
> More fields are to be added in the future.
> 
> Classification flows are added through the ethtool interface, using
> the
> newly introduced flow_rule infrastructure as an internal rule
> representation, allowing to more easily implement tc flower rules if
> need be.
> 
> Flows are separated based on their flow-type (supported flow-types
> are
> tcp4, udp4, tcp6, udp6, ip4, ip6 and ether), each type can contain up
> to
> 7 classification rules. Note that the most specific flow-type is
> always
> used, meaning that a rule in the tcp4 flow will superseed a rule in
> the
> ip4 flow matching the same key)
> 
> When inserting a rule in a given flow, the location given is relative
> to
> the flow :
> 
> ethtool -N eth0 flow-type udp4 dst-port 1234 action 2 loc 0
> 
> ethtool -N eth0 flow-type tcp4 dst-port 1234 action 3 loc 0
> 
> However when removing a rule, the global location is to be used. This
> location can be retrieved by using ethtool -n <interface>.
> 

I am not sure what you mean by "the location given is relative to the
flow", it seems like the rule will end up in a different location than
the user intended, but looking at ethtool documentation it clearly says
that the location the user provides is an absolute rule id/location,
which will be used to delete this rule.

from "man ethtool":
loc N:
Specify the location/ID to insert the rule. This will overwrite any
rule present in that location and will not go through any of the rule
ordering process.

delete N
Deletes the RX classification rule with the given ID.

So the above example should result in one flow rule in your hardware.
but according the code below the calculated index in
mvpp2_ethtool_cls_rule_ins might end up different than the requested
location, which will confuse the user.


> This commit only introduces support for the "steer to rxq" action.
> 
> Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
> ---
> V2: - Re-arranged struct mvpp2_rfs_rule to minimize gaps on 64 bits
> systems,
>     as suggested by David.
> 
>     - Fixed a smatch error indicating a possible out-of-bound array
> access
>     when accessing the internal list of rules.
> 
>     - Other smatch warnings saying that some indices can't be
> negatived were
>       not fixed on purpose, even though no path leads to these
> indicies being
>       negative right now, future work could lead to cases where
> indices aren't
>       valid.
> 
>  drivers/net/ethernet/marvell/mvpp2/mvpp2.h    |  50 ++
>  .../net/ethernet/marvell/mvpp2/mvpp2_cls.c    | 458
> +++++++++++++++++-
>  .../net/ethernet/marvell/mvpp2/mvpp2_cls.h    |  56 ++-
>  .../net/ethernet/marvell/mvpp2/mvpp2_main.c   |  24 +-
>  4 files changed, 573 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
> b/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
> index 67cce2736806..dcb93afb6569 100644
> --- a/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
> +++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
> @@ -14,6 +14,7 @@
>  #include <linux/netdevice.h>
>  #include <linux/phy.h>
>  #include <linux/phylink.h>
> +#include <net/flow_offload.h>
>  
>  /* Fifo Registers */
>  #define MVPP2_RX_DATA_FIFO_SIZE_REG(port)	(0x00 + 4 * (port))
> @@ -126,6 +127,7 @@
>  #define MVPP22_CLS_C2_TCAM_DATA4		0x1b20
>  #define     MVPP22_CLS_C2_LU_TYPE(lu)		((lu) & 0x3f)
>  #define     MVPP22_CLS_C2_PORT_ID(port)		((port) << 8)
> +#define     MVPP22_CLS_C2_PORT_MASK		(0xff << 8)
>  #define MVPP22_CLS_C2_TCAM_INV			0x1b24
>  #define     MVPP22_CLS_C2_TCAM_INV_BIT		BIT(31)
>  #define MVPP22_CLS_C2_HIT_CTR			0x1b50
> @@ -615,6 +617,10 @@
>  #define MVPP2_BIT_IN_WORD(bit)		((bit) % 32)
>  
>  #define MVPP2_N_PRS_FLOWS		52
> +#define MVPP2_N_RFS_ENTRIES_PER_FLOW	7
> +
> +/* There are 7 supported high-level flows */
> +#define MVPP2_N_RFS_RULES		(MVPP2_N_RFS_ENTRIES_PER_FLOW *
> 7)
>  
>  /* RSS constants */
>  #define MVPP22_RSS_TABLE_ENTRIES	32
> @@ -812,6 +818,46 @@ struct mvpp2_queue_vector {
>  	struct cpumask *mask;
>  };
>  
> +/* Internal represention of a Flow Steering rule */
> +struct mvpp2_rfs_rule {
> +	/* Rule location inside the flow*/
> +	int loc;
> +
> +	/* Flow type, such as TCP_V4_FLOW, IP6_FLOW, etc. */
> +	int flow_type;
> +
> +	/* The lookup type used by this flow. This allows the
> classification
> +	 * engines to differentiate between entries that match the same
> +	 * parameters
> +	 */
> +	int lu_type;
> +
> +	/* Index of the entry handling this rule in the flow table */
> +	int flt_index;
> +
> +	/* Index of the C2 TCAM entry handling this rule */
> +	int c2_index;
> +
> +	/* Header fields that needs to be extracted to match this flow
> */
> +	u16 hek_fields;
> +
> +	/* CLS engine : only c2 is supported for now. */
> +	u8 engine;
> +
> +	/* TCAM key and mask for C2-based steering. These fields should
> be
> +	 * encapsulated in a union should we add more engines.
> +	 */
> +	u64 c2_tcam;
> +	u64 c2_tcam_mask;
> +
> +	struct flow_rule *flow;
> +};
> +
> +struct mvpp2_ethtool_fs {
> +	struct mvpp2_rfs_rule rule;
> +	struct ethtool_rxnfc rxnfc;
> +};
> +
>  struct mvpp2_port {
>  	u8 id;
>  
> @@ -883,6 +929,10 @@ struct mvpp2_port {
>  
>  	/* RSS indirection table */
>  	u32 indir[MVPP22_RSS_TABLE_ENTRIES];
> +
> +	/* List of steering rules active on that port */
> +	struct mvpp2_ethtool_fs *rfs_rules[MVPP2_N_RFS_RULES];
> +	int n_rfs_rules;
>  };
>  
>  /* The mvpp2_tx_desc and mvpp2_rx_desc structures describe the
> diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
> b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
> index 1087974d3b98..7457e2ab57eb 100644
> --- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
> +++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
> @@ -448,6 +448,17 @@ static void mvpp2_cls_flow_port_add(struct
> mvpp2_cls_flow_entry *fe,
>  	fe->data[0] |= MVPP2_CLS_FLOW_TBL0_PORT_ID(port);
>  }
>  
> +static void mvpp2_cls_flow_port_remove(struct mvpp2_cls_flow_entry
> *fe,
> +				       u32 port)
> +{
> +	fe->data[0] &= ~MVPP2_CLS_FLOW_TBL0_PORT_ID(port);
> +}
> +
> +static u8 mvpp2_cls_flow_pmap_get(struct mvpp2_cls_flow_entry *fe)
> +{
> +	return (fe->data[0] >> 4) & MVPP2_CLS_FLOW_TBL0_PORT_ID_MASK;
> +}
> +
>  static void mvpp2_cls_flow_lu_type_set(struct mvpp2_cls_flow_entry
> *fe,
>  				       u8 lu_type)
>  {
> @@ -455,6 +466,12 @@ static void mvpp2_cls_flow_lu_type_set(struct
> mvpp2_cls_flow_entry *fe,
>  	fe->data[1] |= MVPP2_CLS_FLOW_TBL1_LU_TYPE(lu_type);
>  }
>  
> +static u8 mvpp2_cls_flow_lu_type_get(struct mvpp2_cls_flow_entry
> *fe)
> +{
> +	return (fe->data[1] &
> +		MVPP2_CLS_FLOW_TBL1_LU_TYPE(MVPP2_CLS_LU_TYPE_MASK)) >>
> 3;
> +}
> +
>  /* Initialize the parser entry for the given flow */
>  static void mvpp2_cls_flow_prs_init(struct mvpp2 *priv,
>  				    const struct mvpp2_cls_flow *flow)
> @@ -539,6 +556,35 @@ void mvpp2_cls_c2_read(struct mvpp2 *priv, int
> index,
>  	c2->valid = !(val & MVPP22_CLS_C2_TCAM_INV_BIT);
>  }
>  
> +static int mvpp2_cls_c2_port_flow_get_type(int flow_type)
> +{
> +	switch (flow_type) {
> +	case TCP_V4_FLOW:
> +		return MVPP22_FLOW_TCP4;
> +	case TCP_V6_FLOW:
> +		return MVPP22_FLOW_TCP6;
> +	case UDP_V4_FLOW:
> +		return MVPP22_FLOW_UDP4;
> +	case UDP_V6_FLOW:
> +		return MVPP22_FLOW_UDP6;
> +	case IPV4_FLOW:
> +		return MVPP22_FLOW_IP4;
> +	case IPV6_FLOW:
> +		return MVPP22_FLOW_IP6;
> +	default:
> +		return -EOPNOTSUPP;
> +	}
> +}
> +
> +static int mvpp2_cls_c2_port_flow_index(struct mvpp2_port *port, int
> lu_type,
> +					int loc)
> +{
> +	if (loc >= MVPP22_CLS_C2_PORT_N_RFS)
> +		return -EINVAL;
> +
> +	return MVPP22_CLS_C2_RFS_LOC(port->id, lu_type, loc);
> +}
> +
>  /* Initialize the flow table entries for the given flow */
>  static void mvpp2_cls_flow_init(struct mvpp2 *priv,
>  				const struct mvpp2_cls_flow *flow)
> @@ -565,7 +611,7 @@ static void mvpp2_cls_flow_init(struct mvpp2
> *priv,
>  
>  	mvpp2_cls_flow_eng_set(&fe, MVPP22_CLS_ENGINE_C2);
>  	mvpp2_cls_flow_port_id_sel(&fe, true);
> -	mvpp2_cls_flow_lu_type_set(&fe, MVPP2_CLS_LU_ALL);
> +	mvpp2_cls_flow_lu_type_set(&fe, MVPP22_FLOW_ETHER);
>  
>  	/* Add all ports */
>  	for (i = 0; i < MVPP2_MAX_PORTS; i++)
> @@ -652,6 +698,26 @@ static int mvpp2_flow_set_hek_fields(struct
> mvpp2_cls_flow_entry *fe,
>  	return 0;
>  }
>  
> +/* Returns the size, in bits, of the corresponding HEK field */
> +static int mvpp2_cls_hek_field_size(u32 field)
> +{
> +	switch (field) {
> +	case MVPP22_CLS_HEK_OPT_MAC_DA:
> +		return 48;
> +	case MVPP22_CLS_HEK_OPT_IP4SA:
> +	case MVPP22_CLS_HEK_OPT_IP4DA:
> +		return 32;
> +	case MVPP22_CLS_HEK_OPT_IP6SA:
> +	case MVPP22_CLS_HEK_OPT_IP6DA:
> +		return 128;
> +	case MVPP22_CLS_HEK_OPT_L4SIP:
> +	case MVPP22_CLS_HEK_OPT_L4DIP:
> +		return 16;
> +	default:
> +		return -1;
> +	}
> +}
> +
>  const struct mvpp2_cls_flow *mvpp2_cls_flow_get(int flow)
>  {
>  	if (flow >= MVPP2_N_PRS_FLOWS)
> @@ -810,7 +876,7 @@ static void mvpp2_port_c2_cls_init(struct
> mvpp2_port *port)
>  
>  	/* Match on Lookup Type */
>  	c2.tcam[4] |=
> MVPP22_CLS_C2_TCAM_EN(MVPP22_CLS_C2_LU_TYPE(MVPP2_CLS_LU_TYPE_MASK));
> -	c2.tcam[4] |= MVPP22_CLS_C2_LU_TYPE(MVPP2_CLS_LU_ALL);
> +	c2.tcam[4] |= MVPP22_CLS_C2_LU_TYPE(MVPP22_FLOW_ETHER);
>  
>  	/* Update RSS status after matching this entry */
>  	c2.act = MVPP22_CLS_C2_ACT_RSS_EN(MVPP22_C2_UPD_LOCK);
> @@ -944,6 +1010,18 @@ void mvpp22_port_rss_disable(struct mvpp2_port
> *port)
>  	mvpp2_rss_port_c2_disable(port);
>  }
>  
> +static void mvpp22_port_c2_lookup_disable(struct mvpp2_port *port,
> int entry)
> +{
> +	struct mvpp2_cls_c2_entry c2;
> +
> +	mvpp2_cls_c2_read(port->priv, entry, &c2);
> +
> +	/* Clear the port map so that the entry doesn't match anymore
> */
> +	c2.tcam[4] &= ~(MVPP22_CLS_C2_PORT_ID(BIT(port->id)));
> +
> +	mvpp2_cls_c2_write(port->priv, &c2);
> +}
> +
>  /* Set CPU queue number for oversize packets */
>  void mvpp2_cls_oversize_rxq_set(struct mvpp2_port *port)
>  {
> @@ -960,6 +1038,382 @@ void mvpp2_cls_oversize_rxq_set(struct
> mvpp2_port *port)
>  	mvpp2_write(port->priv, MVPP2_CLS_SWFWD_PCTRL_REG, val);
>  }
>  
> +static int mvpp2_port_c2_tcam_rule_add(struct mvpp2_port *port,
> +				       struct mvpp2_rfs_rule *rule)
> +{
> +	struct flow_action_entry *act;
> +	struct mvpp2_cls_c2_entry c2;
> +	u8 qh, ql, pmap;
> +
> +	memset(&c2, 0, sizeof(c2));
> +
> +	c2.index = mvpp2_cls_c2_port_flow_index(port, rule->lu_type,
> rule->loc);
> +	if (c2.index < 0)
> +		return -EINVAL;
> +
> +	act = &rule->flow->action.entries[0];
> +
> +	rule->c2_index = c2.index;
> +
> +	c2.tcam[0] = (rule->c2_tcam & 0xffff) |
> +		     ((rule->c2_tcam_mask & 0xffff) << 16);
> +	c2.tcam[1] = ((rule->c2_tcam >> 16) & 0xffff) |
> +		     (((rule->c2_tcam_mask >> 16) & 0xffff) << 16);
> +	c2.tcam[2] = ((rule->c2_tcam >> 32) & 0xffff) |
> +		     (((rule->c2_tcam_mask >> 32) & 0xffff) << 16);
> +	c2.tcam[3] = ((rule->c2_tcam >> 48) & 0xffff) |
> +		     (((rule->c2_tcam_mask >> 48) & 0xffff) << 16);
> +
> +	pmap = BIT(port->id);
> +	c2.tcam[4] = MVPP22_CLS_C2_PORT_ID(pmap);
> +	c2.tcam[4] |=
> MVPP22_CLS_C2_TCAM_EN(MVPP22_CLS_C2_PORT_ID(pmap));
> +
> +	/* Match on Lookup Type */
> +	c2.tcam[4] |=
> MVPP22_CLS_C2_TCAM_EN(MVPP22_CLS_C2_LU_TYPE(MVPP2_CLS_LU_TYPE_MASK));
> +	c2.tcam[4] |= MVPP22_CLS_C2_LU_TYPE(rule->lu_type);
> +
> +	/* Mark packet as "forwarded to software", needed for RSS */
> +	c2.act |= MVPP22_CLS_C2_ACT_FWD(MVPP22_C2_FWD_SW_LOCK);
> +
> +	c2.act |= MVPP22_CLS_C2_ACT_QHIGH(MVPP22_C2_UPD_LOCK) |
> +		   MVPP22_CLS_C2_ACT_QLOW(MVPP22_C2_UPD_LOCK);
> +
> +	qh = ((act->queue.index + port->first_rxq) >> 3) &
> MVPP22_CLS_C2_ATTR0_QHIGH_MASK;
> +	ql = (act->queue.index + port->first_rxq) &
> MVPP22_CLS_C2_ATTR0_QLOW_MASK;
> +
> +	c2.attr[0] = MVPP22_CLS_C2_ATTR0_QHIGH(qh) |
> +		      MVPP22_CLS_C2_ATTR0_QLOW(ql);
> +
> +	c2.valid = true;
> +
> +	mvpp2_cls_c2_write(port->priv, &c2);
> +
> +	return 0;
> +}
> +
> +static int mvpp2_port_c2_rfs_rule_insert(struct mvpp2_port *port,
> +					 struct mvpp2_rfs_rule *rule)
> +{
> +	return mvpp2_port_c2_tcam_rule_add(port, rule);
> +}
> +
> +static int mvpp2_port_flow_table_find(struct mvpp2 *priv, int
> flow_id,
> +				      u8 engine, u8 lu_type, u16
> hek_fields)
> +{
> +	struct mvpp2_cls_flow_entry fe;
> +	u16 flow_hek_fields;
> +	int i, flow_engine;
> +	u8 flow_lu_type;
> +
> +	/* Iterate on all flow table entries for this ID */
> +	for (i = MVPP2_CLS_FLT_FIRST(flow_id);
> +	     i < MVPP2_CLS_FLT_LAST(flow_id); i++) {
> +		mvpp2_cls_flow_read(priv, i, &fe);
> +
> +		flow_hek_fields = mvpp2_flow_get_hek_fields(&fe);
> +		if (flow_hek_fields != hek_fields)
> +			continue;
> +
> +		flow_engine = mvpp2_cls_flow_eng_get(&fe);
> +		if (flow_engine != engine)
> +			continue;
> +
> +		flow_lu_type = mvpp2_cls_flow_lu_type_get(&fe);
> +		if (flow_lu_type != lu_type)
> +			continue;
> +
> +		return i;
> +	}
> +
> +	return -1;
> +}
> +
> +static int mvpp2_port_flow_table_find_free(struct mvpp2 *priv, int
> flow_id)
> +{
> +	struct mvpp2_cls_flow_entry fe;
> +	u8 pmap;
> +	int i;
> +
> +	memset(&fe, 0, sizeof(fe));
> +	for (i = 0; i < MVPP2_N_RFS_ENTRIES_PER_FLOW; i++) {
> +		mvpp2_cls_flow_read(priv, MVPP2_CLS_FLT_C2_RFS(flow_id,
> i),
> +				    &fe);
> +
> +		/* get pmap */
> +		pmap = mvpp2_cls_flow_pmap_get(&fe);
> +		if (!pmap)
> +			return MVPP2_CLS_FLT_C2_RFS(flow_id, i);
> +	}
> +
> +	return -1;
> +}
> +
> +/**
> + * It's possible that multiple different rfs rules use the same
> flow_table
> + * entry, if the rules use the same key and the same engine. In that
> case,
> + * we must make sure not to delete this flow_table entry when
> removing the
> + * rule, unless it's the last rule that uses this entry.
> + */
> +static bool mvpp2_port_flt_has_siblings(struct mvpp2_port *port,
> +					struct mvpp2_rfs_rule *rule)
> +{
> +	struct mvpp2_ethtool_fs *efs;
> +	int i;
> +
> +	for (i = 0; i < MVPP2_N_RFS_RULES; i++) {
> +		efs = port->rfs_rules[i];
> +		if (!efs)
> +			continue;
> +		/* Skip the entry we are trying to find the sibling of
> */
> +		if (efs->rule.loc == rule->loc)
> +			continue;
> +		if (efs->rule.flt_index == rule->flt_index)
> +			return true;
> +	}
> +
> +	return false;
> +}
> +
> +static int mvpp2_port_cls_rfs_rule_remove(struct mvpp2_port *port,
> +					  struct mvpp2_rfs_rule *rule)
> +{
> +	struct mvpp2_cls_flow_entry fe;
> +
> +	if (rule->flt_index >= 0 &&
> +	    !mvpp2_port_flt_has_siblings(port, rule)) {
> +		mvpp2_cls_flow_read(port->priv, rule->flt_index, &fe);
> +		mvpp2_cls_flow_port_remove(&fe, BIT(port->id));
> +		mvpp2_cls_flow_write(port->priv, &fe);
> +	}
> +
> +	if (rule->c2_index >= 0)
> +		mvpp22_port_c2_lookup_disable(port, rule->c2_index);
> +
> +	return 0;
> +}
> +
> +static int mvpp2_port_flt_rfs_rule_insert(struct mvpp2_port *port,
> +					  struct mvpp2_rfs_rule *rule)
> +{
> +	const struct mvpp2_cls_flow *flow;
> +	struct mvpp2 *priv = port->priv;
> +	struct mvpp2_cls_flow_entry fe;
> +	int index, ret, i;
> +
> +	if (rule->engine != MVPP22_CLS_ENGINE_C2)
> +		return -EOPNOTSUPP;
> +
> +	ret = mvpp2_port_c2_rfs_rule_insert(port, rule);
> +	if (ret)
> +		return ret;
> +
> +	for_each_cls_flow_id_with_type(i, rule->flow_type) {
> +		flow = mvpp2_cls_flow_get(i);
> +		if (!flow)
> +			return 0;
> +
> +		/* Don't perform steering on HEK fields that aren't
> relevant
> +		 * for this flow, e.g. steering on VLAN tag when in an
> untagged
> +		 * flow.
> +		 */
> +		if ((rule->hek_fields | flow->supported_hash_opts) !=
> +		    flow->supported_hash_opts)
> +			return 0;
> +
> +		index = mvpp2_port_flow_table_find(priv, flow->flow_id,
> +						   rule->engine, rule-
> >lu_type,
> +						   rule->hek_fields);
> +
> +		/* If there already is a flow entry that matches our
> +		 * parameters
> +		 */
> +		if (index >= 0) {
> +			rule->flt_index = index;
> +			mvpp2_cls_flow_read(priv, index, &fe);
> +			mvpp2_cls_flow_port_add(&fe, BIT(port->id));
> +			mvpp2_cls_flow_write(priv, &fe);
> +			continue;
> +		} else {
> +			index = mvpp2_port_flow_table_find_free(priv,
> +								flow-
> >flow_id);
> +			if (index < 0)
> +				return -EOPNOTSUPP;
> +		}
> +
> +		/* Create entry */
> +		rule->flt_index = index;
> +		mvpp2_cls_flow_read(priv, index, &fe);
> +
> +		mvpp2_cls_flow_eng_set(&fe, rule->engine);
> +		mvpp2_cls_flow_port_id_sel(&fe, true);
> +		mvpp2_flow_set_hek_fields(&fe, rule->hek_fields);
> +		mvpp2_cls_flow_lu_type_set(&fe, rule->lu_type);
> +		mvpp2_cls_flow_port_add(&fe, 0xf);
> +
> +		mvpp2_cls_flow_write(priv, &fe);
> +	}
> +
> +	return 0;
> +}
> +
> +static int mvpp2_cls_c2_build_match(struct mvpp2_rfs_rule *rule)
> +{
> +	struct flow_rule *flow = rule->flow;
> +	struct flow_action_entry *act;
> +	int offs = 64;
> +
> +	act = &flow->action.entries[0];
> +
> +	if (flow_rule_match_key(flow, FLOW_DISSECTOR_KEY_PORTS)) {
> +		struct flow_match_ports match;
> +
> +		flow_rule_match_ports(flow, &match);
> +		if (match.mask->src) {
> +			rule->hek_fields |= MVPP22_CLS_HEK_OPT_L4SIP;
> +			offs -=
> mvpp2_cls_hek_field_size(MVPP22_CLS_HEK_OPT_L4SIP);
> +
> +			rule->c2_tcam |= ((u64)ntohs(match.key->src))
> << offs;
> +			rule->c2_tcam_mask |= ((u64)ntohs(match.mask-
> >src)) << offs;
> +		}
> +
> +		if (match.mask->dst) {
> +			rule->hek_fields |= MVPP22_CLS_HEK_OPT_L4DIP;
> +			offs -=
> mvpp2_cls_hek_field_size(MVPP22_CLS_HEK_OPT_L4DIP);
> +
> +			rule->c2_tcam |= ((u64)ntohs(match.key->dst))
> << offs;
> +			rule->c2_tcam_mask |= ((u64)ntohs(match.mask-
> >dst)) << offs;
> +		}
> +	}
> +
> +	if (hweight16(rule->hek_fields) > MVPP2_FLOW_N_FIELDS)
> +		return -EOPNOTSUPP;
> +
> +	return 0;
> +}
> +
> +static int mvpp2_cls_rfs_parse_rule(struct mvpp2_rfs_rule *rule)
> +{
> +	struct flow_rule *flow = rule->flow;
> +	struct flow_action_entry *act;
> +
> +	act = &flow->action.entries[0];
> +	if (act->id != FLOW_ACTION_QUEUE)
> +		return -EOPNOTSUPP;
> +
> +	/* For now, only use the C2 engine which has a HEK size limited
> to 64
> +	 * bits for TCAM matching.
> +	 */
> +	rule->engine = MVPP22_CLS_ENGINE_C2;
> +
> +	if (mvpp2_cls_c2_build_match(rule))
> +		return -EINVAL;
> +
> +	return 0;
> +}
> +
> +int mvpp2_ethtool_cls_rule_get(struct mvpp2_port *port,
> +			       struct ethtool_rxnfc *rxnfc)
> +{
> +	struct mvpp2_ethtool_fs *efs;
> +
> +	if (rxnfc->fs.location >= MVPP2_N_RFS_RULES)
> +		return -EINVAL;
> +
> +	efs = port->rfs_rules[rxnfc->fs.location];
> +	if (!efs)
> +		return -ENOENT;
> +
> +	memcpy(rxnfc, &efs->rxnfc, sizeof(efs->rxnfc));
> +
> +	return 0;
> +}
> +
> +int mvpp2_ethtool_cls_rule_ins(struct mvpp2_port *port,
> +			       struct ethtool_rxnfc *info)
> +{
> +	struct ethtool_rx_flow_spec_input input = {};
> +	struct ethtool_rx_flow_rule *ethtool_rule;
> +	struct mvpp2_ethtool_fs *efs, *old_efs;
> +	int ret = 0, index;
> +
> +	if (info->fs.location >= MVPP22_CLS_C2_PORT_N_RFS ||
> +	    info->fs.location < 0)
> +		return -EINVAL;
> +
> +	efs = kzalloc(sizeof(*efs), GFP_KERNEL);
> +	if (!efs)
> +		return -ENOMEM;
> +
> +	input.fs = &info->fs;
> +
> +	ethtool_rule = ethtool_rx_flow_rule_create(&input);
> +	if (IS_ERR(ethtool_rule)) {
> +		ret = PTR_ERR(ethtool_rule);
> +		goto clean_rule;
> +	}
> +
> +	efs->rule.flow = ethtool_rule->rule;
> +	efs->rule.flow_type = info->fs.flow_type & ~(FLOW_EXT |
> FLOW_MAC_EXT |
> +						     FLOW_RSS);
> +	efs->rule.lu_type = mvpp2_cls_c2_port_flow_get_type(efs-
> >rule.flow_type);
> +
> +	ret = mvpp2_cls_rfs_parse_rule(&efs->rule);
> +	if (ret)
> +		goto clean_eth_rule;
> +
> +	efs->rule.loc = info->fs.location;
> +
> +	index = (efs->rule.lu_type - 1) * MVPP22_CLS_C2_PORT_N_FLOWS +
> +		efs->rule.loc;
> +
> +	/* Replace an already existing rule */
> +	if (port->rfs_rules[index]) {
> +		old_efs = port->rfs_rules[index];
> +		ret = mvpp2_port_cls_rfs_rule_remove(port, &old_efs-
> >rule);
> +		if (ret)
> +			goto clean_eth_rule;
> +		kfree(old_efs);
> +	}
> +
> +	ret = mvpp2_port_flt_rfs_rule_insert(port, &efs->rule);
> +	if (ret)
> +		goto clean_eth_rule;
> +
> +	info->fs.location = index;
> +
> +	memcpy(&efs->rxnfc, info, sizeof(*info));
> +	port->rfs_rules[index] = efs;
> +	port->n_rfs_rules++;
> +
> +clean_eth_rule:
> +	ethtool_rx_flow_rule_destroy(ethtool_rule);
> +clean_rule:
> +	kfree(efs);
> +	return ret;
> +}
> +
> +int mvpp2_ethtool_cls_rule_del(struct mvpp2_port *port,
> +			       struct ethtool_rxnfc *info)
> +{
> +	struct mvpp2_ethtool_fs *efs;
> +	int ret;
> +
> +	efs = port->rfs_rules[info->fs.location];
> +	if (!efs)
> +		return -EINVAL;
> +
> +	/* Remove the rule from the engines. */
> +	ret = mvpp2_port_cls_rfs_rule_remove(port, &efs->rule);
> +	if (ret)
> +		return ret;
> +
> +	port->n_rfs_rules--;
> +	port->rfs_rules[info->fs.location] = NULL;
> +	kfree(efs);
> +
> +	return 0;
> +}
> +
>  static inline u32 mvpp22_rxfh_indir(struct mvpp2_port *port, u32
> rxq)
>  {
>  	int nrxqs, cpu, cpus = num_possible_cpus();
> diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
> b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
> index 96304ffc5d49..1d91d5d3b0e1 100644
> --- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
> +++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
> @@ -107,15 +107,43 @@ struct mvpp2_cls_c2_entry {
>  	u8 valid;
>  };
>  
> +enum mvpp2_lu_type {
> +	/* LU Type 0 is the default value, it shouldn't be used, to
> avoid
> +	 * spurious matches.
> +	 */
> +	MVPP22_FLOW_TCP4 = 1,
> +	MVPP22_FLOW_TCP6,
> +	MVPP22_FLOW_UDP4,
> +	MVPP22_FLOW_UDP6,
> +	MVPP22_FLOW_IP4,
> +	MVPP22_FLOW_IP6,
> +	MVPP22_FLOW_ETHER,
> +	MVPP22_FLOW_LAST = MVPP22_FLOW_ETHER,
> +};
> +
>  /* Classifier C2 engine entries */
>  #define MVPP22_CLS_C2_N_ENTRIES		256
>  
>  /* Number of per-port dedicated entries in the C2 TCAM */
> -#define MVPP22_CLS_C2_PORT_RANGE	8
> +#define MVPP22_CLS_C2_PORT_FLOW_RANGE	8
> +#define MVPP22_CLS_C2_PORT_N_RFS	MVPP2_N_RFS_ENTRIES_PER_FLOW
> +
> +/* We have one C2 range per supported flow type : ip4, i6, tcp4,
> tcp6, udp4, udp6 */
> +#define MVPP22_CLS_C2_PORT_N_FLOWS	(MVPP22_FLOW_LAST)
> +
> +/* Each port has oen range per flow type + one entry controling the
> global RSS
> + * setting and the default rx queue
> + */
> +#define MVPP22_CLS_C2_PORT_RANGE	(MVPP22_CLS_C2_PORT_FLOW_RANGE
> * \
> +					 MVPP22_CLS_C2_PORT_N_FLOWS +
> 1)
> +
> +#define MVPP22_CLS_C2_PORT_FIRST(p)	((p) *
> MVPP22_CLS_C2_PORT_RANGE)
> +#define MVPP22_CLS_C2_RSS_ENTRY(p)	(MVPP22_CLS_C2_PORT_FIRST((p) +
> 1) - 1)
>  
> -#define MVPP22_CLS_C2_PORT_FIRST(p)	(MVPP22_CLS_C2_N_ENTRIES - \
> -					((p) *
> MVPP22_CLS_C2_PORT_RANGE))
> -#define MVPP22_CLS_C2_RSS_ENTRY(p)	(MVPP22_CLS_C2_PORT_FIRST(p) -
> 1)
> +#define MVPP22_CLS_C2_PORT_FLOW_FIRST(p, f)	(MVPP22_CLS_C2_PORT_FIR
> ST(p) + \
> +						 ((f) - 1) *
> MVPP22_CLS_C2_PORT_FLOW_RANGE)
> +
> +#define MVPP22_CLS_C2_RFS_LOC(p, f, loc)	(MVPP22_CLS_C2_PORT_FLO
> W_FIRST(p, f) + (loc))
>  
>  /* Packet flow ID */
>  enum mvpp2_prs_flow {
> @@ -145,10 +173,6 @@ enum mvpp2_prs_flow {
>  	MVPP2_FL_LAST,
>  };
>  
> -enum mvpp2_cls_lu_type {
> -	MVPP2_CLS_LU_ALL = 0,
> -};
> -
>  /* LU Type defined for all engines, and specified in the flow table
> */
>  #define MVPP2_CLS_LU_TYPE_MASK			0x3f
>  
> @@ -168,11 +192,12 @@ struct mvpp2_cls_flow {
>  	struct mvpp2_prs_result_info prs_ri;
>  };
>  
> -#define MVPP2_CLS_FLT_ENTRIES_PER_FLOW		(MVPP2_MAX_PORT
> S + 1)
> +#define MVPP2_CLS_FLT_ENTRIES_PER_FLOW		(MVPP2_MAX_PORT
> S + 1 + MVPP2_N_RFS_ENTRIES_PER_FLOW)
>  #define MVPP2_CLS_FLT_FIRST(id)			(((id) -
> MVPP2_FL_START) * \
>  						 MVPP2_CLS_FLT_ENTRIES_
> PER_FLOW)
> -#define MVPP2_CLS_FLT_C2_RSS_ENTRY(id)		(MVPP2_CLS_FLT_
> FIRST(id))
> -#define MVPP2_CLS_FLT_HASH_ENTRY(port, id)	(MVPP2_CLS_FLT_C2_RSS_E
> NTRY(id) + (port) + 1)
> +#define MVPP2_CLS_FLT_C2_RFS(id, rfs_n)		(MVPP2_CLS_FLT_
> FIRST(id) + (rfs_n))
> +#define MVPP2_CLS_FLT_C2_RSS_ENTRY(id)		(MVPP2_CLS_FLT_
> C2_RFS(id, MVPP2_N_RFS_ENTRIES_PER_FLOW))
> +#define MVPP2_CLS_FLT_HASH_ENTRY(port, id)	(MVPP2_CLS_FLT_C2_RSS_E
> NTRY(id) + 1 + (port))
>  #define MVPP2_CLS_FLT_LAST(id)			(MVPP2_CLS_FLT_
> FIRST(id) + \
>  						 MVPP2_CLS_FLT_ENTRIES_
> PER_FLOW - 1)
>  
> @@ -246,4 +271,13 @@ u32 mvpp2_cls_c2_hit_count(struct mvpp2 *priv,
> int c2_index);
>  void mvpp2_cls_c2_read(struct mvpp2 *priv, int index,
>  		       struct mvpp2_cls_c2_entry *c2);
>  
> +int mvpp2_ethtool_cls_rule_get(struct mvpp2_port *port,
> +			       struct ethtool_rxnfc *rxnfc);
> +
> +int mvpp2_ethtool_cls_rule_ins(struct mvpp2_port *port,
> +			       struct ethtool_rxnfc *info);
> +
> +int mvpp2_ethtool_cls_rule_del(struct mvpp2_port *port,
> +			       struct ethtool_rxnfc *info);
> +
>  #endif
> diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> index f128ea22b339..d38952eb7aa9 100644
> --- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> +++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> @@ -3937,7 +3937,7 @@ static int mvpp2_ethtool_get_rxnfc(struct
> net_device *dev,
>  				   struct ethtool_rxnfc *info, u32
> *rules)
>  {
>  	struct mvpp2_port *port = netdev_priv(dev);
> -	int ret = 0;
> +	int ret = 0, i, loc = 0;
>  
>  	if (!mvpp22_rss_is_supported())
>  		return -EOPNOTSUPP;
> @@ -3949,6 +3949,18 @@ static int mvpp2_ethtool_get_rxnfc(struct
> net_device *dev,
>  	case ETHTOOL_GRXRINGS:
>  		info->data = port->nrxqs;
>  		break;
> +	case ETHTOOL_GRXCLSRLCNT:
> +		info->rule_cnt = port->n_rfs_rules;
> +		break;
> +	case ETHTOOL_GRXCLSRULE:
> +		ret = mvpp2_ethtool_cls_rule_get(port, info);
> +		break;
> +	case ETHTOOL_GRXCLSRLALL:
> +		for (i = 0; i < MVPP2_N_RFS_RULES; i++) {
> +			if (port->rfs_rules[i])
> +				rules[loc++] = i;
> +		}
> +		break;
>  	default:
>  		return -ENOTSUPP;
>  	}
> @@ -3969,6 +3981,12 @@ static int mvpp2_ethtool_set_rxnfc(struct
> net_device *dev,
>  	case ETHTOOL_SRXFH:
>  		ret = mvpp2_ethtool_rxfh_set(port, info);
>  		break;
> +	case ETHTOOL_SRXCLSRLINS:
> +		ret = mvpp2_ethtool_cls_rule_ins(port, info);
> +		break;
> +	case ETHTOOL_SRXCLSRLDEL:
> +		ret = mvpp2_ethtool_cls_rule_del(port, info);
> +		break;
>  	default:
>  		return -EOPNOTSUPP;
>  	}
> @@ -5040,8 +5058,10 @@ static int mvpp2_port_probe(struct
> platform_device *pdev,
>  	dev->hw_features |= features | NETIF_F_RXCSUM | NETIF_F_GRO |
>  			    NETIF_F_HW_VLAN_CTAG_FILTER;
>  
> -	if (mvpp22_rss_is_supported())
> +	if (mvpp22_rss_is_supported()) {
>  		dev->hw_features |= NETIF_F_RXHASH;
> +		dev->features |= NETIF_F_NTUPLE;
> +	}
>  

I don't think you need this flag, this flag is for a different feature 
needed for netdev ndo ".ndo_rx_flow_steer", Accelerated Receive Flow
Steering which is totally different from user controlled ethtool flow
steering feature you added in this patch.

https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/performance_tuning_guide/network-acc-rfs


>  	if (port->pool_long->id == MVPP2_BM_JUMBO && port->id != 0) {
>  		dev->features &= ~(NETIF_F_IP_CSUM |
> NETIF_F_IPV6_CSUM);
Maxime Chevallier April 24, 2019, 7:01 a.m. UTC | #2
Hello Saeed,

Thanks for the review,

>> When inserting a rule in a given flow, the location given is relative
>> to
>> the flow :
>> 
>> ethtool -N eth0 flow-type udp4 dst-port 1234 action 2 loc 0
>> 
>> ethtool -N eth0 flow-type tcp4 dst-port 1234 action 3 loc 0
>> 
>> However when removing a rule, the global location is to be used. This
>> location can be retrieved by using ethtool -n <interface>.
>>   
>
>I am not sure what you mean by "the location given is relative to the
>flow", it seems like the rule will end up in a different location than
>the user intended, but looking at ethtool documentation it clearly says
>that the location the user provides is an absolute rule id/location,
>which will be used to delete this rule.
>
>from "man ethtool":
>loc N:
>Specify the location/ID to insert the rule. This will overwrite any
>rule present in that location and will not go through any of the rule
>ordering process.
>
>delete N
>Deletes the RX classification rule with the given ID.

I was unsure about this, so I'm glad you commented. One thing
that made me think what I did could be okay is that the documentation
for ETHTOOL_SRXCLSRLINS in ethtool.h says :

"For %ETHTOOL_SRXCLSRLINS, @fs specifies the rule to add or update.
 @fs.@location either specifies the location to use or is a special
 location value with %RX_CLS_LOC_SPECIAL flag set.  On return,
 @fs.@location is the actual rule location."

I interpreted the "On return [...]" part as if we could rewrite the
location if needed when inserting a rule (although it seems ethtool
doesn't do anything with this return value)

The point for doing so is that we have a clear separation in our
classification tables between different traffic classes, so we have a
range of entries for tcp4, one for udp4, one for tcp6, etc.

Having a "global" location numbering scheme would, I think, also be
confusing, since it would make the user use loc 0->7 for tcp4, loc
8->15 for udp4 and so on.

Maybe in this case I should stick with insertions thay rely on
RX_CLS_LOC_SPECIAL (such as "first", "last", "any") and have a scheme
where priorisation is based strictly on the rule insertion order ?

>So the above example should result in one flow rule in your hardware.
>but according the code below the calculated index in
>mvpp2_ethtool_cls_rule_ins might end up different than the requested
>location, which will confuse the user.

I'm also working on writing a proper documentation for this driver,
including the behaviour of the classifier implementation, hopefully
this would help.

Thanks again for the review,

Maxime
Saeed Mahameed April 24, 2019, 6:05 p.m. UTC | #3
On Wed, 2019-04-24 at 09:01 +0200, Maxime Chevallier wrote:
> Hello Saeed,
> 
> Thanks for the review,
> 
> > > When inserting a rule in a given flow, the location given is
> > > relative
> > > to
> > > the flow :
> > > 
> > > ethtool -N eth0 flow-type udp4 dst-port 1234 action 2 loc 0
> > > 
> > > ethtool -N eth0 flow-type tcp4 dst-port 1234 action 3 loc 0
> > > 
> > > However when removing a rule, the global location is to be used.
> > > This
> > > location can be retrieved by using ethtool -n <interface>.
> > >   
> > 
> > I am not sure what you mean by "the location given is relative to
> > the
> > flow", it seems like the rule will end up in a different location
> > than
> > the user intended, but looking at ethtool documentation it clearly
> > says
> > that the location the user provides is an absolute rule
> > id/location,
> > which will be used to delete this rule.
> > 
> > from "man ethtool":
> > loc N:
> > Specify the location/ID to insert the rule. This will overwrite any
> > rule present in that location and will not go through any of the
> > rule
> > ordering process.
> > 
> > delete N
> > Deletes the RX classification rule with the given ID.
> 
> I was unsure about this, so I'm glad you commented. One thing
> that made me think what I did could be okay is that the documentation
> for ETHTOOL_SRXCLSRLINS in ethtool.h says :
> 
> "For %ETHTOOL_SRXCLSRLINS, @fs specifies the rule to add or update.
>  @fs.@location either specifies the location to use or is a special
>  location value with %RX_CLS_LOC_SPECIAL flag set.  On return,
>  @fs.@location is the actual rule location."
> 
> I interpreted the "On return [...]" part as if we could rewrite the
> location if needed when inserting a rule (although it seems ethtool
> doesn't do anything with this return value)
> 

 * For %ETHTOOL_GRXCLSRLCNT, @rule_cnt is set to the number of defined
 * rules on return.  If @data is non-zero on return then it is the
 * size of the rule table, plus the flag % if the
 * driver supports any special location values.  If that flag is not
 * set in @data then special location values should not be used.

Maybe ethtool doesn't do anything with the return value, but if the
user is not using any special flag, then the interpretation should be
absolute location/ID as provided by the user, see below scenario
example

> The point for doing so is that we have a clear separation in our
> classification tables between different traffic classes, so we have a
> range of entries for tcp4, one for udp4, one for tcp6, etc.
> 
> Having a "global" location numbering scheme would, I think, also be
> confusing, since it would make the user use loc 0->7 for tcp4, loc
> 8->15 for udp4 and so on.
> 

why ? even with your hw clear class separation, user can use any loc
for udp4 and tcp4 or any flow for that matter, in case they won't
overlap.

And in case they do overlap, then I think you must have a global 
location scheme! take this scenario for instance:

scenario 1:
loc 0 ip4 action 2
loc 1 udp4 action -1
loc 2 tcp4 action -1

This should result of all udp4, tcp4, and ip4 traffic to go to rx ring
2, even if the user asked to drop udp/tcp4. once rule at location 0 is
deleted then udp/tcp4 traffic will be dropped.

scenario 2:
loc 0 udp4 action -1
loc 1 tcp4 action -1
loc 2 ip4 action 2

should result in dropping all upd4/tcp4 but allow receiving ip4 on ring
4.

User doesn't see and should not see your hw tables scheme, i feel that
for scenario 1 your implementation will drop udp4 and tcp4 since they
will be separated from ip4 rule at loc 0, which is not what the user
expects, please correct me if i am wrong.

that being said, i think you should keep the global location scheme at
least from user perspective and respect the prioritization of the user
inserted rules especially when there are overlapping.

even if there is no overlapping, location could mean: priorities rules
at lower locations in hw processing so they can get higher performance.

> Maybe in this case I should stick with insertions thay rely on
>  (such as "first", "last", "any") and have a scheme
> where priorisation is based strictly on the rule insertion order ?
> 

Sure for when the special flags are set, but you will have to report 
RX_CLS_LOC_SPECIAL on ETHTOOL_GRXCLSRLCNT.

also if you don't want to support the global location scheme then
return -EOPNOTSUPP/-EINVAL when user specifies a non special location 
?

> > So the above example should result in one flow rule in your
> > hardware.
> > but according the code below the calculated index in
> > mvpp2_ethtool_cls_rule_ins might end up different than the
> > requested
> > location, which will confuse the user.
> 
> I'm also working on writing a proper documentation for this driver,
> including the behaviour of the classifier implementation, hopefully
> this would help.
> 

hmm, i think all driver should be aligned and provide same behavior, at
least for the non special flag use case,
vendors must report -EOPNOTSUPPORT if a specific use case operation is
not supported.

> Thanks again for the review,
> 
> Maxime
Maxime Chevallier April 25, 2019, 7:12 a.m. UTC | #4
Hello Saeed,

On Wed, 24 Apr 2019 18:05:51 +0000
Saeed Mahameed <saeedm@mellanox.com> wrote:

>Maybe ethtool doesn't do anything with the return value, but if the
>user is not using any special flag, then the interpretation should be
>absolute location/ID as provided by the user, see below scenario
>example
>
>> The point for doing so is that we have a clear separation in our
>> classification tables between different traffic classes, so we have a
>> range of entries for tcp4, one for udp4, one for tcp6, etc.
>> 
>> Having a "global" location numbering scheme would, I think, also be
>> confusing, since it would make the user use loc 0->7 for tcp4, loc
>> 8->15 for udp4 and so on.
>>   
>
>why ? even with your hw clear class separation, user can use any loc
>for udp4 and tcp4 or any flow for that matter, in case they won't
>overlap.
>
>And in case they do overlap, then I think you must have a global 
>location scheme! take this scenario for instance:
>
>scenario 1:
>loc 0 ip4 action 2
>loc 1 udp4 action -1
>loc 2 tcp4 action -1
>
>This should result of all udp4, tcp4, and ip4 traffic to go to rx ring
>2, even if the user asked to drop udp/tcp4. once rule at location 0 is
>deleted then udp/tcp4 traffic will be dropped.
>
>scenario 2:
>loc 0 udp4 action -1
>loc 1 tcp4 action -1
>loc 2 ip4 action 2
>
>should result in dropping all upd4/tcp4 but allow receiving ip4 on ring
>4.
>
>User doesn't see and should not see your hw tables scheme, i feel that
>for scenario 1 your implementation will drop udp4 and tcp4 since they
>will be separated from ip4 rule at loc 0, which is not what the user
>expects, please correct me if i am wrong.

You're correct, this is what's going to happen with the current
implementation.

>that being said, i think you should keep the global location scheme at
>least from user perspective and respect the prioritization of the user
>inserted rules especially when there are overlapping.
>
>even if there is no overlapping, location could mean: priorities rules
>at lower locations in hw processing so they can get higher performance.

Ok, I'll have to rework the design of the tables a bit to be compliant
with this, but this is achievable.

I'll make sure to CC you on next revisions.

>> Maybe in this case I should stick with insertions thay rely on
>>  (such as "first", "last", "any") and have a scheme
>> where priorisation is based strictly on the rule insertion order ?
>>   
>
>Sure for when the special flags are set, but you will have to report 
>RX_CLS_LOC_SPECIAL on ETHTOOL_GRXCLSRLCNT.
>
>also if you don't want to support the global location scheme then
>return -EOPNOTSUPP/-EINVAL when user specifies a non special location 
>?

Given your review, I'll keep support for the global location scheme.

>> > So the above example should result in one flow rule in your
>> > hardware.
>> > but according the code below the calculated index in
>> > mvpp2_ethtool_cls_rule_ins might end up different than the
>> > requested
>> > location, which will confuse the user.  
>> 
>> I'm also working on writing a proper documentation for this driver,
>> including the behaviour of the classifier implementation, hopefully
>> this would help.
>>   
>
>hmm, i think all driver should be aligned and provide same behavior, at
>least for the non special flag use case,
>vendors must report -EOPNOTSUPPORT if a specific use case operation is
>not supported.

I agree, I'll however document what the limitations are in terms of
supported features, etc.

Thanks for the clarifications, it really helps a lot.

Maxime

Patch
diff mbox series

diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2.h b/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
index 67cce2736806..dcb93afb6569 100644
--- a/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
+++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2.h
@@ -14,6 +14,7 @@ 
 #include <linux/netdevice.h>
 #include <linux/phy.h>
 #include <linux/phylink.h>
+#include <net/flow_offload.h>
 
 /* Fifo Registers */
 #define MVPP2_RX_DATA_FIFO_SIZE_REG(port)	(0x00 + 4 * (port))
@@ -126,6 +127,7 @@ 
 #define MVPP22_CLS_C2_TCAM_DATA4		0x1b20
 #define     MVPP22_CLS_C2_LU_TYPE(lu)		((lu) & 0x3f)
 #define     MVPP22_CLS_C2_PORT_ID(port)		((port) << 8)
+#define     MVPP22_CLS_C2_PORT_MASK		(0xff << 8)
 #define MVPP22_CLS_C2_TCAM_INV			0x1b24
 #define     MVPP22_CLS_C2_TCAM_INV_BIT		BIT(31)
 #define MVPP22_CLS_C2_HIT_CTR			0x1b50
@@ -615,6 +617,10 @@ 
 #define MVPP2_BIT_IN_WORD(bit)		((bit) % 32)
 
 #define MVPP2_N_PRS_FLOWS		52
+#define MVPP2_N_RFS_ENTRIES_PER_FLOW	7
+
+/* There are 7 supported high-level flows */
+#define MVPP2_N_RFS_RULES		(MVPP2_N_RFS_ENTRIES_PER_FLOW * 7)
 
 /* RSS constants */
 #define MVPP22_RSS_TABLE_ENTRIES	32
@@ -812,6 +818,46 @@  struct mvpp2_queue_vector {
 	struct cpumask *mask;
 };
 
+/* Internal represention of a Flow Steering rule */
+struct mvpp2_rfs_rule {
+	/* Rule location inside the flow*/
+	int loc;
+
+	/* Flow type, such as TCP_V4_FLOW, IP6_FLOW, etc. */
+	int flow_type;
+
+	/* The lookup type used by this flow. This allows the classification
+	 * engines to differentiate between entries that match the same
+	 * parameters
+	 */
+	int lu_type;
+
+	/* Index of the entry handling this rule in the flow table */
+	int flt_index;
+
+	/* Index of the C2 TCAM entry handling this rule */
+	int c2_index;
+
+	/* Header fields that needs to be extracted to match this flow */
+	u16 hek_fields;
+
+	/* CLS engine : only c2 is supported for now. */
+	u8 engine;
+
+	/* TCAM key and mask for C2-based steering. These fields should be
+	 * encapsulated in a union should we add more engines.
+	 */
+	u64 c2_tcam;
+	u64 c2_tcam_mask;
+
+	struct flow_rule *flow;
+};
+
+struct mvpp2_ethtool_fs {
+	struct mvpp2_rfs_rule rule;
+	struct ethtool_rxnfc rxnfc;
+};
+
 struct mvpp2_port {
 	u8 id;
 
@@ -883,6 +929,10 @@  struct mvpp2_port {
 
 	/* RSS indirection table */
 	u32 indir[MVPP22_RSS_TABLE_ENTRIES];
+
+	/* List of steering rules active on that port */
+	struct mvpp2_ethtool_fs *rfs_rules[MVPP2_N_RFS_RULES];
+	int n_rfs_rules;
 };
 
 /* The mvpp2_tx_desc and mvpp2_rx_desc structures describe the
diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
index 1087974d3b98..7457e2ab57eb 100644
--- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
+++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c
@@ -448,6 +448,17 @@  static void mvpp2_cls_flow_port_add(struct mvpp2_cls_flow_entry *fe,
 	fe->data[0] |= MVPP2_CLS_FLOW_TBL0_PORT_ID(port);
 }
 
+static void mvpp2_cls_flow_port_remove(struct mvpp2_cls_flow_entry *fe,
+				       u32 port)
+{
+	fe->data[0] &= ~MVPP2_CLS_FLOW_TBL0_PORT_ID(port);
+}
+
+static u8 mvpp2_cls_flow_pmap_get(struct mvpp2_cls_flow_entry *fe)
+{
+	return (fe->data[0] >> 4) & MVPP2_CLS_FLOW_TBL0_PORT_ID_MASK;
+}
+
 static void mvpp2_cls_flow_lu_type_set(struct mvpp2_cls_flow_entry *fe,
 				       u8 lu_type)
 {
@@ -455,6 +466,12 @@  static void mvpp2_cls_flow_lu_type_set(struct mvpp2_cls_flow_entry *fe,
 	fe->data[1] |= MVPP2_CLS_FLOW_TBL1_LU_TYPE(lu_type);
 }
 
+static u8 mvpp2_cls_flow_lu_type_get(struct mvpp2_cls_flow_entry *fe)
+{
+	return (fe->data[1] &
+		MVPP2_CLS_FLOW_TBL1_LU_TYPE(MVPP2_CLS_LU_TYPE_MASK)) >> 3;
+}
+
 /* Initialize the parser entry for the given flow */
 static void mvpp2_cls_flow_prs_init(struct mvpp2 *priv,
 				    const struct mvpp2_cls_flow *flow)
@@ -539,6 +556,35 @@  void mvpp2_cls_c2_read(struct mvpp2 *priv, int index,
 	c2->valid = !(val & MVPP22_CLS_C2_TCAM_INV_BIT);
 }
 
+static int mvpp2_cls_c2_port_flow_get_type(int flow_type)
+{
+	switch (flow_type) {
+	case TCP_V4_FLOW:
+		return MVPP22_FLOW_TCP4;
+	case TCP_V6_FLOW:
+		return MVPP22_FLOW_TCP6;
+	case UDP_V4_FLOW:
+		return MVPP22_FLOW_UDP4;
+	case UDP_V6_FLOW:
+		return MVPP22_FLOW_UDP6;
+	case IPV4_FLOW:
+		return MVPP22_FLOW_IP4;
+	case IPV6_FLOW:
+		return MVPP22_FLOW_IP6;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static int mvpp2_cls_c2_port_flow_index(struct mvpp2_port *port, int lu_type,
+					int loc)
+{
+	if (loc >= MVPP22_CLS_C2_PORT_N_RFS)
+		return -EINVAL;
+
+	return MVPP22_CLS_C2_RFS_LOC(port->id, lu_type, loc);
+}
+
 /* Initialize the flow table entries for the given flow */
 static void mvpp2_cls_flow_init(struct mvpp2 *priv,
 				const struct mvpp2_cls_flow *flow)
@@ -565,7 +611,7 @@  static void mvpp2_cls_flow_init(struct mvpp2 *priv,
 
 	mvpp2_cls_flow_eng_set(&fe, MVPP22_CLS_ENGINE_C2);
 	mvpp2_cls_flow_port_id_sel(&fe, true);
-	mvpp2_cls_flow_lu_type_set(&fe, MVPP2_CLS_LU_ALL);
+	mvpp2_cls_flow_lu_type_set(&fe, MVPP22_FLOW_ETHER);
 
 	/* Add all ports */
 	for (i = 0; i < MVPP2_MAX_PORTS; i++)
@@ -652,6 +698,26 @@  static int mvpp2_flow_set_hek_fields(struct mvpp2_cls_flow_entry *fe,
 	return 0;
 }
 
+/* Returns the size, in bits, of the corresponding HEK field */
+static int mvpp2_cls_hek_field_size(u32 field)
+{
+	switch (field) {
+	case MVPP22_CLS_HEK_OPT_MAC_DA:
+		return 48;
+	case MVPP22_CLS_HEK_OPT_IP4SA:
+	case MVPP22_CLS_HEK_OPT_IP4DA:
+		return 32;
+	case MVPP22_CLS_HEK_OPT_IP6SA:
+	case MVPP22_CLS_HEK_OPT_IP6DA:
+		return 128;
+	case MVPP22_CLS_HEK_OPT_L4SIP:
+	case MVPP22_CLS_HEK_OPT_L4DIP:
+		return 16;
+	default:
+		return -1;
+	}
+}
+
 const struct mvpp2_cls_flow *mvpp2_cls_flow_get(int flow)
 {
 	if (flow >= MVPP2_N_PRS_FLOWS)
@@ -810,7 +876,7 @@  static void mvpp2_port_c2_cls_init(struct mvpp2_port *port)
 
 	/* Match on Lookup Type */
 	c2.tcam[4] |= MVPP22_CLS_C2_TCAM_EN(MVPP22_CLS_C2_LU_TYPE(MVPP2_CLS_LU_TYPE_MASK));
-	c2.tcam[4] |= MVPP22_CLS_C2_LU_TYPE(MVPP2_CLS_LU_ALL);
+	c2.tcam[4] |= MVPP22_CLS_C2_LU_TYPE(MVPP22_FLOW_ETHER);
 
 	/* Update RSS status after matching this entry */
 	c2.act = MVPP22_CLS_C2_ACT_RSS_EN(MVPP22_C2_UPD_LOCK);
@@ -944,6 +1010,18 @@  void mvpp22_port_rss_disable(struct mvpp2_port *port)
 	mvpp2_rss_port_c2_disable(port);
 }
 
+static void mvpp22_port_c2_lookup_disable(struct mvpp2_port *port, int entry)
+{
+	struct mvpp2_cls_c2_entry c2;
+
+	mvpp2_cls_c2_read(port->priv, entry, &c2);
+
+	/* Clear the port map so that the entry doesn't match anymore */
+	c2.tcam[4] &= ~(MVPP22_CLS_C2_PORT_ID(BIT(port->id)));
+
+	mvpp2_cls_c2_write(port->priv, &c2);
+}
+
 /* Set CPU queue number for oversize packets */
 void mvpp2_cls_oversize_rxq_set(struct mvpp2_port *port)
 {
@@ -960,6 +1038,382 @@  void mvpp2_cls_oversize_rxq_set(struct mvpp2_port *port)
 	mvpp2_write(port->priv, MVPP2_CLS_SWFWD_PCTRL_REG, val);
 }
 
+static int mvpp2_port_c2_tcam_rule_add(struct mvpp2_port *port,
+				       struct mvpp2_rfs_rule *rule)
+{
+	struct flow_action_entry *act;
+	struct mvpp2_cls_c2_entry c2;
+	u8 qh, ql, pmap;
+
+	memset(&c2, 0, sizeof(c2));
+
+	c2.index = mvpp2_cls_c2_port_flow_index(port, rule->lu_type, rule->loc);
+	if (c2.index < 0)
+		return -EINVAL;
+
+	act = &rule->flow->action.entries[0];
+
+	rule->c2_index = c2.index;
+
+	c2.tcam[0] = (rule->c2_tcam & 0xffff) |
+		     ((rule->c2_tcam_mask & 0xffff) << 16);
+	c2.tcam[1] = ((rule->c2_tcam >> 16) & 0xffff) |
+		     (((rule->c2_tcam_mask >> 16) & 0xffff) << 16);
+	c2.tcam[2] = ((rule->c2_tcam >> 32) & 0xffff) |
+		     (((rule->c2_tcam_mask >> 32) & 0xffff) << 16);
+	c2.tcam[3] = ((rule->c2_tcam >> 48) & 0xffff) |
+		     (((rule->c2_tcam_mask >> 48) & 0xffff) << 16);
+
+	pmap = BIT(port->id);
+	c2.tcam[4] = MVPP22_CLS_C2_PORT_ID(pmap);
+	c2.tcam[4] |= MVPP22_CLS_C2_TCAM_EN(MVPP22_CLS_C2_PORT_ID(pmap));
+
+	/* Match on Lookup Type */
+	c2.tcam[4] |= MVPP22_CLS_C2_TCAM_EN(MVPP22_CLS_C2_LU_TYPE(MVPP2_CLS_LU_TYPE_MASK));
+	c2.tcam[4] |= MVPP22_CLS_C2_LU_TYPE(rule->lu_type);
+
+	/* Mark packet as "forwarded to software", needed for RSS */
+	c2.act |= MVPP22_CLS_C2_ACT_FWD(MVPP22_C2_FWD_SW_LOCK);
+
+	c2.act |= MVPP22_CLS_C2_ACT_QHIGH(MVPP22_C2_UPD_LOCK) |
+		   MVPP22_CLS_C2_ACT_QLOW(MVPP22_C2_UPD_LOCK);
+
+	qh = ((act->queue.index + port->first_rxq) >> 3) & MVPP22_CLS_C2_ATTR0_QHIGH_MASK;
+	ql = (act->queue.index + port->first_rxq) & MVPP22_CLS_C2_ATTR0_QLOW_MASK;
+
+	c2.attr[0] = MVPP22_CLS_C2_ATTR0_QHIGH(qh) |
+		      MVPP22_CLS_C2_ATTR0_QLOW(ql);
+
+	c2.valid = true;
+
+	mvpp2_cls_c2_write(port->priv, &c2);
+
+	return 0;
+}
+
+static int mvpp2_port_c2_rfs_rule_insert(struct mvpp2_port *port,
+					 struct mvpp2_rfs_rule *rule)
+{
+	return mvpp2_port_c2_tcam_rule_add(port, rule);
+}
+
+static int mvpp2_port_flow_table_find(struct mvpp2 *priv, int flow_id,
+				      u8 engine, u8 lu_type, u16 hek_fields)
+{
+	struct mvpp2_cls_flow_entry fe;
+	u16 flow_hek_fields;
+	int i, flow_engine;
+	u8 flow_lu_type;
+
+	/* Iterate on all flow table entries for this ID */
+	for (i = MVPP2_CLS_FLT_FIRST(flow_id);
+	     i < MVPP2_CLS_FLT_LAST(flow_id); i++) {
+		mvpp2_cls_flow_read(priv, i, &fe);
+
+		flow_hek_fields = mvpp2_flow_get_hek_fields(&fe);
+		if (flow_hek_fields != hek_fields)
+			continue;
+
+		flow_engine = mvpp2_cls_flow_eng_get(&fe);
+		if (flow_engine != engine)
+			continue;
+
+		flow_lu_type = mvpp2_cls_flow_lu_type_get(&fe);
+		if (flow_lu_type != lu_type)
+			continue;
+
+		return i;
+	}
+
+	return -1;
+}
+
+static int mvpp2_port_flow_table_find_free(struct mvpp2 *priv, int flow_id)
+{
+	struct mvpp2_cls_flow_entry fe;
+	u8 pmap;
+	int i;
+
+	memset(&fe, 0, sizeof(fe));
+	for (i = 0; i < MVPP2_N_RFS_ENTRIES_PER_FLOW; i++) {
+		mvpp2_cls_flow_read(priv, MVPP2_CLS_FLT_C2_RFS(flow_id, i),
+				    &fe);
+
+		/* get pmap */
+		pmap = mvpp2_cls_flow_pmap_get(&fe);
+		if (!pmap)
+			return MVPP2_CLS_FLT_C2_RFS(flow_id, i);
+	}
+
+	return -1;
+}
+
+/**
+ * It's possible that multiple different rfs rules use the same flow_table
+ * entry, if the rules use the same key and the same engine. In that case,
+ * we must make sure not to delete this flow_table entry when removing the
+ * rule, unless it's the last rule that uses this entry.
+ */
+static bool mvpp2_port_flt_has_siblings(struct mvpp2_port *port,
+					struct mvpp2_rfs_rule *rule)
+{
+	struct mvpp2_ethtool_fs *efs;
+	int i;
+
+	for (i = 0; i < MVPP2_N_RFS_RULES; i++) {
+		efs = port->rfs_rules[i];
+		if (!efs)
+			continue;
+		/* Skip the entry we are trying to find the sibling of */
+		if (efs->rule.loc == rule->loc)
+			continue;
+		if (efs->rule.flt_index == rule->flt_index)
+			return true;
+	}
+
+	return false;
+}
+
+static int mvpp2_port_cls_rfs_rule_remove(struct mvpp2_port *port,
+					  struct mvpp2_rfs_rule *rule)
+{
+	struct mvpp2_cls_flow_entry fe;
+
+	if (rule->flt_index >= 0 &&
+	    !mvpp2_port_flt_has_siblings(port, rule)) {
+		mvpp2_cls_flow_read(port->priv, rule->flt_index, &fe);
+		mvpp2_cls_flow_port_remove(&fe, BIT(port->id));
+		mvpp2_cls_flow_write(port->priv, &fe);
+	}
+
+	if (rule->c2_index >= 0)
+		mvpp22_port_c2_lookup_disable(port, rule->c2_index);
+
+	return 0;
+}
+
+static int mvpp2_port_flt_rfs_rule_insert(struct mvpp2_port *port,
+					  struct mvpp2_rfs_rule *rule)
+{
+	const struct mvpp2_cls_flow *flow;
+	struct mvpp2 *priv = port->priv;
+	struct mvpp2_cls_flow_entry fe;
+	int index, ret, i;
+
+	if (rule->engine != MVPP22_CLS_ENGINE_C2)
+		return -EOPNOTSUPP;
+
+	ret = mvpp2_port_c2_rfs_rule_insert(port, rule);
+	if (ret)
+		return ret;
+
+	for_each_cls_flow_id_with_type(i, rule->flow_type) {
+		flow = mvpp2_cls_flow_get(i);
+		if (!flow)
+			return 0;
+
+		/* Don't perform steering on HEK fields that aren't relevant
+		 * for this flow, e.g. steering on VLAN tag when in an untagged
+		 * flow.
+		 */
+		if ((rule->hek_fields | flow->supported_hash_opts) !=
+		    flow->supported_hash_opts)
+			return 0;
+
+		index = mvpp2_port_flow_table_find(priv, flow->flow_id,
+						   rule->engine, rule->lu_type,
+						   rule->hek_fields);
+
+		/* If there already is a flow entry that matches our
+		 * parameters
+		 */
+		if (index >= 0) {
+			rule->flt_index = index;
+			mvpp2_cls_flow_read(priv, index, &fe);
+			mvpp2_cls_flow_port_add(&fe, BIT(port->id));
+			mvpp2_cls_flow_write(priv, &fe);
+			continue;
+		} else {
+			index = mvpp2_port_flow_table_find_free(priv,
+								flow->flow_id);
+			if (index < 0)
+				return -EOPNOTSUPP;
+		}
+
+		/* Create entry */
+		rule->flt_index = index;
+		mvpp2_cls_flow_read(priv, index, &fe);
+
+		mvpp2_cls_flow_eng_set(&fe, rule->engine);
+		mvpp2_cls_flow_port_id_sel(&fe, true);
+		mvpp2_flow_set_hek_fields(&fe, rule->hek_fields);
+		mvpp2_cls_flow_lu_type_set(&fe, rule->lu_type);
+		mvpp2_cls_flow_port_add(&fe, 0xf);
+
+		mvpp2_cls_flow_write(priv, &fe);
+	}
+
+	return 0;
+}
+
+static int mvpp2_cls_c2_build_match(struct mvpp2_rfs_rule *rule)
+{
+	struct flow_rule *flow = rule->flow;
+	struct flow_action_entry *act;
+	int offs = 64;
+
+	act = &flow->action.entries[0];
+
+	if (flow_rule_match_key(flow, FLOW_DISSECTOR_KEY_PORTS)) {
+		struct flow_match_ports match;
+
+		flow_rule_match_ports(flow, &match);
+		if (match.mask->src) {
+			rule->hek_fields |= MVPP22_CLS_HEK_OPT_L4SIP;
+			offs -= mvpp2_cls_hek_field_size(MVPP22_CLS_HEK_OPT_L4SIP);
+
+			rule->c2_tcam |= ((u64)ntohs(match.key->src)) << offs;
+			rule->c2_tcam_mask |= ((u64)ntohs(match.mask->src)) << offs;
+		}
+
+		if (match.mask->dst) {
+			rule->hek_fields |= MVPP22_CLS_HEK_OPT_L4DIP;
+			offs -= mvpp2_cls_hek_field_size(MVPP22_CLS_HEK_OPT_L4DIP);
+
+			rule->c2_tcam |= ((u64)ntohs(match.key->dst)) << offs;
+			rule->c2_tcam_mask |= ((u64)ntohs(match.mask->dst)) << offs;
+		}
+	}
+
+	if (hweight16(rule->hek_fields) > MVPP2_FLOW_N_FIELDS)
+		return -EOPNOTSUPP;
+
+	return 0;
+}
+
+static int mvpp2_cls_rfs_parse_rule(struct mvpp2_rfs_rule *rule)
+{
+	struct flow_rule *flow = rule->flow;
+	struct flow_action_entry *act;
+
+	act = &flow->action.entries[0];
+	if (act->id != FLOW_ACTION_QUEUE)
+		return -EOPNOTSUPP;
+
+	/* For now, only use the C2 engine which has a HEK size limited to 64
+	 * bits for TCAM matching.
+	 */
+	rule->engine = MVPP22_CLS_ENGINE_C2;
+
+	if (mvpp2_cls_c2_build_match(rule))
+		return -EINVAL;
+
+	return 0;
+}
+
+int mvpp2_ethtool_cls_rule_get(struct mvpp2_port *port,
+			       struct ethtool_rxnfc *rxnfc)
+{
+	struct mvpp2_ethtool_fs *efs;
+
+	if (rxnfc->fs.location >= MVPP2_N_RFS_RULES)
+		return -EINVAL;
+
+	efs = port->rfs_rules[rxnfc->fs.location];
+	if (!efs)
+		return -ENOENT;
+
+	memcpy(rxnfc, &efs->rxnfc, sizeof(efs->rxnfc));
+
+	return 0;
+}
+
+int mvpp2_ethtool_cls_rule_ins(struct mvpp2_port *port,
+			       struct ethtool_rxnfc *info)
+{
+	struct ethtool_rx_flow_spec_input input = {};
+	struct ethtool_rx_flow_rule *ethtool_rule;
+	struct mvpp2_ethtool_fs *efs, *old_efs;
+	int ret = 0, index;
+
+	if (info->fs.location >= MVPP22_CLS_C2_PORT_N_RFS ||
+	    info->fs.location < 0)
+		return -EINVAL;
+
+	efs = kzalloc(sizeof(*efs), GFP_KERNEL);
+	if (!efs)
+		return -ENOMEM;
+
+	input.fs = &info->fs;
+
+	ethtool_rule = ethtool_rx_flow_rule_create(&input);
+	if (IS_ERR(ethtool_rule)) {
+		ret = PTR_ERR(ethtool_rule);
+		goto clean_rule;
+	}
+
+	efs->rule.flow = ethtool_rule->rule;
+	efs->rule.flow_type = info->fs.flow_type & ~(FLOW_EXT | FLOW_MAC_EXT |
+						     FLOW_RSS);
+	efs->rule.lu_type = mvpp2_cls_c2_port_flow_get_type(efs->rule.flow_type);
+
+	ret = mvpp2_cls_rfs_parse_rule(&efs->rule);
+	if (ret)
+		goto clean_eth_rule;
+
+	efs->rule.loc = info->fs.location;
+
+	index = (efs->rule.lu_type - 1) * MVPP22_CLS_C2_PORT_N_FLOWS +
+		efs->rule.loc;
+
+	/* Replace an already existing rule */
+	if (port->rfs_rules[index]) {
+		old_efs = port->rfs_rules[index];
+		ret = mvpp2_port_cls_rfs_rule_remove(port, &old_efs->rule);
+		if (ret)
+			goto clean_eth_rule;
+		kfree(old_efs);
+	}
+
+	ret = mvpp2_port_flt_rfs_rule_insert(port, &efs->rule);
+	if (ret)
+		goto clean_eth_rule;
+
+	info->fs.location = index;
+
+	memcpy(&efs->rxnfc, info, sizeof(*info));
+	port->rfs_rules[index] = efs;
+	port->n_rfs_rules++;
+
+clean_eth_rule:
+	ethtool_rx_flow_rule_destroy(ethtool_rule);
+clean_rule:
+	kfree(efs);
+	return ret;
+}
+
+int mvpp2_ethtool_cls_rule_del(struct mvpp2_port *port,
+			       struct ethtool_rxnfc *info)
+{
+	struct mvpp2_ethtool_fs *efs;
+	int ret;
+
+	efs = port->rfs_rules[info->fs.location];
+	if (!efs)
+		return -EINVAL;
+
+	/* Remove the rule from the engines. */
+	ret = mvpp2_port_cls_rfs_rule_remove(port, &efs->rule);
+	if (ret)
+		return ret;
+
+	port->n_rfs_rules--;
+	port->rfs_rules[info->fs.location] = NULL;
+	kfree(efs);
+
+	return 0;
+}
+
 static inline u32 mvpp22_rxfh_indir(struct mvpp2_port *port, u32 rxq)
 {
 	int nrxqs, cpu, cpus = num_possible_cpus();
diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
index 96304ffc5d49..1d91d5d3b0e1 100644
--- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
+++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.h
@@ -107,15 +107,43 @@  struct mvpp2_cls_c2_entry {
 	u8 valid;
 };
 
+enum mvpp2_lu_type {
+	/* LU Type 0 is the default value, it shouldn't be used, to avoid
+	 * spurious matches.
+	 */
+	MVPP22_FLOW_TCP4 = 1,
+	MVPP22_FLOW_TCP6,
+	MVPP22_FLOW_UDP4,
+	MVPP22_FLOW_UDP6,
+	MVPP22_FLOW_IP4,
+	MVPP22_FLOW_IP6,
+	MVPP22_FLOW_ETHER,
+	MVPP22_FLOW_LAST = MVPP22_FLOW_ETHER,
+};
+
 /* Classifier C2 engine entries */
 #define MVPP22_CLS_C2_N_ENTRIES		256
 
 /* Number of per-port dedicated entries in the C2 TCAM */
-#define MVPP22_CLS_C2_PORT_RANGE	8
+#define MVPP22_CLS_C2_PORT_FLOW_RANGE	8
+#define MVPP22_CLS_C2_PORT_N_RFS	MVPP2_N_RFS_ENTRIES_PER_FLOW
+
+/* We have one C2 range per supported flow type : ip4, i6, tcp4, tcp6, udp4, udp6 */
+#define MVPP22_CLS_C2_PORT_N_FLOWS	(MVPP22_FLOW_LAST)
+
+/* Each port has oen range per flow type + one entry controling the global RSS
+ * setting and the default rx queue
+ */
+#define MVPP22_CLS_C2_PORT_RANGE	(MVPP22_CLS_C2_PORT_FLOW_RANGE * \
+					 MVPP22_CLS_C2_PORT_N_FLOWS + 1)
+
+#define MVPP22_CLS_C2_PORT_FIRST(p)	((p) * MVPP22_CLS_C2_PORT_RANGE)
+#define MVPP22_CLS_C2_RSS_ENTRY(p)	(MVPP22_CLS_C2_PORT_FIRST((p) + 1) - 1)
 
-#define MVPP22_CLS_C2_PORT_FIRST(p)	(MVPP22_CLS_C2_N_ENTRIES - \
-					((p) * MVPP22_CLS_C2_PORT_RANGE))
-#define MVPP22_CLS_C2_RSS_ENTRY(p)	(MVPP22_CLS_C2_PORT_FIRST(p) - 1)
+#define MVPP22_CLS_C2_PORT_FLOW_FIRST(p, f)	(MVPP22_CLS_C2_PORT_FIRST(p) + \
+						 ((f) - 1) * MVPP22_CLS_C2_PORT_FLOW_RANGE)
+
+#define MVPP22_CLS_C2_RFS_LOC(p, f, loc)	(MVPP22_CLS_C2_PORT_FLOW_FIRST(p, f) + (loc))
 
 /* Packet flow ID */
 enum mvpp2_prs_flow {
@@ -145,10 +173,6 @@  enum mvpp2_prs_flow {
 	MVPP2_FL_LAST,
 };
 
-enum mvpp2_cls_lu_type {
-	MVPP2_CLS_LU_ALL = 0,
-};
-
 /* LU Type defined for all engines, and specified in the flow table */
 #define MVPP2_CLS_LU_TYPE_MASK			0x3f
 
@@ -168,11 +192,12 @@  struct mvpp2_cls_flow {
 	struct mvpp2_prs_result_info prs_ri;
 };
 
-#define MVPP2_CLS_FLT_ENTRIES_PER_FLOW		(MVPP2_MAX_PORTS + 1)
+#define MVPP2_CLS_FLT_ENTRIES_PER_FLOW		(MVPP2_MAX_PORTS + 1 + MVPP2_N_RFS_ENTRIES_PER_FLOW)
 #define MVPP2_CLS_FLT_FIRST(id)			(((id) - MVPP2_FL_START) * \
 						 MVPP2_CLS_FLT_ENTRIES_PER_FLOW)
-#define MVPP2_CLS_FLT_C2_RSS_ENTRY(id)		(MVPP2_CLS_FLT_FIRST(id))
-#define MVPP2_CLS_FLT_HASH_ENTRY(port, id)	(MVPP2_CLS_FLT_C2_RSS_ENTRY(id) + (port) + 1)
+#define MVPP2_CLS_FLT_C2_RFS(id, rfs_n)		(MVPP2_CLS_FLT_FIRST(id) + (rfs_n))
+#define MVPP2_CLS_FLT_C2_RSS_ENTRY(id)		(MVPP2_CLS_FLT_C2_RFS(id, MVPP2_N_RFS_ENTRIES_PER_FLOW))
+#define MVPP2_CLS_FLT_HASH_ENTRY(port, id)	(MVPP2_CLS_FLT_C2_RSS_ENTRY(id) + 1 + (port))
 #define MVPP2_CLS_FLT_LAST(id)			(MVPP2_CLS_FLT_FIRST(id) + \
 						 MVPP2_CLS_FLT_ENTRIES_PER_FLOW - 1)
 
@@ -246,4 +271,13 @@  u32 mvpp2_cls_c2_hit_count(struct mvpp2 *priv, int c2_index);
 void mvpp2_cls_c2_read(struct mvpp2 *priv, int index,
 		       struct mvpp2_cls_c2_entry *c2);
 
+int mvpp2_ethtool_cls_rule_get(struct mvpp2_port *port,
+			       struct ethtool_rxnfc *rxnfc);
+
+int mvpp2_ethtool_cls_rule_ins(struct mvpp2_port *port,
+			       struct ethtool_rxnfc *info);
+
+int mvpp2_ethtool_cls_rule_del(struct mvpp2_port *port,
+			       struct ethtool_rxnfc *info);
+
 #endif
diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
index f128ea22b339..d38952eb7aa9 100644
--- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
+++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
@@ -3937,7 +3937,7 @@  static int mvpp2_ethtool_get_rxnfc(struct net_device *dev,
 				   struct ethtool_rxnfc *info, u32 *rules)
 {
 	struct mvpp2_port *port = netdev_priv(dev);
-	int ret = 0;
+	int ret = 0, i, loc = 0;
 
 	if (!mvpp22_rss_is_supported())
 		return -EOPNOTSUPP;
@@ -3949,6 +3949,18 @@  static int mvpp2_ethtool_get_rxnfc(struct net_device *dev,
 	case ETHTOOL_GRXRINGS:
 		info->data = port->nrxqs;
 		break;
+	case ETHTOOL_GRXCLSRLCNT:
+		info->rule_cnt = port->n_rfs_rules;
+		break;
+	case ETHTOOL_GRXCLSRULE:
+		ret = mvpp2_ethtool_cls_rule_get(port, info);
+		break;
+	case ETHTOOL_GRXCLSRLALL:
+		for (i = 0; i < MVPP2_N_RFS_RULES; i++) {
+			if (port->rfs_rules[i])
+				rules[loc++] = i;
+		}
+		break;
 	default:
 		return -ENOTSUPP;
 	}
@@ -3969,6 +3981,12 @@  static int mvpp2_ethtool_set_rxnfc(struct net_device *dev,
 	case ETHTOOL_SRXFH:
 		ret = mvpp2_ethtool_rxfh_set(port, info);
 		break;
+	case ETHTOOL_SRXCLSRLINS:
+		ret = mvpp2_ethtool_cls_rule_ins(port, info);
+		break;
+	case ETHTOOL_SRXCLSRLDEL:
+		ret = mvpp2_ethtool_cls_rule_del(port, info);
+		break;
 	default:
 		return -EOPNOTSUPP;
 	}
@@ -5040,8 +5058,10 @@  static int mvpp2_port_probe(struct platform_device *pdev,
 	dev->hw_features |= features | NETIF_F_RXCSUM | NETIF_F_GRO |
 			    NETIF_F_HW_VLAN_CTAG_FILTER;
 
-	if (mvpp22_rss_is_supported())
+	if (mvpp22_rss_is_supported()) {
 		dev->hw_features |= NETIF_F_RXHASH;
+		dev->features |= NETIF_F_NTUPLE;
+	}
 
 	if (port->pool_long->id == MVPP2_BM_JUMBO && port->id != 0) {
 		dev->features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM);