diff mbox series

[v2,05/24] net/nfp: add the offload support of basic actions

Message ID 1665382142-21684-6-git-send-email-chaoyong.he@corigine.com (mailing list archive)
State Superseded, archived
Delegated to: Ferruh Yigit
Headers
Series add the basic rte_flow offload support of nfp PMD |

Checks

Context Check Description
ci/checkpatch success coding style OK

Commit Message

Chaoyong He Oct. 10, 2022, 6:08 a.m. UTC 
  Add the offload support of very basic actions: mark, rss,
count, drop and output.

Signed-off-by: Chaoyong He <chaoyong.he@corigine.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@corigine.com>
---
 doc/guides/nics/features/nfp.ini         |   6 ++
 doc/guides/rel_notes/release_22_11.rst   |   6 ++
 drivers/net/nfp/flower/nfp_flower_cmsg.h |  11 +++
 drivers/net/nfp/nfp_flow.c               | 112 +++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_flow.h               |  37 ++++++++++
 5 files changed, 172 insertions(+)

Comments

Ferruh Yigit Oct. 10, 2022, 2:52 p.m. UTC | #1 
On 10/10/2022 7:08 AM, Chaoyong He wrote:
> Add the offload support of very basic actions: mark, rss,
> count, drop and output.
> 
> Signed-off-by: Chaoyong He <chaoyong.he@corigine.com>
> Reviewed-by: Niklas Söderlund <niklas.soderlund@corigine.com>
> ---
>   doc/guides/nics/features/nfp.ini         |   6 ++
>   doc/guides/rel_notes/release_22_11.rst   |   6 ++
>   drivers/net/nfp/flower/nfp_flower_cmsg.h |  11 +++
>   drivers/net/nfp/nfp_flow.c               | 112 +++++++++++++++++++++++++++++++
>   drivers/net/nfp/nfp_flow.h               |  37 ++++++++++
>   5 files changed, 172 insertions(+)
> 
> diff --git a/doc/guides/nics/features/nfp.ini b/doc/guides/nics/features/nfp.ini
> index 4460cf0..194fe64 100644
> --- a/doc/guides/nics/features/nfp.ini
> +++ b/doc/guides/nics/features/nfp.ini
> @@ -29,3 +29,9 @@ Usage doc            = Y
>   [rte_flow items]
>   eth                  = Y
>   port_id              = Y
> +
> +[rte_flow actions]
> +count                = Y
> +dec_ttl              = Y
> +drop                 = Y
> +port_id              = Y

Are these actions supported? As far as I can see all done in the code is 
print log when these actions detected.

For 'dec_ttl' even logging seems missing.

I think better to document them as supported when actual support is added.

> diff --git a/doc/guides/rel_notes/release_22_11.rst b/doc/guides/rel_notes/release_22_11.rst
> index cd22d95..ac0223f 100644
> --- a/doc/guides/rel_notes/release_22_11.rst
> +++ b/doc/guides/rel_notes/release_22_11.rst
> @@ -155,6 +155,12 @@ New Features
>   
>       * Ethernet
>   
> +  Add the support of rte_flow actions as follow:
> +
> +    * Count
> +    * Drop
> +    * TTL decrement

Not sure if this level of details is required in release notes, it can 
be sufficient that driver flow API support is added.
Chaoyong He Oct. 19, 2022, 11:32 a.m. UTC | #2 
> On 10/10/2022 7:08 AM, Chaoyong He wrote:
> > Add the offload support of very basic actions: mark, rss, count, drop
> > and output.
> >
> > Signed-off-by: Chaoyong He <chaoyong.he@corigine.com>
> > Reviewed-by: Niklas Söderlund <niklas.soderlund@corigine.com>
> > ---
> >   doc/guides/nics/features/nfp.ini         |   6 ++
> >   doc/guides/rel_notes/release_22_11.rst   |   6 ++
> >   drivers/net/nfp/flower/nfp_flower_cmsg.h |  11 +++
> >   drivers/net/nfp/nfp_flow.c               | 112
> +++++++++++++++++++++++++++++++
> >   drivers/net/nfp/nfp_flow.h               |  37 ++++++++++
> >   5 files changed, 172 insertions(+)
> >
> > diff --git a/doc/guides/nics/features/nfp.ini
> > b/doc/guides/nics/features/nfp.ini
> > index 4460cf0..194fe64 100644
> > --- a/doc/guides/nics/features/nfp.ini
> > +++ b/doc/guides/nics/features/nfp.ini
> > @@ -29,3 +29,9 @@ Usage doc            = Y
> >   [rte_flow items]
> >   eth                  = Y
> >   port_id              = Y
> > +
> > +[rte_flow actions]
> > +count                = Y
> > +dec_ttl              = Y
> > +drop                 = Y
> > +port_id              = Y
> 
> Are these actions supported? As far as I can see all done in the code is print
> log when these actions detected.
> 
> For 'dec_ttl' even logging seems missing.
> 
> I think better to document them as supported when actual support is added.
> 

I will delete the 'dec_ttl' here, and we does support the offload of other ones, thanks.

> > diff --git a/doc/guides/rel_notes/release_22_11.rst
> > b/doc/guides/rel_notes/release_22_11.rst
> > index cd22d95..ac0223f 100644
> > --- a/doc/guides/rel_notes/release_22_11.rst
> > +++ b/doc/guides/rel_notes/release_22_11.rst
> > @@ -155,6 +155,12 @@ New Features
> >
> >       * Ethernet
> >
> > +  Add the support of rte_flow actions as follow:
> > +
> > +    * Count
> > +    * Drop
> > +    * TTL decrement
> 
> Not sure if this level of details is required in release notes, it can be sufficient
> that driver flow API support is added.

Okay, I will revise them at next version patch, thanks.
diff mbox series

Patch

diff --git a/doc/guides/nics/features/nfp.ini b/doc/guides/nics/features/nfp.ini
index 4460cf0..194fe64 100644
--- a/doc/guides/nics/features/nfp.ini
+++ b/doc/guides/nics/features/nfp.ini
@@ -29,3 +29,9 @@  Usage doc            = Y
 [rte_flow items]
 eth                  = Y
 port_id              = Y
+
+[rte_flow actions]
+count                = Y
+dec_ttl              = Y
+drop                 = Y
+port_id              = Y
diff --git a/doc/guides/rel_notes/release_22_11.rst b/doc/guides/rel_notes/release_22_11.rst
index cd22d95..ac0223f 100644
--- a/doc/guides/rel_notes/release_22_11.rst
+++ b/doc/guides/rel_notes/release_22_11.rst
@@ -155,6 +155,12 @@  New Features
 
     * Ethernet
 
+  Add the support of rte_flow actions as follow:
+
+    * Count
+    * Drop
+    * TTL decrement
+
 * **Updated NXP dpaa2 driver.**
 
   * Added support for flow action REPRESENTED_PORT.
diff --git a/drivers/net/nfp/flower/nfp_flower_cmsg.h b/drivers/net/nfp/flower/nfp_flower_cmsg.h
index 75a3b91..d776e37 100644
--- a/drivers/net/nfp/flower/nfp_flower_cmsg.h
+++ b/drivers/net/nfp/flower/nfp_flower_cmsg.h
@@ -253,6 +253,17 @@  struct nfp_flower_mac_mpls {
 	rte_be32_t mpls_lse;
 };
 
+struct nfp_fl_act_head {
+	uint8_t jump_id;
+	uint8_t len_lw;
+};
+
+struct nfp_fl_act_output {
+	struct nfp_fl_act_head head;
+	rte_be16_t flags;
+	rte_be32_t port;
+};
+
 int nfp_flower_cmsg_mac_repr(struct nfp_app_fw_flower *app_fw_flower);
 int nfp_flower_cmsg_repr_reify(struct nfp_app_fw_flower *app_fw_flower,
 		struct nfp_flower_representor *repr);
diff --git a/drivers/net/nfp/nfp_flow.c b/drivers/net/nfp/nfp_flow.c
index a077423..ec021f0 100644
--- a/drivers/net/nfp/nfp_flow.c
+++ b/drivers/net/nfp/nfp_flow.c
@@ -558,6 +558,22 @@  struct nfp_mask_id_entry {
 		case RTE_FLOW_ACTION_TYPE_VOID:
 			PMD_DRV_LOG(DEBUG, "RTE_FLOW_ACTION_TYPE_VOID detected");
 			break;
+		case RTE_FLOW_ACTION_TYPE_MARK:
+			PMD_DRV_LOG(DEBUG, "RTE_FLOW_ACTION_TYPE_MARK detected");
+			break;
+		case RTE_FLOW_ACTION_TYPE_DROP:
+			PMD_DRV_LOG(DEBUG, "RTE_FLOW_ACTION_TYPE_DROP detected");
+			break;
+		case RTE_FLOW_ACTION_TYPE_COUNT:
+			PMD_DRV_LOG(DEBUG, "RTE_FLOW_ACTION_TYPE_COUNT detected");
+			break;
+		case RTE_FLOW_ACTION_TYPE_RSS:
+			PMD_DRV_LOG(DEBUG, "RTE_FLOW_ACTION_TYPE_RSS detected");
+			break;
+		case RTE_FLOW_ACTION_TYPE_PORT_ID:
+			PMD_DRV_LOG(DEBUG, "RTE_FLOW_ACTION_TYPE_PORT_ID detected");
+			key_ls->act_size += sizeof(struct nfp_fl_act_output);
+			break;
 		default:
 			PMD_DRV_LOG(ERR, "Action type %d not supported.", action->type);
 			return -ENOTSUP;
@@ -785,6 +801,96 @@  struct nfp_mask_id_entry {
 	return 0;
 }
 
+static int
+nfp_flow_action_output(char *act_data,
+		const struct rte_flow_action *action,
+		struct nfp_fl_rule_metadata *nfp_flow_meta)
+{
+	size_t act_size;
+	struct rte_eth_dev *ethdev;
+	struct nfp_fl_act_output *output;
+	struct nfp_flower_representor *representor;
+	const struct rte_flow_action_port_id *port_id;
+
+	port_id = action->conf;
+	if (port_id == NULL || port_id->id >= RTE_MAX_ETHPORTS)
+		return -ERANGE;
+
+	ethdev = &rte_eth_devices[port_id->id];
+	representor = (struct nfp_flower_representor *)ethdev->data->dev_private;
+	act_size = sizeof(struct nfp_fl_act_output);
+
+	output = (struct nfp_fl_act_output *)act_data;
+	output->head.jump_id = NFP_FL_ACTION_OPCODE_OUTPUT;
+	output->head.len_lw  = act_size >> NFP_FL_LW_SIZ;
+	output->flags        = rte_cpu_to_be_16(NFP_FL_OUT_FLAGS_LAST);
+	output->port         = rte_cpu_to_be_32(representor->port_id);
+
+	nfp_flow_meta->shortcut = rte_cpu_to_be_32(representor->port_id);
+
+	return 0;
+}
+
+static int
+nfp_flow_compile_action(__rte_unused struct nfp_flower_representor *representor,
+		const struct rte_flow_action actions[],
+		struct rte_flow *nfp_flow)
+{
+	int ret = 0;
+	char *position;
+	char *action_data;
+	bool drop_flag = false;
+	uint32_t total_actions = 0;
+	const struct rte_flow_action *action;
+	struct nfp_fl_rule_metadata *nfp_flow_meta;
+
+	nfp_flow_meta = nfp_flow->payload.meta;
+	action_data   = nfp_flow->payload.action_data;
+	position      = action_data;
+
+	for (action = actions; action->type != RTE_FLOW_ACTION_TYPE_END; ++action) {
+		switch (action->type) {
+		case RTE_FLOW_ACTION_TYPE_VOID:
+			break;
+		case RTE_FLOW_ACTION_TYPE_MARK:
+			PMD_DRV_LOG(DEBUG, "Process RTE_FLOW_ACTION_TYPE_MARK");
+			break;
+		case RTE_FLOW_ACTION_TYPE_DROP:
+			PMD_DRV_LOG(DEBUG, "Process RTE_FLOW_ACTION_TYPE_DROP");
+			drop_flag = true;
+			break;
+		case RTE_FLOW_ACTION_TYPE_COUNT:
+			PMD_DRV_LOG(DEBUG, "Process RTE_FLOW_ACTION_TYPE_COUNT");
+			break;
+		case RTE_FLOW_ACTION_TYPE_RSS:
+			PMD_DRV_LOG(DEBUG, "Process RTE_FLOW_ACTION_TYPE_RSS");
+			break;
+		case RTE_FLOW_ACTION_TYPE_PORT_ID:
+			PMD_DRV_LOG(DEBUG, "Process RTE_FLOW_ACTION_TYPE_PORT_ID");
+			ret = nfp_flow_action_output(position, action, nfp_flow_meta);
+			if (ret != 0) {
+				PMD_DRV_LOG(ERR, "Failed when process"
+						" RTE_FLOW_ACTION_TYPE_PORT_ID");
+				return ret;
+			}
+
+			position += sizeof(struct nfp_fl_act_output);
+			break;
+		default:
+			PMD_DRV_LOG(ERR, "Unsupported action type: %d", action->type);
+			return -ENOTSUP;
+		}
+		total_actions++;
+	}
+
+	if (drop_flag)
+		nfp_flow_meta->shortcut = rte_cpu_to_be_32(NFP_FL_SC_ACT_DROP);
+	else if (total_actions > 1)
+		nfp_flow_meta->shortcut = rte_cpu_to_be_32(NFP_FL_SC_ACT_NULL);
+
+	return 0;
+}
+
 static struct rte_flow *
 nfp_flow_process(struct nfp_flower_representor *representor,
 		const struct rte_flow_item items[],
@@ -833,6 +939,12 @@  struct nfp_mask_id_entry {
 		goto free_flow;
 	}
 
+	ret = nfp_flow_compile_action(representor, actions, nfp_flow);
+	if (ret != 0) {
+		PMD_DRV_LOG(ERR, "nfp flow action process failed.");
+		goto free_flow;
+	}
+
 	nfp_flow_meta = nfp_flow->payload.meta;
 	mask_data = nfp_flow->payload.mask_data;
 	mask_len = key_layer.key_size;
diff --git a/drivers/net/nfp/nfp_flow.h b/drivers/net/nfp/nfp_flow.h
index 83f9f29..f1fbada 100644
--- a/drivers/net/nfp/nfp_flow.h
+++ b/drivers/net/nfp/nfp_flow.h
@@ -33,6 +33,43 @@ 
 /* The firmware expects lengths in units of long words */
 #define NFP_FL_LW_SIZ                   2
 
+#define NFP_FL_SC_ACT_DROP      0x80000000
+#define NFP_FL_SC_ACT_USER      0x7D000000
+#define NFP_FL_SC_ACT_POPV      0x6A000000
+#define NFP_FL_SC_ACT_NULL      0x00000000
+
+/* Action opcodes */
+#define NFP_FL_ACTION_OPCODE_OUTPUT             0
+#define NFP_FL_ACTION_OPCODE_PUSH_VLAN          1
+#define NFP_FL_ACTION_OPCODE_POP_VLAN           2
+#define NFP_FL_ACTION_OPCODE_PUSH_MPLS          3
+#define NFP_FL_ACTION_OPCODE_POP_MPLS           4
+#define NFP_FL_ACTION_OPCODE_USERSPACE          5
+#define NFP_FL_ACTION_OPCODE_SET_TUNNEL         6
+#define NFP_FL_ACTION_OPCODE_SET_ETHERNET       7
+#define NFP_FL_ACTION_OPCODE_SET_MPLS           8
+#define NFP_FL_ACTION_OPCODE_SET_IPV4_ADDRS     9
+#define NFP_FL_ACTION_OPCODE_SET_IPV4_TTL_TOS   10
+#define NFP_FL_ACTION_OPCODE_SET_IPV6_SRC       11
+#define NFP_FL_ACTION_OPCODE_SET_IPV6_DST       12
+#define NFP_FL_ACTION_OPCODE_SET_IPV6_TC_HL_FL  13
+#define NFP_FL_ACTION_OPCODE_SET_UDP            14
+#define NFP_FL_ACTION_OPCODE_SET_TCP            15
+#define NFP_FL_ACTION_OPCODE_PRE_LAG            16
+#define NFP_FL_ACTION_OPCODE_PRE_TUNNEL         17
+#define NFP_FL_ACTION_OPCODE_PRE_GS             18
+#define NFP_FL_ACTION_OPCODE_GS                 19
+#define NFP_FL_ACTION_OPCODE_PUSH_NSH           20
+#define NFP_FL_ACTION_OPCODE_POP_NSH            21
+#define NFP_FL_ACTION_OPCODE_SET_QUEUE          22
+#define NFP_FL_ACTION_OPCODE_CONNTRACK          23
+#define NFP_FL_ACTION_OPCODE_METER              24
+#define NFP_FL_ACTION_OPCODE_CT_NAT_EXT         25
+#define NFP_FL_ACTION_OPCODE_PUSH_GENEVE        26
+#define NFP_FL_ACTION_OPCODE_NUM                32
+
+#define NFP_FL_OUT_FLAGS_LAST            (1 << 15)
+
 /* Tunnel ports */
 #define NFP_FL_PORT_TYPE_TUN            0x50000000