> -----Original Message-----
> From: Iremonger, Bernard
> Sent: Wednesday, December 6, 2017 12:41 PM
> To: Iremonger, Bernard <bernard.iremon...@intel.com>; Singh, Jasvinder
> <jasvinder.si...@intel.com>; dev@dpdk.org
> Subject: RE: [PATCH 2/3] lib/librte_flow_classy: add run api for flow
> classification
>
> Hi Jasvinder,
>
> <snip>
>
> > > > -----Original Message-----
> > > > From: Singh, Jasvinder
> > > > Sent: Thursday, November 23, 2017 11:32 AM
> > > > To: dev@dpdk.org
> > > > Cc: Iremonger, Bernard <bernard.iremon...@intel.com>
> > > > Subject: [PATCH 2/3] lib/librte_flow_classy: add run api for flow
> > > > classification
> > > >
> > > > This patch extends the flow classification library by adding run api.
> > > > This function classifies the packets based on the flow rules
> > > > stored in the classifier table.
> > > > During lookup operation, the table entry is identified on lookup
> > > > hit and based on meta-data stored at table entry, actions are
> > > > performed on the current packet.
> > > > The meta-information about the actions stored in the table entry
> > > > is determined from the actions fields specified in flow rules.
> > > >
> > > > Signed-off-by: Jasvinder Singh <jasvinder.si...@intel.com>
> > > > ---
> > > > lib/librte_flow_classify/rte_flow_classify.c | 56
> > > > ++++++++++++++++++++++
> > > > lib/librte_flow_classify/rte_flow_classify.h | 24 ++++++++++
> > > > .../rte_flow_classify_version.map | 1 +
> > > > 3 files changed, 81 insertions(+)
> > > >
> > > > diff --git a/lib/librte_flow_classify/rte_flow_classify.c
> > > > b/lib/librte_flow_classify/rte_flow_classify.c
> > > > index ff1bc86..5433bfe 100644
> > > > --- a/lib/librte_flow_classify/rte_flow_classify.c
> > > > +++ b/lib/librte_flow_classify/rte_flow_classify.c
> > > > @@ -37,6 +37,9 @@
> > > > #include <rte_table_acl.h>
> > > > #include <stdbool.h>
> > > >
> > > > +#define RTE_PKT_METADATA_PTR(pkt, offset) \
> > > > + (&((uint32_t *)(pkt))[offset])
> > > > +
> > > > int librte_flow_classify_logtype;
> > > >
> > > > static uint32_t unique_id = 1;
> > > > @@ -674,6 +677,59 @@ action_apply(struct rte_flow_classifier *cls,
> > > > }
> > > >
> > > > int
> > > > +rte_flow_classifier_run(struct rte_flow_classifier *cls,
> > > > + struct rte_mbuf **pkts,
> > > > + const uint16_t nb_pkts,
> > > > + uint32_t pkt_offset)
> > > > +{
> > > > + struct rte_flow_action_mark *mark;
> > > > + struct classify_action *action;
> > > > + uint64_t pkts_mask = RTE_LEN2MASK(nb_pkts, uint64_t);
> > > > + uint64_t action_mask;
> > > > + uint32_t *ptr, i, j;
> > > > + int ret = -EINVAL;
> > > > +
> > > > + if (!cls || !pkts || nb_pkts == 0)
> > > > + return ret;
> > > > +
> > > > + for (i = 0; i < cls->num_tables; i++) {
> > > > + if (cls->table_mask & (1LU << i)) {
> > > > + struct rte_cls_table *table = &cls->tables[i];
> > > > + uint64_t lookup_hit_mask;
> > > > +
> > > > + ret = table->ops.f_lookup(table->h_table,
> > > > + pkts, pkts_mask, &lookup_hit_mask,
> > > > + (void **)cls->entries);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + if (lookup_hit_mask) {
> > > > + for (j = 0; j < nb_pkts; j++) {
> > > > + uint64_t pkt_mask = 1LLU << j;
> > > > +
> > > > + if ((lookup_hit_mask &
> > > > pkt_mask) ==
> > > > 0)
> > > > + continue;
> > > > + /* Meta-data */
> > > > + enum rte_flow_action_type
> > > > act_type
> > > > =
> > > > +
> > > > RTE_FLOW_ACTION_TYPE_MARK;
> > > > + action =
> > > > &cls->entries[j]->action;
> > > > + action_mask =
> > > > action->action_mask;
> > > > +
> > > > + if (action_mask & (1LLU <<
> > > > act_type))
> > > > {
> > > > + mark =
> > > > &action->act.mark;
> > > > + ptr =
> > > > RTE_PKT_METADATA_PTR(
> > > > + pkts[j],
> > > > pkt_offset);
> > > > + *ptr = mark->id;
> > > > + }
> > > > + }
> > > > + }
> > > > + }
> > > > + }
> > > > +
> > > > + return 0;
> > > > +}
> > > > +
> > > > +int
> > > > rte_flow_classifier_query(struct rte_flow_classifier *cls,
> > > > struct rte_mbuf **pkts,
> > > > const uint16_t nb_pkts,
> > > > diff --git a/lib/librte_flow_classify/rte_flow_classify.h
> > > > b/lib/librte_flow_classify/rte_flow_classify.h
> > > > index b9b669f..b74bd11 100644
> > > > --- a/lib/librte_flow_classify/rte_flow_classify.h
> > > > +++ b/lib/librte_flow_classify/rte_flow_classify.h
> > > > @@ -273,6 +273,30 @@ rte_flow_classify_table_entry_delete(struct
> > > > rte_flow_classifier *cls,
> > > > struct rte_flow_classify_rule *rule);
> > > >
> > > > /**
> > > > + * Flow classifier run.
> > > > + *
> > > > + * As a result of lookup operation, flow classifer idenfies the
> > > > + * table entries that are hit and executes the actions on the packets.
> > > > + *
> > > > + * @param[in] cls
> > > > + * Flow classifier handle
> > > > + * @param[in] pkts
> > > > + * Pointer to packets to process
> > > > + * @param[in] nb_pkts
> > > > + * Number of packets to process
> > > > + * @param[in] pkt_offset
> > > > + * Offset to store action metadata in the mbuf headroom
> > > > + *
> > > > + * @return
> > > > + * 0 on success, error code otherwise.
> > > > + */
> > > > +int
> > > > +rte_flow_classifier_run(struct rte_flow_classifier *cls,
> > > > + struct rte_mbuf **pkts,
> > > > + const uint16_t nb_pkts,
> > > > + uint32_t pkt_offset);
> > > > +
> > > > +/**
> > > > * Query flow classifier for given rule.
> > > > *
> > > > * @param[in] cls
> > > > diff --git
> > > > a/lib/librte_flow_classify/rte_flow_classify_version.map
> > > > b/lib/librte_flow_classify/rte_flow_classify_version.map
> > > > index 49bc25c..b51cb1a 100644
> > > > --- a/lib/librte_flow_classify/rte_flow_classify_version.map
> > > > +++ b/lib/librte_flow_classify/rte_flow_classify_version.map
> > > > @@ -4,6 +4,7 @@ EXPERIMENTAL {
> > > > rte_flow_classifier_create;
> > > > rte_flow_classifier_free;
> > > > rte_flow_classifier_query;
> > > > + rte_flow_classifier_run;
> > > > rte_flow_classify_table_create;
> > > > rte_flow_classify_table_entry_add;
> > > > rte_flow_classify_table_entry_delete;
> > > > --
> > > > 2.9.3
> > >
> <snip>
>
> Tests should be added to test_flow_classify.c for the
> rte_flow_classifier_run() API.
> The flow_classify sample application should exercise the
> rte_flow_classifier_run() API.
Ok, will do in v2.
Jasvinder
