Hi,
> -----Original Message-----
> From: Alexander Kozyrev <akozy...@nvidia.com>
> Sent: Tuesday, January 12, 2021 4:16 PM
> To: Ori Kam <or...@nvidia.com>; dev@dpdk.org
> Cc: Slava Ovsiienko <viachesl...@nvidia.com>; NBU-Contact-Thomas Monjalon
> <tho...@monjalon.net>; ferruh.yi...@intel.com;
> andrew.rybche...@oktetlabs.ru
> Subject: RE: [PATCH] ethdev: introduce generic copy rte flow action
>
> > From: Ori Kam <or...@nvidia.com> on Sunday, January 10, 2021 3:01
> > Hi Alexander,
> >
> > I guess that the test-pmd part will be available later right?
> Yes, please take a look at v2 version for testpmd implementation.
>
> > > -----Original Message-----
> > > From: Alexander Kozyrev <akozy...@nvidia.com>
> > > Sent: Friday, January 8, 2021 8:33 AM
> > > Subject: [PATCH] ethdev: introduce generic copy rte flow action
> > >
> > > Implement a generic copy flow API to allow copying of an arbitrary
> > > header field (as well as mark, metadata or tag) to another item.
> > >
> > > This generic copy mechanism removes the necessity to implement a
> > > separate RTE Flow action every time we need to modify a new packet
> > > field in the future. A user-provided value can be used from a
> > > specified tag/metadata or directly copied from other packet field.
> > >
> > > The number of bits to copy as well as the offset to start from can
> > > be specified to allow a partial copy or copy into an arbitrary
> > > place in a packet for greater flexibility.
> > >
> >
> > Since the question why you are using enum and not just offset from
> > the start of the packet, was discussed and raised by number of people it
> > will
> be
> > best
> > if it will appear in the commit log, at least the advantages to this
> > implementation.
> Ok, will add to v3 commit message.
>
> > > RFC:
> > >
> > https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatches.
> > d
> > >
> >
> pdk.org%2Fpatch%2F85384%2F&data=04%7C01%7Corika%40nvidia.com%
> > >
> >
> 7Cd04c2e49c3a840994da408d8b39f3304%7C43083d15727340c1b7db39efd9cc
> > >
> >
> c17a%7C0%7C0%7C637456843629116269%7CUnknown%7CTWFpbGZsb3d8eyJ
> > >
> >
> WIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C
> > >
> 1000&sdata=85096rASBtzbjU42pV6sPkl3nVt5HlR6%2BL9nxI3qgFA%3D&a
> > > mp;reserved=0
> > >
> > > Signed-off-by: Alexander Kozyrev <akozy...@nvidia.com>
> > > ---
> > > doc/guides/prog_guide/rte_flow.rst | 35 ++++++++++++++++++
> > > lib/librte_ethdev/rte_flow.c | 1 +
> > > lib/librte_ethdev/rte_flow.h | 59 ++++++++++++++++++++++++++++++
> > > 3 files changed, 95 insertions(+)
> > >
> > > diff --git a/doc/guides/prog_guide/rte_flow.rst
> > > b/doc/guides/prog_guide/rte_flow.rst
> > > index 86b3444803..b737ff9dad 100644
> > > --- a/doc/guides/prog_guide/rte_flow.rst
> > > +++ b/doc/guides/prog_guide/rte_flow.rst
> > > @@ -2766,6 +2766,41 @@ The behaviour of the shared action defined by
> > > ``action`` argument of type
> > > | no properties |
> > > +---------------+
> > >
> > > +Action: ``COPY_ITEM``
> > > +^^^^^^^^^^^^^^^^^^^^^
> > > +
> > > +Copy ``width`` bits from ``src`` item to ``dst`` item.
> > > +
> > > +An arbitrary header field (as well as mark, metadata or tag values)
> > > +can be used as both source and destination items as set by ``item``.
> > > +
> >
> > For tag I think you should also use the index right?
> Right, index is here to access any element in the tag array in addition to
> outermost or any inner packet header fields. Will specify explicitly in the
> doc.
>
> > > +Inner packet header fields can be accessed using the ``index`` and
> > > +it is possible to start the copy from the ``offset`` bits in an item.
> >
> > Please specify what means index 0 /1 ... 0 is outer most? Inner most?
> > You can look at the RSS level for reference.
> > I think it will be best to use the same values.
> 0 is outer most, of course, I'Il add some clarification about that.
>
Not necessary, like I said please look at the RSS, 0 can mean the inner most.
> > What happens if we want to copy between different sizes?
> > for example copy IPV6 src to number of tags? I assume we will be using
> > offset
> > and
> > split the copy command to number of actions right?
> That is the correct understanding. We can utilize 4 copy_item actions in this
> case to
> copy 32-bits of an IPv6 SRC at the time (specifying different offsets) to 4
> different Tag fields.
>
> > > +
> > > +.. _table_rte_flow_action_copy_item:
> > > +
> > > +.. table:: COPY_ITEM
> > > +
> > > + +-----------------------------------------+
> > > + | Field | Value |
> > > + +===============+=========================+
> > > + | ``dst`` | destination item |
> > > + | ``src`` | source item |
> > > + | ``width`` | number of bits to copy |
> > > + +---------------+-------------------------+
> > > +
> > > +.. _table_rte_flow_action_copy_data:
> > > +
> > > +.. table:: destination/source item definition
> > > +
> > > + +----------------------------------------------------------+
> > > + | Field | Value |
> > > +
> +===============+==========================================+
> > > + | ``item`` | ID of a packet field/mark/metadata/tag |
> > > + | ``index`` | index of outer/inner header or tag array |
> > > + | ``offset`` | number of bits to skip during the copy |
> > > + +---------------+------------------------------------------+
> > > +
> > > Negative types
> > > ~~~~~~~~~~~~~~
> > >
> > > diff --git a/lib/librte_ethdev/rte_flow.c b/lib/librte_ethdev/rte_flow.c
> > > index a06f64c271..fdbabefc47 100644
> > > --- a/lib/librte_ethdev/rte_flow.c
> > > +++ b/lib/librte_ethdev/rte_flow.c
> > > @@ -176,6 +176,7 @@ static const struct rte_flow_desc_data
> > > rte_flow_desc_action[] = {
> > > MK_FLOW_ACTION(SET_IPV6_DSCP, sizeof(struct
> > > rte_flow_action_set_dscp)),
> > > MK_FLOW_ACTION(AGE, sizeof(struct rte_flow_action_age)),
> > > MK_FLOW_ACTION(SAMPLE, sizeof(struct rte_flow_action_sample)),
> > > + MK_FLOW_ACTION(COPY_ITEM, sizeof(struct
> > > rte_flow_action_copy_item)),
> > > /**
> > > * Shared action represented as handle of type
> > > * (struct rte_flow_shared action *) stored in conf field (see
> > > diff --git a/lib/librte_ethdev/rte_flow.h b/lib/librte_ethdev/rte_flow.h
> > > index 0977a78270..0540c861fb 100644
> > > --- a/lib/librte_ethdev/rte_flow.h
> > > +++ b/lib/librte_ethdev/rte_flow.h
> > > @@ -2198,6 +2198,16 @@ enum rte_flow_action_type {
> > > * struct rte_flow_shared_action).
> > > */
> > > RTE_FLOW_ACTION_TYPE_SHARED,
> > > +
> > > + /**
> > > + * Copy a packet header field, tag, mark or metadata.
> > > + *
> > > + * Allow saving an arbitrary header field by copying its value
> > > + * to a tag/mark/metadata or copy it into another header field.
> > > + *
> > > + * See struct rte_flow_action_copy_item.
> > > + */
> > > + RTE_FLOW_ACTION_TYPE_COPY_ITEM,
> > > };
> > >
> > > /**
> > > @@ -2791,6 +2801,55 @@ struct rte_flow_action_set_dscp {
> > > */
> > > struct rte_flow_shared_action;
> > >
> > > +enum rte_flow_item_id {
> > > + RTE_FLOW_ITEM_NONE = 0,
> > > + RTE_FLOW_ITEM_MAC_DST,
> > > + RTE_FLOW_ITEM_MAC_SRC,
> > > + RTE_FLOW_ITEM_VLAN_TYPE,
> > > + RTE_FLOW_ITEM_VLAN_ID,
> > > + RTE_FLOW_ITEM_MAC_TYPE,
> > > + RTE_FLOW_ITEM_IPV4_DSCP,
> > > + RTE_FLOW_ITEM_IPV4_TTL,
> > > + RTE_FLOW_ITEM_IPV4_SRC,
> > > + RTE_FLOW_ITEM_IPV4_DST,
> > > + RTE_FLOW_ITEM_IPV6_HOPLIMIT,
> > > + RTE_FLOW_ITEM_IPV6_SRC,
> > > + RTE_FLOW_ITEM_IPV6_DST,
> > > + RTE_FLOW_ITEM_TCP_PORT_SRC,
> > > + RTE_FLOW_ITEM_TCP_PORT_DST,
> > > + RTE_FLOW_ITEM_TCP_SEQ_NUM,
> > > + RTE_FLOW_ITEM_TCP_ACK_NUM,
> > > + RTE_FLOW_ITEM_TCP_FLAGS,
> > > + RTE_FLOW_ITEM_UDP_PORT_SRC,
> > > + RTE_FLOW_ITEM_UDP_PORT_DST,
> > > + RTE_FLOW_ITEM_VXLAN_VNI,
> > > + RTE_FLOW_ITEM_GENEVE_VNI,
> > > + RTE_FLOW_ITEM_GTP_TEID,
> > > + RTE_FLOW_ITEM_TAG,
> > > + RTE_FLOW_ITEM_MARK,
> > > + RTE_FLOW_ITEM_META,
> > > +};
> >
> > I don't think this name is good since it not rte_flow_item this is just
> > internal
> > enumeration
> > for this action.
> Will rename to rte_flow_action_copy_item_id in v3? Is this ok?
>
Yes better,
Or rte_flow_copy_item_id?
> > > +
> > > +struct rte_flow_action_copy_data {
> > > + enum rte_flow_item_id item;
> > > + uint32_t index;
> > > + uint32_t offset;
> >
> > Why use 32 bits? Since this copy only one register with max len of 32 bit.
> > The max offset can 31? Same for the index.
> This extends the flexibility of the copy API. This way you can modify any
> place
> in a packet as you desire by specifying RTE_FLOW_ITEM_NONE to start with
> the
> beginning of a packet and choosing a particular offset value to point in a
> desired place.
> And since packet length can be pretty big we need to accommodate this.
> Do you think this flexibility is not necessary and we should bound the offset
> to
> the
> length of a selected packet field only? Index is 32 bit for the same field in
> RSS.
>
Nice idea but then I would add ITEM_OFFSET or something
around this name, I don't think NONE is a valid value.
> > > +};
> > > +
> > > +/**
> > > + * RTE_FLOW_ACTION_TYPE_COPY_ITEM
> > > + *
> > > + * Copies a specified number of bits from a source header field
> > > + * to a destination header field. Tag, mark or metadata can also
> > > + * be used as a source/destination to allow saving/overwriting
> > > + * an arbituary header field with a user-specified value.
> > > + */
> > > +struct rte_flow_action_copy_item {
> > > + struct rte_flow_action_copy_data dst;
> > > + struct rte_flow_action_copy_data src;
> > > + uint32_t width;
> >
> > Why use 32 bit register?
> Again, to make API as generic as possible in case there will be a PMD driver
> that can copy a huge chunk of a packet into the another place of this packet.
> What is your opinion on that?
>
Nice thinking,
> >
> > > +};
> > > +
> > > /* Mbuf dynamic field offset for metadata. */
> > > extern int32_t rte_flow_dynf_metadata_offs;
> > >
> > > --
> > > 2.24.1
> >
> > Best,
> > Ori
