Hi,
This patchset introduces a kernel intermediate representation (IR) to
express ACL hardware offloads, this is heavily based on the existing
flow dissector infrastructure and the TC actions. This IR can be used by
different frontend ACL interfaces such as ethtool_rxnfc, tc and
netfilter to represent ACL hardware offloads.
The main goals of this patchset are:
* Provide an unified representation that can be passed to the driver
for translation to HW IR. This consolidates the code to be maintained
in the driver and it also simplifies the development of ACL hardware
offloads. Driver developers do not need to add one specific parser for
each supported ACL frontend, instead each frontend just generates
this flow_rule IR and pass it to drivers to populate the hardware IR.
* Do not expose TC software frontend details to the drivers anymore,
such as structure layouts. Hence, if TC needs to be updated to support
a new software feature, no changes to the drivers will be required.
In handcrafted ascii art, the idea is the following:
. ethtool_rxnfc tc netfilter
| (ioctl) (netlink) (netlink)
| | | | translate native
Frontend | | | | interface representation
| | | | to flow_rule IR
| | | |
. \/ \/ \/
. ----- flow_rule IR ------
| |
Drivers | | parsing of flow_rule IR
| | to populate hardware IR
| \/
. hardware IR (driver)
This patchset only converts tc and ethtool_rxnfc to use this
infrastructure. Therefore. this patchset comes with no netfilter changes
at this stage.
The proposed object that represents rules is the following:
struct flow_rule {
struct flow_match match;
struct flow_action action;
};
The flow_match structure wraps Jiri Pirko's existing representation
available in cls_flower based on flow dissectors to represent the
matching side:
struct flow_match {
struct flow_dissector *dissector;
void *mask;
void *key;
};
The mask and key layouts are opaque, given the dissector object provides
the used_keys flags - to check for rule selectors that are being used -
and the offset to the corresponding key and mask in the opaque
container structures.
Then, the actions to be performed on the matching packets is represented
through the flow_action object:
struct flow_action {
unsigned int num_entries;
struct flow_action_entry entries[0];
};
This object comes with a num_entries field that specifies the number of
actions - this supports an arbitrary number of actions, the driver will
impose its own restrictions on this - and the array that stores
flow_action_entries structures (entries). Each flow action entry is
defined as it follows:
struct flow_action_entry {
enum flow_action_id id;
union {
u32 chain_index; /* FLOW_ACTION_GOTO */
struct net_device *dev; /* FLOW_ACTION_REDIRECT
*/
struct { /* FLOW_ACTION_VLAN */
u16 vid;
__be16 proto;
u8 prio;
} vlan;
struct { /*
FLOW_ACTION_PACKET_EDIT */
enum flow_action_mangle_base htype;
u32 offset;
u32 mask;
u32 val;
} mangle;
const struct ip_tunnel_info *tunnel; /*
FLOW_ACTION_TUNNEL_ENCAP */
u32 csum_flags; /* FLOW_ACTION_CSUM */
u32 mark; /* FLOW_ACTION_MARK */
struct { /* FLOW_ACTION_QUEUE */
u32 ctx;
u32 index;
u8 vf;
} queue;
};
};
Possible actions are extracted from what existing drivers are already
supporting through tc-flower and ethtool_rxnfc interfaces:
enum flow_action_id {
FLOW_ACTION_ACCEPT = 0,
FLOW_ACTION_DROP,
FLOW_ACTION_TRAP,
FLOW_ACTION_GOTO,
FLOW_ACTION_REDIRECT,
FLOW_ACTION_MIRRED,
FLOW_ACTION_VLAN_PUSH,
FLOW_ACTION_VLAN_POP,
FLOW_ACTION_VLAN_MANGLE,
FLOW_ACTION_TUNNEL_ENCAP,
FLOW_ACTION_TUNNEL_DECAP,
FLOW_ACTION_MANGLE,
FLOW_ACTION_ADD,
FLOW_ACTION_CSUM,
FLOW_ACTION_MARK,
FLOW_ACTION_WAKE,
FLOW_ACTION_QUEUE,
};
Common code pattern from the drivers to populate the hardware
intermediate representation looks like this:
if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
struct flow_match_ipv4_addrs match;
flow_rule_match_ipv4_addrs(rule, &match);
flow->l3_key.ipv4.daddr.s_addr = match.key->dst;
flow->l3_mask.ipv4.daddr.s_addr = match.mask->dst;
flow->l3_key.ipv4.saddr.s_addr = match.key->src;
flow->l3_mask.ipv4.saddr.s_addr = match.mask->src;
}
Then, flow action code parser should look like:
flow_action_for_each(i, act, flow_action) {
switch (act->id) {
case FLOW_ACTION_DROP:
actions->flags |= DRIVER_XYZ_ACTION_FLAG_DROP;
break;
case ...:
break;
default:
return -EOPNOTSUPP;
}
}
This new infrastructure resides in:
net/core/flow_offload.c
include/net/flow_offload.h
net/core/ethtool.c (translator from ethtool_rxnfc to flow_rule)
This patchset is composed of 12 patches:
Patch #1 adds the flow_match structure, this includes the
flow_rule_match_key() interface to check for existing selectors
that are in used in the rule and the flow_rule_match_*()
functions to fetch the selector value and the mask. This
also introduces the initial flow_rule structure skeleton to
avoid a follow up patch that would update the same LoCs.
Patch #2 makes changes to packet edit parser of mlx5e driver, to prepare
introduction of the new flow_action to mangle packets.
Patch #3 Introduce flow_action infrastructure. This infrastructure is
based on the TC actions. Patch #8 extends it so it also
supports two new actions that are only available through the
ethtool_rxnfc interface.
Patch #4 Add function to translate TC action to flow_action from
cls_flower.
Patch #5 Add infrastructure to fetch statistics into container structure
and synchronize them to TC actions from cls_flower. Another
preparation patch before patch #7, so we can stop exposing the
TC action native layout to the drivers.
Patch #6 Use flow_action infrastructure from drivers.
Patch #7 Do not expose TC actions to drivers anymore, now that drivers
have been converted to use the flow_action infrastructure after
patch #6.
Patch #8 Support for wake-up-on-lan and queue actions for the flow_action
infrastructure, two actions supported by the ethtool_rxnfc
interface.
Patch #9 Add a function to translate from ethtool_rx_flow_spec structure
to the flow_action structure. This function maps the ethtool_rxnfc
structure to flow_rule. This function resides in net/core/ethtool.c
Patch #10 Use ethtool_rxnfc to flow_rule from bcm_sf2 driver.
Patch #11 A preparation patch for the qlogic/qede driver coming at #12.
Patch #12 Update qlogic/qede driver to use the flow_rule infrastructure.
This removes the duplicated parser for ethtool_rxnfc by using
flow_rule structure.
Please apply, thanks!
P.S: Moving forward, my rough plan is to propose and discuss the
following changes to use this infrastructure from netfilter for ACL
hardware offload:
* Rename tc_setup ndo to flow_offload (or similar), so it can be used from
netfilter too. Otherwise, I'm open for alternatives.
* Add a conversion function from native netfilter representation to
flow_rule, including extra new glue code from the two-phase commit
protocol to integrate this infrastructure.
Pablo Neira Ayuso (12):
flow_offload: add flow_rule and flow_match structures and use them
net/mlx5e: support for two independent packet edit actions
flow_offload: add flow action infrastructure
cls_api: add translator to flow_action representation
flow_offload: add statistics retrieval infrastructure and use it
drivers: net: use flow action infrastructure
cls_flower: don't expose TC actions to drivers anymore
flow_offload: add wake-up-on-lan and queue to flow_action
ethtool: add ethtool_rx_flow_spec to flow_rule structure translator
dsa: bcm_sf2: use flow_rule infrastructure
qede: place ethtool_rx_flow_spec after code after TC flower codebase
qede: use ethtool_rx_flow_rule() to remove duplicated parser code
drivers/net/dsa/bcm_sf2_cfp.c | 102 +--
drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c | 252 ++++----
.../net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c | 450 ++++++-------
drivers/net/ethernet/intel/i40e/i40e_main.c | 178 ++----
drivers/net/ethernet/intel/iavf/iavf_main.c | 195 +++---
drivers/net/ethernet/intel/igb/igb_main.c | 64 +-
.../net/ethernet/mellanox/mlx5/core/en/tc_tun.c | 68 +-
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 697 ++++++++++-----------
drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c | 2 +-
.../net/ethernet/mellanox/mlxsw/spectrum_flower.c | 258 ++++----
drivers/net/ethernet/netronome/nfp/flower/action.c | 198 +++---
drivers/net/ethernet/netronome/nfp/flower/match.c | 417 ++++++------
.../net/ethernet/netronome/nfp/flower/offload.c | 150 ++---
drivers/net/ethernet/qlogic/qede/qede_filter.c | 572 +++++++----------
include/linux/ethtool.h | 15 +
include/net/flow_offload.h | 203 ++++++
include/net/pkt_cls.h | 18 +-
net/core/Makefile | 2 +-
net/core/ethtool.c | 241 +++++++
net/core/flow_offload.c | 153 +++++
net/sched/cls_api.c | 116 ++++
net/sched/cls_flower.c | 71 ++-
22 files changed, 2416 insertions(+), 2006 deletions(-)
create mode 100644 include/net/flow_offload.h
create mode 100644 net/core/flow_offload.c
--
2.11.0
