Skip to contentSkip to conversation

cxgb4: add TC-MATCHALL classifier offload

PATCH Superseded Rahul Lakkireddy 3 participants 11 messages Nov 15, 2019

Subsystems (3)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v3 cover cxgb4: add TC-MATCHALL classifier offload

This series of patches add support to offload TC-MATCHALL classifier to hardware to classify all outgoing and incoming traffic on the underlying port. Only 1 egress and 1 ingress rule each can be offloaded on the underlying port.

Patch 1 adds support for TC-MATCHALL classifier offload on the egress side. TC-POLICE is the only action that can be offloaded on the egress side and is used to rate limit all outgoing traffic to specified max rate.

Patch 2 adds support for TC-MATCHALL classifier offload on the ingress side. The same set of actions supported by existing TC-FLOWER classifier offload can be applied on all the incoming traffic.

Thanks,
Rahul

---
v3:
- Added check in patch 1 to reject police offload if prio is not 1.
- Assign block_shared variable only for TC_SETUP_BLOCK in patch 1.

v2:
- Added check to reject flow block sharing for policers in patch 1.
- Removed logic to fetch free index from end of TCAM in patch 2. Must maintain the same ordering as in kernel.

Rahul Lakkireddy (2):
cxgb4: add TC-MATCHALL classifier egress offload
cxgb4: add TC-MATCHALL classifier ingress offload

 drivers/net/ethernet/chelsio/cxgb4/Makefile   |   3 +-
 drivers/net/ethernet/chelsio/cxgb4/cxgb4.h    |   6 +
 .../net/ethernet/chelsio/cxgb4/cxgb4_main.c   |  79 +++-
 .../ethernet/chelsio/cxgb4/cxgb4_tc_flower.c  |  21 +-
 .../ethernet/chelsio/cxgb4/cxgb4_tc_flower.h  |   6 +
 .../chelsio/cxgb4/cxgb4_tc_matchall.c         | 345 ++++++++++++++++++
 .../chelsio/cxgb4/cxgb4_tc_matchall.h         |  50 +++
 .../ethernet/chelsio/cxgb4/cxgb4_tc_mqprio.c  |   5 +-
 drivers/net/ethernet/chelsio/cxgb4/sched.c    |  56 ++-
 drivers/net/ethernet/chelsio/cxgb4/sched.h    |   1 +
 drivers/net/ethernet/chelsio/cxgb4/t4_hw.c    |  11 +-
 11 files changed, 543 insertions(+), 40 deletions(-)
 create mode 100644 drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
 create mode 100644 drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
Signature · 1 line
2.24.0
#1 lore raw

PATCH v3 1/2 cxgb4: add TC-MATCHALL classifier egress offload

Add TC-MATCHALL classifier offload with TC-POLICE action applied for all outgoing traffic on the underlying interface. Split flow block offload to support both egress and ingress classification.

For example, to rate limit all outgoing traffic to 1 Gbps:

$ tc qdisc add dev enp2s0f4 clsact
$ tc filter add dev enp2s0f4 egress prio 1 matchall skip_sw \
	action police rate 1Gbit burst 8Kbit

Note that skip_sw is important. Otherwise, both stack and hardware will end up doing policing. Priority must be set to 1 and policing can't be shared across flow blocks. Only 1 egress matchall rule can be active at a time on the underlying interface.

---
v3:
- Added check to reject police offload if prio is not 1.
- Assign block_shared variable only for TC_SETUP_BLOCK.

v2:
- Added check to reject flow block sharing for policers.

 drivers/net/ethernet/chelsio/cxgb4/Makefile   |   3 +-
 drivers/net/ethernet/chelsio/cxgb4/cxgb4.h    |   6 +
 .../net/ethernet/chelsio/cxgb4/cxgb4_main.c   |  72 +++++-
 .../chelsio/cxgb4/cxgb4_tc_matchall.c         | 217 ++++++++++++++++++
 .../chelsio/cxgb4/cxgb4_tc_matchall.h         |  35 +++
 .../ethernet/chelsio/cxgb4/cxgb4_tc_mqprio.c  |   5 +-
 drivers/net/ethernet/chelsio/cxgb4/sched.c    |  56 +++--
 drivers/net/ethernet/chelsio/cxgb4/sched.h    |   1 +
 drivers/net/ethernet/chelsio/cxgb4/t4_hw.c    |  11 +-
 9 files changed, 377 insertions(+), 29 deletions(-)
 create mode 100644 drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
 create mode 100644 drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
drivers/net/ethernet/chelsio/cxgb4/Makefile +2 -1
index 49a19308073b..a4b4d475abf8 100644
@@ -8,7 +8,8 @@ obj-$(CONFIG_CHELSIO_T4) += cxgb4.o
 cxgb4-objs := cxgb4_main.o l2t.o smt.o t4_hw.o sge.o clip_tbl.o cxgb4_ethtool.o \
 	      cxgb4_uld.o srq.o sched.o cxgb4_filter.o cxgb4_tc_u32.o \
 	      cxgb4_ptp.o cxgb4_tc_flower.o cxgb4_cudbg.o cxgb4_mps.o \
-	      cudbg_common.o cudbg_lib.o cudbg_zlib.o cxgb4_tc_mqprio.o
+	      cudbg_common.o cudbg_lib.o cudbg_zlib.o cxgb4_tc_mqprio.o \
+	      cxgb4_tc_matchall.o
 cxgb4-$(CONFIG_CHELSIO_T4_DCB) +=  cxgb4_dcb.o
 cxgb4-$(CONFIG_CHELSIO_T4_FCOE) +=  cxgb4_fcoe.o
 cxgb4-$(CONFIG_DEBUG_FS) += cxgb4_debugfs.o
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h +6 -0
index 1fb273b294a1..9376be6e51a3 100644
@@ -603,6 +603,8 @@ struct port_info {
 	u8 vivld;
 	u8 smt_idx;
 	u8 rx_cchan;
+
+	bool tc_block_shared;
 };
 
 struct dentry;
@@ -1101,6 +1103,9 @@ struct adapter {
 
 	/* TC MQPRIO offload */
 	struct cxgb4_tc_mqprio *tc_mqprio;
+
+	/* TC MATCHALL classifier offload */
+	struct cxgb4_tc_matchall *tc_matchall;
 };
 
 /* Support for "sched-class" command to allow a TX Scheduling Class to be
@@ -1130,6 +1135,7 @@ enum {
 
 enum {
 	SCHED_CLASS_LEVEL_CL_RL = 0,    /* class rate limiter */
+	SCHED_CLASS_LEVEL_CH_RL = 2,    /* channel rate limiter */
 };
 
 enum {
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c +68 -4

116 lines not on this page · raw message

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +217 -0

217 lines not on this page · raw message

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h new file +35 -0
new file mode 100644
index 000000000000..ce5b0800f0d8
@@ -0,0 +1,35 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2019 Chelsio Communications.  All rights reserved. */
+
+#ifndef __CXGB4_TC_MATCHALL_H__
+#define __CXGB4_TC_MATCHALL_H__
+
+#include <net/pkt_cls.h>
+
+enum cxgb4_matchall_state {
+	CXGB4_MATCHALL_STATE_DISABLED = 0,
+	CXGB4_MATCHALL_STATE_ENABLED,
+};
+
+struct cxgb4_matchall_egress_entry {
+	enum cxgb4_matchall_state state; /* Current MATCHALL offload state */
+	u8 hwtc; /* Traffic class bound to port */
+	u64 cookie; /* Used to identify the MATCHALL rule offloaded */
+};
+
+struct cxgb4_tc_port_matchall {
+	struct cxgb4_matchall_egress_entry egress; /* Egress offload info */
+};
+
+struct cxgb4_tc_matchall {
+	struct cxgb4_tc_port_matchall *port_matchall; /* Per port entry */
+};
+
+int cxgb4_tc_matchall_replace(struct net_device *dev,
+			      struct tc_cls_matchall_offload *cls_matchall);
+int cxgb4_tc_matchall_destroy(struct net_device *dev,
+			      struct tc_cls_matchall_offload *cls_matchall);
+
+int cxgb4_init_tc_matchall(struct adapter *adap);
+void cxgb4_cleanup_tc_matchall(struct adapter *adap);
+#endif /* __CXGB4_TC_MATCHALL_H__ */
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_mqprio.c +2 -3
index ce442c63f496..db55673b77bd 100644
@@ -11,8 +11,7 @@ static int cxgb4_mqprio_validate(struct net_device *dev,
 	u64 min_rate = 0, max_rate = 0, max_link_rate;
 	struct port_info *pi = netdev2pinfo(dev);
 	struct adapter *adap = netdev2adap(dev);
-	u32 qcount = 0, qoffset = 0;
-	u32 link_ok, speed, mtu;
+	u32 speed, qcount = 0, qoffset = 0;
 	int ret;
 	u8 i;
 
@@ -35,7 +34,7 @@ static int cxgb4_mqprio_validate(struct net_device *dev,
 		return -ERANGE;
 	}
 
-	ret = t4_get_link_params(pi, &link_ok, &speed, &mtu);
+	ret = t4_get_link_params(pi, NULL, &speed, NULL);
 	if (ret) {
 		netdev_err(dev, "Failed to get link speed, ret: %d\n", ret);
 		return -EINVAL;
drivers/net/ethernet/chelsio/cxgb4/sched.c +39 -17

102 lines not on this page · raw message

drivers/net/ethernet/chelsio/cxgb4/sched.h +1 -0
index 80bed8e59362..e92ff68bdd0a 100644
@@ -52,6 +52,7 @@ enum {
 
 enum sched_fw_ops {
 	SCHED_FW_OP_ADD,
+	SCHED_FW_OP_DEL,
 };
 
 enum sched_bind_type {
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c +7 -4
index f2a7824da42b..19d18acfc9a6 100644
@@ -8777,8 +8777,8 @@ int t4_get_link_params(struct port_info *pi, unsigned int *link_okp,
 		       unsigned int *speedp, unsigned int *mtup)
 {
 	unsigned int fw_caps = pi->adapter->params.fw_caps_support;
-	struct fw_port_cmd port_cmd;
 	unsigned int action, link_ok, mtu;
+	struct fw_port_cmd port_cmd;
 	fw_port_cap32_t linkattr;
 	int ret;
 
@@ -8813,9 +8813,12 @@ int t4_get_link_params(struct port_info *pi, unsigned int *link_okp,
 			be32_to_cpu(port_cmd.u.info32.auxlinfo32_mtu32));
 	}
 
-	*link_okp = link_ok;
-	*speedp = fwcap_to_speed(linkattr);
-	*mtup = mtu;
+	if (link_okp)
+		*link_okp = link_ok;
+	if (speedp)
+		*speedp = fwcap_to_speed(linkattr);
+	if (mtup)
+		*mtup = mtu;
 
 	return 0;
 }
Signature · 1 line
2.24.0
#2 lore raw

Re: [PATCH net-next v3 1/2] cxgb4: add TC-MATCHALL classifier egress offload

Fri, Nov 15, 2019 at 01:14:20PM CET, [email protected] wrote:

[...]

Quoted 4 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +4 -0
+static int cxgb4_matchall_egress_validate(struct net_device *dev,
+					  struct tc_cls_matchall_offload *cls)
+{
+	struct netlink_ext_ack *extack = cls->common.extack;

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c line 21

Quoted 10 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +10 -0
+	struct flow_action *actions = &cls->rule->action;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct flow_action_entry *entry;
+	u64 max_link_rate;
+	u32 i, speed;
+	int ret;
+
+	if (cls->common.prio != 1) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Egress MATCHALL offload must have prio 1");

I don't understand why you need it to be prio 1.

#3 lore raw

Re: [PATCH net-next v3 1/2] cxgb4: add TC-MATCHALL classifier egress offload

Quoted 5 lines · Jiri Pirko

Fri, Nov 15, 2019 at 01:14:20PM CET, [email protected] wrote:

[...]

Quoted 16 lines
Quoted 4 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +4 -0
+static int cxgb4_matchall_egress_validate(struct net_device *dev,
+					  struct tc_cls_matchall_offload *cls)
+{
+	struct netlink_ext_ack *extack = cls->common.extack;

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c line 21

Quoted 10 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +10 -0
+	struct flow_action *actions = &cls->rule->action;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct flow_action_entry *entry;
+	u64 max_link_rate;
+	u32 i, speed;
+	int ret;
+
+	if (cls->common.prio != 1) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Egress MATCHALL offload must have prio 1");

I don't understand why you need it to be prio 1.

This is to maintain rule ordering with the kernel. Jakub has suggested this in my earlier series [1][2]. I see similar checks in various drivers (mlx5 and nfp), while offloading matchall with policer.

[1] http://patchwork.ozlabs.org/patch/1194936/#2304413
[2] http://patchwork.ozlabs.org/patch/1194301/#2303749

Thanks,
Rahul

#4 lore raw

Re: [PATCH net-next v3 1/2] cxgb4: add TC-MATCHALL classifier egress offload

Quoted 23 lines · Fri, Nov 15, 2019 at 04:08:30PM CET, [email protected] wrote:
Quoted 5 lines · Jiri Pirko

Fri, Nov 15, 2019 at 01:14:20PM CET, [email protected] wrote:

[...]

Quoted 16 lines
Quoted 4 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +4 -0
+static int cxgb4_matchall_egress_validate(struct net_device *dev,
+					  struct tc_cls_matchall_offload *cls)
+{
+	struct netlink_ext_ack *extack = cls->common.extack;

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c line 21

Quoted 10 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +10 -0
+	struct flow_action *actions = &cls->rule->action;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct flow_action_entry *entry;
+	u64 max_link_rate;
+	u32 i, speed;
+	int ret;
+
+	if (cls->common.prio != 1) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Egress MATCHALL offload must have prio 1");

I don't understand why you need it to be prio 1.

Quoted 3 lines

This is to maintain rule ordering with the kernel. Jakub has suggested this in my earlier series [1][2]. I see similar checks in various drivers (mlx5 and nfp), while offloading matchall with policer.

I don't think that is correct. If matchall is the only filter there, it does not matter which prio is it. It matters only in case there are other filters.

The code should just check for other filters and forbid to insert the rule if other filters have higher prio (lower number).

Quoted 6 lines

[1] http://patchwork.ozlabs.org/patch/1194936/#2304413
[2] http://patchwork.ozlabs.org/patch/1194301/#2303749

Thanks,
Rahul

#5 lore raw

Re: [PATCH net-next v3 1/2] cxgb4: add TC-MATCHALL classifier egress offload

Quoted 20 lines · Jiri Pirko
Quoted 19 lines · Fri, Nov 15, 2019 at 04:08:30PM CET, [email protected] wrote:
Quoted 17 lines · Jiri Pirko
Quoted 4 lines · Fri, Nov 15, 2019 at 01:14:20PM CET, [email protected] wrote:
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +4 -0
+static int cxgb4_matchall_egress_validate(struct net_device *dev,
+					  struct tc_cls_matchall_offload *cls)
+{
+	struct netlink_ext_ack *extack = cls->common.extack;

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c line 21

Quoted 10 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +10 -0
+	struct flow_action *actions = &cls->rule->action;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct flow_action_entry *entry;
+	u64 max_link_rate;
+	u32 i, speed;
+	int ret;
+
+	if (cls->common.prio != 1) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Egress MATCHALL offload must have prio 1");  

I don't understand why you need it to be prio 1.

Quoted 7 lines
Quoted 3 lines

This is to maintain rule ordering with the kernel. Jakub has suggested this in my earlier series [1][2]. I see similar checks in various drivers (mlx5 and nfp), while offloading matchall with policer.

I don't think that is correct. If matchall is the only filter there, it does not matter which prio is it. It matters only in case there are other filters.

Yup, the ingress side is the one that matters.

Quoted 2 lines

The code should just check for other filters and forbid to insert the rule if other filters have higher prio (lower number).

Ack as well, that'd work even better.

I've capitulated to the prio == 1 condition as "good enough" when netronome was adding the policer offload for OvS.

#6 lore raw

Re: [PATCH net-next v3 1/2] cxgb4: add TC-MATCHALL classifier egress offload

Quoted 31 lines · Jakub Kicinski
Quoted 20 lines · Jiri Pirko
Quoted 19 lines · Fri, Nov 15, 2019 at 04:08:30PM CET, [email protected] wrote:
Quoted 17 lines · Jiri Pirko
Quoted 4 lines · Fri, Nov 15, 2019 at 01:14:20PM CET, [email protected] wrote:
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +4 -0
+static int cxgb4_matchall_egress_validate(struct net_device *dev,
+					  struct tc_cls_matchall_offload *cls)
+{
+	struct netlink_ext_ack *extack = cls->common.extack;

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c line 21

Quoted 10 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c new file +10 -0
+	struct flow_action *actions = &cls->rule->action;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct flow_action_entry *entry;
+	u64 max_link_rate;
+	u32 i, speed;
+	int ret;
+
+	if (cls->common.prio != 1) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Egress MATCHALL offload must have prio 1");  

I don't understand why you need it to be prio 1.

Quoted 7 lines
Quoted 3 lines

This is to maintain rule ordering with the kernel. Jakub has suggested this in my earlier series [1][2]. I see similar checks in various drivers (mlx5 and nfp), while offloading matchall with policer.

I don't think that is correct. If matchall is the only filter there, it does not matter which prio is it. It matters only in case there are other filters.

Yup, the ingress side is the one that matters.

Quoted 7 lines
Quoted 2 lines

The code should just check for other filters and forbid to insert the rule if other filters have higher prio (lower number).

Ack as well, that'd work even better.

I've capitulated to the prio == 1 condition as "good enough" when netronome was adding the policer offload for OvS.

I see. I thought there was some sort of mutual agreement, that to offload police, then prio must be 1, when I saw several drivers do it. I don't have a police offload on ingress side yet. So, I'm guessing this check for prio is not needed at all for my series? Please confirm again so that I'm on the same page. :)

Thanks,
Rahul

#7 lore raw

Re: [PATCH net-next v3 1/2] cxgb4: add TC-MATCHALL classifier egress offload

Quoted 11 lines · Rahul Lakkireddy
Quoted 11 lines
Quoted 3 lines

I don't think that is correct. If matchall is the only filter there, it does not matter which prio is it. It matters only in case there are other filters.

Yup, the ingress side is the one that matters.

Quoted 2 lines

The code should just check for other filters and forbid to insert the rule if other filters have higher prio (lower number).

Ack as well, that'd work even better.

Quoted 8 lines
Quoted 2 lines

I've capitulated to the prio == 1 condition as "good enough" when netronome was adding the policer offload for OvS.

I see. I thought there was some sort of mutual agreement, that to offload police, then prio must be 1, when I saw several drivers do it. I don't have a police offload on ingress side yet. So, I'm guessing this check for prio is not needed at all for my series? Please confirm again so that I'm on the same page. :)

You still need to make sure that Hw orders the rules the same as SW would. If there are lower prio flower offloads added first and then high prio matchall rule, the flower filters should no longer match. And vice versa if higher prio flower rules are inserted after match all they should take precedence.

If that's how it currently works then all is good, but I don't see the relevant checks anywhere.

As Jiri suggested, you can narrow the scope to whatever you're actual use case requires. For example simply refuse to offload the filters if it would require reordering or ejecting rules from HW.

#8 lore raw

PATCH v3 2/2 cxgb4: add TC-MATCHALL classifier ingress offload

In reply to Rahul Lakkireddy

Add TC-MATCHALL classifier ingress offload support. The same actions supported by existing TC-FLOWER offload can be applied to all incoming traffic on the underlying interface.

Only 1 ingress matchall rule can be active at a time on the underlying interface.

---
v3:
- No change.

v2:
- Removed logic to fetch free index from end of TCAM. Must maintain same ordering as in kernel.

 .../net/ethernet/chelsio/cxgb4/cxgb4_main.c   |  15 +-
 .../ethernet/chelsio/cxgb4/cxgb4_tc_flower.c  |  21 ++-
 .../ethernet/chelsio/cxgb4/cxgb4_tc_flower.h  |   6 +
 .../chelsio/cxgb4/cxgb4_tc_matchall.c         | 132 +++++++++++++++++-
 .../chelsio/cxgb4/cxgb4_tc_matchall.h         |  19 ++-
 5 files changed, 174 insertions(+), 19 deletions(-)
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c +11 -4
index 5509a4436d8f..cbe411932cbc 100644
@@ -3236,7 +3236,8 @@ static int cxgb_setup_tc_cls_u32(struct net_device *dev,
 }
 
 static int cxgb_setup_tc_matchall(struct net_device *dev,
-				  struct tc_cls_matchall_offload *cls_matchall)
+				  struct tc_cls_matchall_offload *cls_matchall,
+				  bool ingress)
 {
 	struct adapter *adap = netdev2adap(dev);
 
@@ -3245,9 +3246,13 @@ static int cxgb_setup_tc_matchall(struct net_device *dev,
 
 	switch (cls_matchall->command) {
 	case TC_CLSMATCHALL_REPLACE:
-		return cxgb4_tc_matchall_replace(dev, cls_matchall);
+		return cxgb4_tc_matchall_replace(dev, cls_matchall, ingress);
 	case TC_CLSMATCHALL_DESTROY:
-		return cxgb4_tc_matchall_destroy(dev, cls_matchall);
+		return cxgb4_tc_matchall_destroy(dev, cls_matchall, ingress);
+	case TC_CLSMATCHALL_STATS:
+		if (ingress)
+			return cxgb4_tc_matchall_stats(dev, cls_matchall);
+		break;
 	default:
 		break;
 	}
@@ -3277,6 +3282,8 @@ static int cxgb_setup_tc_block_ingress_cb(enum tc_setup_type type,
 		return cxgb_setup_tc_cls_u32(dev, type_data);
 	case TC_SETUP_CLSFLOWER:
 		return cxgb_setup_tc_flower(dev, type_data);
+	case TC_SETUP_CLSMATCHALL:
+		return cxgb_setup_tc_matchall(dev, type_data, true);
 	default:
 		return -EOPNOTSUPP;
 	}
@@ -3301,7 +3308,7 @@ static int cxgb_setup_tc_block_egress_cb(enum tc_setup_type type,
 
 	switch (type) {
 	case TC_SETUP_CLSMATCHALL:
-		return cxgb_setup_tc_matchall(dev, type_data);
+		return cxgb_setup_tc_matchall(dev, type_data, false);
 	default:
 		break;
 	}
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c +10 -11
index e447976bdd3e..6b9cf6535a57 100644
@@ -378,15 +378,14 @@ static void process_pedit_field(struct ch_filter_specification *fs, u32 val,
 	}
 }
 
-static void cxgb4_process_flow_actions(struct net_device *in,
-				       struct flow_cls_offload *cls,
-				       struct ch_filter_specification *fs)
+void cxgb4_process_flow_actions(struct net_device *in,
+				struct flow_action *actions,
+				struct ch_filter_specification *fs)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
 	struct flow_action_entry *act;
 	int i;
 
-	flow_action_for_each(i, act, &rule->action) {
+	flow_action_for_each(i, act, actions) {
 		switch (act->id) {
 		case FLOW_ACTION_ACCEPT:
 			fs->action = FILTER_PASS;
@@ -544,17 +543,16 @@ static bool valid_pedit_action(struct net_device *dev,
 	return true;
 }
 
-static int cxgb4_validate_flow_actions(struct net_device *dev,
-				       struct flow_cls_offload *cls)
+int cxgb4_validate_flow_actions(struct net_device *dev,
+				struct flow_action *actions)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
 	struct flow_action_entry *act;
 	bool act_redir = false;
 	bool act_pedit = false;
 	bool act_vlan = false;
 	int i;
 
-	flow_action_for_each(i, act, &rule->action) {
+	flow_action_for_each(i, act, actions) {
 		switch (act->id) {
 		case FLOW_ACTION_ACCEPT:
 		case FLOW_ACTION_DROP:
@@ -636,6 +634,7 @@ static int cxgb4_validate_flow_actions(struct net_device *dev,
 int cxgb4_tc_flower_replace(struct net_device *dev,
 			    struct flow_cls_offload *cls)
 {
+	struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
 	struct adapter *adap = netdev2adap(dev);
 	struct ch_tc_flower_entry *ch_flower;
 	struct ch_filter_specification *fs;
@@ -643,7 +642,7 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
 	int fidx;
 	int ret;
 
-	if (cxgb4_validate_flow_actions(dev, cls))
+	if (cxgb4_validate_flow_actions(dev, &rule->action))
 		return -EOPNOTSUPP;
 
 	if (cxgb4_validate_flow_match(dev, cls))
@@ -658,7 +657,7 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
 	fs = &ch_flower->fs;
 	fs->hitcnts = 1;
 	cxgb4_process_flow_match(dev, cls, fs);
-	cxgb4_process_flow_actions(dev, cls, fs);
+	cxgb4_process_flow_actions(dev, &rule->action, fs);
 
 	fs->hash = is_filter_exact_match(adap, fs);
 	if (fs->hash) {
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h +6 -0
index eb4c95248baf..e132516e9868 100644
@@ -108,6 +108,12 @@ struct ch_tc_pedit_fields {
 #define PEDIT_TCP_SPORT_DPORT		0x0
 #define PEDIT_UDP_SPORT_DPORT		0x0
 
+void cxgb4_process_flow_actions(struct net_device *in,
+				struct flow_action *actions,
+				struct ch_filter_specification *fs);
+int cxgb4_validate_flow_actions(struct net_device *dev,
+				struct flow_action *actions);
+
 int cxgb4_tc_flower_replace(struct net_device *dev,
 			    struct flow_cls_offload *cls);
 int cxgb4_tc_flower_destroy(struct net_device *dev,
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c +130 -2

175 lines not on this page · raw message

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h +17 -2
index ce5b0800f0d8..c62d8990a99f 100644
@@ -17,8 +17,19 @@ struct cxgb4_matchall_egress_entry {
 	u64 cookie; /* Used to identify the MATCHALL rule offloaded */
 };
 
+struct cxgb4_matchall_ingress_entry {
+	enum cxgb4_matchall_state state; /* Current MATCHALL offload state */
+	u32 tid; /* Index to hardware filter entry */
+	struct ch_filter_specification fs; /* Filter entry */
+	u64 bytes; /* # of bytes hitting the filter */
+	u64 packets; /* # of packets hitting the filter */
+	u64 last_used; /* Last updated jiffies time */
+	u64 cookie; /* Used to identify the MATCHALL rule offloaded */
+};
+
 struct cxgb4_tc_port_matchall {
 	struct cxgb4_matchall_egress_entry egress; /* Egress offload info */
+	struct cxgb4_matchall_ingress_entry ingress; /* Ingress offload info */
 };
 
 struct cxgb4_tc_matchall {
@@ -26,9 +37,13 @@ struct cxgb4_tc_matchall {
 };
 
 int cxgb4_tc_matchall_replace(struct net_device *dev,
-			      struct tc_cls_matchall_offload *cls_matchall);
+			      struct tc_cls_matchall_offload *cls_matchall,
+			      bool ingress);
 int cxgb4_tc_matchall_destroy(struct net_device *dev,
-			      struct tc_cls_matchall_offload *cls_matchall);
+			      struct tc_cls_matchall_offload *cls_matchall,
+			      bool ingress);
+int cxgb4_tc_matchall_stats(struct net_device *dev,
+			    struct tc_cls_matchall_offload *cls_matchall);
 
 int cxgb4_init_tc_matchall(struct adapter *adap);
 void cxgb4_cleanup_tc_matchall(struct adapter *adap);
Signature · 1 line
2.24.0
#9 lore raw

Re: [PATCH net-next v3 2/2] cxgb4: add TC-MATCHALL classifier ingress offload

Fri, Nov 15, 2019 at 01:14:21PM CET, [email protected] wrote:

[...]

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h struct cxgb4_tc_matchall { · line 46

Quoted 13 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h +6 -2
@@ -26,9 +37,13 @@ struct cxgb4_tc_matchall {
 };
 
 int cxgb4_tc_matchall_replace(struct net_device *dev,
-			      struct tc_cls_matchall_offload *cls_matchall);
+			      struct tc_cls_matchall_offload *cls_matchall,
+			      bool ingress);
 int cxgb4_tc_matchall_destroy(struct net_device *dev,
-			      struct tc_cls_matchall_offload *cls_matchall);
+			      struct tc_cls_matchall_offload *cls_matchall,
+			      bool ingress);
+int cxgb4_tc_matchall_stats(struct net_device *dev,
+			    struct tc_cls_matchall_offload *cls_matchall);

Hmm, you only add stats function in this second patch. Does that mean you don't care for stats in egress?
From looking at cxgb_setup_tc_matchall() looks like I'm right. Why?

Quoted 6 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h +0 -0
 int cxgb4_init_tc_matchall(struct adapter *adap);
 void cxgb4_cleanup_tc_matchall(struct adapter *adap);
Signature · 1 line
2.24.0
#10 lore raw

Re: [PATCH net-next v3 2/2] cxgb4: add TC-MATCHALL classifier ingress offload

Quoted 8 lines · Jiri Pirko

Fri, Nov 15, 2019 at 01:14:21PM CET, [email protected] wrote:

[...]

Quoted 3 lines
(quoted diff) +0 -0
@@ -26,9 +37,13 @@ struct cxgb4_tc_matchall {
 };
 
Quoted 16 lines

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h struct cxgb4_tc_matchall { · line 46

Quoted 10 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h +6 -2
 int cxgb4_tc_matchall_replace(struct net_device *dev,
-			      struct tc_cls_matchall_offload *cls_matchall);
+			      struct tc_cls_matchall_offload *cls_matchall,
+			      bool ingress);
 int cxgb4_tc_matchall_destroy(struct net_device *dev,
-			      struct tc_cls_matchall_offload *cls_matchall);
+			      struct tc_cls_matchall_offload *cls_matchall,
+			      bool ingress);
+int cxgb4_tc_matchall_stats(struct net_device *dev,
+			    struct tc_cls_matchall_offload *cls_matchall);

Hmm, you only add stats function in this second patch. Does that mean you don't care for stats in egress?
From looking at cxgb_setup_tc_matchall() looks like I'm right. Why?

We're currently missing support to fetch these stats from hardware/ firmware on egress side. So, I could only implement it for the ingress side for now.

Quoted 6 lines
Quoted 6 lines
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h +0 -0
 int cxgb4_init_tc_matchall(struct adapter *adap);
 void cxgb4_cleanup_tc_matchall(struct adapter *adap);
Signature · 1 line
2.24.0

Thanks,
Rahul