[PATCH net-next 10/10] sparx5: add support for configuring PSFP via tc

From: Daniel Machon
Date: Thu Feb 02 2023 - 05:45:45 EST

Next message: Rafael J. Wysocki: "Re: [PATCH] ACPI: Make custom_method use per-open state"
Previous message: Daniel Machon: "[PATCH net-next 09/10] net: microchip: sparx5: initialize PSFP"
In reply to: Simon Horman: "Re: [PATCH net-next 09/10] net: microchip: sparx5: initialize PSFP"
Next in thread: Simon Horman: "Re: [PATCH net-next 10/10] sparx5: add support for configuring PSFP via tc"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Add support for tc actions gate and police, in order to implement
support for configuring PSFP through tc.

Signed-off-by: Daniel Machon <daniel.machon@xxxxxxxxxxxxx>
---
.../microchip/sparx5/sparx5_tc_flower.c | 237 +++++++++++++++++-
.../net/ethernet/microchip/vcap/vcap_api.c | 3 +-
.../ethernet/microchip/vcap/vcap_api_client.h | 3 +
3 files changed, 240 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/microchip/sparx5/sparx5_tc_flower.c b/drivers/net/ethernet/microchip/sparx5/sparx5_tc_flower.c
index 217ff127e3c7..f962304272c2 100644
--- a/drivers/net/ethernet/microchip/sparx5/sparx5_tc_flower.c
+++ b/drivers/net/ethernet/microchip/sparx5/sparx5_tc_flower.c
@@ -4,6 +4,7 @@
* Copyright (c) 2022 Microchip Technology Inc. and its subsidiaries.
*/

+#include <net/tc_act/tc_gate.h>
#include <net/tcp.h>

#include "sparx5_tc.h"
@@ -989,19 +990,156 @@ static int sparx5_tc_add_rule_link(struct vcap_control *vctrl,
return err;
}

+static int sparx5_tc_flower_parse_act_gate(struct sparx5_psfp_sg *sg,
+ struct flow_action_entry *act,
+ struct netlink_ext_ack *extack)
+{
+ int i;
+
+ if (act->gate.prio < -1 || act->gate.prio > SPX5_PSFP_SG_MAX_IPV) {
+ NL_SET_ERR_MSG_MOD(extack, "Invalid gate priority");
+ return -EINVAL;
+ }
+
+ if (act->gate.cycletime < SPX5_PSFP_SG_MIN_CYCLE_TIME_NS ||
+ act->gate.cycletime > SPX5_PSFP_SG_MAX_CYCLE_TIME_NS) {
+ NL_SET_ERR_MSG_MOD(extack, "Invalid gate cycletime");
+ return -EINVAL;
+ }
+
+ if (act->gate.cycletimeext > SPX5_PSFP_SG_MAX_CYCLE_TIME_NS) {
+ NL_SET_ERR_MSG_MOD(extack, "Invalid gate cycletimeext");
+ return -EINVAL;
+ }
+
+ if (act->gate.num_entries >= SPX5_PSFP_GCE_CNT) {
+ NL_SET_ERR_MSG_MOD(extack, "Invalid number of gate entries");
+ return -EINVAL;
+ }
+
+ sg->gate_state = true;
+ sg->ipv = act->gate.prio;
+ sg->num_entries = act->gate.num_entries;
+ sg->cycletime = act->gate.cycletime;
+ sg->cycletimeext = act->gate.cycletimeext;
+
+ for (i = 0; i < sg->num_entries; i++) {
+ sg->gce[i].gate_state = !!act->gate.entries[i].gate_state;
+ sg->gce[i].interval = act->gate.entries[i].interval;
+ sg->gce[i].ipv = act->gate.entries[i].ipv;
+ sg->gce[i].maxoctets = act->gate.entries[i].maxoctets;
+ }
+
+ return 0;
+}
+
+static int sparx5_tc_flower_parse_act_police(struct sparx5_policer *pol,
+ struct flow_action_entry *act,
+ struct netlink_ext_ack *extack)
+{
+ pol->type = SPX5_POL_SERVICE;
+ pol->rate = div_u64(act->police.rate_bytes_ps, 1000) * 8;
+ pol->burst = act->police.burst;
+ pol->idx = act->hw_index;
+
+ /* rate is now in kbit */
+ if (pol->rate > DIV_ROUND_UP(SPX5_SDLB_GROUP_RATE_MAX, 1000)) {
+ NL_SET_ERR_MSG_MOD(extack, "Maximum rate exceeded");
+ return -EINVAL;
+ }
+
+ if (act->police.exceed.act_id != FLOW_ACTION_DROP) {
+ NL_SET_ERR_MSG_MOD(extack, "Offload not supported when exceed action is not drop");
+ return -EOPNOTSUPP;
+ }
+
+ if (act->police.notexceed.act_id != FLOW_ACTION_PIPE &&
+ act->police.notexceed.act_id != FLOW_ACTION_ACCEPT) {
+ NL_SET_ERR_MSG_MOD(extack, "Offload not supported when conform action is not pipe or ok");
+ return -EOPNOTSUPP;
+ }
+
+ return 0;
+}
+
+static int sparx5_tc_flower_psfp_setup(struct sparx5 *sparx5,
+ struct vcap_rule *vrule, int sg_idx,
+ int pol_idx, struct sparx5_psfp_sg *sg,
+ struct sparx5_psfp_fm *fm,
+ struct sparx5_psfp_sf *sf)
+{
+ u32 psfp_sfid = 0, psfp_fmid = 0, psfp_sgid = 0;
+ int ret;
+
+ /* Must always have a stream gate - max sdu (filter option) is evaluated
+ * after frames have passed the gate, so in case of only a policer, we
+ * allocate a stream gate that is always open.
+ */
+ if (sg_idx < 0) {
+ sg_idx = sparx5_pool_idx_to_id(SPX5_PSFP_SG_OPEN);
+ sg->ipv = 0; /* Disabled */
+ sg->cycletime = SPX5_PSFP_SG_CYCLE_TIME_DEFAULT;
+ sg->num_entries = 1;
+ sg->gate_state = 1; /* Open */
+ sg->gate_enabled = 1;
+ sg->gce[0].gate_state = 1;
+ sg->gce[0].interval = SPX5_PSFP_SG_CYCLE_TIME_DEFAULT;
+ sg->gce[0].ipv = 0;
+ sg->gce[0].maxoctets = 0; /* Disabled */
+ }
+
+ ret = sparx5_psfp_sg_add(sparx5, sg_idx, sg, &psfp_sgid);
+ if (ret < 0)
+ return ret;
+
+ if (pol_idx >= 0) {
+ /* Add new flow-meter */
+ ret = sparx5_psfp_fm_add(sparx5, pol_idx, fm, &psfp_fmid);
+ if (ret < 0)
+ return ret;
+ }
+
+ /* Map stream filter to stream gate */
+ sf->sgid = psfp_sgid;
+
+ /* Add new stream-filter and map it to a steam gate */
+ ret = sparx5_psfp_sf_add(sparx5, sf, &psfp_sfid);
+ if (ret < 0)
+ return ret;
+
+ /* Streams are classified by ISDX - map ISDX 1:1 to sfid for now. */
+ sparx5_isdx_conf_set(sparx5, psfp_sfid, psfp_sfid, psfp_fmid);
+
+ ret = vcap_rule_add_action_bit(vrule, VCAP_AF_ISDX_ADD_REPLACE_SEL,
+ VCAP_BIT_1);
+ if (ret)
+ return ret;
+
+ ret = vcap_rule_add_action_u32(vrule, VCAP_AF_ISDX_VAL, psfp_sfid);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
static int sparx5_tc_flower_replace(struct net_device *ndev,
struct flow_cls_offload *fco,
struct vcap_admin *admin,
bool ingress)
{
+ struct sparx5_psfp_sf sf = { .max_sdu = SPX5_PSFP_SF_MAX_SDU };
+ struct netlink_ext_ack *extack = fco->common.extack;
+ int err, idx, tc_sg_idx = -1, tc_pol_idx = -1;
struct sparx5_port *port = netdev_priv(ndev);
struct sparx5_multiple_rules multi = {};
+ struct sparx5 *sparx5 = port->sparx5;
+ struct sparx5_psfp_sg sg = { 0 };
+ struct sparx5_psfp_fm fm = { 0 };
struct flow_action_entry *act;
struct vcap_control *vctrl;
struct flow_rule *frule;
struct vcap_rule *vrule;
u16 l3_proto;
- int err, idx;

vctrl = port->sparx5->vcap_ctrl;

@@ -1033,6 +1171,26 @@ static int sparx5_tc_flower_replace(struct net_device *ndev,
frule = flow_cls_offload_flow_rule(fco);
flow_action_for_each(idx, act, &frule->action) {
switch (act->id) {
+ case FLOW_ACTION_GATE: {
+ err = sparx5_tc_flower_parse_act_gate(&sg, act, extack);
+ if (err < 0)
+ goto out;
+
+ tc_sg_idx = act->hw_index;
+
+ break;
+ }
+ case FLOW_ACTION_POLICE: {
+ err = sparx5_tc_flower_parse_act_police(&fm.pol, act,
+ extack);
+ if (err < 0)
+ goto out;
+
+ tc_pol_idx = fm.pol.idx;
+ sf.max_sdu = act->police.mtu;
+
+ break;
+ }
case FLOW_ACTION_TRAP:
if (admin->vtype != VCAP_TYPE_IS2 &&
admin->vtype != VCAP_TYPE_ES2) {
@@ -1079,6 +1237,14 @@ static int sparx5_tc_flower_replace(struct net_device *ndev,
}
}

+ /* Setup PSFP */
+ if (tc_sg_idx >= 0 || tc_pol_idx >= 0) {
+ err = sparx5_tc_flower_psfp_setup(sparx5, vrule, tc_sg_idx,
+ tc_pol_idx, &sg, &fm, &sf);
+ if (err)
+ goto out;
+ }
+
err = sparx5_tc_select_protocol_keyset(ndev, vrule, admin, l3_proto,
&multi);
if (err) {
@@ -1107,19 +1273,86 @@ static int sparx5_tc_flower_replace(struct net_device *ndev,
return err;
}

+static void sparx5_tc_free_psfp_resources(struct sparx5 *sparx5,
+ struct vcap_rule *vrule)
+{
+ struct vcap_client_actionfield *afield;
+ u32 isdx, sfid, sgid, fmid;
+
+ /* Check if VCAP_AF_ISDX_VAL action is set for this rule - and if
+ * it is used for stream and/or flow-meter classification.
+ */
+ afield = vcap_find_actionfield(vrule, VCAP_AF_ISDX_VAL);
+ if (!afield)
+ return;
+
+ isdx = afield->data.u32.value;
+ sfid = sparx5_psfp_isdx_get_sf(sparx5, isdx);
+
+ if (!sfid)
+ return;
+
+ fmid = sparx5_psfp_isdx_get_fm(sparx5, isdx);
+ sgid = sparx5_psfp_sf_get_sg(sparx5, sfid);
+
+ if (fmid && sparx5_psfp_fm_del(sparx5, fmid) < 0)
+ pr_err("%s:%d Could not delete invalid fmid: %d", __func__,
+ __LINE__, fmid);
+
+ if (sgid && sparx5_psfp_sg_del(sparx5, sgid) < 0)
+ pr_err("%s:%d Could not delete invalid sgid: %d", __func__,
+ __LINE__, sgid);
+
+ if (sparx5_psfp_sf_del(sparx5, sfid) < 0)
+ pr_err("%s:%d Could not delete invalid sfid: %d", __func__,
+ __LINE__, sfid);
+
+ sparx5_isdx_conf_set(sparx5, isdx, 0, 0);
+}
+
+static int sparx5_tc_free_rule_resources(struct net_device *ndev,
+ struct vcap_control *vctrl,
+ int rule_id)
+{
+ struct sparx5_port *port = netdev_priv(ndev);
+ struct sparx5 *sparx5 = port->sparx5;
+ struct vcap_rule *vrule;
+ int ret = 0;
+
+ vrule = vcap_get_rule(vctrl, rule_id);
+ if (!vrule || IS_ERR(vrule))
+ return -EINVAL;
+
+ sparx5_tc_free_psfp_resources(sparx5, vrule);
+
+ vcap_free_rule(vrule);
+ return ret;
+}
+
static int sparx5_tc_flower_destroy(struct net_device *ndev,
struct flow_cls_offload *fco,
struct vcap_admin *admin)
{
struct sparx5_port *port = netdev_priv(ndev);
+ int err = -ENOENT, count = 0, rule_id;
struct vcap_control *vctrl;
- int err = -ENOENT, rule_id;

vctrl = port->sparx5->vcap_ctrl;
while (true) {
rule_id = vcap_lookup_rule_by_cookie(vctrl, fco->cookie);
if (rule_id <= 0)
break;
+ if (count == 0) {
+ /* Resources are attached to the first rule of
+ * a set of rules. Only works if the rules are
+ * in the correct order.
+ */
+ err = sparx5_tc_free_rule_resources(ndev, vctrl,
+ rule_id);
+ if (err)
+ pr_err("%s:%d: could not free resources %d\n",
+ __func__, __LINE__, rule_id);
+ }
err = vcap_del_rule(vctrl, ndev, rule_id);
if (err) {
pr_err("%s:%d: could not delete rule %d\n",
diff --git a/drivers/net/ethernet/microchip/vcap/vcap_api.c b/drivers/net/ethernet/microchip/vcap/vcap_api.c
index 660d7cd92fcc..6307d59f23da 100644
--- a/drivers/net/ethernet/microchip/vcap/vcap_api.c
+++ b/drivers/net/ethernet/microchip/vcap/vcap_api.c
@@ -2755,7 +2755,7 @@ int vcap_rule_get_key_u32(struct vcap_rule *rule, enum vcap_key_field key,
EXPORT_SYMBOL_GPL(vcap_rule_get_key_u32);

/* Find a client action field in a rule */
-static struct vcap_client_actionfield *
+struct vcap_client_actionfield *
vcap_find_actionfield(struct vcap_rule *rule, enum vcap_action_field act)
{
struct vcap_rule_internal *ri = (struct vcap_rule_internal *)rule;
@@ -2766,6 +2766,7 @@ vcap_find_actionfield(struct vcap_rule *rule, enum vcap_action_field act)
return caf;
return NULL;
}
+EXPORT_SYMBOL_GPL(vcap_find_actionfield);

/* Check if the actionfield is already in the rule */
static bool vcap_actionfield_unique(struct vcap_rule *rule,
diff --git a/drivers/net/ethernet/microchip/vcap/vcap_api_client.h b/drivers/net/ethernet/microchip/vcap/vcap_api_client.h
index de29540fd190..417af9754bcc 100644
--- a/drivers/net/ethernet/microchip/vcap/vcap_api_client.h
+++ b/drivers/net/ethernet/microchip/vcap/vcap_api_client.h
@@ -268,4 +268,7 @@ int vcap_rule_mod_action_u32(struct vcap_rule *rule,
/* Get a 32 bit key field value and mask from the rule */
int vcap_rule_get_key_u32(struct vcap_rule *rule, enum vcap_key_field key,
u32 *value, u32 *mask);
+
+struct vcap_client_actionfield *
+vcap_find_actionfield(struct vcap_rule *rule, enum vcap_action_field act);
#endif /* __VCAP_API_CLIENT__ */
--
2.34.1

Next message: Rafael J. Wysocki: "Re: [PATCH] ACPI: Make custom_method use per-open state"
Previous message: Daniel Machon: "[PATCH net-next 09/10] net: microchip: sparx5: initialize PSFP"
In reply to: Simon Horman: "Re: [PATCH net-next 09/10] net: microchip: sparx5: initialize PSFP"
Next in thread: Simon Horman: "Re: [PATCH net-next 10/10] sparx5: add support for configuring PSFP via tc"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]