offload.c 53.3 KB
Newer Older
1 2
// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
/* Copyright (C) 2017-2018 Netronome Systems, Inc. */
3 4 5 6 7 8 9

#include <linux/skbuff.h>
#include <net/devlink.h>
#include <net/pkt_cls.h>

#include "cmsg.h"
#include "main.h"
10
#include "conntrack.h"
11 12 13 14 15 16 17
#include "../nfpcore/nfp_cpp.h"
#include "../nfpcore/nfp_nsp.h"
#include "../nfp_app.h"
#include "../nfp_main.h"
#include "../nfp_net.h"
#include "../nfp_port.h"

18 19 20 21
#define NFP_FLOWER_SUPPORTED_TCPFLAGS \
	(TCPHDR_FIN | TCPHDR_SYN | TCPHDR_RST | \
	 TCPHDR_PSH | TCPHDR_URG)

22 23 24 25
#define NFP_FLOWER_SUPPORTED_CTLFLAGS \
	(FLOW_DIS_IS_FRAGMENT | \
	 FLOW_DIS_FIRST_FRAG)

26 27 28 29 30
#define NFP_FLOWER_WHITELIST_DISSECTOR \
	(BIT(FLOW_DISSECTOR_KEY_CONTROL) | \
	 BIT(FLOW_DISSECTOR_KEY_BASIC) | \
	 BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) | \
	 BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) | \
31
	 BIT(FLOW_DISSECTOR_KEY_TCP) | \
32 33 34
	 BIT(FLOW_DISSECTOR_KEY_PORTS) | \
	 BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) | \
	 BIT(FLOW_DISSECTOR_KEY_VLAN) | \
35
	 BIT(FLOW_DISSECTOR_KEY_CVLAN) | \
36 37 38 39 40
	 BIT(FLOW_DISSECTOR_KEY_ENC_KEYID) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_CONTROL) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_PORTS) | \
41
	 BIT(FLOW_DISSECTOR_KEY_ENC_OPTS) | \
42
	 BIT(FLOW_DISSECTOR_KEY_ENC_IP) | \
43
	 BIT(FLOW_DISSECTOR_KEY_MPLS) | \
44 45
	 BIT(FLOW_DISSECTOR_KEY_CT) | \
	 BIT(FLOW_DISSECTOR_KEY_META) | \
46 47
	 BIT(FLOW_DISSECTOR_KEY_IP))

48 49 50 51 52
#define NFP_FLOWER_WHITELIST_TUN_DISSECTOR \
	(BIT(FLOW_DISSECTOR_KEY_ENC_CONTROL) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_KEYID) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) | \
53
	 BIT(FLOW_DISSECTOR_KEY_ENC_OPTS) | \
54 55
	 BIT(FLOW_DISSECTOR_KEY_ENC_PORTS) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_IP))
56 57 58

#define NFP_FLOWER_WHITELIST_TUN_DISSECTOR_R \
	(BIT(FLOW_DISSECTOR_KEY_ENC_CONTROL) | \
59
	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS))
60

61 62 63 64
#define NFP_FLOWER_WHITELIST_TUN_DISSECTOR_V6_R \
	(BIT(FLOW_DISSECTOR_KEY_ENC_CONTROL) | \
	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS))

65 66 67 68 69 70 71
#define NFP_FLOWER_MERGE_FIELDS \
	(NFP_FLOWER_LAYER_PORT | \
	 NFP_FLOWER_LAYER_MAC | \
	 NFP_FLOWER_LAYER_TP | \
	 NFP_FLOWER_LAYER_IPV4 | \
	 NFP_FLOWER_LAYER_IPV6)

72
#define NFP_FLOWER_PRE_TUN_RULE_FIELDS \
73 74
	(NFP_FLOWER_LAYER_EXT_META | \
	 NFP_FLOWER_LAYER_PORT | \
75
	 NFP_FLOWER_LAYER_MAC | \
76 77
	 NFP_FLOWER_LAYER_IPV4 | \
	 NFP_FLOWER_LAYER_IPV6)
78

79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
struct nfp_flower_merge_check {
	union {
		struct {
			__be16 tci;
			struct nfp_flower_mac_mpls l2;
			struct nfp_flower_tp_ports l4;
			union {
				struct nfp_flower_ipv4 ipv4;
				struct nfp_flower_ipv6 ipv6;
			};
		};
		unsigned long vals[8];
	};
};

94
int
95 96
nfp_flower_xmit_flow(struct nfp_app *app, struct nfp_fl_payload *nfp_flow,
		     u8 mtype)
97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
{
	u32 meta_len, key_len, mask_len, act_len, tot_len;
	struct sk_buff *skb;
	unsigned char *msg;

	meta_len =  sizeof(struct nfp_fl_rule_metadata);
	key_len = nfp_flow->meta.key_len;
	mask_len = nfp_flow->meta.mask_len;
	act_len = nfp_flow->meta.act_len;

	tot_len = meta_len + key_len + mask_len + act_len;

	/* Convert to long words as firmware expects
	 * lengths in units of NFP_FL_LW_SIZ.
	 */
	nfp_flow->meta.key_len >>= NFP_FL_LW_SIZ;
	nfp_flow->meta.mask_len >>= NFP_FL_LW_SIZ;
	nfp_flow->meta.act_len >>= NFP_FL_LW_SIZ;

116
	skb = nfp_flower_cmsg_alloc(app, tot_len, mtype, GFP_KERNEL);
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
	if (!skb)
		return -ENOMEM;

	msg = nfp_flower_cmsg_get_data(skb);
	memcpy(msg, &nfp_flow->meta, meta_len);
	memcpy(&msg[meta_len], nfp_flow->unmasked_data, key_len);
	memcpy(&msg[meta_len + key_len], nfp_flow->mask_data, mask_len);
	memcpy(&msg[meta_len + key_len + mask_len],
	       nfp_flow->action_data, act_len);

	/* Convert back to bytes as software expects
	 * lengths in units of bytes.
	 */
	nfp_flow->meta.key_len <<= NFP_FL_LW_SIZ;
	nfp_flow->meta.mask_len <<= NFP_FL_LW_SIZ;
	nfp_flow->meta.act_len <<= NFP_FL_LW_SIZ;

134
	nfp_ctrl_tx(app->ctrl, skb);
135 136 137 138

	return 0;
}

139
static bool nfp_flower_check_higher_than_mac(struct flow_rule *rule)
140
{
141 142 143 144
	return flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV4_ADDRS) ||
	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV6_ADDRS) ||
	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS) ||
	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ICMP);
145 146
}

147
static bool nfp_flower_check_higher_than_l3(struct flow_rule *rule)
148 149 150 151 152
{
	return flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS) ||
	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ICMP);
}

153
static int
154
nfp_flower_calc_opt_layer(struct flow_dissector_key_enc_opts *enc_opts,
155
			  u32 *key_layer_two, int *key_size, bool ipv6,
156
			  struct netlink_ext_ack *extack)
157
{
158 159
	if (enc_opts->len > NFP_FL_MAX_GENEVE_OPT_KEY ||
	    (ipv6 && enc_opts->len > NFP_FL_MAX_GENEVE_OPT_KEY_V6)) {
160
		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: geneve options exceed maximum length");
161
		return -EOPNOTSUPP;
162
	}
163

164
	if (enc_opts->len > 0) {
165 166 167 168 169 170 171
		*key_layer_two |= NFP_FLOWER_LAYER2_GENEVE_OP;
		*key_size += sizeof(struct nfp_flower_geneve_options);
	}

	return 0;
}

172 173 174 175 176
static int
nfp_flower_calc_udp_tun_layer(struct flow_dissector_key_ports *enc_ports,
			      struct flow_dissector_key_enc_opts *enc_op,
			      u32 *key_layer_two, u8 *key_layer, int *key_size,
			      struct nfp_flower_priv *priv,
177
			      enum nfp_flower_tun_type *tun_type, bool ipv6,
178 179 180 181 182 183 184 185
			      struct netlink_ext_ack *extack)
{
	int err;

	switch (enc_ports->dst) {
	case htons(IANA_VXLAN_UDP_PORT):
		*tun_type = NFP_FL_TUNNEL_VXLAN;
		*key_layer |= NFP_FLOWER_LAYER_VXLAN;
186 187 188 189 190 191 192 193 194

		if (ipv6) {
			*key_layer |= NFP_FLOWER_LAYER_EXT_META;
			*key_size += sizeof(struct nfp_flower_ext_meta);
			*key_layer_two |= NFP_FLOWER_LAYER2_TUN_IPV6;
			*key_size += sizeof(struct nfp_flower_ipv6_udp_tun);
		} else {
			*key_size += sizeof(struct nfp_flower_ipv4_udp_tun);
		}
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209

		if (enc_op) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: encap options not supported on vxlan tunnels");
			return -EOPNOTSUPP;
		}
		break;
	case htons(GENEVE_UDP_PORT):
		if (!(priv->flower_ext_feats & NFP_FL_FEATS_GENEVE)) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: loaded firmware does not support geneve offload");
			return -EOPNOTSUPP;
		}
		*tun_type = NFP_FL_TUNNEL_GENEVE;
		*key_layer |= NFP_FLOWER_LAYER_EXT_META;
		*key_size += sizeof(struct nfp_flower_ext_meta);
		*key_layer_two |= NFP_FLOWER_LAYER2_GENEVE;
210 211 212 213 214 215 216

		if (ipv6) {
			*key_layer_two |= NFP_FLOWER_LAYER2_TUN_IPV6;
			*key_size += sizeof(struct nfp_flower_ipv6_udp_tun);
		} else {
			*key_size += sizeof(struct nfp_flower_ipv4_udp_tun);
		}
217 218 219 220 221 222 223

		if (!enc_op)
			break;
		if (!(priv->flower_ext_feats & NFP_FL_FEATS_GENEVE_OPT)) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: loaded firmware does not support geneve option offload");
			return -EOPNOTSUPP;
		}
224 225
		err = nfp_flower_calc_opt_layer(enc_op, key_layer_two, key_size,
						ipv6, extack);
226 227 228 229 230 231 232 233 234 235 236
		if (err)
			return err;
		break;
	default:
		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: tunnel type unknown");
		return -EOPNOTSUPP;
	}

	return 0;
}

237
int
238
nfp_flower_calculate_key_layers(struct nfp_app *app,
239
				struct net_device *netdev,
240
				struct nfp_fl_key_ls *ret_key_ls,
241
				struct flow_rule *rule,
242 243
				enum nfp_flower_tun_type *tun_type,
				struct netlink_ext_ack *extack)
244
{
245 246
	struct flow_dissector *dissector = rule->match.dissector;
	struct flow_match_basic basic = { NULL, NULL};
247
	struct nfp_flower_priv *priv = app->priv;
248 249 250
	u32 key_layer_two;
	u8 key_layer;
	int key_size;
251
	int err;
252

253 254
	if (dissector->used_keys & ~NFP_FLOWER_WHITELIST_DISSECTOR) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: match not supported");
255
		return -EOPNOTSUPP;
256
	}
257

258
	/* If any tun dissector is used then the required set must be used. */
259
	if (dissector->used_keys & NFP_FLOWER_WHITELIST_TUN_DISSECTOR &&
260 261
	    (dissector->used_keys & NFP_FLOWER_WHITELIST_TUN_DISSECTOR_V6_R)
	    != NFP_FLOWER_WHITELIST_TUN_DISSECTOR_V6_R &&
262
	    (dissector->used_keys & NFP_FLOWER_WHITELIST_TUN_DISSECTOR_R)
263 264
	    != NFP_FLOWER_WHITELIST_TUN_DISSECTOR_R) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: tunnel match not supported");
265
		return -EOPNOTSUPP;
266
	}
267 268

	key_layer_two = 0;
269
	key_layer = NFP_FLOWER_LAYER_PORT;
J
John Hurley 已提交
270
	key_size = sizeof(struct nfp_flower_meta_tci) +
271 272
		   sizeof(struct nfp_flower_in_port);

273 274
	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS) ||
	    flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_MPLS)) {
275 276 277
		key_layer |= NFP_FLOWER_LAYER_MAC;
		key_size += sizeof(struct nfp_flower_mac_mpls);
	}
278

279 280
	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
		struct flow_match_vlan vlan;
281

282
		flow_rule_match_vlan(rule, &vlan);
283
		if (!(priv->flower_ext_feats & NFP_FL_FEATS_VLAN_PCP) &&
284 285
		    vlan.key->vlan_priority) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: loaded firmware does not support VLAN PCP offload");
286
			return -EOPNOTSUPP;
287
		}
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311
		if (priv->flower_ext_feats & NFP_FL_FEATS_VLAN_QINQ &&
		    !(key_layer_two & NFP_FLOWER_LAYER2_QINQ)) {
			key_layer |= NFP_FLOWER_LAYER_EXT_META;
			key_size += sizeof(struct nfp_flower_ext_meta);
			key_size += sizeof(struct nfp_flower_vlan);
			key_layer_two |= NFP_FLOWER_LAYER2_QINQ;
		}
	}

	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CVLAN)) {
		struct flow_match_vlan cvlan;

		if (!(priv->flower_ext_feats & NFP_FL_FEATS_VLAN_QINQ)) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: loaded firmware does not support VLAN QinQ offload");
			return -EOPNOTSUPP;
		}

		flow_rule_match_vlan(rule, &cvlan);
		if (!(key_layer_two & NFP_FLOWER_LAYER2_QINQ)) {
			key_layer |= NFP_FLOWER_LAYER_EXT_META;
			key_size += sizeof(struct nfp_flower_ext_meta);
			key_size += sizeof(struct nfp_flower_vlan);
			key_layer_two |= NFP_FLOWER_LAYER2_QINQ;
		}
312 313
	}

314 315 316
	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_CONTROL)) {
		struct flow_match_enc_opts enc_op = { NULL, NULL };
		struct flow_match_ipv4_addrs ipv4_addrs;
317
		struct flow_match_ipv6_addrs ipv6_addrs;
318 319
		struct flow_match_control enc_ctl;
		struct flow_match_ports enc_ports;
320
		bool ipv6_tun = false;
321 322 323

		flow_rule_match_enc_control(rule, &enc_ctl);

324 325 326 327
		if (enc_ctl.mask->addr_type != 0xffff) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: wildcarded protocols on tunnels are not supported");
			return -EOPNOTSUPP;
		}
328 329 330 331 332 333

		ipv6_tun = enc_ctl.key->addr_type ==
				FLOW_DISSECTOR_KEY_IPV6_ADDRS;
		if (ipv6_tun &&
		    !(priv->flower_ext_feats & NFP_FL_FEATS_IPV6_TUN)) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: firmware does not support IPv6 tunnels");
334
			return -EOPNOTSUPP;
335
		}
336

337 338 339
		if (!ipv6_tun &&
		    enc_ctl.key->addr_type != FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: tunnel address type not IPv4 or IPv6");
340
			return -EOPNOTSUPP;
341
		}
342

343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
		if (ipv6_tun) {
			flow_rule_match_enc_ipv6_addrs(rule, &ipv6_addrs);
			if (memchr_inv(&ipv6_addrs.mask->dst, 0xff,
				       sizeof(ipv6_addrs.mask->dst))) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: only an exact match IPv6 destination address is supported");
				return -EOPNOTSUPP;
			}
		} else {
			flow_rule_match_enc_ipv4_addrs(rule, &ipv4_addrs);
			if (ipv4_addrs.mask->dst != cpu_to_be32(~0)) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: only an exact match IPv4 destination address is supported");
				return -EOPNOTSUPP;
			}
		}

358 359
		if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_OPTS))
			flow_rule_match_enc_opts(rule, &enc_op);
360

361 362
		if (!flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
			/* check if GRE, which has no enc_ports */
363 364 365 366
			if (!netif_is_gretap(netdev)) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: an exact match on L4 destination port is required for non-GRE tunnels");
				return -EOPNOTSUPP;
			}
367

368 369 370 371 372 373 374 375 376
			*tun_type = NFP_FL_TUNNEL_GRE;
			key_layer |= NFP_FLOWER_LAYER_EXT_META;
			key_size += sizeof(struct nfp_flower_ext_meta);
			key_layer_two |= NFP_FLOWER_LAYER2_GRE;

			if (ipv6_tun) {
				key_layer_two |= NFP_FLOWER_LAYER2_TUN_IPV6;
				key_size +=
					sizeof(struct nfp_flower_ipv6_udp_tun);
377
			} else {
378 379 380 381 382 383
				key_size +=
					sizeof(struct nfp_flower_ipv4_udp_tun);
			}

			if (enc_op.key) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: encap options not supported on GRE tunnels");
384 385 386 387 388 389 390 391 392 393 394 395 396 397
				return -EOPNOTSUPP;
			}
		} else {
			flow_rule_match_enc_ports(rule, &enc_ports);
			if (enc_ports.mask->dst != cpu_to_be16(~0)) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: only an exact match L4 destination port is supported");
				return -EOPNOTSUPP;
			}

			err = nfp_flower_calc_udp_tun_layer(enc_ports.key,
							    enc_op.key,
							    &key_layer_two,
							    &key_layer,
							    &key_size, priv,
398 399
							    tun_type, ipv6_tun,
							    extack);
400 401 402 403 404 405 406 407 408 409
			if (err)
				return err;

			/* Ensure the ingress netdev matches the expected
			 * tun type.
			 */
			if (!nfp_fl_netdev_is_tunnel_type(netdev, *tun_type)) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: ingress netdev does not match the expected tunnel type");
				return -EOPNOTSUPP;
			}
410
		}
411
	}
412

413 414
	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC))
		flow_rule_match_basic(rule, &basic);
415

416
	if (basic.mask && basic.mask->n_proto) {
417
		/* Ethernet type is present in the key. */
418
		switch (basic.key->n_proto) {
419 420 421 422 423 424 425 426 427 428 429 430 431 432
		case cpu_to_be16(ETH_P_IP):
			key_layer |= NFP_FLOWER_LAYER_IPV4;
			key_size += sizeof(struct nfp_flower_ipv4);
			break;

		case cpu_to_be16(ETH_P_IPV6):
			key_layer |= NFP_FLOWER_LAYER_IPV6;
			key_size += sizeof(struct nfp_flower_ipv6);
			break;

		/* Currently we do not offload ARP
		 * because we rely on it to get to the host.
		 */
		case cpu_to_be16(ETH_P_ARP):
433
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: ARP not supported");
434 435
			return -EOPNOTSUPP;

436 437 438 439 440 441 442 443
		case cpu_to_be16(ETH_P_MPLS_UC):
		case cpu_to_be16(ETH_P_MPLS_MC):
			if (!(key_layer & NFP_FLOWER_LAYER_MAC)) {
				key_layer |= NFP_FLOWER_LAYER_MAC;
				key_size += sizeof(struct nfp_flower_mac_mpls);
			}
			break;

444 445 446 447 448
		/* Will be included in layer 2. */
		case cpu_to_be16(ETH_P_8021Q):
			break;

		default:
449 450
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: match on given EtherType is not supported");
			return -EOPNOTSUPP;
451
		}
452
	} else if (nfp_flower_check_higher_than_mac(rule)) {
453 454
		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: cannot match above L2 without specified EtherType");
		return -EOPNOTSUPP;
455 456
	}

457 458
	if (basic.mask && basic.mask->ip_proto) {
		switch (basic.key->ip_proto) {
459 460 461 462 463 464 465 466 467 468 469
		case IPPROTO_TCP:
		case IPPROTO_UDP:
		case IPPROTO_SCTP:
		case IPPROTO_ICMP:
		case IPPROTO_ICMPV6:
			key_layer |= NFP_FLOWER_LAYER_TP;
			key_size += sizeof(struct nfp_flower_tp_ports);
			break;
		}
	}

470
	if (!(key_layer & NFP_FLOWER_LAYER_TP) &&
471
	    nfp_flower_check_higher_than_l3(rule)) {
472 473 474 475
		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: cannot match on L4 information without specified IP protocol type");
		return -EOPNOTSUPP;
	}

476 477
	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_TCP)) {
		struct flow_match_tcp tcp;
478 479
		u32 tcp_flags;

480 481
		flow_rule_match_tcp(rule, &tcp);
		tcp_flags = be16_to_cpu(tcp.key->flags);
482

483 484
		if (tcp_flags & ~NFP_FLOWER_SUPPORTED_TCPFLAGS) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: no match support for selected TCP flags");
485
			return -EOPNOTSUPP;
486
		}
487 488 489 490 491

		/* We only support PSH and URG flags when either
		 * FIN, SYN or RST is present as well.
		 */
		if ((tcp_flags & (TCPHDR_PSH | TCPHDR_URG)) &&
492 493
		    !(tcp_flags & (TCPHDR_FIN | TCPHDR_SYN | TCPHDR_RST))) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: PSH and URG is only supported when used with FIN, SYN or RST");
494
			return -EOPNOTSUPP;
495
		}
496

497 498 499
		/* We need to store TCP flags in the either the IPv4 or IPv6 key
		 * space, thus we need to ensure we include a IPv4/IPv6 key
		 * layer if we have not done so already.
500
		 */
501 502
		if (!basic.key) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: match on TCP flags requires a match on L3 protocol");
503
			return -EOPNOTSUPP;
504
		}
505 506 507

		if (!(key_layer & NFP_FLOWER_LAYER_IPV4) &&
		    !(key_layer & NFP_FLOWER_LAYER_IPV6)) {
508
			switch (basic.key->n_proto) {
509 510 511 512 513 514
			case cpu_to_be16(ETH_P_IP):
				key_layer |= NFP_FLOWER_LAYER_IPV4;
				key_size += sizeof(struct nfp_flower_ipv4);
				break;

			case cpu_to_be16(ETH_P_IPV6):
J
John Hurley 已提交
515
					key_layer |= NFP_FLOWER_LAYER_IPV6;
516 517 518 519
				key_size += sizeof(struct nfp_flower_ipv6);
				break;

			default:
520
				NL_SET_ERR_MSG_MOD(extack, "unsupported offload: match on TCP flags requires a match on IPv4/IPv6");
521 522
				return -EOPNOTSUPP;
			}
523 524 525
		}
	}

526 527
	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CONTROL)) {
		struct flow_match_control ctl;
528

529
		flow_rule_match_control(rule, &ctl);
530 531
		if (ctl.key->flags & ~NFP_FLOWER_SUPPORTED_CTLFLAGS) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: match on unknown control flag");
532
			return -EOPNOTSUPP;
533
		}
534 535
	}

536 537 538 539 540 541 542
	ret_key_ls->key_layer = key_layer;
	ret_key_ls->key_layer_two = key_layer_two;
	ret_key_ls->key_size = key_size;

	return 0;
}

543
struct nfp_fl_payload *
544
nfp_flower_allocate_new(struct nfp_fl_key_ls *key_layer)
545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
{
	struct nfp_fl_payload *flow_pay;

	flow_pay = kmalloc(sizeof(*flow_pay), GFP_KERNEL);
	if (!flow_pay)
		return NULL;

	flow_pay->meta.key_len = key_layer->key_size;
	flow_pay->unmasked_data = kmalloc(key_layer->key_size, GFP_KERNEL);
	if (!flow_pay->unmasked_data)
		goto err_free_flow;

	flow_pay->meta.mask_len = key_layer->key_size;
	flow_pay->mask_data = kmalloc(key_layer->key_size, GFP_KERNEL);
	if (!flow_pay->mask_data)
		goto err_free_unmasked;

562 563 564 565
	flow_pay->action_data = kmalloc(NFP_FL_MAX_A_SIZ, GFP_KERNEL);
	if (!flow_pay->action_data)
		goto err_free_mask;

566
	flow_pay->nfp_tun_ipv4_addr = 0;
567
	flow_pay->nfp_tun_ipv6 = NULL;
568
	flow_pay->meta.flags = 0;
569
	INIT_LIST_HEAD(&flow_pay->linked_flows);
J
John Hurley 已提交
570
	flow_pay->in_hw = false;
571
	flow_pay->pre_tun_rule.dev = NULL;
572

573 574
	return flow_pay;

575 576
err_free_mask:
	kfree(flow_pay->mask_data);
577 578 579 580 581 582 583
err_free_unmasked:
	kfree(flow_pay->unmasked_data);
err_free_flow:
	kfree(flow_pay);
	return NULL;
}

584 585 586 587 588 589 590 591 592 593
static int
nfp_flower_update_merge_with_actions(struct nfp_fl_payload *flow,
				     struct nfp_flower_merge_check *merge,
				     u8 *last_act_id, int *act_out)
{
	struct nfp_fl_set_ipv6_tc_hl_fl *ipv6_tc_hl_fl;
	struct nfp_fl_set_ip4_ttl_tos *ipv4_ttl_tos;
	struct nfp_fl_set_ip4_addrs *ipv4_add;
	struct nfp_fl_set_ipv6_addr *ipv6_add;
	struct nfp_fl_push_vlan *push_vlan;
594
	struct nfp_fl_pre_tunnel *pre_tun;
595 596 597 598
	struct nfp_fl_set_tport *tport;
	struct nfp_fl_set_eth *eth;
	struct nfp_fl_act_head *a;
	unsigned int act_off = 0;
599
	bool ipv6_tun = false;
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
	u8 act_id = 0;
	u8 *ports;
	int i;

	while (act_off < flow->meta.act_len) {
		a = (struct nfp_fl_act_head *)&flow->action_data[act_off];
		act_id = a->jump_id;

		switch (act_id) {
		case NFP_FL_ACTION_OPCODE_OUTPUT:
			if (act_out)
				(*act_out)++;
			break;
		case NFP_FL_ACTION_OPCODE_PUSH_VLAN:
			push_vlan = (struct nfp_fl_push_vlan *)a;
			if (push_vlan->vlan_tci)
				merge->tci = cpu_to_be16(0xffff);
			break;
		case NFP_FL_ACTION_OPCODE_POP_VLAN:
			merge->tci = cpu_to_be16(0);
			break;
621
		case NFP_FL_ACTION_OPCODE_SET_TUNNEL:
622 623 624 625 626
			/* New tunnel header means l2 to l4 can be matched. */
			eth_broadcast_addr(&merge->l2.mac_dst[0]);
			eth_broadcast_addr(&merge->l2.mac_src[0]);
			memset(&merge->l4, 0xff,
			       sizeof(struct nfp_flower_tp_ports));
627 628 629 630 631 632
			if (ipv6_tun)
				memset(&merge->ipv6, 0xff,
				       sizeof(struct nfp_flower_ipv6));
			else
				memset(&merge->ipv4, 0xff,
				       sizeof(struct nfp_flower_ipv4));
633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679
			break;
		case NFP_FL_ACTION_OPCODE_SET_ETHERNET:
			eth = (struct nfp_fl_set_eth *)a;
			for (i = 0; i < ETH_ALEN; i++)
				merge->l2.mac_dst[i] |= eth->eth_addr_mask[i];
			for (i = 0; i < ETH_ALEN; i++)
				merge->l2.mac_src[i] |=
					eth->eth_addr_mask[ETH_ALEN + i];
			break;
		case NFP_FL_ACTION_OPCODE_SET_IPV4_ADDRS:
			ipv4_add = (struct nfp_fl_set_ip4_addrs *)a;
			merge->ipv4.ipv4_src |= ipv4_add->ipv4_src_mask;
			merge->ipv4.ipv4_dst |= ipv4_add->ipv4_dst_mask;
			break;
		case NFP_FL_ACTION_OPCODE_SET_IPV4_TTL_TOS:
			ipv4_ttl_tos = (struct nfp_fl_set_ip4_ttl_tos *)a;
			merge->ipv4.ip_ext.ttl |= ipv4_ttl_tos->ipv4_ttl_mask;
			merge->ipv4.ip_ext.tos |= ipv4_ttl_tos->ipv4_tos_mask;
			break;
		case NFP_FL_ACTION_OPCODE_SET_IPV6_SRC:
			ipv6_add = (struct nfp_fl_set_ipv6_addr *)a;
			for (i = 0; i < 4; i++)
				merge->ipv6.ipv6_src.in6_u.u6_addr32[i] |=
					ipv6_add->ipv6[i].mask;
			break;
		case NFP_FL_ACTION_OPCODE_SET_IPV6_DST:
			ipv6_add = (struct nfp_fl_set_ipv6_addr *)a;
			for (i = 0; i < 4; i++)
				merge->ipv6.ipv6_dst.in6_u.u6_addr32[i] |=
					ipv6_add->ipv6[i].mask;
			break;
		case NFP_FL_ACTION_OPCODE_SET_IPV6_TC_HL_FL:
			ipv6_tc_hl_fl = (struct nfp_fl_set_ipv6_tc_hl_fl *)a;
			merge->ipv6.ip_ext.ttl |=
				ipv6_tc_hl_fl->ipv6_hop_limit_mask;
			merge->ipv6.ip_ext.tos |= ipv6_tc_hl_fl->ipv6_tc_mask;
			merge->ipv6.ipv6_flow_label_exthdr |=
				ipv6_tc_hl_fl->ipv6_label_mask;
			break;
		case NFP_FL_ACTION_OPCODE_SET_UDP:
		case NFP_FL_ACTION_OPCODE_SET_TCP:
			tport = (struct nfp_fl_set_tport *)a;
			ports = (u8 *)&merge->l4.port_src;
			for (i = 0; i < 4; i++)
				ports[i] |= tport->tp_port_mask[i];
			break;
		case NFP_FL_ACTION_OPCODE_PRE_TUNNEL:
680 681 682 683
			pre_tun = (struct nfp_fl_pre_tunnel *)a;
			ipv6_tun = be16_to_cpu(pre_tun->flags) &
					NFP_FL_PRE_TUN_IPV6;
			break;
684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793
		case NFP_FL_ACTION_OPCODE_PRE_LAG:
		case NFP_FL_ACTION_OPCODE_PUSH_GENEVE:
			break;
		default:
			return -EOPNOTSUPP;
		}

		act_off += a->len_lw << NFP_FL_LW_SIZ;
	}

	if (last_act_id)
		*last_act_id = act_id;

	return 0;
}

static int
nfp_flower_populate_merge_match(struct nfp_fl_payload *flow,
				struct nfp_flower_merge_check *merge,
				bool extra_fields)
{
	struct nfp_flower_meta_tci *meta_tci;
	u8 *mask = flow->mask_data;
	u8 key_layer, match_size;

	memset(merge, 0, sizeof(struct nfp_flower_merge_check));

	meta_tci = (struct nfp_flower_meta_tci *)mask;
	key_layer = meta_tci->nfp_flow_key_layer;

	if (key_layer & ~NFP_FLOWER_MERGE_FIELDS && !extra_fields)
		return -EOPNOTSUPP;

	merge->tci = meta_tci->tci;
	mask += sizeof(struct nfp_flower_meta_tci);

	if (key_layer & NFP_FLOWER_LAYER_EXT_META)
		mask += sizeof(struct nfp_flower_ext_meta);

	mask += sizeof(struct nfp_flower_in_port);

	if (key_layer & NFP_FLOWER_LAYER_MAC) {
		match_size = sizeof(struct nfp_flower_mac_mpls);
		memcpy(&merge->l2, mask, match_size);
		mask += match_size;
	}

	if (key_layer & NFP_FLOWER_LAYER_TP) {
		match_size = sizeof(struct nfp_flower_tp_ports);
		memcpy(&merge->l4, mask, match_size);
		mask += match_size;
	}

	if (key_layer & NFP_FLOWER_LAYER_IPV4) {
		match_size = sizeof(struct nfp_flower_ipv4);
		memcpy(&merge->ipv4, mask, match_size);
	}

	if (key_layer & NFP_FLOWER_LAYER_IPV6) {
		match_size = sizeof(struct nfp_flower_ipv6);
		memcpy(&merge->ipv6, mask, match_size);
	}

	return 0;
}

static int
nfp_flower_can_merge(struct nfp_fl_payload *sub_flow1,
		     struct nfp_fl_payload *sub_flow2)
{
	/* Two flows can be merged if sub_flow2 only matches on bits that are
	 * either matched by sub_flow1 or set by a sub_flow1 action. This
	 * ensures that every packet that hits sub_flow1 and recirculates is
	 * guaranteed to hit sub_flow2.
	 */
	struct nfp_flower_merge_check sub_flow1_merge, sub_flow2_merge;
	int err, act_out = 0;
	u8 last_act_id = 0;

	err = nfp_flower_populate_merge_match(sub_flow1, &sub_flow1_merge,
					      true);
	if (err)
		return err;

	err = nfp_flower_populate_merge_match(sub_flow2, &sub_flow2_merge,
					      false);
	if (err)
		return err;

	err = nfp_flower_update_merge_with_actions(sub_flow1, &sub_flow1_merge,
						   &last_act_id, &act_out);
	if (err)
		return err;

	/* Must only be 1 output action and it must be the last in sequence. */
	if (act_out != 1 || last_act_id != NFP_FL_ACTION_OPCODE_OUTPUT)
		return -EOPNOTSUPP;

	/* Reject merge if sub_flow2 matches on something that is not matched
	 * on or set in an action by sub_flow1.
	 */
	err = bitmap_andnot(sub_flow2_merge.vals, sub_flow2_merge.vals,
			    sub_flow1_merge.vals,
			    sizeof(struct nfp_flower_merge_check) * 8);
	if (err)
		return -EINVAL;

	return 0;
}

794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810
static unsigned int
nfp_flower_copy_pre_actions(char *act_dst, char *act_src, int len,
			    bool *tunnel_act)
{
	unsigned int act_off = 0, act_len;
	struct nfp_fl_act_head *a;
	u8 act_id = 0;

	while (act_off < len) {
		a = (struct nfp_fl_act_head *)&act_src[act_off];
		act_len = a->len_lw << NFP_FL_LW_SIZ;
		act_id = a->jump_id;

		switch (act_id) {
		case NFP_FL_ACTION_OPCODE_PRE_TUNNEL:
			if (tunnel_act)
				*tunnel_act = true;
811
			fallthrough;
812 813 814 815 816 817 818 819 820 821 822 823 824
		case NFP_FL_ACTION_OPCODE_PRE_LAG:
			memcpy(act_dst + act_off, act_src + act_off, act_len);
			break;
		default:
			return act_off;
		}

		act_off += act_len;
	}

	return act_off;
}

825 826
static int
nfp_fl_verify_post_tun_acts(char *acts, int len, struct nfp_fl_push_vlan **vlan)
827 828 829 830 831 832
{
	struct nfp_fl_act_head *a;
	unsigned int act_off = 0;

	while (act_off < len) {
		a = (struct nfp_fl_act_head *)&acts[act_off];
833 834 835 836

		if (a->jump_id == NFP_FL_ACTION_OPCODE_PUSH_VLAN && !act_off)
			*vlan = (struct nfp_fl_push_vlan *)a;
		else if (a->jump_id != NFP_FL_ACTION_OPCODE_OUTPUT)
837 838 839 840 841
			return -EOPNOTSUPP;

		act_off += a->len_lw << NFP_FL_LW_SIZ;
	}

842 843 844 845
	/* Ensure any VLAN push also has an egress action. */
	if (*vlan && act_off <= sizeof(struct nfp_fl_push_vlan))
		return -EOPNOTSUPP;

846 847 848
	return 0;
}

849 850 851
static int
nfp_fl_push_vlan_after_tun(char *acts, int len, struct nfp_fl_push_vlan *vlan)
{
852
	struct nfp_fl_set_tun *tun;
853 854 855 856 857 858
	struct nfp_fl_act_head *a;
	unsigned int act_off = 0;

	while (act_off < len) {
		a = (struct nfp_fl_act_head *)&acts[act_off];

859 860
		if (a->jump_id == NFP_FL_ACTION_OPCODE_SET_TUNNEL) {
			tun = (struct nfp_fl_set_tun *)a;
861 862 863 864 865 866 867 868 869 870 871 872 873
			tun->outer_vlan_tpid = vlan->vlan_tpid;
			tun->outer_vlan_tci = vlan->vlan_tci;

			return 0;
		}

		act_off += a->len_lw << NFP_FL_LW_SIZ;
	}

	/* Return error if no tunnel action is found. */
	return -EOPNOTSUPP;
}

874 875 876 877 878 879
static int
nfp_flower_merge_action(struct nfp_fl_payload *sub_flow1,
			struct nfp_fl_payload *sub_flow2,
			struct nfp_fl_payload *merge_flow)
{
	unsigned int sub1_act_len, sub2_act_len, pre_off1, pre_off2;
880
	struct nfp_fl_push_vlan *post_tun_push_vlan = NULL;
881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916
	bool tunnel_act = false;
	char *merge_act;
	int err;

	/* The last action of sub_flow1 must be output - do not merge this. */
	sub1_act_len = sub_flow1->meta.act_len - sizeof(struct nfp_fl_output);
	sub2_act_len = sub_flow2->meta.act_len;

	if (!sub2_act_len)
		return -EINVAL;

	if (sub1_act_len + sub2_act_len > NFP_FL_MAX_A_SIZ)
		return -EINVAL;

	/* A shortcut can only be applied if there is a single action. */
	if (sub1_act_len)
		merge_flow->meta.shortcut = cpu_to_be32(NFP_FL_SC_ACT_NULL);
	else
		merge_flow->meta.shortcut = sub_flow2->meta.shortcut;

	merge_flow->meta.act_len = sub1_act_len + sub2_act_len;
	merge_act = merge_flow->action_data;

	/* Copy any pre-actions to the start of merge flow action list. */
	pre_off1 = nfp_flower_copy_pre_actions(merge_act,
					       sub_flow1->action_data,
					       sub1_act_len, &tunnel_act);
	merge_act += pre_off1;
	sub1_act_len -= pre_off1;
	pre_off2 = nfp_flower_copy_pre_actions(merge_act,
					       sub_flow2->action_data,
					       sub2_act_len, NULL);
	merge_act += pre_off2;
	sub2_act_len -= pre_off2;

	/* FW does a tunnel push when egressing, therefore, if sub_flow 1 pushes
917 918
	 * a tunnel, there are restrictions on what sub_flow 2 actions lead to a
	 * valid merge.
919 920 921 922
	 */
	if (tunnel_act) {
		char *post_tun_acts = &sub_flow2->action_data[pre_off2];

923 924
		err = nfp_fl_verify_post_tun_acts(post_tun_acts, sub2_act_len,
						  &post_tun_push_vlan);
925 926
		if (err)
			return err;
927 928 929 930 931

		if (post_tun_push_vlan) {
			pre_off2 += sizeof(*post_tun_push_vlan);
			sub2_act_len -= sizeof(*post_tun_push_vlan);
		}
932 933 934 935
	}

	/* Copy remaining actions from sub_flows 1 and 2. */
	memcpy(merge_act, sub_flow1->action_data + pre_off1, sub1_act_len);
936 937 938 939 940 941 942 943 944 945 946

	if (post_tun_push_vlan) {
		/* Update tunnel action in merge to include VLAN push. */
		err = nfp_fl_push_vlan_after_tun(merge_act, sub1_act_len,
						 post_tun_push_vlan);
		if (err)
			return err;

		merge_flow->meta.act_len -= sizeof(*post_tun_push_vlan);
	}

947 948 949 950 951 952
	merge_act += sub1_act_len;
	memcpy(merge_act, sub_flow2->action_data + pre_off2, sub2_act_len);

	return 0;
}

953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
/* Flow link code should only be accessed under RTNL. */
static void nfp_flower_unlink_flow(struct nfp_fl_payload_link *link)
{
	list_del(&link->merge_flow.list);
	list_del(&link->sub_flow.list);
	kfree(link);
}

static void nfp_flower_unlink_flows(struct nfp_fl_payload *merge_flow,
				    struct nfp_fl_payload *sub_flow)
{
	struct nfp_fl_payload_link *link;

	list_for_each_entry(link, &merge_flow->linked_flows, merge_flow.list)
		if (link->sub_flow.flow == sub_flow) {
			nfp_flower_unlink_flow(link);
			return;
		}
}

static int nfp_flower_link_flows(struct nfp_fl_payload *merge_flow,
				 struct nfp_fl_payload *sub_flow)
{
	struct nfp_fl_payload_link *link;

	link = kmalloc(sizeof(*link), GFP_KERNEL);
	if (!link)
		return -ENOMEM;

	link->merge_flow.flow = merge_flow;
	list_add_tail(&link->merge_flow.list, &merge_flow->linked_flows);
	link->sub_flow.flow = sub_flow;
	list_add_tail(&link->sub_flow.list, &sub_flow->linked_flows);

	return 0;
}

990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
/**
 * nfp_flower_merge_offloaded_flows() - Merge 2 existing flows to single flow.
 * @app:	Pointer to the APP handle
 * @sub_flow1:	Initial flow matched to produce merge hint
 * @sub_flow2:	Post recirculation flow matched in merge hint
 *
 * Combines 2 flows (if valid) to a single flow, removing the initial from hw
 * and offloading the new, merged flow.
 *
 * Return: negative value on error, 0 in success.
 */
int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
				     struct nfp_fl_payload *sub_flow1,
				     struct nfp_fl_payload *sub_flow2)
{
J
John Hurley 已提交
1005
	struct nfp_flower_priv *priv = app->priv;
1006 1007
	struct nfp_fl_payload *merge_flow;
	struct nfp_fl_key_ls merge_key_ls;
1008 1009
	struct nfp_merge_info *merge_info;
	u64 parent_ctx = 0;
1010 1011
	int err;

1012 1013 1014 1015 1016 1017 1018
	ASSERT_RTNL();

	if (sub_flow1 == sub_flow2 ||
	    nfp_flower_is_merge_flow(sub_flow1) ||
	    nfp_flower_is_merge_flow(sub_flow2))
		return -EINVAL;

1019 1020 1021 1022 1023 1024 1025 1026 1027
	/* check if the two flows are already merged */
	parent_ctx = (u64)(be32_to_cpu(sub_flow1->meta.host_ctx_id)) << 32;
	parent_ctx |= (u64)(be32_to_cpu(sub_flow2->meta.host_ctx_id));
	if (rhashtable_lookup_fast(&priv->merge_table,
				   &parent_ctx, merge_table_params)) {
		nfp_flower_cmsg_warn(app, "The two flows are already merged.\n");
		return 0;
	}

1028 1029 1030 1031
	err = nfp_flower_can_merge(sub_flow1, sub_flow2);
	if (err)
		return err;

1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	merge_key_ls.key_size = sub_flow1->meta.key_len;

	merge_flow = nfp_flower_allocate_new(&merge_key_ls);
	if (!merge_flow)
		return -ENOMEM;

	merge_flow->tc_flower_cookie = (unsigned long)merge_flow;
	merge_flow->ingress_dev = sub_flow1->ingress_dev;

	memcpy(merge_flow->unmasked_data, sub_flow1->unmasked_data,
	       sub_flow1->meta.key_len);
	memcpy(merge_flow->mask_data, sub_flow1->mask_data,
	       sub_flow1->meta.mask_len);

	err = nfp_flower_merge_action(sub_flow1, sub_flow2, merge_flow);
	if (err)
		goto err_destroy_merge_flow;

1050 1051 1052 1053 1054 1055 1056 1057
	err = nfp_flower_link_flows(merge_flow, sub_flow1);
	if (err)
		goto err_destroy_merge_flow;

	err = nfp_flower_link_flows(merge_flow, sub_flow2);
	if (err)
		goto err_unlink_sub_flow1;

1058 1059
	err = nfp_compile_flow_metadata(app, merge_flow->tc_flower_cookie, merge_flow,
					merge_flow->ingress_dev, NULL);
J
John Hurley 已提交
1060 1061
	if (err)
		goto err_unlink_sub_flow2;
1062

J
John Hurley 已提交
1063 1064 1065 1066 1067
	err = rhashtable_insert_fast(&priv->flow_table, &merge_flow->fl_node,
				     nfp_flower_table_params);
	if (err)
		goto err_release_metadata;

1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
	merge_info = kmalloc(sizeof(*merge_info), GFP_KERNEL);
	if (!merge_info) {
		err = -ENOMEM;
		goto err_remove_rhash;
	}
	merge_info->parent_ctx = parent_ctx;
	err = rhashtable_insert_fast(&priv->merge_table, &merge_info->ht_node,
				     merge_table_params);
	if (err)
		goto err_destroy_merge_info;

J
John Hurley 已提交
1079 1080 1081
	err = nfp_flower_xmit_flow(app, merge_flow,
				   NFP_FLOWER_CMSG_TYPE_FLOW_MOD);
	if (err)
1082
		goto err_remove_merge_info;
J
John Hurley 已提交
1083 1084 1085 1086 1087 1088

	merge_flow->in_hw = true;
	sub_flow1->in_hw = false;

	return 0;

1089 1090 1091 1092 1093 1094
err_remove_merge_info:
	WARN_ON_ONCE(rhashtable_remove_fast(&priv->merge_table,
					    &merge_info->ht_node,
					    merge_table_params));
err_destroy_merge_info:
	kfree(merge_info);
J
John Hurley 已提交
1095 1096 1097 1098 1099 1100 1101
err_remove_rhash:
	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
					    &merge_flow->fl_node,
					    nfp_flower_table_params));
err_release_metadata:
	nfp_modify_flow_metadata(app, merge_flow);
err_unlink_sub_flow2:
1102 1103 1104
	nfp_flower_unlink_flows(merge_flow, sub_flow2);
err_unlink_sub_flow1:
	nfp_flower_unlink_flows(merge_flow, sub_flow1);
1105 1106 1107 1108 1109 1110
err_destroy_merge_flow:
	kfree(merge_flow->action_data);
	kfree(merge_flow->mask_data);
	kfree(merge_flow->unmasked_data);
	kfree(merge_flow);
	return err;
1111 1112
}

1113 1114 1115 1116
/**
 * nfp_flower_validate_pre_tun_rule()
 * @app:	Pointer to the APP handle
 * @flow:	Pointer to NFP flow representation of rule
1117
 * @key_ls:	Pointer to NFP key layers structure
1118 1119 1120 1121 1122 1123 1124 1125 1126
 * @extack:	Netlink extended ACK report
 *
 * Verifies the flow as a pre-tunnel rule.
 *
 * Return: negative value on error, 0 if verified.
 */
static int
nfp_flower_validate_pre_tun_rule(struct nfp_app *app,
				 struct nfp_fl_payload *flow,
1127
				 struct nfp_fl_key_ls *key_ls,
1128 1129
				 struct netlink_ext_ack *extack)
{
1130
	struct nfp_flower_priv *priv = app->priv;
1131 1132
	struct nfp_flower_meta_tci *meta_tci;
	struct nfp_flower_mac_mpls *mac;
1133
	u8 *ext = flow->unmasked_data;
1134 1135 1136 1137 1138 1139 1140
	struct nfp_fl_act_head *act;
	u8 *mask = flow->mask_data;
	bool vlan = false;
	int act_offset;
	u8 key_layer;

	meta_tci = (struct nfp_flower_meta_tci *)flow->unmasked_data;
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151
	key_layer = key_ls->key_layer;
	if (!(priv->flower_ext_feats & NFP_FL_FEATS_VLAN_QINQ)) {
		if (meta_tci->tci & cpu_to_be16(NFP_FLOWER_MASK_VLAN_PRESENT)) {
			u16 vlan_tci = be16_to_cpu(meta_tci->tci);

			vlan_tci &= ~NFP_FLOWER_MASK_VLAN_PRESENT;
			flow->pre_tun_rule.vlan_tci = cpu_to_be16(vlan_tci);
			vlan = true;
		} else {
			flow->pre_tun_rule.vlan_tci = cpu_to_be16(0xffff);
		}
1152 1153 1154 1155 1156
	}

	if (key_layer & ~NFP_FLOWER_PRE_TUN_RULE_FIELDS) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: too many match fields");
		return -EOPNOTSUPP;
1157 1158 1159
	} else if (key_ls->key_layer_two & ~NFP_FLOWER_LAYER2_QINQ) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: non-vlan in extended match fields");
		return -EOPNOTSUPP;
1160 1161 1162 1163 1164 1165 1166
	}

	if (!(key_layer & NFP_FLOWER_LAYER_MAC)) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: MAC fields match required");
		return -EOPNOTSUPP;
	}

1167 1168 1169 1170 1171 1172
	if (!(key_layer & NFP_FLOWER_LAYER_IPV4) &&
	    !(key_layer & NFP_FLOWER_LAYER_IPV6)) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: match on ipv4/ipv6 eth_type must be present");
		return -EOPNOTSUPP;
	}

1173 1174
	/* Skip fields known to exist. */
	mask += sizeof(struct nfp_flower_meta_tci);
1175 1176 1177 1178 1179
	ext += sizeof(struct nfp_flower_meta_tci);
	if (key_ls->key_layer_two) {
		mask += sizeof(struct nfp_flower_ext_meta);
		ext += sizeof(struct nfp_flower_ext_meta);
	}
1180
	mask += sizeof(struct nfp_flower_in_port);
1181
	ext += sizeof(struct nfp_flower_in_port);
1182

1183 1184 1185 1186 1187 1188 1189
	/* Ensure destination MAC address matches pre_tun_dev. */
	mac = (struct nfp_flower_mac_mpls *)ext;
	if (memcmp(&mac->mac_dst[0], flow->pre_tun_rule.dev->dev_addr, 6)) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: dest MAC must match output dev MAC");
		return -EOPNOTSUPP;
	}

1190 1191 1192 1193 1194 1195 1196
	/* Ensure destination MAC address is fully matched. */
	mac = (struct nfp_flower_mac_mpls *)mask;
	if (!is_broadcast_ether_addr(&mac->mac_dst[0])) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: dest MAC field must not be masked");
		return -EOPNOTSUPP;
	}

1197 1198 1199 1200 1201
	if (mac->mpls_lse) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: MPLS not supported");
		return -EOPNOTSUPP;
	}

1202 1203
	mask += sizeof(struct nfp_flower_mac_mpls);
	ext += sizeof(struct nfp_flower_mac_mpls);
1204 1205 1206
	if (key_layer & NFP_FLOWER_LAYER_IPV4 ||
	    key_layer & NFP_FLOWER_LAYER_IPV6) {
		/* Flags and proto fields have same offset in IPv4 and IPv6. */
1207 1208
		int ip_flags = offsetof(struct nfp_flower_ipv4, ip_ext.flags);
		int ip_proto = offsetof(struct nfp_flower_ipv4, ip_ext.proto);
1209
		int size;
1210 1211
		int i;

1212 1213 1214 1215
		size = key_layer & NFP_FLOWER_LAYER_IPV4 ?
			sizeof(struct nfp_flower_ipv4) :
			sizeof(struct nfp_flower_ipv6);

1216 1217

		/* Ensure proto and flags are the only IP layer fields. */
1218
		for (i = 0; i < size; i++)
1219 1220 1221 1222
			if (mask[i] && i != ip_flags && i != ip_proto) {
				NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: only flags and proto can be matched in ip header");
				return -EOPNOTSUPP;
			}
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
		ext += size;
		mask += size;
	}

	if ((priv->flower_ext_feats & NFP_FL_FEATS_VLAN_QINQ)) {
		if (key_ls->key_layer_two & NFP_FLOWER_LAYER2_QINQ) {
			struct nfp_flower_vlan *vlan_tags;
			u16 vlan_tci;

			vlan_tags = (struct nfp_flower_vlan *)ext;

			vlan_tci = be16_to_cpu(vlan_tags->outer_tci);

			vlan_tci &= ~NFP_FLOWER_MASK_VLAN_PRESENT;
			flow->pre_tun_rule.vlan_tci = cpu_to_be16(vlan_tci);
			vlan = true;
		} else {
			flow->pre_tun_rule.vlan_tci = cpu_to_be16(0xffff);
		}
1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
	}

	/* Action must be a single egress or pop_vlan and egress. */
	act_offset = 0;
	act = (struct nfp_fl_act_head *)&flow->action_data[act_offset];
	if (vlan) {
		if (act->jump_id != NFP_FL_ACTION_OPCODE_POP_VLAN) {
			NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: match on VLAN must have VLAN pop as first action");
			return -EOPNOTSUPP;
		}

		act_offset += act->len_lw << NFP_FL_LW_SIZ;
		act = (struct nfp_fl_act_head *)&flow->action_data[act_offset];
	}

	if (act->jump_id != NFP_FL_ACTION_OPCODE_OUTPUT) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: non egress action detected where egress was expected");
		return -EOPNOTSUPP;
	}

	act_offset += act->len_lw << NFP_FL_LW_SIZ;

	/* Ensure there are no more actions after egress. */
	if (act_offset != flow->meta.act_len) {
		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: egress is not the last action");
		return -EOPNOTSUPP;
	}

	return 0;
1271 1272
}

1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
static bool offload_pre_check(struct flow_cls_offload *flow)
{
	struct flow_rule *rule = flow_cls_offload_flow_rule(flow);
	struct flow_dissector *dissector = rule->match.dissector;

	if (dissector->used_keys & BIT(FLOW_DISSECTOR_KEY_CT))
		return false;

	if (flow->common.chain_index)
		return false;

	return true;
}

1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
/**
 * nfp_flower_add_offload() - Adds a new flow to hardware.
 * @app:	Pointer to the APP handle
 * @netdev:	netdev structure.
 * @flow:	TC flower classifier offload structure.
 *
 * Adds a new flow to the repeated hash structure and action payload.
 *
 * Return: negative value on error, 0 if configured successfully.
 */
static int
nfp_flower_add_offload(struct nfp_app *app, struct net_device *netdev,
1299
		       struct flow_cls_offload *flow)
1300
{
1301
	struct flow_rule *rule = flow_cls_offload_flow_rule(flow);
1302
	enum nfp_flower_tun_type tun_type = NFP_FL_TUNNEL_NONE;
1303
	struct nfp_flower_priv *priv = app->priv;
1304
	struct netlink_ext_ack *extack = NULL;
1305 1306
	struct nfp_fl_payload *flow_pay;
	struct nfp_fl_key_ls *key_layer;
1307
	struct nfp_port *port = NULL;
1308 1309
	int err;

1310
	extack = flow->common.extack;
1311 1312 1313
	if (nfp_netdev_is_nfp_repr(netdev))
		port = nfp_port_from_netdev(netdev);

1314 1315 1316 1317 1318 1319
	if (is_pre_ct_flow(flow))
		return nfp_fl_ct_handle_pre_ct(priv, netdev, flow, extack);

	if (is_post_ct_flow(flow))
		return nfp_fl_ct_handle_post_ct(priv, netdev, flow, extack);

1320 1321 1322
	if (!offload_pre_check(flow))
		return -EOPNOTSUPP;

1323 1324 1325 1326
	key_layer = kmalloc(sizeof(*key_layer), GFP_KERNEL);
	if (!key_layer)
		return -ENOMEM;

1327
	err = nfp_flower_calculate_key_layers(app, netdev, key_layer, rule,
1328
					      &tun_type, extack);
1329 1330 1331
	if (err)
		goto err_free_key_ls;

1332
	flow_pay = nfp_flower_allocate_new(key_layer);
1333 1334 1335 1336 1337
	if (!flow_pay) {
		err = -ENOMEM;
		goto err_free_key_ls;
	}

1338
	err = nfp_flower_compile_flow_match(app, rule, key_layer, netdev,
1339
					    flow_pay, tun_type, extack);
1340 1341 1342
	if (err)
		goto err_destroy_flow;

1343
	err = nfp_flower_compile_action(app, rule, netdev, flow_pay, extack);
1344 1345 1346
	if (err)
		goto err_destroy_flow;

1347
	if (flow_pay->pre_tun_rule.dev) {
1348
		err = nfp_flower_validate_pre_tun_rule(app, flow_pay, key_layer, extack);
1349 1350 1351 1352
		if (err)
			goto err_destroy_flow;
	}

1353
	err = nfp_compile_flow_metadata(app, flow->cookie, flow_pay, netdev, extack);
1354 1355 1356 1357
	if (err)
		goto err_destroy_flow;

	flow_pay->tc_flower_cookie = flow->cookie;
1358 1359
	err = rhashtable_insert_fast(&priv->flow_table, &flow_pay->fl_node,
				     nfp_flower_table_params);
1360 1361
	if (err) {
		NL_SET_ERR_MSG_MOD(extack, "invalid entry: cannot insert flow into tables for offloads");
1362
		goto err_release_metadata;
1363
	}
1364

1365 1366 1367 1368 1369
	if (flow_pay->pre_tun_rule.dev)
		err = nfp_flower_xmit_pre_tun_flow(app, flow_pay);
	else
		err = nfp_flower_xmit_flow(app, flow_pay,
					   NFP_FLOWER_CMSG_TYPE_FLOW_ADD);
1370 1371
	if (err)
		goto err_remove_rhash;
1372

1373 1374
	if (port)
		port->tc_offload_cnt++;
1375

J
John Hurley 已提交
1376 1377
	flow_pay->in_hw = true;

1378 1379 1380 1381
	/* Deallocate flow payload when flower rule has been destroyed. */
	kfree(key_layer);

	return 0;
1382

1383 1384 1385 1386
err_remove_rhash:
	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
					    &flow_pay->fl_node,
					    nfp_flower_table_params));
1387 1388
err_release_metadata:
	nfp_modify_flow_metadata(app, flow_pay);
1389
err_destroy_flow:
1390 1391
	if (flow_pay->nfp_tun_ipv6)
		nfp_tunnel_put_ipv6_off(app, flow_pay->nfp_tun_ipv6);
1392
	kfree(flow_pay->action_data);
1393 1394 1395 1396 1397 1398
	kfree(flow_pay->mask_data);
	kfree(flow_pay->unmasked_data);
	kfree(flow_pay);
err_free_key_ls:
	kfree(key_layer);
	return err;
1399 1400
}

J
John Hurley 已提交
1401 1402 1403 1404 1405 1406 1407
static void
nfp_flower_remove_merge_flow(struct nfp_app *app,
			     struct nfp_fl_payload *del_sub_flow,
			     struct nfp_fl_payload *merge_flow)
{
	struct nfp_flower_priv *priv = app->priv;
	struct nfp_fl_payload_link *link, *temp;
1408
	struct nfp_merge_info *merge_info;
J
John Hurley 已提交
1409
	struct nfp_fl_payload *origin;
1410
	u64 parent_ctx = 0;
J
John Hurley 已提交
1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
	bool mod = false;
	int err;

	link = list_first_entry(&merge_flow->linked_flows,
				struct nfp_fl_payload_link, merge_flow.list);
	origin = link->sub_flow.flow;

	/* Re-add rule the merge had overwritten if it has not been deleted. */
	if (origin != del_sub_flow)
		mod = true;

	err = nfp_modify_flow_metadata(app, merge_flow);
	if (err) {
		nfp_flower_cmsg_warn(app, "Metadata fail for merge flow delete.\n");
		goto err_free_links;
	}

	if (!mod) {
		err = nfp_flower_xmit_flow(app, merge_flow,
					   NFP_FLOWER_CMSG_TYPE_FLOW_DEL);
		if (err) {
			nfp_flower_cmsg_warn(app, "Failed to delete merged flow.\n");
			goto err_free_links;
		}
	} else {
		__nfp_modify_flow_metadata(priv, origin);
		err = nfp_flower_xmit_flow(app, origin,
					   NFP_FLOWER_CMSG_TYPE_FLOW_MOD);
		if (err)
			nfp_flower_cmsg_warn(app, "Failed to revert merge flow.\n");
		origin->in_hw = true;
	}

err_free_links:
	/* Clean any links connected with the merged flow. */
	list_for_each_entry_safe(link, temp, &merge_flow->linked_flows,
1447 1448 1449 1450
				 merge_flow.list) {
		u32 ctx_id = be32_to_cpu(link->sub_flow.flow->meta.host_ctx_id);

		parent_ctx = (parent_ctx << 32) | (u64)(ctx_id);
J
John Hurley 已提交
1451
		nfp_flower_unlink_flow(link);
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
	}

	merge_info = rhashtable_lookup_fast(&priv->merge_table,
					    &parent_ctx,
					    merge_table_params);
	if (merge_info) {
		WARN_ON_ONCE(rhashtable_remove_fast(&priv->merge_table,
						    &merge_info->ht_node,
						    merge_table_params));
		kfree(merge_info);
	}
J
John Hurley 已提交
1463 1464 1465 1466 1467 1468 1469 1470 1471 1472

	kfree(merge_flow->action_data);
	kfree(merge_flow->mask_data);
	kfree(merge_flow->unmasked_data);
	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
					    &merge_flow->fl_node,
					    nfp_flower_table_params));
	kfree_rcu(merge_flow, rcu);
}

1473
void
J
John Hurley 已提交
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
nfp_flower_del_linked_merge_flows(struct nfp_app *app,
				  struct nfp_fl_payload *sub_flow)
{
	struct nfp_fl_payload_link *link, *temp;

	/* Remove any merge flow formed from the deleted sub_flow. */
	list_for_each_entry_safe(link, temp, &sub_flow->linked_flows,
				 sub_flow.list)
		nfp_flower_remove_merge_flow(app, sub_flow,
					     link->merge_flow.flow);
}

1486 1487 1488 1489 1490 1491 1492
/**
 * nfp_flower_del_offload() - Removes a flow from hardware.
 * @app:	Pointer to the APP handle
 * @netdev:	netdev structure.
 * @flow:	TC flower classifier offload structure
 *
 * Removes a flow from the repeated hash structure and clears the
J
John Hurley 已提交
1493
 * action payload. Any flows merged from this are also deleted.
1494 1495 1496 1497 1498
 *
 * Return: negative value on error, 0 if removed successfully.
 */
static int
nfp_flower_del_offload(struct nfp_app *app, struct net_device *netdev,
1499
		       struct flow_cls_offload *flow)
1500
{
1501
	struct nfp_flower_priv *priv = app->priv;
1502
	struct nfp_fl_ct_map_entry *ct_map_ent;
1503
	struct netlink_ext_ack *extack = NULL;
1504
	struct nfp_fl_payload *nfp_flow;
1505
	struct nfp_port *port = NULL;
1506 1507
	int err;

1508
	extack = flow->common.extack;
1509 1510 1511
	if (nfp_netdev_is_nfp_repr(netdev))
		port = nfp_port_from_netdev(netdev);

1512 1513 1514 1515 1516 1517 1518 1519
	/* Check ct_map_table */
	ct_map_ent = rhashtable_lookup_fast(&priv->ct_map_table, &flow->cookie,
					    nfp_ct_map_params);
	if (ct_map_ent) {
		err = nfp_fl_ct_del_flow(ct_map_ent);
		return err;
	}

1520
	nfp_flow = nfp_flower_search_fl_table(app, flow->cookie, netdev);
1521 1522
	if (!nfp_flow) {
		NL_SET_ERR_MSG_MOD(extack, "invalid entry: cannot remove flow that does not exist");
1523
		return -ENOENT;
1524
	}
1525 1526

	err = nfp_modify_flow_metadata(app, nfp_flow);
1527
	if (err)
J
John Hurley 已提交
1528
		goto err_free_merge_flow;
1529

1530 1531 1532
	if (nfp_flow->nfp_tun_ipv4_addr)
		nfp_tunnel_del_ipv4_off(app, nfp_flow->nfp_tun_ipv4_addr);

1533 1534 1535
	if (nfp_flow->nfp_tun_ipv6)
		nfp_tunnel_put_ipv6_off(app, nfp_flow->nfp_tun_ipv6);

J
John Hurley 已提交
1536 1537 1538 1539 1540
	if (!nfp_flow->in_hw) {
		err = 0;
		goto err_free_merge_flow;
	}

1541 1542 1543 1544 1545
	if (nfp_flow->pre_tun_rule.dev)
		err = nfp_flower_xmit_pre_tun_del_flow(app, nfp_flow);
	else
		err = nfp_flower_xmit_flow(app, nfp_flow,
					   NFP_FLOWER_CMSG_TYPE_FLOW_DEL);
J
John Hurley 已提交
1546
	/* Fall through on error. */
1547

J
John Hurley 已提交
1548 1549
err_free_merge_flow:
	nfp_flower_del_linked_merge_flows(app, nfp_flow);
1550 1551
	if (port)
		port->tc_offload_cnt--;
1552 1553 1554
	kfree(nfp_flow->action_data);
	kfree(nfp_flow->mask_data);
	kfree(nfp_flow->unmasked_data);
1555 1556 1557
	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
					    &nfp_flow->fl_node,
					    nfp_flower_table_params));
1558 1559
	kfree_rcu(nfp_flow, rcu);
	return err;
1560 1561
}

1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592
static void
__nfp_flower_update_merge_stats(struct nfp_app *app,
				struct nfp_fl_payload *merge_flow)
{
	struct nfp_flower_priv *priv = app->priv;
	struct nfp_fl_payload_link *link;
	struct nfp_fl_payload *sub_flow;
	u64 pkts, bytes, used;
	u32 ctx_id;

	ctx_id = be32_to_cpu(merge_flow->meta.host_ctx_id);
	pkts = priv->stats[ctx_id].pkts;
	/* Do not cycle subflows if no stats to distribute. */
	if (!pkts)
		return;
	bytes = priv->stats[ctx_id].bytes;
	used = priv->stats[ctx_id].used;

	/* Reset stats for the merge flow. */
	priv->stats[ctx_id].pkts = 0;
	priv->stats[ctx_id].bytes = 0;

	/* The merge flow has received stats updates from firmware.
	 * Distribute these stats to all subflows that form the merge.
	 * The stats will collected from TC via the subflows.
	 */
	list_for_each_entry(link, &merge_flow->linked_flows, merge_flow.list) {
		sub_flow = link->sub_flow.flow;
		ctx_id = be32_to_cpu(sub_flow->meta.host_ctx_id);
		priv->stats[ctx_id].pkts += pkts;
		priv->stats[ctx_id].bytes += bytes;
1593 1594
		priv->stats[ctx_id].used = max_t(u64, used,
						 priv->stats[ctx_id].used);
1595 1596 1597
	}
}

1598
void
1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
nfp_flower_update_merge_stats(struct nfp_app *app,
			      struct nfp_fl_payload *sub_flow)
{
	struct nfp_fl_payload_link *link;

	/* Get merge flows that the subflow forms to distribute their stats. */
	list_for_each_entry(link, &sub_flow->linked_flows, sub_flow.list)
		__nfp_flower_update_merge_stats(app, link->merge_flow.flow);
}

1609 1610 1611
/**
 * nfp_flower_get_stats() - Populates flow stats obtained from hardware.
 * @app:	Pointer to the APP handle
1612
 * @netdev:	Netdev structure.
1613 1614 1615 1616 1617 1618 1619 1620
 * @flow:	TC flower classifier offload structure
 *
 * Populates a flow statistics structure which which corresponds to a
 * specific flow.
 *
 * Return: negative value on error, 0 if stats populated successfully.
 */
static int
1621
nfp_flower_get_stats(struct nfp_app *app, struct net_device *netdev,
1622
		     struct flow_cls_offload *flow)
1623
{
1624
	struct nfp_flower_priv *priv = app->priv;
1625
	struct nfp_fl_ct_map_entry *ct_map_ent;
1626
	struct netlink_ext_ack *extack = NULL;
1627
	struct nfp_fl_payload *nfp_flow;
1628
	u32 ctx_id;
1629

1630 1631 1632 1633 1634 1635
	/* Check ct_map table first */
	ct_map_ent = rhashtable_lookup_fast(&priv->ct_map_table, &flow->cookie,
					    nfp_ct_map_params);
	if (ct_map_ent)
		return nfp_fl_ct_stats(flow, ct_map_ent);

1636
	extack = flow->common.extack;
1637
	nfp_flow = nfp_flower_search_fl_table(app, flow->cookie, netdev);
1638 1639
	if (!nfp_flow) {
		NL_SET_ERR_MSG_MOD(extack, "invalid entry: cannot dump stats for flow that does not exist");
1640
		return -EINVAL;
1641
	}
1642

1643 1644 1645
	ctx_id = be32_to_cpu(nfp_flow->meta.host_ctx_id);

	spin_lock_bh(&priv->stats_lock);
1646 1647 1648 1649
	/* If request is for a sub_flow, update stats from merged flows. */
	if (!list_empty(&nfp_flow->linked_flows))
		nfp_flower_update_merge_stats(app, nfp_flow);

1650
	flow_stats_update(&flow->stats, priv->stats[ctx_id].bytes,
1651
			  priv->stats[ctx_id].pkts, 0, priv->stats[ctx_id].used,
1652
			  FLOW_ACTION_HW_STATS_DELAYED);
1653

1654 1655 1656
	priv->stats[ctx_id].pkts = 0;
	priv->stats[ctx_id].bytes = 0;
	spin_unlock_bh(&priv->stats_lock);
1657 1658

	return 0;
1659 1660 1661 1662
}

static int
nfp_flower_repr_offload(struct nfp_app *app, struct net_device *netdev,
1663
			struct flow_cls_offload *flower)
1664
{
1665
	if (!eth_proto_is_802_3(flower->common.protocol))
1666 1667
		return -EOPNOTSUPP;

1668
	switch (flower->command) {
1669
	case FLOW_CLS_REPLACE:
1670
		return nfp_flower_add_offload(app, netdev, flower);
1671
	case FLOW_CLS_DESTROY:
1672
		return nfp_flower_del_offload(app, netdev, flower);
1673
	case FLOW_CLS_STATS:
1674
		return nfp_flower_get_stats(app, netdev, flower);
1675 1676 1677
	default:
		return -EOPNOTSUPP;
	}
1678 1679
}

1680 1681 1682
static int nfp_flower_setup_tc_block_cb(enum tc_setup_type type,
					void *type_data, void *cb_priv)
{
1683
	struct flow_cls_common_offload *common = type_data;
1684
	struct nfp_repr *repr = cb_priv;
1685

1686
	if (!tc_can_offload_extack(repr->netdev, common->extack))
1687 1688
		return -EOPNOTSUPP;

1689 1690
	switch (type) {
	case TC_SETUP_CLSFLOWER:
1691
		return nfp_flower_repr_offload(repr->app, repr->netdev,
1692
					       type_data);
1693 1694 1695
	case TC_SETUP_CLSMATCHALL:
		return nfp_flower_setup_qos_offload(repr->app, repr->netdev,
						    type_data);
1696 1697 1698 1699 1700
	default:
		return -EOPNOTSUPP;
	}
}

1701 1702
static LIST_HEAD(nfp_block_cb_list);

1703
static int nfp_flower_setup_tc_block(struct net_device *netdev,
1704
				     struct flow_block_offload *f)
1705
{
1706
	struct nfp_repr *repr = netdev_priv(netdev);
1707
	struct nfp_flower_repr_priv *repr_priv;
1708
	struct flow_block_cb *block_cb;
1709

1710
	if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
1711 1712
		return -EOPNOTSUPP;

1713
	repr_priv = repr->app_priv;
1714 1715
	repr_priv->block_shared = f->block_shared;
	f->driver_block_list = &nfp_block_cb_list;
1716

1717
	switch (f->command) {
1718
	case FLOW_BLOCK_BIND:
1719 1720 1721 1722
		if (flow_block_cb_is_busy(nfp_flower_setup_tc_block_cb, repr,
					  &nfp_block_cb_list))
			return -EBUSY;

1723
		block_cb = flow_block_cb_alloc(nfp_flower_setup_tc_block_cb,
1724 1725 1726 1727 1728 1729 1730
					       repr, repr, NULL);
		if (IS_ERR(block_cb))
			return PTR_ERR(block_cb);

		flow_block_cb_add(block_cb, f);
		list_add_tail(&block_cb->driver_list, &nfp_block_cb_list);
		return 0;
1731
	case FLOW_BLOCK_UNBIND:
1732 1733
		block_cb = flow_block_cb_lookup(f->block,
						nfp_flower_setup_tc_block_cb,
1734 1735 1736 1737 1738 1739
						repr);
		if (!block_cb)
			return -ENOENT;

		flow_block_cb_remove(block_cb, f);
		list_del(&block_cb->driver_list);
1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754
		return 0;
	default:
		return -EOPNOTSUPP;
	}
}

int nfp_flower_setup_tc(struct nfp_app *app, struct net_device *netdev,
			enum tc_setup_type type, void *type_data)
{
	switch (type) {
	case TC_SETUP_BLOCK:
		return nfp_flower_setup_tc_block(netdev, type_data);
	default:
		return -EOPNOTSUPP;
	}
1755
}
1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779

struct nfp_flower_indr_block_cb_priv {
	struct net_device *netdev;
	struct nfp_app *app;
	struct list_head list;
};

static struct nfp_flower_indr_block_cb_priv *
nfp_flower_indr_block_cb_priv_lookup(struct nfp_app *app,
				     struct net_device *netdev)
{
	struct nfp_flower_indr_block_cb_priv *cb_priv;
	struct nfp_flower_priv *priv = app->priv;

	/* All callback list access should be protected by RTNL. */
	ASSERT_RTNL();

	list_for_each_entry(cb_priv, &priv->indr_block_cb_priv, list)
		if (cb_priv->netdev == netdev)
			return cb_priv;

	return NULL;
}

1780 1781
static int nfp_flower_setup_indr_block_cb(enum tc_setup_type type,
					  void *type_data, void *cb_priv)
1782 1783 1784 1785 1786 1787
{
	struct nfp_flower_indr_block_cb_priv *priv = cb_priv;

	switch (type) {
	case TC_SETUP_CLSFLOWER:
		return nfp_flower_repr_offload(priv->app, priv->netdev,
1788
					       type_data);
1789 1790 1791 1792 1793
	default:
		return -EOPNOTSUPP;
	}
}

1794
void nfp_flower_setup_indr_tc_release(void *cb_priv)
1795 1796 1797 1798 1799 1800 1801
{
	struct nfp_flower_indr_block_cb_priv *priv = cb_priv;

	list_del(&priv->list);
	kfree(priv);
}

1802
static int
1803
nfp_flower_setup_indr_tc_block(struct net_device *netdev, struct Qdisc *sch, struct nfp_app *app,
1804 1805
			       struct flow_block_offload *f, void *data,
			       void (*cleanup)(struct flow_block_cb *block_cb))
1806 1807 1808
{
	struct nfp_flower_indr_block_cb_priv *cb_priv;
	struct nfp_flower_priv *priv = app->priv;
1809
	struct flow_block_cb *block_cb;
1810

1811 1812 1813 1814
	if ((f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS &&
	     !nfp_flower_internal_port_can_offload(app, netdev)) ||
	    (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_EGRESS &&
	     nfp_flower_internal_port_can_offload(app, netdev)))
1815 1816 1817
		return -EOPNOTSUPP;

	switch (f->command) {
1818
	case FLOW_BLOCK_BIND:
1819 1820 1821 1822 1823 1824 1825
		cb_priv = nfp_flower_indr_block_cb_priv_lookup(app, netdev);
		if (cb_priv &&
		    flow_block_cb_is_busy(nfp_flower_setup_indr_block_cb,
					  cb_priv,
					  &nfp_block_cb_list))
			return -EBUSY;

1826 1827 1828 1829 1830 1831 1832 1833
		cb_priv = kmalloc(sizeof(*cb_priv), GFP_KERNEL);
		if (!cb_priv)
			return -ENOMEM;

		cb_priv->netdev = netdev;
		cb_priv->app = app;
		list_add(&cb_priv->list, &priv->indr_block_cb_priv);

1834 1835 1836
		block_cb = flow_indr_block_cb_alloc(nfp_flower_setup_indr_block_cb,
						    cb_priv, cb_priv,
						    nfp_flower_setup_indr_tc_release,
1837
						    f, netdev, sch, data, app, cleanup);
1838
		if (IS_ERR(block_cb)) {
1839 1840
			list_del(&cb_priv->list);
			kfree(cb_priv);
1841
			return PTR_ERR(block_cb);
1842 1843
		}

1844 1845 1846
		flow_block_cb_add(block_cb, f);
		list_add_tail(&block_cb->driver_list, &nfp_block_cb_list);
		return 0;
1847
	case FLOW_BLOCK_UNBIND:
1848
		cb_priv = nfp_flower_indr_block_cb_priv_lookup(app, netdev);
1849 1850 1851
		if (!cb_priv)
			return -ENOENT;

1852
		block_cb = flow_block_cb_lookup(f->block,
1853 1854 1855 1856
						nfp_flower_setup_indr_block_cb,
						cb_priv);
		if (!block_cb)
			return -ENOENT;
1857

1858
		flow_indr_block_cb_remove(block_cb, f);
1859
		list_del(&block_cb->driver_list);
1860 1861 1862 1863 1864 1865 1866
		return 0;
	default:
		return -EOPNOTSUPP;
	}
	return 0;
}

1867
int
1868
nfp_flower_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch, void *cb_priv,
1869 1870 1871
			    enum tc_setup_type type, void *type_data,
			    void *data,
			    void (*cleanup)(struct flow_block_cb *block_cb))
1872
{
1873 1874 1875
	if (!nfp_fl_is_netdev_to_offload(netdev))
		return -EOPNOTSUPP;

1876 1877
	switch (type) {
	case TC_SETUP_BLOCK:
1878
		return nfp_flower_setup_indr_tc_block(netdev, sch, cb_priv,
1879
						      type_data, data, cleanup);
1880 1881 1882 1883
	default:
		return -EOPNOTSUPP;
	}
}