cxgb4_tc_flower.c

   1/*
   2 * This file is part of the Chelsio T4/T5/T6 Ethernet driver for Linux.
   3 *
   4 * Copyright (c) 2017 Chelsio Communications, Inc. All rights reserved.
   5 *
   6 * This software is available to you under a choice of one of two
   7 * licenses.  You may choose to be licensed under the terms of the GNU
   8 * General Public License (GPL) Version 2, available from the file
   9 * COPYING in the main directory of this source tree, or the
  10 * OpenIB.org BSD license below:
  11 *
  12 *     Redistribution and use in source and binary forms, with or
  13 *     without modification, are permitted provided that the following
  14 *     conditions are met:
  15 *
  16 *      - Redistributions of source code must retain the above
  17 *        copyright notice, this list of conditions and the following
  18 *        disclaimer.
  19 *
  20 *      - Redistributions in binary form must reproduce the above
  21 *        copyright notice, this list of conditions and the following
  22 *        disclaimer in the documentation and/or other materials
  23 *        provided with the distribution.
  24 *
  25 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  26 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  27 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  28 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  29 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  30 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  31 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  32 * SOFTWARE.
  33 */
  34
  35#include <net/tc_act/tc_mirred.h>
  36#include <net/tc_act/tc_pedit.h>
  37#include <net/tc_act/tc_gact.h>
  38#include <net/tc_act/tc_vlan.h>
  39
  40#include "cxgb4.h"
  41#include "cxgb4_filter.h"
  42#include "cxgb4_tc_flower.h"
  43
  44#define STATS_CHECK_PERIOD (HZ / 2)
  45
  46static struct ch_tc_pedit_fields pedits[] = {
  47	PEDIT_FIELDS(ETH_, DMAC_31_0, 4, dmac, 0),
  48	PEDIT_FIELDS(ETH_, DMAC_47_32, 2, dmac, 4),
  49	PEDIT_FIELDS(ETH_, SMAC_15_0, 2, smac, 0),
  50	PEDIT_FIELDS(ETH_, SMAC_47_16, 4, smac, 2),
  51	PEDIT_FIELDS(IP4_, SRC, 4, nat_fip, 0),
  52	PEDIT_FIELDS(IP4_, DST, 4, nat_lip, 0),
  53	PEDIT_FIELDS(IP6_, SRC_31_0, 4, nat_fip, 0),
  54	PEDIT_FIELDS(IP6_, SRC_63_32, 4, nat_fip, 4),
  55	PEDIT_FIELDS(IP6_, SRC_95_64, 4, nat_fip, 8),
  56	PEDIT_FIELDS(IP6_, SRC_127_96, 4, nat_fip, 12),
  57	PEDIT_FIELDS(IP6_, DST_31_0, 4, nat_lip, 0),
  58	PEDIT_FIELDS(IP6_, DST_63_32, 4, nat_lip, 4),
  59	PEDIT_FIELDS(IP6_, DST_95_64, 4, nat_lip, 8),
  60	PEDIT_FIELDS(IP6_, DST_127_96, 4, nat_lip, 12),
  61};
  62
  63static const struct cxgb4_natmode_config cxgb4_natmode_config_array[] = {
  64	/* Default supported NAT modes */
  65	{
  66		.chip = CHELSIO_T5,
  67		.flags = CXGB4_ACTION_NATMODE_NONE,
  68		.natmode = NAT_MODE_NONE,
  69	},
  70	{
  71		.chip = CHELSIO_T5,
  72		.flags = CXGB4_ACTION_NATMODE_DIP,
  73		.natmode = NAT_MODE_DIP,
  74	},
  75	{
  76		.chip = CHELSIO_T5,
  77		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_DPORT,
  78		.natmode = NAT_MODE_DIP_DP,
  79	},
  80	{
  81		.chip = CHELSIO_T5,
  82		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_DPORT |
  83			 CXGB4_ACTION_NATMODE_SIP,
  84		.natmode = NAT_MODE_DIP_DP_SIP,
  85	},
  86	{
  87		.chip = CHELSIO_T5,
  88		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_DPORT |
  89			 CXGB4_ACTION_NATMODE_SPORT,
  90		.natmode = NAT_MODE_DIP_DP_SP,
  91	},
  92	{
  93		.chip = CHELSIO_T5,
  94		.flags = CXGB4_ACTION_NATMODE_SIP | CXGB4_ACTION_NATMODE_SPORT,
  95		.natmode = NAT_MODE_SIP_SP,
  96	},
  97	{
  98		.chip = CHELSIO_T5,
  99		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_SIP |
 100			 CXGB4_ACTION_NATMODE_SPORT,
 101		.natmode = NAT_MODE_DIP_SIP_SP,
 102	},
 103	{
 104		.chip = CHELSIO_T5,
 105		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_SIP |
 106			 CXGB4_ACTION_NATMODE_DPORT |
 107			 CXGB4_ACTION_NATMODE_SPORT,
 108		.natmode = NAT_MODE_ALL,
 109	},
 110	/* T6+ can ignore L4 ports when they're disabled. */
 111	{
 112		.chip = CHELSIO_T6,
 113		.flags = CXGB4_ACTION_NATMODE_SIP,
 114		.natmode = NAT_MODE_SIP_SP,
 115	},
 116	{
 117		.chip = CHELSIO_T6,
 118		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_SPORT,
 119		.natmode = NAT_MODE_DIP_DP_SP,
 120	},
 121	{
 122		.chip = CHELSIO_T6,
 123		.flags = CXGB4_ACTION_NATMODE_DIP | CXGB4_ACTION_NATMODE_SIP,
 124		.natmode = NAT_MODE_ALL,
 125	},
 126};
 127
 128static void cxgb4_action_natmode_tweak(struct ch_filter_specification *fs,
 129				       u8 natmode_flags)
 130{
 131	u8 i = 0;
 132
 133	/* Translate the enabled NAT 4-tuple fields to one of the
 134	 * hardware supported NAT mode configurations. This ensures
 135	 * that we pick a valid combination, where the disabled fields
 136	 * do not get overwritten to 0.
 137	 */
 138	for (i = 0; i < ARRAY_SIZE(cxgb4_natmode_config_array); i++) {
 139		if (cxgb4_natmode_config_array[i].flags == natmode_flags) {
 140			fs->nat_mode = cxgb4_natmode_config_array[i].natmode;
 141			return;
 142		}
 143	}
 144}
 145
 146static struct ch_tc_flower_entry *allocate_flower_entry(void)
 147{
 148	struct ch_tc_flower_entry *new = kzalloc(sizeof(*new), GFP_KERNEL);
 149	if (new)
 150		spin_lock_init(&new->lock);
 151	return new;
 152}
 153
 154/* Must be called with either RTNL or rcu_read_lock */
 155static struct ch_tc_flower_entry *ch_flower_lookup(struct adapter *adap,
 156						   unsigned long flower_cookie)
 157{
 158	return rhashtable_lookup_fast(&adap->flower_tbl, &flower_cookie,
 159				      adap->flower_ht_params);
 160}
 161
 162static void cxgb4_process_flow_match(struct net_device *dev,
 163				     struct flow_rule *rule,
 164				     u16 addr_type,
 165				     struct ch_filter_specification *fs)
 166{
 167
 168	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
 169		struct flow_match_basic match;
 170		u16 ethtype_key, ethtype_mask;
 171
 172		flow_rule_match_basic(rule, &match);
 173		ethtype_key = ntohs(match.key->n_proto);
 174		ethtype_mask = ntohs(match.mask->n_proto);
 175
 176		if (ethtype_key == ETH_P_ALL) {
 177			ethtype_key = 0;
 178			ethtype_mask = 0;
 179		}
 180
 181		if (ethtype_key == ETH_P_IPV6)
 182			fs->type = 1;
 183
 184		fs->val.ethtype = ethtype_key;
 185		fs->mask.ethtype = ethtype_mask;
 186		fs->val.proto = match.key->ip_proto;
 187		fs->mask.proto = match.mask->ip_proto;
 188	}
 189
 190	if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
 191		struct flow_match_ipv4_addrs match;
 192
 193		flow_rule_match_ipv4_addrs(rule, &match);
 194		fs->type = 0;
 195		memcpy(&fs->val.lip[0], &match.key->dst, sizeof(match.key->dst));
 196		memcpy(&fs->val.fip[0], &match.key->src, sizeof(match.key->src));
 197		memcpy(&fs->mask.lip[0], &match.mask->dst, sizeof(match.mask->dst));
 198		memcpy(&fs->mask.fip[0], &match.mask->src, sizeof(match.mask->src));
 199
 200		/* also initialize nat_lip/fip to same values */
 201		memcpy(&fs->nat_lip[0], &match.key->dst, sizeof(match.key->dst));
 202		memcpy(&fs->nat_fip[0], &match.key->src, sizeof(match.key->src));
 203	}
 204
 205	if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
 206		struct flow_match_ipv6_addrs match;
 207
 208		flow_rule_match_ipv6_addrs(rule, &match);
 209		fs->type = 1;
 210		memcpy(&fs->val.lip[0], match.key->dst.s6_addr,
 211		       sizeof(match.key->dst));
 212		memcpy(&fs->val.fip[0], match.key->src.s6_addr,
 213		       sizeof(match.key->src));
 214		memcpy(&fs->mask.lip[0], match.mask->dst.s6_addr,
 215		       sizeof(match.mask->dst));
 216		memcpy(&fs->mask.fip[0], match.mask->src.s6_addr,
 217		       sizeof(match.mask->src));
 218
 219		/* also initialize nat_lip/fip to same values */
 220		memcpy(&fs->nat_lip[0], match.key->dst.s6_addr,
 221		       sizeof(match.key->dst));
 222		memcpy(&fs->nat_fip[0], match.key->src.s6_addr,
 223		       sizeof(match.key->src));
 224	}
 225
 226	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS)) {
 227		struct flow_match_ports match;
 228
 229		flow_rule_match_ports(rule, &match);
 230		fs->val.lport = be16_to_cpu(match.key->dst);
 231		fs->mask.lport = be16_to_cpu(match.mask->dst);
 232		fs->val.fport = be16_to_cpu(match.key->src);
 233		fs->mask.fport = be16_to_cpu(match.mask->src);
 234
 235		/* also initialize nat_lport/fport to same values */
 236		fs->nat_lport = fs->val.lport;
 237		fs->nat_fport = fs->val.fport;
 238	}
 239
 240	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IP)) {
 241		struct flow_match_ip match;
 242
 243		flow_rule_match_ip(rule, &match);
 244		fs->val.tos = match.key->tos;
 245		fs->mask.tos = match.mask->tos;
 246	}
 247
 248	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
 249		struct flow_match_enc_keyid match;
 250
 251		flow_rule_match_enc_keyid(rule, &match);
 252		fs->val.vni = be32_to_cpu(match.key->keyid);
 253		fs->mask.vni = be32_to_cpu(match.mask->keyid);
 254		if (fs->mask.vni) {
 255			fs->val.encap_vld = 1;
 256			fs->mask.encap_vld = 1;
 257		}
 258	}
 259
 260	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
 261		struct flow_match_vlan match;
 262		u16 vlan_tci, vlan_tci_mask;
 263
 264		flow_rule_match_vlan(rule, &match);
 265		vlan_tci = match.key->vlan_id | (match.key->vlan_priority <<
 266					       VLAN_PRIO_SHIFT);
 267		vlan_tci_mask = match.mask->vlan_id | (match.mask->vlan_priority <<
 268						     VLAN_PRIO_SHIFT);
 269		fs->val.ivlan = vlan_tci;
 270		fs->mask.ivlan = vlan_tci_mask;
 271
 272		fs->val.ivlan_vld = 1;
 273		fs->mask.ivlan_vld = 1;
 274
 275		/* Chelsio adapters use ivlan_vld bit to match vlan packets
 276		 * as 802.1Q. Also, when vlan tag is present in packets,
 277		 * ethtype match is used then to match on ethtype of inner
 278		 * header ie. the header following the vlan header.
 279		 * So, set the ivlan_vld based on ethtype info supplied by
 280		 * TC for vlan packets if its 802.1Q. And then reset the
 281		 * ethtype value else, hw will try to match the supplied
 282		 * ethtype value with ethtype of inner header.
 283		 */
 284		if (fs->val.ethtype == ETH_P_8021Q) {
 285			fs->val.ethtype = 0;
 286			fs->mask.ethtype = 0;
 287		}
 288	}
 289
 290	/* Match only packets coming from the ingress port where this
 291	 * filter will be created.
 292	 */
 293	fs->val.iport = netdev2pinfo(dev)->port_id;
 294	fs->mask.iport = ~0;
 295}
 296
 297static int cxgb4_validate_flow_match(struct netlink_ext_ack *extack,
 298				     struct flow_rule *rule)
 299{
 300	struct flow_dissector *dissector = rule->match.dissector;
 301	u16 ethtype_mask = 0;
 302	u16 ethtype_key = 0;
 303
 304	if (dissector->used_keys &
 305	    ~(BIT_ULL(FLOW_DISSECTOR_KEY_CONTROL) |
 306	      BIT_ULL(FLOW_DISSECTOR_KEY_BASIC) |
 307	      BIT_ULL(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
 308	      BIT_ULL(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
 309	      BIT_ULL(FLOW_DISSECTOR_KEY_PORTS) |
 310	      BIT_ULL(FLOW_DISSECTOR_KEY_ENC_KEYID) |
 311	      BIT_ULL(FLOW_DISSECTOR_KEY_VLAN) |
 312	      BIT_ULL(FLOW_DISSECTOR_KEY_IP))) {
 313		NL_SET_ERR_MSG_FMT_MOD(extack,
 314				       "Unsupported key used: 0x%llx",
 315				       dissector->used_keys);
 316		return -EOPNOTSUPP;
 317	}
 318
 319	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
 320		struct flow_match_basic match;
 321
 322		flow_rule_match_basic(rule, &match);
 323		ethtype_key = ntohs(match.key->n_proto);
 324		ethtype_mask = ntohs(match.mask->n_proto);
 325	}
 326
 327	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IP)) {
 328		u16 eth_ip_type = ethtype_key & ethtype_mask;
 329		struct flow_match_ip match;
 330
 331		if (eth_ip_type != ETH_P_IP && eth_ip_type != ETH_P_IPV6) {
 332			NL_SET_ERR_MSG_MOD(extack,
 333					   "IP Key supported only with IPv4/v6");
 334			return -EINVAL;
 335		}
 336
 337		flow_rule_match_ip(rule, &match);
 338		if (match.mask->ttl) {
 339			NL_SET_ERR_MSG_MOD(extack,
 340					   "ttl match unsupported for offload");
 341			return -EOPNOTSUPP;
 342		}
 343	}
 344
 345	return 0;
 346}
 347
 348static void offload_pedit(struct ch_filter_specification *fs, u32 val, u32 mask,
 349			  u8 field)
 350{
 351	u32 set_val = val & ~mask;
 352	u32 offset = 0;
 353	u8 size = 1;
 354	int i;
 355
 356	for (i = 0; i < ARRAY_SIZE(pedits); i++) {
 357		if (pedits[i].field == field) {
 358			offset = pedits[i].offset;
 359			size = pedits[i].size;
 360			break;
 361		}
 362	}
 363	memcpy((u8 *)fs + offset, &set_val, size);
 364}
 365
 366static void process_pedit_field(struct ch_filter_specification *fs, u32 val,
 367				u32 mask, u32 offset, u8 htype,
 368				u8 *natmode_flags)
 369{
 370	switch (htype) {
 371	case FLOW_ACT_MANGLE_HDR_TYPE_ETH:
 372		switch (offset) {
 373		case PEDIT_ETH_DMAC_31_0:
 374			fs->newdmac = 1;
 375			offload_pedit(fs, val, mask, ETH_DMAC_31_0);
 376			break;
 377		case PEDIT_ETH_DMAC_47_32_SMAC_15_0:
 378			if (~mask & PEDIT_ETH_DMAC_MASK)
 379				offload_pedit(fs, val, mask, ETH_DMAC_47_32);
 380			else
 381				offload_pedit(fs, val >> 16, mask >> 16,
 382					      ETH_SMAC_15_0);
 383			break;
 384		case PEDIT_ETH_SMAC_47_16:
 385			fs->newsmac = 1;
 386			offload_pedit(fs, val, mask, ETH_SMAC_47_16);
 387		}
 388		break;
 389	case FLOW_ACT_MANGLE_HDR_TYPE_IP4:
 390		switch (offset) {
 391		case PEDIT_IP4_SRC:
 392			offload_pedit(fs, val, mask, IP4_SRC);
 393			*natmode_flags |= CXGB4_ACTION_NATMODE_SIP;
 394			break;
 395		case PEDIT_IP4_DST:
 396			offload_pedit(fs, val, mask, IP4_DST);
 397			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 398		}
 399		break;
 400	case FLOW_ACT_MANGLE_HDR_TYPE_IP6:
 401		switch (offset) {
 402		case PEDIT_IP6_SRC_31_0:
 403			offload_pedit(fs, val, mask, IP6_SRC_31_0);
 404			*natmode_flags |= CXGB4_ACTION_NATMODE_SIP;
 405			break;
 406		case PEDIT_IP6_SRC_63_32:
 407			offload_pedit(fs, val, mask, IP6_SRC_63_32);
 408			*natmode_flags |=  CXGB4_ACTION_NATMODE_SIP;
 409			break;
 410		case PEDIT_IP6_SRC_95_64:
 411			offload_pedit(fs, val, mask, IP6_SRC_95_64);
 412			*natmode_flags |= CXGB4_ACTION_NATMODE_SIP;
 413			break;
 414		case PEDIT_IP6_SRC_127_96:
 415			offload_pedit(fs, val, mask, IP6_SRC_127_96);
 416			*natmode_flags |= CXGB4_ACTION_NATMODE_SIP;
 417			break;
 418		case PEDIT_IP6_DST_31_0:
 419			offload_pedit(fs, val, mask, IP6_DST_31_0);
 420			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 421			break;
 422		case PEDIT_IP6_DST_63_32:
 423			offload_pedit(fs, val, mask, IP6_DST_63_32);
 424			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 425			break;
 426		case PEDIT_IP6_DST_95_64:
 427			offload_pedit(fs, val, mask, IP6_DST_95_64);
 428			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 429			break;
 430		case PEDIT_IP6_DST_127_96:
 431			offload_pedit(fs, val, mask, IP6_DST_127_96);
 432			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 433		}
 434		break;
 435	case FLOW_ACT_MANGLE_HDR_TYPE_TCP:
 436		switch (offset) {
 437		case PEDIT_TCP_SPORT_DPORT:
 438			if (~mask & PEDIT_TCP_UDP_SPORT_MASK) {
 439				fs->nat_fport = val;
 440				*natmode_flags |= CXGB4_ACTION_NATMODE_SPORT;
 441			} else {
 442				fs->nat_lport = val >> 16;
 443				*natmode_flags |= CXGB4_ACTION_NATMODE_DPORT;
 444			}
 445		}
 446		break;
 447	case FLOW_ACT_MANGLE_HDR_TYPE_UDP:
 448		switch (offset) {
 449		case PEDIT_UDP_SPORT_DPORT:
 450			if (~mask & PEDIT_TCP_UDP_SPORT_MASK) {
 451				fs->nat_fport = val;
 452				*natmode_flags |= CXGB4_ACTION_NATMODE_SPORT;
 453			} else {
 454				fs->nat_lport = val >> 16;
 455				*natmode_flags |= CXGB4_ACTION_NATMODE_DPORT;
 456			}
 457		}
 458		break;
 459	}
 460}
 461
 462static int cxgb4_action_natmode_validate(struct adapter *adap, u8 natmode_flags,
 463					 struct netlink_ext_ack *extack)
 464{
 465	u8 i = 0;
 466
 467	/* Extract the NAT mode to enable based on what 4-tuple fields
 468	 * are enabled to be overwritten. This ensures that the
 469	 * disabled fields don't get overwritten to 0.
 470	 */
 471	for (i = 0; i < ARRAY_SIZE(cxgb4_natmode_config_array); i++) {
 472		const struct cxgb4_natmode_config *c;
 473
 474		c = &cxgb4_natmode_config_array[i];
 475		if (CHELSIO_CHIP_VERSION(adap->params.chip) >= c->chip &&
 476		    natmode_flags == c->flags)
 477			return 0;
 478	}
 479	NL_SET_ERR_MSG_MOD(extack, "Unsupported NAT mode 4-tuple combination");
 480	return -EOPNOTSUPP;
 481}
 482
 483void cxgb4_process_flow_actions(struct net_device *in,
 484				struct flow_action *actions,
 485				struct ch_filter_specification *fs)
 486{
 487	struct flow_action_entry *act;
 488	u8 natmode_flags = 0;
 489	int i;
 490
 491	flow_action_for_each(i, act, actions) {
 492		switch (act->id) {
 493		case FLOW_ACTION_ACCEPT:
 494			fs->action = FILTER_PASS;
 495			break;
 496		case FLOW_ACTION_DROP:
 497			fs->action = FILTER_DROP;
 498			break;
 499		case FLOW_ACTION_MIRRED:
 500		case FLOW_ACTION_REDIRECT: {
 501			struct net_device *out = act->dev;
 502			struct port_info *pi = netdev_priv(out);
 503
 504			fs->action = FILTER_SWITCH;
 505			fs->eport = pi->port_id;
 506			}
 507			break;
 508		case FLOW_ACTION_VLAN_POP:
 509		case FLOW_ACTION_VLAN_PUSH:
 510		case FLOW_ACTION_VLAN_MANGLE: {
 511			u8 prio = act->vlan.prio;
 512			u16 vid = act->vlan.vid;
 513			u16 vlan_tci = (prio << VLAN_PRIO_SHIFT) | vid;
 514			switch (act->id) {
 515			case FLOW_ACTION_VLAN_POP:
 516				fs->newvlan |= VLAN_REMOVE;
 517				break;
 518			case FLOW_ACTION_VLAN_PUSH:
 519				fs->newvlan |= VLAN_INSERT;
 520				fs->vlan = vlan_tci;
 521				break;
 522			case FLOW_ACTION_VLAN_MANGLE:
 523				fs->newvlan |= VLAN_REWRITE;
 524				fs->vlan = vlan_tci;
 525				break;
 526			default:
 527				break;
 528			}
 529			}
 530			break;
 531		case FLOW_ACTION_MANGLE: {
 532			u32 mask, val, offset;
 533			u8 htype;
 534
 535			htype = act->mangle.htype;
 536			mask = act->mangle.mask;
 537			val = act->mangle.val;
 538			offset = act->mangle.offset;
 539
 540			process_pedit_field(fs, val, mask, offset, htype,
 541					    &natmode_flags);
 542			}
 543			break;
 544		case FLOW_ACTION_QUEUE:
 545			fs->action = FILTER_PASS;
 546			fs->dirsteer = 1;
 547			fs->iq = act->queue.index;
 548			break;
 549		default:
 550			break;
 551		}
 552	}
 553	if (natmode_flags)
 554		cxgb4_action_natmode_tweak(fs, natmode_flags);
 555
 556}
 557
 558static bool valid_l4_mask(u32 mask)
 559{
 560	u16 hi, lo;
 561
 562	/* Either the upper 16-bits (SPORT) OR the lower
 563	 * 16-bits (DPORT) can be set, but NOT BOTH.
 564	 */
 565	hi = (mask >> 16) & 0xFFFF;
 566	lo = mask & 0xFFFF;
 567
 568	return hi && lo ? false : true;
 569}
 570
 571static bool valid_pedit_action(struct netlink_ext_ack *extack,
 572			       const struct flow_action_entry *act,
 573			       u8 *natmode_flags)
 574{
 575	u32 mask, offset;
 576	u8 htype;
 577
 578	htype = act->mangle.htype;
 579	mask = act->mangle.mask;
 580	offset = act->mangle.offset;
 581
 582	switch (htype) {
 583	case FLOW_ACT_MANGLE_HDR_TYPE_ETH:
 584		switch (offset) {
 585		case PEDIT_ETH_DMAC_31_0:
 586		case PEDIT_ETH_DMAC_47_32_SMAC_15_0:
 587		case PEDIT_ETH_SMAC_47_16:
 588			break;
 589		default:
 590			NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit field");
 591			return false;
 592		}
 593		break;
 594	case FLOW_ACT_MANGLE_HDR_TYPE_IP4:
 595		switch (offset) {
 596		case PEDIT_IP4_SRC:
 597			*natmode_flags |= CXGB4_ACTION_NATMODE_SIP;
 598			break;
 599		case PEDIT_IP4_DST:
 600			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 601			break;
 602		default:
 603			NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit field");
 604			return false;
 605		}
 606		break;
 607	case FLOW_ACT_MANGLE_HDR_TYPE_IP6:
 608		switch (offset) {
 609		case PEDIT_IP6_SRC_31_0:
 610		case PEDIT_IP6_SRC_63_32:
 611		case PEDIT_IP6_SRC_95_64:
 612		case PEDIT_IP6_SRC_127_96:
 613			*natmode_flags |= CXGB4_ACTION_NATMODE_SIP;
 614			break;
 615		case PEDIT_IP6_DST_31_0:
 616		case PEDIT_IP6_DST_63_32:
 617		case PEDIT_IP6_DST_95_64:
 618		case PEDIT_IP6_DST_127_96:
 619			*natmode_flags |= CXGB4_ACTION_NATMODE_DIP;
 620			break;
 621		default:
 622			NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit field");
 623			return false;
 624		}
 625		break;
 626	case FLOW_ACT_MANGLE_HDR_TYPE_TCP:
 627		switch (offset) {
 628		case PEDIT_TCP_SPORT_DPORT:
 629			if (!valid_l4_mask(~mask)) {
 630				NL_SET_ERR_MSG_MOD(extack,
 631						   "Unsupported mask for TCP L4 ports");
 632				return false;
 633			}
 634			if (~mask & PEDIT_TCP_UDP_SPORT_MASK)
 635				*natmode_flags |= CXGB4_ACTION_NATMODE_SPORT;
 636			else
 637				*natmode_flags |= CXGB4_ACTION_NATMODE_DPORT;
 638			break;
 639		default:
 640			NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit field");
 641			return false;
 642		}
 643		break;
 644	case FLOW_ACT_MANGLE_HDR_TYPE_UDP:
 645		switch (offset) {
 646		case PEDIT_UDP_SPORT_DPORT:
 647			if (!valid_l4_mask(~mask)) {
 648				NL_SET_ERR_MSG_MOD(extack,
 649						   "Unsupported mask for UDP L4 ports");
 650				return false;
 651			}
 652			if (~mask & PEDIT_TCP_UDP_SPORT_MASK)
 653				*natmode_flags |= CXGB4_ACTION_NATMODE_SPORT;
 654			else
 655				*natmode_flags |= CXGB4_ACTION_NATMODE_DPORT;
 656			break;
 657		default:
 658			NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit field");
 659			return false;
 660		}
 661		break;
 662	default:
 663		NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit type");
 664		return false;
 665	}
 666	return true;
 667}
 668
 669int cxgb4_validate_flow_actions(struct net_device *dev,
 670				struct flow_action *actions,
 671				struct netlink_ext_ack *extack,
 672				u8 matchall_filter)
 673{
 674	struct adapter *adap = netdev2adap(dev);
 675	struct flow_action_entry *act;
 676	bool act_redir = false;
 677	bool act_pedit = false;
 678	bool act_vlan = false;
 679	u8 natmode_flags = 0;
 680	int i;
 681
 682	if (!flow_action_basic_hw_stats_check(actions, extack))
 683		return -EOPNOTSUPP;
 684
 685	flow_action_for_each(i, act, actions) {
 686		switch (act->id) {
 687		case FLOW_ACTION_ACCEPT:
 688		case FLOW_ACTION_DROP:
 689			/* Do nothing */
 690			break;
 691		case FLOW_ACTION_MIRRED:
 692		case FLOW_ACTION_REDIRECT: {
 693			struct net_device *n_dev, *target_dev;
 694			bool found = false;
 695			unsigned int i;
 696
 697			if (act->id == FLOW_ACTION_MIRRED &&
 698			    !matchall_filter) {
 699				NL_SET_ERR_MSG_MOD(extack,
 700						   "Egress mirror action is only supported for tc-matchall");
 701				return -EOPNOTSUPP;
 702			}
 703
 704			target_dev = act->dev;
 705			for_each_port(adap, i) {
 706				n_dev = adap->port[i];
 707				if (target_dev == n_dev) {
 708					found = true;
 709					break;
 710				}
 711			}
 712
 713			/* If interface doesn't belong to our hw, then
 714			 * the provided output port is not valid
 715			 */
 716			if (!found) {
 717				NL_SET_ERR_MSG_MOD(extack, "Out port invalid");
 718				return -EINVAL;
 719			}
 720			act_redir = true;
 721			}
 722			break;
 723		case FLOW_ACTION_VLAN_POP:
 724		case FLOW_ACTION_VLAN_PUSH:
 725		case FLOW_ACTION_VLAN_MANGLE: {
 726			u16 proto = be16_to_cpu(act->vlan.proto);
 727
 728			switch (act->id) {
 729			case FLOW_ACTION_VLAN_POP:
 730				break;
 731			case FLOW_ACTION_VLAN_PUSH:
 732			case FLOW_ACTION_VLAN_MANGLE:
 733				if (proto != ETH_P_8021Q) {
 734					NL_SET_ERR_MSG_MOD(extack,
 735							   "Unsupported vlan proto");
 736					return -EOPNOTSUPP;
 737				}
 738				break;
 739			default:
 740				NL_SET_ERR_MSG_MOD(extack,
 741						   "Unsupported vlan action");
 742				return -EOPNOTSUPP;
 743			}
 744			act_vlan = true;
 745			}
 746			break;
 747		case FLOW_ACTION_MANGLE: {
 748			bool pedit_valid = valid_pedit_action(extack, act,
 749							      &natmode_flags);
 750
 751			if (!pedit_valid)
 752				return -EOPNOTSUPP;
 753			act_pedit = true;
 754			}
 755			break;
 756		case FLOW_ACTION_QUEUE:
 757			/* Do nothing. cxgb4_set_filter will validate */
 758			break;
 759		default:
 760			NL_SET_ERR_MSG_MOD(extack, "Unsupported action");
 761			return -EOPNOTSUPP;
 762		}
 763	}
 764
 765	if ((act_pedit || act_vlan) && !act_redir) {
 766		NL_SET_ERR_MSG_MOD(extack,
 767				   "pedit/vlan rewrite invalid without egress redirect");
 768		return -EINVAL;
 769	}
 770
 771	if (act_pedit) {
 772		int ret;
 773
 774		ret = cxgb4_action_natmode_validate(adap, natmode_flags,
 775						    extack);
 776		if (ret)
 777			return ret;
 778	}
 779
 780	return 0;
 781}
 782
 783static void cxgb4_tc_flower_hash_prio_add(struct adapter *adap, u32 tc_prio)
 784{
 785	spin_lock_bh(&adap->tids.ftid_lock);
 786	if (adap->tids.tc_hash_tids_max_prio < tc_prio)
 787		adap->tids.tc_hash_tids_max_prio = tc_prio;
 788	spin_unlock_bh(&adap->tids.ftid_lock);
 789}
 790
 791static void cxgb4_tc_flower_hash_prio_del(struct adapter *adap, u32 tc_prio)
 792{
 793	struct tid_info *t = &adap->tids;
 794	struct ch_tc_flower_entry *fe;
 795	struct rhashtable_iter iter;
 796	u32 found = 0;
 797
 798	spin_lock_bh(&t->ftid_lock);
 799	/* Bail if the current rule is not the one with the max
 800	 * prio.
 801	 */
 802	if (t->tc_hash_tids_max_prio != tc_prio)
 803		goto out_unlock;
 804
 805	/* Search for the next rule having the same or next lower
 806	 * max prio.
 807	 */
 808	rhashtable_walk_enter(&adap->flower_tbl, &iter);
 809	do {
 810		rhashtable_walk_start(&iter);
 811
 812		fe = rhashtable_walk_next(&iter);
 813		while (!IS_ERR_OR_NULL(fe)) {
 814			if (fe->fs.hash &&
 815			    fe->fs.tc_prio <= t->tc_hash_tids_max_prio) {
 816				t->tc_hash_tids_max_prio = fe->fs.tc_prio;
 817				found++;
 818
 819				/* Bail if we found another rule
 820				 * having the same prio as the
 821				 * current max one.
 822				 */
 823				if (fe->fs.tc_prio == tc_prio)
 824					break;
 825			}
 826
 827			fe = rhashtable_walk_next(&iter);
 828		}
 829
 830		rhashtable_walk_stop(&iter);
 831	} while (fe == ERR_PTR(-EAGAIN));
 832	rhashtable_walk_exit(&iter);
 833
 834	if (!found)
 835		t->tc_hash_tids_max_prio = 0;
 836
 837out_unlock:
 838	spin_unlock_bh(&t->ftid_lock);
 839}
 840
 841int cxgb4_flow_rule_replace(struct net_device *dev, struct flow_rule *rule,
 842			    u32 tc_prio, struct netlink_ext_ack *extack,
 843			    struct ch_filter_specification *fs, u32 *tid)
 844{
 845	struct adapter *adap = netdev2adap(dev);
 846	struct filter_ctx ctx;
 847	u16 addr_type = 0;
 848	u8 inet_family;
 849	int fidx, ret;
 850
 851	if (cxgb4_validate_flow_actions(dev, &rule->action, extack, 0))
 852		return -EOPNOTSUPP;
 853
 854	if (cxgb4_validate_flow_match(extack, rule))
 855		return -EOPNOTSUPP;
 856
 857	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CONTROL)) {
 858		struct flow_match_control match;
 859
 860		flow_rule_match_control(rule, &match);
 861		addr_type = match.key->addr_type;
 862
 863		if (match.mask->flags & FLOW_DIS_IS_FRAGMENT) {
 864			fs->val.frag = match.key->flags & FLOW_DIS_IS_FRAGMENT;
 865			fs->mask.frag = true;
 866		}
 867
 868		if (!flow_rule_is_supp_control_flags(FLOW_DIS_IS_FRAGMENT,
 869						     match.mask->flags, extack))
 870			return -EOPNOTSUPP;
 871
 872	} else if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
 873		addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
 874	} else if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
 875		addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
 876	}
 877
 878	cxgb4_process_flow_match(dev, rule, addr_type, fs);
 879	cxgb4_process_flow_actions(dev, &rule->action, fs);
 880
 881	fs->hash = is_filter_exact_match(adap, fs);
 882	inet_family = fs->type ? PF_INET6 : PF_INET;
 883
 884	/* Get a free filter entry TID, where we can insert this new
 885	 * rule. Only insert rule if its prio doesn't conflict with
 886	 * existing rules.
 887	 */
 888	fidx = cxgb4_get_free_ftid(dev, inet_family, fs->hash,
 889				   tc_prio);
 890	if (fidx < 0) {
 891		NL_SET_ERR_MSG_MOD(extack,
 892				   "No free LETCAM index available");
 893		return -ENOMEM;
 894	}
 895
 896	if (fidx < adap->tids.nhpftids) {
 897		fs->prio = 1;
 898		fs->hash = 0;
 899	}
 900
 901	/* If the rule can be inserted into HASH region, then ignore
 902	 * the index to normal FILTER region.
 903	 */
 904	if (fs->hash)
 905		fidx = 0;
 906
 907	fs->tc_prio = tc_prio;
 908
 909	init_completion(&ctx.completion);
 910	ret = __cxgb4_set_filter(dev, fidx, fs, &ctx);
 911	if (ret) {
 912		NL_SET_ERR_MSG_FMT_MOD(extack, "filter creation err %d", ret);
 913		return ret;
 914	}
 915
 916	/* Wait for reply */
 917	ret = wait_for_completion_timeout(&ctx.completion, 10 * HZ);
 918	if (!ret)
 919		return -ETIMEDOUT;
 920
 921	/* Check if hw returned error for filter creation */
 922	if (ctx.result)
 923		return ctx.result;
 924
 925	*tid = ctx.tid;
 926
 927	if (fs->hash)
 928		cxgb4_tc_flower_hash_prio_add(adap, tc_prio);
 929
 930	return 0;
 931}
 932
 933int cxgb4_tc_flower_replace(struct net_device *dev,
 934			    struct flow_cls_offload *cls)
 935{
 936	struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
 937	struct netlink_ext_ack *extack = cls->common.extack;
 938	struct adapter *adap = netdev2adap(dev);
 939	struct ch_tc_flower_entry *ch_flower;
 940	struct ch_filter_specification *fs;
 941	int ret;
 942
 943	ch_flower = allocate_flower_entry();
 944	if (!ch_flower) {
 945		netdev_err(dev, "%s: ch_flower alloc failed.\n", __func__);
 946		return -ENOMEM;
 947	}
 948
 949	fs = &ch_flower->fs;
 950	fs->hitcnts = 1;
 951	fs->tc_cookie = cls->cookie;
 952
 953	ret = cxgb4_flow_rule_replace(dev, rule, cls->common.prio, extack, fs,
 954				      &ch_flower->filter_id);
 955	if (ret)
 956		goto free_entry;
 957
 958	ch_flower->tc_flower_cookie = cls->cookie;
 959	ret = rhashtable_insert_fast(&adap->flower_tbl, &ch_flower->node,
 960				     adap->flower_ht_params);
 961	if (ret)
 962		goto del_filter;
 963
 964	return 0;
 965
 966del_filter:
 967	if (fs->hash)
 968		cxgb4_tc_flower_hash_prio_del(adap, cls->common.prio);
 969
 970	cxgb4_del_filter(dev, ch_flower->filter_id, &ch_flower->fs);
 971
 972free_entry:
 973	kfree(ch_flower);
 974	return ret;
 975}
 976
 977int cxgb4_flow_rule_destroy(struct net_device *dev, u32 tc_prio,
 978			    struct ch_filter_specification *fs, int tid)
 979{
 980	struct adapter *adap = netdev2adap(dev);
 981	u8 hash;
 982	int ret;
 983
 984	hash = fs->hash;
 985
 986	ret = cxgb4_del_filter(dev, tid, fs);
 987	if (ret)
 988		return ret;
 989
 990	if (hash)
 991		cxgb4_tc_flower_hash_prio_del(adap, tc_prio);
 992
 993	return ret;
 994}
 995
 996int cxgb4_tc_flower_destroy(struct net_device *dev,
 997			    struct flow_cls_offload *cls)
 998{
 999	struct adapter *adap = netdev2adap(dev);
1000	struct ch_tc_flower_entry *ch_flower;
1001	int ret;
1002
1003	ch_flower = ch_flower_lookup(adap, cls->cookie);
1004	if (!ch_flower)
1005		return -ENOENT;
1006
1007	rhashtable_remove_fast(&adap->flower_tbl, &ch_flower->node,
1008			       adap->flower_ht_params);
1009
1010	ret = cxgb4_flow_rule_destroy(dev, ch_flower->fs.tc_prio,
1011				      &ch_flower->fs, ch_flower->filter_id);
1012	if (ret)
1013		netdev_err(dev, "Flow rule destroy failed for tid: %u, ret: %d",
1014			   ch_flower->filter_id, ret);
1015
1016	kfree_rcu(ch_flower, rcu);
1017	return ret;
1018}
1019
1020static void ch_flower_stats_handler(struct work_struct *work)
1021{
1022	struct adapter *adap = container_of(work, struct adapter,
1023					    flower_stats_work);
1024	struct ch_tc_flower_entry *flower_entry;
1025	struct ch_tc_flower_stats *ofld_stats;
1026	struct rhashtable_iter iter;
1027	u64 packets;
1028	u64 bytes;
1029	int ret;
1030
1031	rhashtable_walk_enter(&adap->flower_tbl, &iter);
1032	do {
1033		rhashtable_walk_start(&iter);
1034
1035		while ((flower_entry = rhashtable_walk_next(&iter)) &&
1036		       !IS_ERR(flower_entry)) {
1037			ret = cxgb4_get_filter_counters(adap->port[0],
1038							flower_entry->filter_id,
1039							&packets, &bytes,
1040							flower_entry->fs.hash);
1041			if (!ret) {
1042				spin_lock(&flower_entry->lock);
1043				ofld_stats = &flower_entry->stats;
1044
1045				if (ofld_stats->prev_packet_count != packets) {
1046					ofld_stats->prev_packet_count = packets;
1047					ofld_stats->last_used = jiffies;
1048				}
1049				spin_unlock(&flower_entry->lock);
1050			}
1051		}
1052
1053		rhashtable_walk_stop(&iter);
1054
1055	} while (flower_entry == ERR_PTR(-EAGAIN));
1056	rhashtable_walk_exit(&iter);
1057	mod_timer(&adap->flower_stats_timer, jiffies + STATS_CHECK_PERIOD);
1058}
1059
1060static void ch_flower_stats_cb(struct timer_list *t)
1061{
1062	struct adapter *adap = timer_container_of(adap, t, flower_stats_timer);
1063
1064	schedule_work(&adap->flower_stats_work);
1065}
1066
1067int cxgb4_tc_flower_stats(struct net_device *dev,
1068			  struct flow_cls_offload *cls)
1069{
1070	struct adapter *adap = netdev2adap(dev);
1071	struct ch_tc_flower_stats *ofld_stats;
1072	struct ch_tc_flower_entry *ch_flower;
1073	u64 packets;
1074	u64 bytes;
1075	int ret;
1076
1077	ch_flower = ch_flower_lookup(adap, cls->cookie);
1078	if (!ch_flower) {
1079		ret = -ENOENT;
1080		goto err;
1081	}
1082
1083	ret = cxgb4_get_filter_counters(dev, ch_flower->filter_id,
1084					&packets, &bytes,
1085					ch_flower->fs.hash);
1086	if (ret < 0)
1087		goto err;
1088
1089	spin_lock_bh(&ch_flower->lock);
1090	ofld_stats = &ch_flower->stats;
1091	if (ofld_stats->packet_count != packets) {
1092		if (ofld_stats->prev_packet_count != packets)
1093			ofld_stats->last_used = jiffies;
1094		flow_stats_update(&cls->stats, bytes - ofld_stats->byte_count,
1095				  packets - ofld_stats->packet_count, 0,
1096				  ofld_stats->last_used,
1097				  FLOW_ACTION_HW_STATS_IMMEDIATE);
1098
1099		ofld_stats->packet_count = packets;
1100		ofld_stats->byte_count = bytes;
1101		ofld_stats->prev_packet_count = packets;
1102	}
1103	spin_unlock_bh(&ch_flower->lock);
1104	return 0;
1105
1106err:
1107	return ret;
1108}
1109
1110static const struct rhashtable_params cxgb4_tc_flower_ht_params = {
1111	.nelem_hint = 384,
1112	.head_offset = offsetof(struct ch_tc_flower_entry, node),
1113	.key_offset = offsetof(struct ch_tc_flower_entry, tc_flower_cookie),
1114	.key_len = sizeof(((struct ch_tc_flower_entry *)0)->tc_flower_cookie),
1115	.max_size = 524288,
1116	.min_size = 512,
1117	.automatic_shrinking = true
1118};
1119
1120int cxgb4_init_tc_flower(struct adapter *adap)
1121{
1122	int ret;
1123
1124	if (adap->tc_flower_initialized)
1125		return -EEXIST;
1126
1127	adap->flower_ht_params = cxgb4_tc_flower_ht_params;
1128	ret = rhashtable_init(&adap->flower_tbl, &adap->flower_ht_params);
1129	if (ret)
1130		return ret;
1131
1132	INIT_WORK(&adap->flower_stats_work, ch_flower_stats_handler);
1133	timer_setup(&adap->flower_stats_timer, ch_flower_stats_cb, 0);
1134	mod_timer(&adap->flower_stats_timer, jiffies + STATS_CHECK_PERIOD);
1135	adap->tc_flower_initialized = true;
1136	return 0;
1137}
1138
1139void cxgb4_cleanup_tc_flower(struct adapter *adap)
1140{
1141	if (!adap->tc_flower_initialized)
1142		return;
1143
1144	if (adap->flower_stats_timer.function)
1145		timer_shutdown_sync(&adap->flower_stats_timer);
1146	cancel_work_sync(&adap->flower_stats_work);
1147	rhashtable_destroy(&adap->flower_tbl);
1148	adap->tc_flower_initialized = false;
1149}