From: Felix Fietkau Date: Sun, 26 Jul 2020 14:03:21 +0200 Subject: [PATCH] net: add support for threaded NAPI polling For some drivers (especially 802.11 drivers), doing a lot of work in the NAPI poll function does not perform well. Since NAPI poll is bound to the CPU it was scheduled from, we can easily end up with a few very busy CPUs spending most of their time in softirq/ksoftirqd and some idle ones. Introduce threaded NAPI for such drivers based on a workqueue. The API is the same except for using netif_threaded_napi_add instead of netif_napi_add. In my tests with mt76 on MT7621 using threaded NAPI + a thread for tx scheduling improves LAN->WLAN bridging throughput by 10-50%. Throughput without threaded NAPI is wildly inconsistent, depending on the CPU that runs the tx scheduling thread. With threaded NAPI, throughput seems stable and consistent (and higher than the best results I got without it). Based on a patch by Hillf Danton Cc: Hillf Danton Signed-off-by: Felix Fietkau --- --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -347,6 +347,7 @@ struct napi_struct { struct list_head dev_list; struct hlist_node napi_hash_node; unsigned int napi_id; + struct work_struct work; }; enum { @@ -357,6 +358,7 @@ enum { NAPI_STATE_LISTED, /* NAPI added to system lists */ NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ + NAPI_STATE_THREADED, /* Use threaded NAPI */ }; enum { @@ -367,6 +369,7 @@ enum { NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), + NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), }; enum gro_result { @@ -2411,6 +2414,26 @@ void netif_napi_add(struct net_device *d int (*poll)(struct napi_struct *, int), int weight); /** + * netif_threaded_napi_add - initialize a NAPI context + * @dev: network device + * @napi: NAPI context + * @poll: polling function + * @weight: default weight + * + * This variant of netif_napi_add() should be used from drivers using NAPI + * with CPU intensive poll functions. + * This will schedule polling from a high priority workqueue + */ +static inline void netif_threaded_napi_add(struct net_device *dev, + struct napi_struct *napi, + int (*poll)(struct napi_struct *, int), + int weight) +{ + set_bit(NAPI_STATE_THREADED, &napi->state); + netif_napi_add(dev, napi, poll, weight); +} + +/** * netif_tx_napi_add - initialize a NAPI context * @dev: network device * @napi: NAPI context --- a/net/core/dev.c +++ b/net/core/dev.c @@ -159,6 +159,7 @@ static DEFINE_SPINLOCK(offload_lock); struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; struct list_head ptype_all __read_mostly; /* Taps */ static struct list_head offload_base __read_mostly; +static struct workqueue_struct *napi_workq __read_mostly; static int netif_rx_internal(struct sk_buff *skb); static int call_netdevice_notifiers_info(unsigned long val, @@ -6407,6 +6408,11 @@ void __napi_schedule(struct napi_struct { unsigned long flags; + if (test_bit(NAPI_STATE_THREADED, &n->state)) { + queue_work(napi_workq, &n->work); + return; + } + local_irq_save(flags); ____napi_schedule(this_cpu_ptr(&softnet_data), n); local_irq_restore(flags); @@ -6454,6 +6460,11 @@ EXPORT_SYMBOL(napi_schedule_prep); */ void __napi_schedule_irqoff(struct napi_struct *n) { + if (test_bit(NAPI_STATE_THREADED, &n->state)) { + queue_work(napi_workq, &n->work); + return; + } + ____napi_schedule(this_cpu_ptr(&softnet_data), n); } EXPORT_SYMBOL(__napi_schedule_irqoff); @@ -6715,6 +6726,86 @@ static void init_gro_hash(struct napi_st napi->gro_bitmask = 0; } +static int __napi_poll(struct napi_struct *n, bool *repoll) +{ + int work, weight; + + weight = n->weight; + + /* This NAPI_STATE_SCHED test is for avoiding a race + * with netpoll's poll_napi(). Only the entity which + * obtains the lock and sees NAPI_STATE_SCHED set will + * actually make the ->poll() call. Therefore we avoid + * accidentally calling ->poll() when NAPI is not scheduled. + */ + work = 0; + if (test_bit(NAPI_STATE_SCHED, &n->state)) { + work = n->poll(n, weight); + trace_napi_poll(n, work, weight); + } + + if (unlikely(work > weight)) + pr_err_once("NAPI poll function %pS returned %d, exceeding its budget of %d.\n", + n->poll, work, weight); + + if (likely(work < weight)) + return work; + + /* Drivers must not modify the NAPI state if they + * consume the entire weight. In such cases this code + * still "owns" the NAPI instance and therefore can + * move the instance around on the list at-will. + */ + if (unlikely(napi_disable_pending(n))) { + napi_complete(n); + return work; + } + + if (n->gro_bitmask) { + /* flush too old packets + * If HZ < 1000, flush all packets. + */ + napi_gro_flush(n, HZ >= 1000); + } + + gro_normal_list(n); + + *repoll = true; + + return work; +} + +static void napi_workfn(struct work_struct *work) +{ + struct napi_struct *n = container_of(work, struct napi_struct, work); + void *have; + + for (;;) { + bool repoll = false; + + local_bh_disable(); + + have = netpoll_poll_lock(n); + __napi_poll(n, &repoll); + netpoll_poll_unlock(have); + + local_bh_enable(); + + if (!repoll) + return; + + if (!need_resched()) + continue; + + /* + * have to pay for the latency of task switch even if + * napi is scheduled + */ + queue_work(napi_workq, work); + return; + } +} + void netif_napi_add(struct net_device *dev, struct napi_struct *napi, int (*poll)(struct napi_struct *, int), int weight) { @@ -6738,6 +6829,7 @@ void netif_napi_add(struct net_device *d #ifdef CONFIG_NETPOLL napi->poll_owner = -1; #endif + INIT_WORK(&napi->work, napi_workfn); set_bit(NAPI_STATE_SCHED, &napi->state); set_bit(NAPI_STATE_NPSVC, &napi->state); list_add_rcu(&napi->dev_list, &dev->napi_list); @@ -6780,6 +6872,7 @@ void __netif_napi_del(struct napi_struct if (!test_and_clear_bit(NAPI_STATE_LISTED, &napi->state)) return; + cancel_work_sync(&napi->work); napi_hash_del(napi); list_del_rcu(&napi->dev_list); napi_free_frags(napi); @@ -6791,53 +6884,19 @@ EXPORT_SYMBOL(__netif_napi_del); static int napi_poll(struct napi_struct *n, struct list_head *repoll) { + bool do_repoll = false; void *have; - int work, weight; + int work; list_del_init(&n->poll_list); have = netpoll_poll_lock(n); - weight = n->weight; + work = __napi_poll(n, &do_repoll); - /* This NAPI_STATE_SCHED test is for avoiding a race - * with netpoll's poll_napi(). Only the entity which - * obtains the lock and sees NAPI_STATE_SCHED set will - * actually make the ->poll() call. Therefore we avoid - * accidentally calling ->poll() when NAPI is not scheduled. - */ - work = 0; - if (test_bit(NAPI_STATE_SCHED, &n->state)) { - work = n->poll(n, weight); - trace_napi_poll(n, work, weight); - } - - if (unlikely(work > weight)) - pr_err_once("NAPI poll function %pS returned %d, exceeding its budget of %d.\n", - n->poll, work, weight); - - if (likely(work < weight)) + if (!do_repoll) goto out_unlock; - /* Drivers must not modify the NAPI state if they - * consume the entire weight. In such cases this code - * still "owns" the NAPI instance and therefore can - * move the instance around on the list at-will. - */ - if (unlikely(napi_disable_pending(n))) { - napi_complete(n); - goto out_unlock; - } - - if (n->gro_bitmask) { - /* flush too old packets - * If HZ < 1000, flush all packets. - */ - napi_gro_flush(n, HZ >= 1000); - } - - gro_normal_list(n); - /* Some drivers may have called napi_schedule * prior to exhausting their budget. */ @@ -11291,6 +11350,10 @@ static int __init net_dev_init(void) sd->backlog.weight = weight_p; } + napi_workq = alloc_workqueue("napi_workq", WQ_UNBOUND | WQ_HIGHPRI, + WQ_UNBOUND_MAX_ACTIVE | WQ_SYSFS); + BUG_ON(!napi_workq); + dev_boot_phase = 0; /* The loopback device is special if any other network devices '#n118'>118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 
--- a/include/uapi/linux/pkt_sched.h
+++ b/include/uapi/linux/pkt_sched.h
@@ -226,6 +226,33 @@ struct tc_sfq_xstats {
 	__s32		allot;
 };
 
+/* ESFQ section */
+
+enum
+{
+        /* traditional */
+	TCA_SFQ_HASH_CLASSIC,
+	TCA_SFQ_HASH_DST,
+	TCA_SFQ_HASH_SRC,
+	TCA_SFQ_HASH_FWMARK,
+	/* conntrack */
+	TCA_SFQ_HASH_CTORIGDST,
+	TCA_SFQ_HASH_CTORIGSRC,
+	TCA_SFQ_HASH_CTREPLDST,
+	TCA_SFQ_HASH_CTREPLSRC,
+	TCA_SFQ_HASH_CTNATCHG,
+};
+
+struct tc_esfq_qopt
+{
+	unsigned	quantum;	/* Bytes per round allocated to flow */
+	int		perturb_period;	/* Period of hash perturbation */
+	__u32		limit;		/* Maximal packets in queue */
+	unsigned	divisor;	/* Hash divisor  */
+	unsigned	flows;		/* Maximal number of flows  */
+	unsigned	hash_kind;	/* Hash function to use for flow identification */
+};
+
 /* RED section */
 
 enum {
--- a/net/sched/Kconfig
+++ b/net/sched/Kconfig
@@ -148,6 +148,37 @@ config NET_SCH_SFQ
 	  To compile this code as a module, choose M here: the
 	  module will be called sch_sfq.
 
+config NET_SCH_ESFQ
+	tristate "Enhanced Stochastic Fairness Queueing (ESFQ)"
+	---help---
+	  Say Y here if you want to use the Enhanced Stochastic Fairness
+	  Queueing (ESFQ) packet scheduling algorithm for some of your network
+	  devices or as a leaf discipline for a classful qdisc such as HTB or
+	  CBQ (see the top of <file:net/sched/sch_esfq.c> for details and
+	  references to the SFQ algorithm).
+
+	  This is an enchanced SFQ version which allows you to control some
+	  hardcoded values in the SFQ scheduler.
+
+	  ESFQ also adds control of the hash function used to identify packet
+	  flows. The original SFQ discipline hashes by connection; ESFQ add
+	  several other hashing methods, such as by src IP or by dst IP, which
+	  can be more fair to users in some networking situations.
+
+	  To compile this code as a module, choose M here: the
+	  module will be called sch_esfq.
+
+config NET_SCH_ESFQ_NFCT
+	bool "Connection Tracking Hash Types"
+	depends on NET_SCH_ESFQ && NF_CONNTRACK
+	---help---
+	  Say Y here to enable support for hashing based on netfilter connection
+	  tracking information. This is useful for a router that is also using
+	  NAT to connect privately-addressed hosts to the Internet. If you want
+	  to provide fair distribution of upstream bandwidth, ESFQ must use
+	  connection tracking information, since all outgoing packets will share
+	  the same source address.
+
 config NET_SCH_TEQL
 	tristate "True Link Equalizer (TEQL)"
 	---help---
--- a/net/sched/Makefile
+++ b/net/sched/Makefile
@@ -26,6 +26,7 @@ obj-$(CONFIG_NET_SCH_INGRESS)	+= sch_ing
 obj-$(CONFIG_NET_SCH_DSMARK)	+= sch_dsmark.o
 obj-$(CONFIG_NET_SCH_SFB)	+= sch_sfb.o
 obj-$(CONFIG_NET_SCH_SFQ)	+= sch_sfq.o
+obj-$(CONFIG_NET_SCH_ESFQ)	+= sch_esfq.o
 obj-$(CONFIG_NET_SCH_TBF)	+= sch_tbf.o
 obj-$(CONFIG_NET_SCH_TEQL)	+= sch_teql.o
 obj-$(CONFIG_NET_SCH_PRIO)	+= sch_prio.o
--- /dev/null
+++ b/net/sched/sch_esfq.c
@@ -0,0 +1,702 @@
+/*
+ * net/sched/sch_esfq.c	Extended Stochastic Fairness Queueing discipline.
+ *
+ *		This program is free software; you can redistribute it and/or
+ *		modify it under the terms of the GNU General Public License
+ *		as published by the Free Software Foundation; either version
+ *		2 of the License, or (at your option) any later version.
+ *
+ * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
+ *
+ * Changes:	Alexander Atanasov, <alex@ssi.bg>
+ *		Added dynamic depth,limit,divisor,hash_kind options.
+ *		Added dst and src hashes.
+ *
+ * 		Alexander Clouter, <alex@digriz.org.uk>
+ *		Ported ESFQ to Linux 2.6.
+ *
+ * 		Corey Hickey, <bugfood-c@fatooh.org>
+ *		Maintenance of the Linux 2.6 port.
+ *		Added fwmark hash (thanks to Robert Kurjata).
+ *		Added usage of jhash.
+ *		Added conntrack support.
+ *		Added ctnatchg hash (thanks to Ben Pfountz).
+ */
+
+#include <linux/module.h>
+#include <asm/uaccess.h>
+#include <linux/bitops.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/jiffies.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/socket.h>
+#include <linux/sockios.h>
+#include <linux/in.h>
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/if_ether.h>
+#include <linux/inet.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/notifier.h>
+#include <linux/init.h>
+#include <net/ip.h>
+#include <net/netlink.h>
+#include <linux/ipv6.h>
+#include <net/route.h>
+#include <linux/skbuff.h>
+#include <net/sock.h>
+#include <net/pkt_sched.h>
+#include <linux/jhash.h>
+#ifdef CONFIG_NET_SCH_ESFQ_NFCT
+#include <net/netfilter/nf_conntrack.h>
+#endif
+
+/*	Stochastic Fairness Queuing algorithm.
+	For more comments look at sch_sfq.c.
+	The difference is that you can change limit, depth,
+	hash table size and choose alternate hash types.
+
+	classic:	same as in sch_sfq.c
+	dst:		destination IP address
+	src:		source IP address
+	fwmark:		netfilter mark value
+	ctorigdst:	original destination IP address
+	ctorigsrc:	original source IP address
+	ctrepldst:	reply destination IP address
+	ctreplsrc:	reply source IP
+
+*/
+
+#define ESFQ_HEAD 0
+#define ESFQ_TAIL 1
+
+/* This type should contain at least SFQ_DEPTH*2 values */
+typedef unsigned int esfq_index;
+
+struct esfq_head
+{
+	esfq_index	next;
+	esfq_index	prev;
+};
+
+struct esfq_sched_data
+{
+/* Parameters */
+	int		perturb_period;
+	unsigned	quantum;	/* Allotment per round: MUST BE >= MTU */
+	int		limit;
+	unsigned	depth;
+	unsigned	hash_divisor;
+	unsigned	hash_kind;
+/* Variables */
+	struct timer_list perturb_timer;
+	int		perturbation;
+	esfq_index	tail;		/* Index of current slot in round */
+	esfq_index	max_depth;	/* Maximal depth */
+
+	esfq_index	*ht;			/* Hash table */
+	esfq_index	*next;			/* Active slots link */
+	short		*allot;			/* Current allotment per slot */
+	unsigned short	*hash;			/* Hash value indexed by slots */
+	struct sk_buff_head	*qs;		/* Slot queue */
+	struct esfq_head	*dep;		/* Linked list of slots, indexed by depth */
+};
+
+/* This contains the info we will hash. */
+struct esfq_packet_info
+{
+	u32	proto;		/* protocol or port */
+	u32	src;		/* source from packet header */
+	u32	dst;		/* destination from packet header */
+	u32	ctorigsrc;	/* original source from conntrack */
+	u32	ctorigdst;	/* original destination from conntrack */
+	u32	ctreplsrc;	/* reply source from conntrack */
+	u32	ctrepldst;	/* reply destination from conntrack */
+	u32	mark;		/* netfilter mark (fwmark) */
+};
+
+static __inline__ unsigned esfq_jhash_1word(struct esfq_sched_data *q,u32 a)
+{
+	return jhash_1word(a, q->perturbation) & (q->hash_divisor-1);
+}
+
+static __inline__ unsigned esfq_jhash_2words(struct esfq_sched_data *q, u32 a, u32 b)
+{
+	return jhash_2words(a, b, q->perturbation) & (q->hash_divisor-1);
+}
+
+static __inline__ unsigned esfq_jhash_3words(struct esfq_sched_data *q, u32 a, u32 b, u32 c)
+{
+	return jhash_3words(a, b, c, q->perturbation) & (q->hash_divisor-1);
+}
+
+static unsigned esfq_hash(struct esfq_sched_data *q, struct sk_buff *skb)
+{
+	struct esfq_packet_info info;
+#ifdef CONFIG_NET_SCH_ESFQ_NFCT
+	enum ip_conntrack_info ctinfo;
+	struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
+#endif
+
+	switch (skb->protocol) {
+	case __constant_htons(ETH_P_IP):
+	{
+		struct iphdr *iph = ip_hdr(skb);
+		info.dst = iph->daddr;
+		info.src = iph->saddr;
+		if (!(iph->frag_off&htons(IP_MF|IP_OFFSET)) &&
+		    (iph->protocol == IPPROTO_TCP ||
+		     iph->protocol == IPPROTO_UDP ||
+		     iph->protocol == IPPROTO_SCTP ||
+		     iph->protocol == IPPROTO_DCCP ||
+		     iph->protocol == IPPROTO_ESP))
+			info.proto = *(((u32*)iph) + iph->ihl);
+		else
+			info.proto = iph->protocol;
+		break;
+	}
+	case __constant_htons(ETH_P_IPV6):
+	{
+		struct ipv6hdr *iph = ipv6_hdr(skb);
+		/* Hash ipv6 addresses into a u32. This isn't ideal,
+		 * but the code is simple. */
+		info.dst = jhash2(iph->daddr.s6_addr32, 4, q->perturbation);
+		info.src = jhash2(iph->saddr.s6_addr32, 4, q->perturbation);
+		if (iph->nexthdr == IPPROTO_TCP ||
+		    iph->nexthdr == IPPROTO_UDP ||
+		    iph->nexthdr == IPPROTO_SCTP ||
+		    iph->nexthdr == IPPROTO_DCCP ||
+		    iph->nexthdr == IPPROTO_ESP)
+			info.proto = *(u32*)&iph[1];
+		else
+			info.proto = iph->nexthdr;
+		break;
+	}
+	default:
+		info.dst   = (u32)(unsigned long)skb_dst(skb);
+		info.src   = (u32)(unsigned long)skb->sk;
+		info.proto = skb->protocol;
+	}
+
+	info.mark = skb->mark;
+
+#ifdef CONFIG_NET_SCH_ESFQ_NFCT
+	/* defaults if there is no conntrack info */
+	info.ctorigsrc = info.src;
+	info.ctorigdst = info.dst;
+	info.ctreplsrc = info.dst;
+	info.ctrepldst = info.src;
+	/* collect conntrack info */
+	if (ct && ct != &nf_conntrack_untracked) {
+		if (skb->protocol == __constant_htons(ETH_P_IP)) {
+			info.ctorigsrc = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip;
+			info.ctorigdst = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.u3.ip;
+			info.ctreplsrc = ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.u3.ip;
+			info.ctrepldst = ct->tuplehash[IP_CT_DIR_REPLY].tuple.dst.u3.ip;
+		}
+		else if (skb->protocol == __constant_htons(ETH_P_IPV6)) {
+			/* Again, hash ipv6 addresses into a single u32. */
+			info.ctorigsrc = jhash2(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip6, 4, q->perturbation);
+			info.ctorigdst = jhash2(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.u3.ip6, 4, q->perturbation);
+			info.ctreplsrc = jhash2(ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.u3.ip6, 4, q->perturbation);
+			info.ctrepldst = jhash2(ct->tuplehash[IP_CT_DIR_REPLY].tuple.dst.u3.ip6, 4, q->perturbation);
+		}
+
+	}
+#endif
+
+	switch(q->hash_kind) {
+	case TCA_SFQ_HASH_CLASSIC:
+		return esfq_jhash_3words(q, info.dst, info.src, info.proto);
+	case TCA_SFQ_HASH_DST:
+		return esfq_jhash_1word(q, info.dst);
+	case TCA_SFQ_HASH_SRC:
+		return esfq_jhash_1word(q, info.src);
+	case TCA_SFQ_HASH_FWMARK:
+		return esfq_jhash_1word(q, info.mark);
+#ifdef CONFIG_NET_SCH_ESFQ_NFCT
+	case TCA_SFQ_HASH_CTORIGDST:
+		return esfq_jhash_1word(q, info.ctorigdst);
+	case TCA_SFQ_HASH_CTORIGSRC:
+		return esfq_jhash_1word(q, info.ctorigsrc);
+	case TCA_SFQ_HASH_CTREPLDST:
+		return esfq_jhash_1word(q, info.ctrepldst);
+	case TCA_SFQ_HASH_CTREPLSRC:
+		return esfq_jhash_1word(q, info.ctreplsrc);
+	case TCA_SFQ_HASH_CTNATCHG:
+	{
+		if (info.ctorigdst == info.ctreplsrc)
+			return esfq_jhash_1word(q, info.ctorigsrc);
+		return esfq_jhash_1word(q, info.ctreplsrc);
+	}
+#endif
+	default:
+		if (net_ratelimit())
+			printk(KERN_WARNING "ESFQ: Unknown hash method. Falling back to classic.\n");
+	}
+	return esfq_jhash_3words(q, info.dst, info.src, info.proto);
+}
+
+static inline void esfq_link(struct esfq_sched_data *q, esfq_index x)
+{
+	esfq_index p, n;
+	int d = q->qs[x].qlen + q->depth;
+
+	p = d;
+	n = q->dep[d].next;
+	q->dep[x].next = n;
+	q->dep[x].prev = p;
+	q->dep[p].next = q->dep[n].prev = x;
+}
+
+static inline void esfq_dec(struct esfq_sched_data *q, esfq_index x)
+{
+	esfq_index p, n;
+
+	n = q->dep[x].next;
+	p = q->dep[x].prev;
+	q->dep[p].next = n;
+	q->dep[n].prev = p;
+
+	if (n == p && q->max_depth == q->qs[x].qlen + 1)
+		q->max_depth--;
+
+	esfq_link(q, x);
+}
+
+static inline void esfq_inc(struct esfq_sched_data *q, esfq_index x)
+{
+	esfq_index p, n;
+	int d;
+
+	n = q->dep[x].next;
+	p = q->dep[x].prev;
+	q->dep[p].next = n;
+	q->dep[n].prev = p;
+	d = q->qs[x].qlen;
+	if (q->max_depth < d)
+		q->max_depth = d;
+
+	esfq_link(q, x);
+}
+
+static unsigned int esfq_drop(struct Qdisc *sch)
+{
+	struct esfq_sched_data *q = qdisc_priv(sch);
+	esfq_index d = q->max_depth;
+	struct sk_buff *skb;
+	unsigned int len;
+
+	/* Queue is full! Find the longest slot and
+	   drop a packet from it */
+
+	if (d > 1) {
+		esfq_index x = q->dep[d+q->depth].next;
+		skb = q->qs[x].prev;
+		len = skb->len;
+		__skb_unlink(skb, &q->qs[x]);
+		kfree_skb(skb);
+		esfq_dec(q, x);
+		sch->q.qlen--;
+		sch->qstats.drops++;
+		sch->qstats.backlog -= len;
+		return len;
+	}
+
+	if (d == 1) {
+		/* It is difficult to believe, but ALL THE SLOTS HAVE LENGTH 1. */
+		d = q->next[q->tail];
+		q->next[q->tail] = q->next[d];
+		q->allot[q->next[d]] += q->quantum;
+		skb = q->qs[d].prev;
+		len = skb->len;
+		__skb_unlink(skb, &q->qs[d]);
+		kfree_skb(skb);
+		esfq_dec(q, d);
+		sch->q.qlen--;
+		q->ht[q->hash[d]] = q->depth;
+		sch->qstats.drops++;
+		sch->qstats.backlog -= len;
+		return len;
+	}
+
+	return 0;
+}
+
+static void esfq_q_enqueue(struct sk_buff *skb, struct esfq_sched_data *q, unsigned int end)
+{
+	unsigned hash = esfq_hash(q, skb);
+	unsigned depth = q->depth;
+	esfq_index x;
+
+	x = q->ht[hash];
+	if (x == depth) {
+		q->ht[hash] = x = q->dep[depth].next;
+		q->hash[x] = hash;
+	}
+
+	if (end == ESFQ_TAIL)
+		__skb_queue_tail(&q->qs[x], skb);
+	else
+		__skb_queue_head(&q->qs[x], skb);
+
+	esfq_inc(q, x);
+	if (q->qs[x].qlen == 1) {		/* The flow is new */
+		if (q->tail == depth) {	/* It is the first flow */
+			q->tail = x;
+			q->next[x] = x;
+			q->allot[x] = q->quantum;
+		} else {
+			q->next[x] = q->next[q->tail];
+			q->next[q->tail] = x;
+			q->tail = x;
+		}
+	}
+}
+
+static int esfq_enqueue(struct sk_buff *skb, struct Qdisc* sch)
+{
+	struct esfq_sched_data *q = qdisc_priv(sch);
+	esfq_q_enqueue(skb, q, ESFQ_TAIL);
+	sch->qstats.backlog += skb->len;
+	if (++sch->q.qlen < q->limit-1) {
+		sch->bstats.bytes += skb->len;
+		sch->bstats.packets++;
+		return 0;
+	}
+
+	sch->qstats.drops++;
+	esfq_drop(sch);
+	return NET_XMIT_CN;
+}
+
+static struct sk_buff *esfq_peek(struct Qdisc* sch)
+{
+	struct esfq_sched_data *q = qdisc_priv(sch);
+	esfq_index a;
+
+	/* No active slots */
+	if (q->tail == q->depth)
+		return NULL;
+
+	a = q->next[q->tail];
+	return skb_peek(&q->qs[a]);
+}
+
+static struct sk_buff *esfq_q_dequeue(struct esfq_sched_data *q)
+{
+	struct sk_buff *skb;
+	unsigned depth = q->depth;
+	esfq_index a, old_a;
+
+	/* No active slots */
+	if (q->tail == depth)
+		return NULL;
+
+	a = old_a = q->next[q->tail];
+
+	/* Grab packet */
+	skb = __skb_dequeue(&q->qs[a]);
+	esfq_dec(q, a);
+
+	/* Is the slot empty? */
+	if (q->qs[a].qlen == 0) {
+		q->ht[q->hash[a]] = depth;
+		a = q->next[a];
+		if (a == old_a) {
+			q->tail = depth;
+			return skb;
+		}
+		q->next[q->tail] = a;
+		q->allot[a] += q->quantum;
+	} else if ((q->allot[a] -= skb->len) <= 0) {
+		q->tail = a;
+		a = q->next[a];
+		q->allot[a] += q->quantum;
+	}
+
+	return skb;
+}
+
+static struct sk_buff *esfq_dequeue(struct Qdisc* sch)
+{
+	struct esfq_sched_data *q = qdisc_priv(sch);
+	struct sk_buff *skb;
+
+	skb = esfq_q_dequeue(q);
+	if (skb == NULL)
+		return NULL;
+	sch->q.qlen--;
+	sch->qstats.backlog -= skb->len;
+	return skb;
+}
+
+static void esfq_q_destroy(struct esfq_sched_data *q)