[PATCH net-next v8 0/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[PATCH net-next v8 0/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Toke Høiland-Jørgensen]

This patch series adds the CAKE qdisc, and has been split up to ease
review.

I have attempted to split out each configurable feature into its own patch.
The first commit adds the base shaper and packet scheduler, while
subsequent commits add the optional features. The full userspace API and
most data structures are included in this commit, but options not
understood in the base version will be ignored.

The result of applying the entire series is identical to the out of tree
version that have seen extensive testing in previous deployments, most
notably as an out of tree patch to OpenWrt. However, note that I have only
compile tested the individual patches; so the whole series should be
considered as a unit.

---
Changelog

v8:
  - Remove inline keyword from function definitions
  - Simplify ACK filter; remove the complex state handling to make the
    logic easier to follow. This will potentially be a bit less efficient,
    but I have not been able to measure a difference.

v7:
  - Split up patch into a series to ease review.
  - Constify the ACK filter.

v6:
  - Fix 6in4 encapsulation checks in ACK filter code
  - Checkpatch fixes

v5:
  - Refactor ACK filter code and hopefully fix the safety issues
    properly this time.

v4:
  - Only split GSO packets if shaping at speeds <= 1Gbps
  - Fix overhead calculation code to also work for GSO packets
  - Don't re-implement kvzalloc()
  - Remove local header include from out-of-tree build (fixes kbuild-bot
    complaint).
  - Several fixes to the ACK filter:
    - Check pskb_may_pull() before deref of transport headers.
    - Don't run ACK filter logic on split GSO packets
    - Fix TCP sequence number compare to deal with wraparounds

v3:
  - Use IS_REACHABLE() macro to fix compilation when sch_cake is
    built-in and conntrack is a module.
  - Switch the stats output to use nested netlink attributes instead
    of a versioned struct.
  - Remove GPL boilerplate.
  - Fix array initialisation style.

v2:
  - Fix kbuild test bot complaint
  - Clean up the netlink ABI
  - Fix checkpatch complaints
  - A few tweaks to the behaviour of cake based on testing carried out
    while writing the paper.

---

Toke Høiland-Jørgensen (7):
      sched: Add Common Applications Kept Enhanced (cake) qdisc
      sch_cake: Add ingress mode
      sch_cake: Add optional ACK filter
      sch_cake: Add NAT awareness to packet classifier
      sch_cake: Add DiffServ handling
      sch_cake: Add overhead compensation support to the rate shaper
      sch_cake: Conditionally split GSO segments


 include/uapi/linux/pkt_sched.h |  105 ++
 net/sched/Kconfig              |   11 
 net/sched/Makefile             |    1 
 net/sched/sch_cake.c           | 2595 ++++++++++++++++++++++++++++++++++++++++
 4 files changed, 2712 insertions(+)
 create mode 100644 net/sched/sch_cake.c

[PATCH net-next v8 3/7] sch_cake: Add optional ACK filter

[Toke Høiland-Jørgensen]

The ACK filter is an optional feature of CAKE which is designed to improve
performance on links with very asymmetrical rate limits. On such links
(which are unfortunately quite prevalent, especially for DSL and cable
subscribers), the downstream throughput can be limited by the number of
ACKs capable of being transmitted in the *upstream* direction.

Filtering ACKs can, in general, have adverse effects on TCP performance
because it interferes with ACK clocking (especially in slow start), and it
reduces the flow's resiliency to ACKs being dropped further along the path.
To alleviate these drawbacks, the ACK filter in CAKE tries its best to
always keep enough ACKs queued to ensure forward progress in the TCP flow
being filtered. It does this by only filtering redundant ACKs. In its
default 'conservative' mode, the filter will always keep at least two
redundant ACKs in the queue, while in 'aggressive' mode, it will filter
down to a single ACK.

The ACK filter works by inspecting the per-flow queue on every packet
enqueue. Starting at the head of the queue, the filter looks for another
eligible packet to drop (so the ACK being dropped is always closer to the
head of the queue than the packet being enqueued). An ACK is eligible only
if it ACKs *fewer* cumulative bytes than the new packet being enqueued.
This prevents duplicate ACKs from being filtered (unless there is also SACK
options present), to avoid interfering with retransmission logic. In
aggressive mode, an eligible packet is always dropped, while in
conservative mode, at least two ACKs are kept in the queue. Only pure ACKs
(with no data segments) are considered eligible for dropping, but when an
ACK with data segments is enqueued, this can cause another pure ACK to
become eligible for dropping.

The approach described above ensures that this ACK filter avoids most of
the drawbacks of a naive filtering mechanism that only keeps flow state but
does not inspect the queue. This is the rationale for including the ACK
filter in CAKE itself rather than as separate module (as the TC filter, for
instance).

Our performance evaluation has shown that on a 30/1 Mbps link with a
bidirectional traffic test (RRUL), turning on the ACK filter on the
upstream link improves downstream throughput by ~20% (both modes) and
upstream throughput by ~12% in conservative mode and ~40% in aggressive
mode, at the cost of ~5ms of inter-flow latency due to the increased
congestion.

In *really* pathological cases, the effect can be a lot more; for instance,
the ACK filter increases the achievable downstream throughput on a link
with 100 Kbps in the upstream direction by an order of magnitude (from ~2.5
Mbps to ~25 Mbps).

Finally, even though we consider the ACK filter to be safer than most, we
do not recommend turning it on everywhere: on more symmetrical link
bandwidths the effect is negligible at best.

Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 net/sched/sch_cake.c |  261 +++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 255 insertions(+), 6 deletions(-)

diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 7ca86e3ed14c..9a70e99afe7e 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -127,7 +127,6 @@ struct cake_flow {
 	/* this stuff is all needed per-flow at dequeue time */
 	struct sk_buff	  *head;
 	struct sk_buff	  *tail;
-	struct sk_buff	  *ackcheck;
 	struct list_head  flowchain;
 	s32		  deficit;
 	struct cobalt_vars cvars;
@@ -712,9 +711,6 @@ static struct sk_buff *dequeue_head(struct cake_flow *flow)
 	if (skb) {
 		flow->head = skb->next;
 		skb->next = NULL;
-
-		if (skb == flow->ackcheck)
-			flow->ackcheck = NULL;
 	}
 
 	return skb;
@@ -732,6 +728,239 @@ static void flow_queue_add(struct cake_flow *flow, struct sk_buff *skb)
 	skb->next = NULL;
 }
 
+static struct iphdr *cake_get_iphdr(const struct sk_buff *skb,
+				    struct ipv6hdr *buf)
+{
+	unsigned int offset = skb_network_offset(skb);
+	struct iphdr *iph;
+
+	iph = skb_header_pointer(skb, offset, sizeof(struct iphdr), buf);
+
+	if (!iph)
+		return NULL;
+
+	if (iph->version == 4 && iph->protocol == IPPROTO_IPV6)
+		return skb_header_pointer(skb, offset + iph->ihl * 4,
+					  sizeof(struct ipv6hdr), buf);
+
+	else if (iph->version == 4)
+		return iph;
+
+	else if (iph->version == 6)
+		return skb_header_pointer(skb, offset, sizeof(struct ipv6hdr),
+					  buf);
+
+	return NULL;
+}
+
+static struct tcphdr *cake_get_tcphdr(const struct sk_buff *skb,
+				      void *buf, unsigned int bufsize)
+{
+	unsigned int offset = skb_network_offset(skb);
+	const struct ipv6hdr *ipv6h;
+	const struct tcphdr *tcph;
+	const struct iphdr *iph;
+	struct ipv6hdr _ipv6h;
+	struct tcphdr _tcph;
+
+	ipv6h = skb_header_pointer(skb, offset, sizeof(_ipv6h), &_ipv6h);
+
+	if (!ipv6h)
+		return NULL;
+
+	if (ipv6h->version == 4) {
+		iph = (struct iphdr *)ipv6h;
+		offset += iph->ihl * 4;
+
+		/* special-case 6in4 tunnelling, as that is a common way to get
+		 * v6 connectivity in the home
+		 */
+		if (iph->protocol == IPPROTO_IPV6) {
+			ipv6h = skb_header_pointer(skb, offset,
+						   sizeof(_ipv6h), &_ipv6h);
+
+			if (!ipv6h || ipv6h->nexthdr != IPPROTO_TCP)
+				return NULL;
+
+			offset += sizeof(struct ipv6hdr);
+
+		} else if (iph->protocol != IPPROTO_TCP) {
+			return NULL;
+		}
+
+	} else if (ipv6h->version == 6) {
+		if (ipv6h->nexthdr != IPPROTO_TCP)
+			return NULL;
+
+		offset += sizeof(struct ipv6hdr);
+	} else {
+		return NULL;
+	}
+
+	tcph = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph);
+	if (!tcph)
+		return NULL;
+
+	return skb_header_pointer(skb, offset,
+				  min(__tcp_hdrlen(tcph), bufsize), buf);
+}
+
+static bool cake_tcph_is_sack(const struct tcphdr *tcph)
+{
+	/* inspired by tcp_parse_options in tcp_input.c */
+	int length = __tcp_hdrlen(tcph) - sizeof(struct tcphdr);
+	const u8 *ptr = (const u8 *)(tcph + 1);
+
+	while (length > 0) {
+		int opcode = *ptr++;
+		int opsize;
+
+		if (opcode == TCPOPT_EOL)
+			break;
+		if (opcode == TCPOPT_NOP) {
+			length--;
+			continue;
+		}
+		opsize = *ptr++;
+		if (opsize < 2 || opsize > length)
+			break;
+		if (opcode == TCPOPT_SACK)
+			return true;
+		ptr += opsize - 2;
+		length -= opsize;
+	}
+
+	return false;
+}
+
+static struct sk_buff *cake_ack_filter(struct cake_sched_data *q,
+				       struct cake_flow *flow)
+{
+	bool aggressive = q->ack_filter == CAKE_ACK_AGGRESSIVE;
+	struct sk_buff *elig_ack = NULL, *elig_ack_prev = NULL;
+	struct sk_buff *skb_check, *skb_prev = NULL;
+	const struct ipv6hdr *ipv6h, *ipv6h_check;
+	const struct tcphdr *tcph, *tcph_check;
+	const struct iphdr *iph, *iph_check;
+	const struct sk_buff *skb;
+	struct ipv6hdr _iph, _iph_check;
+	struct tcphdr _tcph_check;
+	unsigned char _tcph[64]; /* need to hold maximum hdr size */
+	int seglen, num_found = 0;
+
+	/* no other possible ACKs to filter */
+	if (flow->head == flow->tail)
+		return NULL;
+
+	skb = flow->tail;
+	tcph = cake_get_tcphdr(skb, _tcph, sizeof(_tcph));
+	iph = cake_get_iphdr(skb, &_iph);
+	if (!tcph)
+		return NULL;
+
+	/* the 'triggering' packet need only have the ACK flag set.
+	 * also check that SYN is not set, as there won't be any previous ACKs.
+	 */
+	if ((tcp_flag_word(tcph) &
+	     (TCP_FLAG_ACK | TCP_FLAG_SYN)) != TCP_FLAG_ACK)
+		return NULL;
+
+	/* the 'triggering' ACK is at the tail of the queue, we have already
+	 * returned if it is the only packet in the flow. loop through the rest
+	 * of the queue looking for pure ACKs with the same 5-tuple as the
+	 * triggering one.
+	 */
+	for (skb_check = flow->head;
+	     skb_check && skb_check != skb;
+	     skb_prev = skb_check, skb_check = skb_check->next) {
+		iph_check = cake_get_iphdr(skb_check, &_iph_check);
+		tcph_check = cake_get_tcphdr(skb_check, &_tcph_check,
+					     sizeof(_tcph_check));
+
+		/* only TCP packets with matching 5-tuple are eligible */
+		if (!tcph_check || iph->version != iph_check->version ||
+		    tcph_check->source != tcph->source ||
+		    tcph_check->dest != tcph->dest)
+			continue;
+
+		if (iph_check->version == 4) {
+			if (iph_check->saddr != iph->saddr ||
+			    iph_check->daddr != iph->daddr)
+				continue;
+
+			seglen = ntohs(iph_check->tot_len) -
+				       (4 * iph_check->ihl);
+		} else if (iph_check->version == 6) {
+			ipv6h = (struct ipv6hdr *)iph;
+			ipv6h_check = (struct ipv6hdr *)iph_check;
+
+			if (ipv6_addr_cmp(&ipv6h_check->saddr, &ipv6h->saddr) ||
+			    ipv6_addr_cmp(&ipv6h_check->daddr, &ipv6h->daddr))
+				continue;
+
+			seglen = ntohs(ipv6h_check->payload_len);
+		} else {
+			WARN_ON(1);  /* shouldn't happen */
+			continue;
+		}
+
+		/* stricter criteria apply to ACKs that we may filter
+		 * 3 reserved flags must be unset to avoid future breakage
+		 * ECE/CWR/NS can be safely ignored
+		 * ACK must be set
+		 * All other flags URG/PSH/RST/SYN/FIN must be unset
+		 * 0x0FFF0000 = all TCP flags (confirm ACK=1, others zero)
+		 * 0x01C00000 = NS/CWR/ECE (safe to ignore)
+		 * 0x0E3F0000 = 0x0FFF0000 & ~0x01C00000
+		 * must be 'pure' ACK, contain zero bytes of segment data
+		 * options are ignored
+		 */
+		if (((tcp_flag_word(tcph_check) &
+			   cpu_to_be32(0x0E3F0000)) != TCP_FLAG_ACK) ||
+			 ((seglen - __tcp_hdrlen(tcph_check)) != 0))
+			continue;
+
+		/* The triggering packet must ACK more data than the ACK under
+		 * consideration, either because is has a strictly higher ACK
+		 * sequence number or because it is a SACK
+		 */
+		if ((ntohl(tcph_check->ack_seq) == ntohl(tcph->ack_seq) &&
+		     !cake_tcph_is_sack(tcph)) ||
+		    (int32_t)(ntohl(tcph_check->ack_seq) -
+			      ntohl(tcph->ack_seq)) > 0)
+			continue;
+
+		/* At this point we have found an eligible pure ACK to drop; if
+		 * we are in aggressive mode, we are done. Otherwise, keep
+		 * searching unless this is the second eligible ACK we
+		 * found.
+		 *
+		 * Since we want to drop ACK closest to the head of the queue,
+		 * save the first eligible ACK we find, even if we need to loop
+		 * again.
+		 */
+		if (!elig_ack) {
+			elig_ack = skb_check;
+			elig_ack_prev = skb_prev;
+		}
+
+		if (num_found++ > 0 || aggressive)
+			goto found;
+	}
+
+	return NULL;
+
+found:
+	if (elig_ack_prev)
+		elig_ack_prev->next = elig_ack->next;
+	else
+		flow->head = elig_ack->next;
+
+	elig_ack->next = NULL;
+
+	return elig_ack;
+}
+
 static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
 				      u32 shift)
 {
@@ -908,6 +1137,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	/* signed len to handle corner case filtered ACK larger than trigger */
 	int len = qdisc_pkt_len(skb);
 	u64 now = cobalt_get_time();
+	struct sk_buff *ack = NULL;
 
 	tin = 0;
 	b = &q->tins[tin];
@@ -941,8 +1171,24 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	cobalt_set_enqueue_time(skb, now);
 	flow_queue_add(flow, skb);
 
-	sch->q.qlen++;
-	q->buffer_used      += skb->truesize;
+	if (q->ack_filter)
+		ack = cake_ack_filter(q, flow);
+
+	if (ack) {
+		b->ack_drops++;
+		sch->qstats.drops++;
+		b->bytes += qdisc_pkt_len(ack);
+		len -= qdisc_pkt_len(ack);
+		q->buffer_used += skb->truesize - ack->truesize;
+		if (q->rate_flags & CAKE_FLAG_INGRESS)
+			cake_advance_shaper(q, b, ack, now, true);
+
+		qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack));
+		consume_skb(ack);
+	} else {
+		sch->q.qlen++;
+		q->buffer_used      += skb->truesize;
+	}
 
 	/* stats */
 	b->packets++;
@@ -1455,6 +1701,9 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
 			q->rate_flags &= ~CAKE_FLAG_INGRESS;
 	}
 
+	if (tb[TCA_CAKE_ACK_FILTER])
+		q->ack_filter = nla_get_u32(tb[TCA_CAKE_ACK_FILTER]);
+
 	if (tb[TCA_CAKE_MEMORY])
 		q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
 

[PATCH net-next v8 4/7] sch_cake: Add NAT awareness to packet classifier

[Toke Høiland-Jørgensen]

When CAKE is deployed on a gateway that also performs NAT (which is a
common deployment mode), the host fairness mechanism cannot distinguish
internal hosts from each other, and so fails to work correctly.

To fix this, we add an optional NAT awareness mode, which will query the
kernel conntrack mechanism to obtain the pre-NAT addresses for each packet
and use that in the flow and host hashing.

When the shaper is enabled and the host is already performing NAT, the cost
of this lookup is negligible. However, in unlimited mode with no NAT being
performed, there is a significant CPU cost at higher bandwidths. For this
reason, the feature is turned off by default.

Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 net/sched/sch_cake.c |   70 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 70 insertions(+)

diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 9a70e99afe7e..cc45a56d35d6 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -70,6 +70,12 @@
 #include <net/tcp.h>
 #include <net/flow_dissector.h>
 
+#if IS_REACHABLE(CONFIG_NF_CONNTRACK)
+#include <net/netfilter/nf_conntrack_core.h>
+#include <net/netfilter/nf_conntrack_zones.h>
+#include <net/netfilter/nf_conntrack.h>
+#endif
+
 #define CAKE_SET_WAYS (8)
 #define CAKE_MAX_TINS (8)
 #define CAKE_QUEUES (1024)
@@ -519,6 +525,61 @@ static bool cobalt_should_drop(struct cobalt_vars *vars,
 	return drop;
 }
 
+#if IS_REACHABLE(CONFIG_NF_CONNTRACK)
+
+static void cake_update_flowkeys(struct flow_keys *keys,
+				 const struct sk_buff *skb)
+{
+	enum ip_conntrack_info ctinfo;
+	bool rev = false;
+
+	struct nf_conn *ct;
+	const struct nf_conntrack_tuple *tuple;
+
+	if (tc_skb_protocol(skb) != htons(ETH_P_IP))
+		return;
+
+	ct = nf_ct_get(skb, &ctinfo);
+	if (ct) {
+		tuple = nf_ct_tuple(ct, CTINFO2DIR(ctinfo));
+	} else {
+		const struct nf_conntrack_tuple_hash *hash;
+		struct nf_conntrack_tuple srctuple;
+
+		if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb),
+				       NFPROTO_IPV4, dev_net(skb->dev),
+				       &srctuple))
+			return;
+
+		hash = nf_conntrack_find_get(dev_net(skb->dev),
+					     &nf_ct_zone_dflt,
+					     &srctuple);
+		if (!hash)
+			return;
+
+		rev = true;
+		ct = nf_ct_tuplehash_to_ctrack(hash);
+		tuple = nf_ct_tuple(ct, !hash->tuple.dst.dir);
+	}
+
+	keys->addrs.v4addrs.src = rev ? tuple->dst.u3.ip : tuple->src.u3.ip;
+	keys->addrs.v4addrs.dst = rev ? tuple->src.u3.ip : tuple->dst.u3.ip;
+
+	if (keys->ports.ports) {
+		keys->ports.src = rev ? tuple->dst.u.all : tuple->src.u.all;
+		keys->ports.dst = rev ? tuple->src.u.all : tuple->dst.u.all;
+	}
+	if (rev)
+		nf_ct_put(ct);
+}
+#else
+static void cake_update_flowkeys(struct flow_keys *keys,
+				 const struct sk_buff *skb)
+{
+	/* There is nothing we can do here without CONNTRACK */
+}
+#endif
+
 /* Cake has several subtle multiple bit settings. In these cases you
  *  would be matching triple isolate mode as well.
  */
@@ -546,6 +607,9 @@ static u32 cake_hash(struct cake_tin_data *q, const struct sk_buff *skb,
 	skb_flow_dissect_flow_keys(skb, &keys,
 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
 
+	if (flow_mode & CAKE_FLOW_NAT_FLAG)
+		cake_update_flowkeys(&keys, skb);
+
 	/* flow_hash_from_keys() sorts the addresses by value, so we have
 	 * to preserve their order in a separate data structure to treat
 	 * src and dst host addresses as independently selectable.
@@ -1673,6 +1737,12 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
 		q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
 				CAKE_FLOW_MASK);
 
+	if (tb[TCA_CAKE_NAT]) {
+		q->flow_mode &= ~CAKE_FLOW_NAT_FLAG;
+		q->flow_mode |= CAKE_FLOW_NAT_FLAG *
+			!!nla_get_u32(tb[TCA_CAKE_NAT]);
+	}
+
 	if (tb[TCA_CAKE_RTT]) {
 		q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
 

[PATCH net-next v8 2/7] sch_cake: Add ingress mode

[Toke Høiland-Jørgensen]

The ingress mode is meant to be enabled when CAKE runs downlink of the
actual bottleneck (such as on an IFB device). The mode changes the shaper
to also account dropped packets to the shaped rate, as these have already
traversed the bottleneck.

Enabling ingress mode will also tune the AQM to always keep at least two
packets queued *for each flow*. This is done by scaling the minimum queue
occupancy level that will disable the AQM by the number of active bulk
flows. The rationale for this is that retransmits are more expensive in
ingress mode, since dropped packets have to traverse the bottleneck again
when they are retransmitted; thus, being more lenient and keeping a minimum
number of packets queued will improve throughput in cases where the number
of active flows are so large that they saturate the bottleneck even at
their minimum window size.

This commit also adds a separate switch to enable ingress mode rate
autoscaling. If enabled, the autoscaling code will observe the actual
traffic rate and adjust the shaper rate to match it. This can help avoid
latency increases in the case where the actual bottleneck rate decreases
below the shaped rate. The scaling filters out spikes by an EWMA filter.

Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 net/sched/sch_cake.c |   70 +++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 66 insertions(+), 4 deletions(-)

diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 8e2f2ba2ed5d..7ca86e3ed14c 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -438,7 +438,8 @@ static bool cobalt_queue_empty(struct cobalt_vars *vars,
 static bool cobalt_should_drop(struct cobalt_vars *vars,
 			       struct cobalt_params *p,
 			       cobalt_time_t now,
-			       struct sk_buff *skb)
+			       struct sk_buff *skb,
+			       u32 bulk_flows)
 {
 	bool drop = false;
 
@@ -463,6 +464,7 @@ static bool cobalt_should_drop(struct cobalt_vars *vars,
 	cobalt_tdiff_t schedule = now - vars->drop_next;
 
 	bool over_target = sojourn > p->target &&
+			   sojourn > p->mtu_time * bulk_flows * 2 &&
 			   sojourn > p->mtu_time * 4;
 	bool next_due    = vars->count && schedule >= 0;
 
@@ -883,6 +885,9 @@ static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
 	b->tin_dropped++;
 	sch->qstats.drops++;
 
+	if (q->rate_flags & CAKE_FLAG_INGRESS)
+		cake_advance_shaper(q, b, skb, now, true);
+
 	__qdisc_drop(skb, to_free);
 	sch->q.qlen--;
 
@@ -951,8 +956,39 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		cake_heapify_up(q, b->overflow_idx[idx]);
 
 	/* incoming bandwidth capacity estimate */
-	q->avg_window_bytes = 0;
-	q->last_packet_time = now;
+	if (q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS) {
+		u64 packet_interval = now - q->last_packet_time;
+
+		if (packet_interval > NSEC_PER_SEC)
+			packet_interval = NSEC_PER_SEC;
+
+		/* filter out short-term bursts, eg. wifi aggregation */
+		q->avg_packet_interval = cake_ewma(q->avg_packet_interval,
+						   packet_interval,
+			packet_interval > q->avg_packet_interval ? 2 : 8);
+
+		q->last_packet_time = now;
+
+		if (packet_interval > q->avg_packet_interval) {
+			u64 window_interval = now - q->avg_window_begin;
+			u64 b = q->avg_window_bytes * (u64)NSEC_PER_SEC;
+
+			do_div(b, window_interval);
+			q->avg_peak_bandwidth =
+				cake_ewma(q->avg_peak_bandwidth, b,
+					  b > q->avg_peak_bandwidth ? 2 : 8);
+			q->avg_window_bytes = 0;
+			q->avg_window_begin = now;
+
+			if (now - q->last_reconfig_time > (NSEC_PER_SEC / 4)) {
+				q->rate_bps = (q->avg_peak_bandwidth * 15) >> 4;
+				cake_reconfigure(sch);
+			}
+		}
+	} else {
+		q->avg_window_bytes = 0;
+		q->last_packet_time = now;
+	}
 
 	/* flowchain */
 	if (!flow->set || flow->set == CAKE_SET_DECAYING) {
@@ -1207,14 +1243,26 @@ static struct sk_buff *cake_dequeue(struct Qdisc *sch)
 		}
 
 		/* Last packet in queue may be marked, shouldn't be dropped */
-		if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb) ||
+		if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb,
+					(b->bulk_flow_count *
+					 !!(q->rate_flags &
+					    CAKE_FLAG_INGRESS))) ||
 		    !flow->head)
 			break;
 
+		/* drop this packet, get another one */
+		if (q->rate_flags & CAKE_FLAG_INGRESS) {
+			len = cake_advance_shaper(q, b, skb,
+						  now, true);
+			flow->deficit -= len;
+			b->tin_deficit -= len;
+		}
 		b->tin_dropped++;
 		qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb));
 		qdisc_qstats_drop(sch);
 		kfree_skb(skb);
+		if (q->rate_flags & CAKE_FLAG_INGRESS)
+			goto retry;
 	}
 
 	b->tin_ecn_mark += !!flow->cvars.ecn_marked;
@@ -1393,6 +1441,20 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
 			q->target = 1;
 	}
 
+	if (tb[TCA_CAKE_AUTORATE]) {
+		if (!!nla_get_u32(tb[TCA_CAKE_AUTORATE]))
+			q->rate_flags |= CAKE_FLAG_AUTORATE_INGRESS;
+		else
+			q->rate_flags &= ~CAKE_FLAG_AUTORATE_INGRESS;
+	}
+
+	if (tb[TCA_CAKE_INGRESS]) {
+		if (!!nla_get_u32(tb[TCA_CAKE_INGRESS]))
+			q->rate_flags |= CAKE_FLAG_INGRESS;
+		else
+			q->rate_flags &= ~CAKE_FLAG_INGRESS;
+	}
+
 	if (tb[TCA_CAKE_MEMORY])
 		q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
 

[PATCH net-next v8 6/7] sch_cake: Add overhead compensation support to the rate shaper

[Toke Høiland-Jørgensen]

This commit adds configurable overhead compensation support to the rate
shaper. With this feature, userspace can configure the actual bottleneck
link overhead and encapsulation mode used, which will be used by the shaper
to calculate the precise duration of each packet on the wire.

This feature is needed because CAKE is often deployed one or two hops
upstream of the actual bottleneck (which can be, e.g., inside a DSL or
cable modem). In this case, the link layer characteristics and overhead
reported by the kernel does not match the actual bottleneck. Being able to
set the actual values in use makes it possible to configure the shaper rate
much closer to the actual bottleneck rate (our experience shows it is
possible to get with 0.1% of the actual physical bottleneck rate), thus
keeping latency low without sacrificing bandwidth.

The overhead compensation has three tunables: A fixed per-packet overhead
size (which, if set, will be accounted from the IP packet header), a
minimum packet size (MPU) and a framing mode supporting either ATM or PTM
framing. We include a set of common keywords in TC to help users configure
the right parameters. If no overhead value is set, the value reported by
the kernel is used.

Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 net/sched/sch_cake.c |  110 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 109 insertions(+), 1 deletion(-)

diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 1e5951d26ed2..cb978a0f8969 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -273,6 +273,7 @@ enum {
 
 struct cobalt_skb_cb {
 	cobalt_time_t enqueue_time;
+	u32           adjusted_len;
 };
 
 static cobalt_time_t cobalt_get_time(void)
@@ -1095,6 +1096,87 @@ static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
 	return avg;
 }
 
+static u32 cake_overhead(struct cake_sched_data *q, struct sk_buff *skb)
+{
+	const struct skb_shared_info *shinfo = skb_shinfo(skb);
+	u32 off = skb_network_offset(skb);
+	u32 len = qdisc_pkt_len(skb);
+	u16 segs = 1;
+
+	if (unlikely(shinfo->gso_size)) {
+		/* borrowed from qdisc_pkt_len_init() */
+		unsigned int hdr_len;
+
+		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
+
+		/* + transport layer */
+		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 |
+					       SKB_GSO_TCPV6))) {
+			const struct tcphdr *th;
+			struct tcphdr _tcphdr;
+
+			th = skb_header_pointer(skb, skb_transport_offset(skb),
+						sizeof(_tcphdr), &_tcphdr);
+			if (likely(th))
+				hdr_len += __tcp_hdrlen(th);
+		} else {
+			struct udphdr _udphdr;
+
+			if (skb_header_pointer(skb, skb_transport_offset(skb),
+					       sizeof(_udphdr), &_udphdr))
+				hdr_len += sizeof(struct udphdr);
+		}
+
+		if (unlikely(shinfo->gso_type & SKB_GSO_DODGY))
+			segs = DIV_ROUND_UP(skb->len - hdr_len,
+					    shinfo->gso_size);
+		else
+			segs = shinfo->gso_segs;
+
+		/* The last segment may be shorter; we ignore this, which means
+		 * that we will over-estimate the size of the whole GSO segment
+		 * by the difference in size. This is conservative, so we live
+		 * with that to avoid the complexity of dealing with it.
+		 */
+		len = shinfo->gso_size + hdr_len;
+	}
+
+	q->avg_netoff = cake_ewma(q->avg_netoff, off << 16, 8);
+
+	if (q->rate_flags & CAKE_FLAG_OVERHEAD)
+		len -= off;
+
+	if (q->max_netlen < len)
+		q->max_netlen = len;
+	if (q->min_netlen > len)
+		q->min_netlen = len;
+
+	len += q->rate_overhead;
+
+	if (len < q->rate_mpu)
+		len = q->rate_mpu;
+
+	if (q->atm_mode == CAKE_ATM_ATM) {
+		len += 47;
+		len /= 48;
+		len *= 53;
+	} else if (q->atm_mode == CAKE_ATM_PTM) {
+		/* Add one byte per 64 bytes or part thereof.
+		 * This is conservative and easier to calculate than the
+		 * precise value.
+		 */
+		len += (len + 63) / 64;
+	}
+
+	if (q->max_adjlen < len)
+		q->max_adjlen = len;
+	if (q->min_adjlen > len)
+		q->min_adjlen = len;
+
+	get_cobalt_cb(skb)->adjusted_len = len * segs;
+	return len;
+}
+
 static void cake_heap_swap(struct cake_sched_data *q, u16 i, u16 j)
 {
 	struct cake_heap_entry ii = q->overflow_heap[i];
@@ -1172,7 +1254,7 @@ static int cake_advance_shaper(struct cake_sched_data *q,
 			       struct sk_buff *skb,
 			       u64 now, bool drop)
 {
-	u32 len = qdisc_pkt_len(skb);
+	u32 len = get_cobalt_cb(skb)->adjusted_len;
 
 	/* charge packet bandwidth to this tin
 	 * and to the global shaper.
@@ -1347,6 +1429,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		b->max_skblen = len;
 
 	cobalt_set_enqueue_time(skb, now);
+	get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb);
 	flow_queue_add(flow, skb);
 
 	if (q->ack_filter)
@@ -2123,6 +2206,31 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
 			!!nla_get_u32(tb[TCA_CAKE_NAT]);
 	}
 
+	if (tb[TCA_CAKE_ATM])
+		q->atm_mode = nla_get_u32(tb[TCA_CAKE_ATM]);
+
+	if (tb[TCA_CAKE_OVERHEAD]) {
+		q->rate_overhead = nla_get_s32(tb[TCA_CAKE_OVERHEAD]);
+		q->rate_flags |= CAKE_FLAG_OVERHEAD;
+
+		q->max_netlen = 0;
+		q->max_adjlen = 0;
+		q->min_netlen = ~0;
+		q->min_adjlen = ~0;
+	}
+
+	if (tb[TCA_CAKE_RAW]) {
+		q->rate_flags &= ~CAKE_FLAG_OVERHEAD;
+
+		q->max_netlen = 0;
+		q->max_adjlen = 0;
+		q->min_netlen = ~0;
+		q->min_adjlen = ~0;
+	}
+
+	if (tb[TCA_CAKE_MPU])
+		q->rate_mpu = nla_get_u32(tb[TCA_CAKE_MPU]);
+
 	if (tb[TCA_CAKE_RTT]) {
 		q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
 

[PATCH net-next v8 7/7] sch_cake: Conditionally split GSO segments

[Toke Høiland-Jørgensen]

At lower bandwidths, the transmission time of a single GSO segment can add
an unacceptable amount of latency due to HOL blocking. Furthermore, with a
software shaper, any tuning mechanism employed by the kernel to control the
maximum size of GSO segments is thrown off by the artificial limit on
bandwidth. For this reason, we split GSO segments into their individual
packets iff the shaper is active and configured to a bandwidth <= 1 Gbps.

Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 net/sched/sch_cake.c |   95 ++++++++++++++++++++++++++++++++++++--------------
 1 file changed, 69 insertions(+), 26 deletions(-)

diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index cb978a0f8969..6b67a5c1418e 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -81,6 +81,7 @@
 #define CAKE_QUEUES (1024)
 #define CAKE_FLOW_MASK 63
 #define CAKE_FLOW_NAT_FLAG 64
+#define CAKE_SPLIT_GSO_THRESHOLD (125000000) /* 1Gbps */
 #define US2TIME(a) (a * (u64)NSEC_PER_USEC)
 
 typedef u64 cobalt_time_t;
@@ -1428,36 +1429,73 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	if (unlikely(len > b->max_skblen))
 		b->max_skblen = len;
 
-	cobalt_set_enqueue_time(skb, now);
-	get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb);
-	flow_queue_add(flow, skb);
-
-	if (q->ack_filter)
-		ack = cake_ack_filter(q, flow);
+	if (skb_is_gso(skb) && q->rate_flags & CAKE_FLAG_SPLIT_GSO) {
+		struct sk_buff *segs, *nskb;
+		netdev_features_t features = netif_skb_features(skb);
+		unsigned int slen = 0;
+
+		segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK);
+		if (IS_ERR_OR_NULL(segs))
+			return qdisc_drop(skb, sch, to_free);
+
+		while (segs) {
+			nskb = segs->next;
+			segs->next = NULL;
+			qdisc_skb_cb(segs)->pkt_len = segs->len;
+			cobalt_set_enqueue_time(segs, now);
+			get_cobalt_cb(segs)->adjusted_len = cake_overhead(q,
+									  segs);
+			flow_queue_add(flow, segs);
+
+			sch->q.qlen++;
+			slen += segs->len;
+			q->buffer_used += segs->truesize;
+			b->packets++;
+			segs = nskb;
+		}
 
-	if (ack) {
-		b->ack_drops++;
-		sch->qstats.drops++;
-		b->bytes += qdisc_pkt_len(ack);
-		len -= qdisc_pkt_len(ack);
-		q->buffer_used += skb->truesize - ack->truesize;
-		if (q->rate_flags & CAKE_FLAG_INGRESS)
-			cake_advance_shaper(q, b, ack, now, true);
+		/* stats */
+		b->bytes	    += slen;
+		b->backlogs[idx]    += slen;
+		b->tin_backlog      += slen;
+		sch->qstats.backlog += slen;
+		q->avg_window_bytes += slen;
 
-		qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack));
-		consume_skb(ack);
+		qdisc_tree_reduce_backlog(sch, 1, len);
+		consume_skb(skb);
 	} else {
-		sch->q.qlen++;
-		q->buffer_used      += skb->truesize;
-	}
+		/* not splitting */
+		cobalt_set_enqueue_time(skb, now);
+		get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb);
+		flow_queue_add(flow, skb);
+
+		if (q->ack_filter)
+			ack = cake_ack_filter(q, flow);
+
+		if (ack) {
+			b->ack_drops++;
+			sch->qstats.drops++;
+			b->bytes += qdisc_pkt_len(ack);
+			len -= qdisc_pkt_len(ack);
+			q->buffer_used += skb->truesize - ack->truesize;
+			if (q->rate_flags & CAKE_FLAG_INGRESS)
+				cake_advance_shaper(q, b, ack, now, true);
+
+			qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack));
+			consume_skb(ack);
+		} else {
+			sch->q.qlen++;
+			q->buffer_used      += skb->truesize;
+		}
 
-	/* stats */
-	b->packets++;
-	b->bytes	    += len;
-	b->backlogs[idx]    += len;
-	b->tin_backlog      += len;
-	sch->qstats.backlog += len;
-	q->avg_window_bytes += len;
+		/* stats */
+		b->packets++;
+		b->bytes	    += len;
+		b->backlogs[idx]    += len;
+		b->tin_backlog      += len;
+		sch->qstats.backlog += len;
+		q->avg_window_bytes += len;
+	}
 
 	if (q->overflow_timeout)
 		cake_heapify_up(q, b->overflow_idx[idx]);
@@ -2265,6 +2303,11 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
 	if (tb[TCA_CAKE_MEMORY])
 		q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
 
+	if (q->rate_bps && q->rate_bps <= CAKE_SPLIT_GSO_THRESHOLD)
+		q->rate_flags |= CAKE_FLAG_SPLIT_GSO;
+	else
+		q->rate_flags &= ~CAKE_FLAG_SPLIT_GSO;
+
 	if (q->tins) {
 		sch_tree_lock(sch);
 		cake_reconfigure(sch);

[PATCH net-next v8 5/7] sch_cake: Add DiffServ handling

[Toke Høiland-Jørgensen]

This adds support for DiffServ-based priority queueing to CAKE. If the
shaper is in use, each priority tier gets its own virtual clock, which
limits that tier's rate to a fraction of the overall shaped rate, to
discourage trying to game the priority mechanism.

CAKE defaults to a simple, three-tier mode that interprets most code points
as "best effort", but places CS1 traffic into a low-priority "bulk" tier
which is assigned 1/16 of the total rate, and a few code points indicating
latency-sensitive or control traffic (specifically TOS4, VA, EF, CS6, CS7)
into a "latency sensitive" high-priority tier, which is assigned 1/4 rate.
The other supported DiffServ modes are a 4-tier mode matching the 802.11e
precedence rules, as well as two 8-tier modes, one of which implements
strict precedence of the eight priority levels.

This commit also adds an optional DiffServ 'wash' mode, which will zero out
the DSCP fields of any packet passing through CAKE. While this can
technically be done with other mechanisms in the kernel, having the feature
available in CAKE significantly decreases configuration complexity; and the
implementation cost is low on top of the other DiffServ-handling code.

Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 net/sched/sch_cake.c |  394 +++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 387 insertions(+), 7 deletions(-)

diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index cc45a56d35d6..1e5951d26ed2 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -305,6 +305,68 @@ static void cobalt_set_enqueue_time(struct sk_buff *skb,
 
 static u16 quantum_div[CAKE_QUEUES + 1] = {0};
 
+/* Diffserv lookup tables */
+
+static const u8 precedence[] = {
+	0, 0, 0, 0, 0, 0, 0, 0,
+	1, 1, 1, 1, 1, 1, 1, 1,
+	2, 2, 2, 2, 2, 2, 2, 2,
+	3, 3, 3, 3, 3, 3, 3, 3,
+	4, 4, 4, 4, 4, 4, 4, 4,
+	5, 5, 5, 5, 5, 5, 5, 5,
+	6, 6, 6, 6, 6, 6, 6, 6,
+	7, 7, 7, 7, 7, 7, 7, 7,
+};
+
+static const u8 diffserv8[] = {
+	2, 5, 1, 2, 4, 2, 2, 2,
+	0, 2, 1, 2, 1, 2, 1, 2,
+	5, 2, 4, 2, 4, 2, 4, 2,
+	3, 2, 3, 2, 3, 2, 3, 2,
+	6, 2, 3, 2, 3, 2, 3, 2,
+	6, 2, 2, 2, 6, 2, 6, 2,
+	7, 2, 2, 2, 2, 2, 2, 2,
+	7, 2, 2, 2, 2, 2, 2, 2,
+};
+
+static const u8 diffserv4[] = {
+	0, 2, 0, 0, 2, 0, 0, 0,
+	1, 0, 0, 0, 0, 0, 0, 0,
+	2, 0, 2, 0, 2, 0, 2, 0,
+	2, 0, 2, 0, 2, 0, 2, 0,
+	3, 0, 2, 0, 2, 0, 2, 0,
+	3, 0, 0, 0, 3, 0, 3, 0,
+	3, 0, 0, 0, 0, 0, 0, 0,
+	3, 0, 0, 0, 0, 0, 0, 0,
+};
+
+static const u8 diffserv3[] = {
+	0, 0, 0, 0, 2, 0, 0, 0,
+	1, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 2, 0, 2, 0,
+	2, 0, 0, 0, 0, 0, 0, 0,
+	2, 0, 0, 0, 0, 0, 0, 0,
+};
+
+static const u8 besteffort[] = {
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+/* tin priority order for stats dumping */
+
+static const u8 normal_order[] = {0, 1, 2, 3, 4, 5, 6, 7};
+static const u8 bulk_order[] = {1, 0, 2, 3};
+
 #define REC_INV_SQRT_CACHE (16)
 static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0};
 
@@ -1189,6 +1251,46 @@ static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
 	return idx + (tin << 16);
 }
 
+static void cake_wash_diffserv(struct sk_buff *skb)
+{
+	switch (skb->protocol) {
+	case htons(ETH_P_IP):
+		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, 0);
+		break;
+	case htons(ETH_P_IPV6):
+		ipv6_change_dsfield(ipv6_hdr(skb), INET_ECN_MASK, 0);
+		break;
+	default:
+		break;
+	};
+}
+
+static u8 cake_handle_diffserv(struct sk_buff *skb, u16 wash)
+{
+	u8 dscp;
+
+	switch (skb->protocol) {
+	case htons(ETH_P_IP):
+		dscp = ipv4_get_dsfield(ip_hdr(skb)) >> 2;
+		if (wash && dscp)
+			ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, 0);
+		return dscp;
+
+	case htons(ETH_P_IPV6):
+		dscp = ipv6_get_dsfield(ipv6_hdr(skb)) >> 2;
+		if (wash && dscp)
+			ipv6_change_dsfield(ipv6_hdr(skb), INET_ECN_MASK, 0);
+		return dscp;
+
+	case htons(ETH_P_ARP):
+		return 0x38;  /* CS7 - Net Control */
+
+	default:
+		/* If there is no Diffserv field, treat as best-effort */
+		return 0;
+	};
+}
+
 static void cake_reconfigure(struct Qdisc *sch);
 
 static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
@@ -1203,7 +1305,19 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	u64 now = cobalt_get_time();
 	struct sk_buff *ack = NULL;
 
-	tin = 0;
+	/* extract the Diffserv Precedence field, if it exists */
+	/* and clear DSCP bits if washing */
+	if (q->tin_mode != CAKE_DIFFSERV_BESTEFFORT) {
+		tin = q->tin_index[cake_handle_diffserv(skb,
+				q->rate_flags & CAKE_FLAG_WASH)];
+		if (unlikely(tin >= q->tin_cnt))
+			tin = 0;
+	} else {
+		tin = 0;
+		if (q->rate_flags & CAKE_FLAG_WASH)
+			cake_wash_diffserv(skb);
+	}
+
 	b = &q->tins[tin];
 
 	/* choose flow to insert into */
@@ -1678,18 +1792,274 @@ static void cake_set_rate(struct cake_tin_data *b, u64 rate, u32 mtu,
 	b->cparams.p_dec = 1 << 20; /* 1/4096 */
 }
 
-static void cake_reconfigure(struct Qdisc *sch)
+static int cake_config_besteffort(struct Qdisc *sch)
 {
 	struct cake_sched_data *q = qdisc_priv(sch);
 	struct cake_tin_data *b = &q->tins[0];
-	int c, ft = 0;
+	u32 rate = q->rate_bps;
+	u32 mtu = psched_mtu(qdisc_dev(sch));
 
 	q->tin_cnt = 1;
-	cake_set_rate(b, q->rate_bps, psched_mtu(qdisc_dev(sch)),
-		      US2TIME(q->target), US2TIME(q->interval));
+
+	q->tin_index = besteffort;
+	q->tin_order = normal_order;
+
+	cake_set_rate(b, rate, mtu, US2TIME(q->target), US2TIME(q->interval));
 	b->tin_quantum_band = 65535;
 	b->tin_quantum_prio = 65535;
 
+	return 0;
+}
+
+static int cake_config_precedence(struct Qdisc *sch)
+{
+	/* convert high-level (user visible) parameters into internal format */
+	struct cake_sched_data *q = qdisc_priv(sch);
+	u32 rate = q->rate_bps;
+	u32 mtu = psched_mtu(qdisc_dev(sch));
+	u32 quantum1 = 256;
+	u32 quantum2 = 256;
+	u32 i;
+
+	q->tin_cnt = 8;
+	q->tin_index = precedence;
+	q->tin_order = normal_order;
+
+	for (i = 0; i < q->tin_cnt; i++) {
+		struct cake_tin_data *b = &q->tins[i];
+
+		cake_set_rate(b, rate, mtu, US2TIME(q->target),
+			      US2TIME(q->interval));
+
+		b->tin_quantum_prio = max_t(u16, 1U, quantum1);
+		b->tin_quantum_band = max_t(u16, 1U, quantum2);
+
+		/* calculate next class's parameters */
+		rate  *= 7;
+		rate >>= 3;
+
+		quantum1  *= 3;
+		quantum1 >>= 1;
+
+		quantum2  *= 7;
+		quantum2 >>= 3;
+	}
+
+	return 0;
+}
+
+/*	List of known Diffserv codepoints:
+ *
+ *	Least Effort (CS1)
+ *	Best Effort (CS0)
+ *	Max Reliability & LLT "Lo" (TOS1)
+ *	Max Throughput (TOS2)
+ *	Min Delay (TOS4)
+ *	LLT "La" (TOS5)
+ *	Assured Forwarding 1 (AF1x) - x3
+ *	Assured Forwarding 2 (AF2x) - x3
+ *	Assured Forwarding 3 (AF3x) - x3
+ *	Assured Forwarding 4 (AF4x) - x3
+ *	Precedence Class 2 (CS2)
+ *	Precedence Class 3 (CS3)
+ *	Precedence Class 4 (CS4)
+ *	Precedence Class 5 (CS5)
+ *	Precedence Class 6 (CS6)
+ *	Precedence Class 7 (CS7)
+ *	Voice Admit (VA)
+ *	Expedited Forwarding (EF)
+
+ *	Total 25 codepoints.
+ */
+
+/*	List of traffic classes in RFC 4594:
+ *		(roughly descending order of contended priority)
+ *		(roughly ascending order of uncontended throughput)
+ *
+ *	Network Control (CS6,CS7)      - routing traffic
+ *	Telephony (EF,VA)         - aka. VoIP streams
+ *	Signalling (CS5)               - VoIP setup
+ *	Multimedia Conferencing (AF4x) - aka. video calls
+ *	Realtime Interactive (CS4)     - eg. games
+ *	Multimedia Streaming (AF3x)    - eg. YouTube, NetFlix, Twitch
+ *	Broadcast Video (CS3)
+ *	Low Latency Data (AF2x,TOS4)      - eg. database
+ *	Ops, Admin, Management (CS2,TOS1) - eg. ssh
+ *	Standard Service (CS0 & unrecognised codepoints)
+ *	High Throughput Data (AF1x,TOS2)  - eg. web traffic
+ *	Low Priority Data (CS1)           - eg. BitTorrent
+
+ *	Total 12 traffic classes.
+ */
+
+static int cake_config_diffserv8(struct Qdisc *sch)
+{
+/*	Pruned list of traffic classes for typical applications:
+ *
+ *		Network Control          (CS6, CS7)
+ *		Minimum Latency          (EF, VA, CS5, CS4)
+ *		Interactive Shell        (CS2, TOS1)
+ *		Low Latency Transactions (AF2x, TOS4)
+ *		Video Streaming          (AF4x, AF3x, CS3)
+ *		Bog Standard             (CS0 etc.)
+ *		High Throughput          (AF1x, TOS2)
+ *		Background Traffic       (CS1)
+ *
+ *		Total 8 traffic classes.
+ */
+
+	struct cake_sched_data *q = qdisc_priv(sch);
+	u32 rate = q->rate_bps;
+	u32 mtu = psched_mtu(qdisc_dev(sch));
+	u32 quantum1 = 256;
+	u32 quantum2 = 256;
+	u32 i;
+
+	q->tin_cnt = 8;
+
+	/* codepoint to class mapping */
+	q->tin_index = diffserv8;
+	q->tin_order = normal_order;
+
+	/* class characteristics */
+	for (i = 0; i < q->tin_cnt; i++) {
+		struct cake_tin_data *b = &q->tins[i];
+
+		cake_set_rate(b, rate, mtu, US2TIME(q->target),
+			      US2TIME(q->interval));
+
+		b->tin_quantum_prio = max_t(u16, 1U, quantum1);
+		b->tin_quantum_band = max_t(u16, 1U, quantum2);
+
+		/* calculate next class's parameters */
+		rate  *= 7;
+		rate >>= 3;
+
+		quantum1  *= 3;
+		quantum1 >>= 1;
+
+		quantum2  *= 7;
+		quantum2 >>= 3;
+	}
+
+	return 0;
+}
+
+static int cake_config_diffserv4(struct Qdisc *sch)
+{
+/*  Further pruned list of traffic classes for four-class system:
+ *
+ *	    Latency Sensitive  (CS7, CS6, EF, VA, CS5, CS4)
+ *	    Streaming Media    (AF4x, AF3x, CS3, AF2x, TOS4, CS2, TOS1)
+ *	    Best Effort        (CS0, AF1x, TOS2, and those not specified)
+ *	    Background Traffic (CS1)
+ *
+ *		Total 4 traffic classes.
+ */
+
+	struct cake_sched_data *q = qdisc_priv(sch);
+	u32 rate = q->rate_bps;
+	u32 mtu = psched_mtu(qdisc_dev(sch));
+	u32 quantum = 1024;
+
+	q->tin_cnt = 4;
+
+	/* codepoint to class mapping */
+	q->tin_index = diffserv4;
+	q->tin_order = bulk_order;
+
+	/* class characteristics */
+	cake_set_rate(&q->tins[0], rate, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+	cake_set_rate(&q->tins[1], rate >> 4, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+	cake_set_rate(&q->tins[2], rate >> 1, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+	cake_set_rate(&q->tins[3], rate >> 2, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+
+	/* priority weights */
+	q->tins[0].tin_quantum_prio = quantum;
+	q->tins[1].tin_quantum_prio = quantum >> 4;
+	q->tins[2].tin_quantum_prio = quantum << 2;
+	q->tins[3].tin_quantum_prio = quantum << 4;
+
+	/* bandwidth-sharing weights */
+	q->tins[0].tin_quantum_band = quantum;
+	q->tins[1].tin_quantum_band = quantum >> 4;
+	q->tins[2].tin_quantum_band = quantum >> 1;
+	q->tins[3].tin_quantum_band = quantum >> 2;
+
+	return 0;
+}
+
+static int cake_config_diffserv3(struct Qdisc *sch)
+{
+/*  Simplified Diffserv structure with 3 tins.
+ *		Low Priority		(CS1)
+ *		Best Effort
+ *		Latency Sensitive	(TOS4, VA, EF, CS6, CS7)
+ */
+	struct cake_sched_data *q = qdisc_priv(sch);
+	u32 rate = q->rate_bps;
+	u32 mtu = psched_mtu(qdisc_dev(sch));
+	u32 quantum = 1024;
+
+	q->tin_cnt = 3;
+
+	/* codepoint to class mapping */
+	q->tin_index = diffserv3;
+	q->tin_order = bulk_order;
+
+	/* class characteristics */
+	cake_set_rate(&q->tins[0], rate, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+	cake_set_rate(&q->tins[1], rate >> 4, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+	cake_set_rate(&q->tins[2], rate >> 2, mtu,
+		      US2TIME(q->target), US2TIME(q->interval));
+
+	/* priority weights */
+	q->tins[0].tin_quantum_prio = quantum;
+	q->tins[1].tin_quantum_prio = quantum >> 4;
+	q->tins[2].tin_quantum_prio = quantum << 4;
+
+	/* bandwidth-sharing weights */
+	q->tins[0].tin_quantum_band = quantum;
+	q->tins[1].tin_quantum_band = quantum >> 4;
+	q->tins[2].tin_quantum_band = quantum >> 2;
+
+	return 0;
+}
+
+static void cake_reconfigure(struct Qdisc *sch)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	int c, ft;
+
+	switch (q->tin_mode) {
+	case CAKE_DIFFSERV_BESTEFFORT:
+		ft = cake_config_besteffort(sch);
+		break;
+
+	case CAKE_DIFFSERV_PRECEDENCE:
+		ft = cake_config_precedence(sch);
+		break;
+
+	case CAKE_DIFFSERV_DIFFSERV8:
+		ft = cake_config_diffserv8(sch);
+		break;
+
+	case CAKE_DIFFSERV_DIFFSERV4:
+		ft = cake_config_diffserv4(sch);
+		break;
+
+	case CAKE_DIFFSERV_DIFFSERV3:
+	default:
+		ft = cake_config_diffserv3(sch);
+		break;
+	};
+
 	for (c = q->tin_cnt; c < CAKE_MAX_TINS; c++) {
 		cake_clear_tin(sch, c);
 		q->tins[c].cparams.mtu_time = q->tins[ft].cparams.mtu_time;
@@ -1733,6 +2103,16 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
 	if (tb[TCA_CAKE_BASE_RATE])
 		q->rate_bps = nla_get_u32(tb[TCA_CAKE_BASE_RATE]);
 
+	if (tb[TCA_CAKE_DIFFSERV_MODE])
+		q->tin_mode = nla_get_u32(tb[TCA_CAKE_DIFFSERV_MODE]);
+
+	if (tb[TCA_CAKE_WASH]) {
+		if (!!nla_get_u32(tb[TCA_CAKE_WASH]))
+			q->rate_flags |= CAKE_FLAG_WASH;
+		else
+			q->rate_flags &= ~CAKE_FLAG_WASH;
+	}
+
 	if (tb[TCA_CAKE_FLOW_MODE])
 		q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
 				CAKE_FLOW_MASK);
@@ -1809,7 +2189,7 @@ static int cake_init(struct Qdisc *sch, struct nlattr *opt,
 	int i, j;
 
 	sch->limit = 10240;
-	q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
+	q->tin_mode = CAKE_DIFFSERV_DIFFSERV3;
 	q->flow_mode  = CAKE_FLOW_TRIPLE;
 
 	q->rate_bps = 0; /* unlimited by default */
@@ -1985,7 +2365,7 @@ static int cake_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 	} while (0)
 
 	for (i = 0; i < q->tin_cnt; i++) {
-		struct cake_tin_data *b = &q->tins[i];
+		struct cake_tin_data *b = &q->tins[q->tin_order[i]];
 
 		ts = nla_nest_start(d->skb, i + 1);
 		if (!ts)

[PATCH net-next v8 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Toke Høiland-Jørgensen]

sch_cake targets the home router use case and is intended to squeeze the
most bandwidth and latency out of even the slowest ISP links and routers,
while presenting an API simple enough that even an ISP can configure it.

Example of use on a cable ISP uplink:

tc qdisc add dev eth0 cake bandwidth 20Mbit nat docsis ack-filter

To shape a cable download link (ifb and tc-mirred setup elided)

tc qdisc add dev ifb0 cake bandwidth 200mbit nat docsis ingress wash

CAKE is filled with:

* A hybrid Codel/Blue AQM algorithm, "Cobalt", tied to an FQ_Codel
  derived Flow Queuing system, which autoconfigures based on the bandwidth.
* A novel "triple-isolate" mode (the default) which balances per-host
  and per-flow FQ even through NAT.
* An deficit based shaper, that can also be used in an unlimited mode.
* 8 way set associative hashing to reduce flow collisions to a minimum.
* A reasonable interpretation of various diffserv latency/loss tradeoffs.
* Support for zeroing diffserv markings for entering and exiting traffic.
* Support for interacting well with Docsis 3.0 shaper framing.
* Extensive support for DSL framing types.
* Support for ack filtering.
* Extensive statistics for measuring, loss, ecn markings, latency
  variation.

A paper describing the design of CAKE is available at
https://arxiv.org/abs/1804.07617

This patch adds the base shaper and packet scheduler, while subsequent
commits add the optional (configurable) features. The full userspace API
and most data structures are included in this commit, but options not
understood in the base version will be ignored.

Various versions baking have been available as an out of tree build for
kernel versions going back to 3.10, as the embedded router world has been
running a few years behind mainline Linux. A stable version has been
generally available on lede-17.01 and later.

sch_cake replaces a combination of iptables, tc filter, htb and fq_codel
in the sqm-scripts, with sane defaults and vastly simpler configuration.

CAKE's principal author is Jonathan Morton, with contributions from
Kevin Darbyshire-Bryant, Toke Høiland-Jørgensen, Sebastian Moeller,
Ryan Mounce, Guido Sarducci, Dean Scarff, Nils Andreas Svee, Dave Täht,
and Loganaden Velvindron.

Testing from Pete Heist, Georgios Amanakis, and the many other members of
the cake@lists.bufferbloat.net mailing list.

tc -s qdisc show dev eth2
qdisc cake 1: root refcnt 2 bandwidth 100Mbit diffserv3 triple-isolate rtt 100.0ms raw overhead 0
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
 memory used: 0b of 5000000b
 capacity estimate: 100Mbit
 min/max network layer size:        65535 /       0
 min/max overhead-adjusted size:    65535 /       0
 average network hdr offset:            0

                   Bulk  Best Effort        Voice
  thresh       6250Kbit      100Mbit       25Mbit
  target          5.0ms        5.0ms        5.0ms
  interval      100.0ms      100.0ms      100.0ms
  pk_delay          0us          0us          0us
  av_delay          0us          0us          0us
  sp_delay          0us          0us          0us
  pkts                0            0            0
  bytes               0            0            0
  way_inds            0            0            0
  way_miss            0            0            0
  way_cols            0            0            0
  drops               0            0            0
  marks               0            0            0
  ack_drop            0            0            0
  sp_flows            0            0            0
  bk_flows            0            0            0
  un_flows            0            0            0
  max_len             0            0            0
  quantum           300         1514          762

Tested-by: Pete Heist <peteheist@gmail.com>
Tested-by: Georgios Amanakis <gamanakis@gmail.com>
Signed-off-by: Dave Taht <dave.taht@gmail.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
 include/uapi/linux/pkt_sched.h |  105 ++
 net/sched/Kconfig              |   11 
 net/sched/Makefile             |    1 
 net/sched/sch_cake.c           | 1683 ++++++++++++++++++++++++++++++++++++++++
 4 files changed, 1800 insertions(+)
 create mode 100644 net/sched/sch_cake.c

diff --git a/include/uapi/linux/pkt_sched.h b/include/uapi/linux/pkt_sched.h
index 37b5096ae97b..bc581473c0b0 100644
--- a/include/uapi/linux/pkt_sched.h
+++ b/include/uapi/linux/pkt_sched.h
@@ -934,4 +934,109 @@ enum {
 
 #define TCA_CBS_MAX (__TCA_CBS_MAX - 1)
 
+/* CAKE */
+enum {
+	TCA_CAKE_UNSPEC,
+	TCA_CAKE_BASE_RATE,
+	TCA_CAKE_DIFFSERV_MODE,
+	TCA_CAKE_ATM,
+	TCA_CAKE_FLOW_MODE,
+	TCA_CAKE_OVERHEAD,
+	TCA_CAKE_RTT,
+	TCA_CAKE_TARGET,
+	TCA_CAKE_AUTORATE,
+	TCA_CAKE_MEMORY,
+	TCA_CAKE_NAT,
+	TCA_CAKE_RAW,
+	TCA_CAKE_WASH,
+	TCA_CAKE_MPU,
+	TCA_CAKE_INGRESS,
+	TCA_CAKE_ACK_FILTER,
+	TCA_CAKE_SPLIT_GSO,
+	__TCA_CAKE_MAX
+};
+#define TCA_CAKE_MAX	(__TCA_CAKE_MAX - 1)
+
+enum {
+	__TCA_CAKE_STATS_INVALID,
+	TCA_CAKE_STATS_CAPACITY_ESTIMATE,
+	TCA_CAKE_STATS_MEMORY_LIMIT,
+	TCA_CAKE_STATS_MEMORY_USED,
+	TCA_CAKE_STATS_AVG_NETOFF,
+	TCA_CAKE_STATS_MIN_NETLEN,
+	TCA_CAKE_STATS_MAX_NETLEN,
+	TCA_CAKE_STATS_MIN_ADJLEN,
+	TCA_CAKE_STATS_MAX_ADJLEN,
+	TCA_CAKE_STATS_TIN_STATS,
+	__TCA_CAKE_STATS_MAX
+};
+#define TCA_CAKE_STATS_MAX (__TCA_CAKE_STATS_MAX - 1)
+
+enum {
+	__TCA_CAKE_TIN_STATS_INVALID,
+	TCA_CAKE_TIN_STATS_PAD,
+	TCA_CAKE_TIN_STATS_SENT_PACKETS,
+	TCA_CAKE_TIN_STATS_SENT_BYTES64,
+	TCA_CAKE_TIN_STATS_DROPPED_PACKETS,
+	TCA_CAKE_TIN_STATS_DROPPED_BYTES64,
+	TCA_CAKE_TIN_STATS_ACKS_DROPPED_PACKETS,
+	TCA_CAKE_TIN_STATS_ACKS_DROPPED_BYTES64,
+	TCA_CAKE_TIN_STATS_ECN_MARKED_PACKETS,
+	TCA_CAKE_TIN_STATS_ECN_MARKED_BYTES64,
+	TCA_CAKE_TIN_STATS_BACKLOG_PACKETS,
+	TCA_CAKE_TIN_STATS_BACKLOG_BYTES64,
+	TCA_CAKE_TIN_STATS_THRESHOLD_RATE,
+	TCA_CAKE_TIN_STATS_TARGET_US,
+	TCA_CAKE_TIN_STATS_INTERVAL_US,
+	TCA_CAKE_TIN_STATS_WAY_INDIRECT_HITS,
+	TCA_CAKE_TIN_STATS_WAY_MISSES,
+	TCA_CAKE_TIN_STATS_WAY_COLLISIONS,
+	TCA_CAKE_TIN_STATS_PEAK_DELAY_US,
+	TCA_CAKE_TIN_STATS_AVG_DELAY_US,
+	TCA_CAKE_TIN_STATS_BASE_DELAY_US,
+	TCA_CAKE_TIN_STATS_SPARSE_FLOWS,
+	TCA_CAKE_TIN_STATS_BULK_FLOWS,
+	TCA_CAKE_TIN_STATS_UNRESPONSIVE_FLOWS,
+	TCA_CAKE_TIN_STATS_MAX_SKBLEN,
+	TCA_CAKE_TIN_STATS_FLOW_QUANTUM,
+	__TCA_CAKE_TIN_STATS_MAX
+};
+#define TCA_CAKE_TIN_STATS_MAX (__TCA_CAKE_TIN_STATS_MAX - 1)
+#define TC_CAKE_MAX_TINS (8)
+
+enum {
+	CAKE_FLOW_NONE = 0,
+	CAKE_FLOW_SRC_IP,
+	CAKE_FLOW_DST_IP,
+	CAKE_FLOW_HOSTS,    /* = CAKE_FLOW_SRC_IP | CAKE_FLOW_DST_IP */
+	CAKE_FLOW_FLOWS,
+	CAKE_FLOW_DUAL_SRC, /* = CAKE_FLOW_SRC_IP | CAKE_FLOW_FLOWS */
+	CAKE_FLOW_DUAL_DST, /* = CAKE_FLOW_DST_IP | CAKE_FLOW_FLOWS */
+	CAKE_FLOW_TRIPLE,   /* = CAKE_FLOW_HOSTS  | CAKE_FLOW_FLOWS */
+	CAKE_FLOW_MAX,
+};
+
+enum {
+	CAKE_DIFFSERV_DIFFSERV3 = 0,
+	CAKE_DIFFSERV_DIFFSERV4,
+	CAKE_DIFFSERV_DIFFSERV8,
+	CAKE_DIFFSERV_BESTEFFORT,
+	CAKE_DIFFSERV_PRECEDENCE,
+	CAKE_DIFFSERV_MAX
+};
+
+enum {
+	CAKE_ACK_NONE = 0,
+	CAKE_ACK_FILTER,
+	CAKE_ACK_AGGRESSIVE,
+	CAKE_ACK_MAX
+};
+
+enum {
+	CAKE_ATM_NONE = 0,
+	CAKE_ATM_ATM,
+	CAKE_ATM_PTM,
+	CAKE_ATM_MAX
+};
+
 #endif
diff --git a/net/sched/Kconfig b/net/sched/Kconfig
index a01169fb5325..6e7d614b5757 100644
--- a/net/sched/Kconfig
+++ b/net/sched/Kconfig
@@ -284,6 +284,17 @@ config NET_SCH_FQ_CODEL
 
 	  If unsure, say N.
 
+config NET_SCH_CAKE
+	tristate "Common Applications Kept Enhanced (CAKE)"
+	help
+	  Say Y here if you want to use the Common Applications Kept Enhanced
+          (CAKE) queue management algorithm.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called sch_cake.
+
+	  If unsure, say N.
+
 config NET_SCH_FQ
 	tristate "Fair Queue"
 	help
diff --git a/net/sched/Makefile b/net/sched/Makefile
index 8811d3804878..435054cee32c 100644
--- a/net/sched/Makefile
+++ b/net/sched/Makefile
@@ -50,6 +50,7 @@ obj-$(CONFIG_NET_SCH_CHOKE)	+= sch_choke.o
 obj-$(CONFIG_NET_SCH_QFQ)	+= sch_qfq.o
 obj-$(CONFIG_NET_SCH_CODEL)	+= sch_codel.o
 obj-$(CONFIG_NET_SCH_FQ_CODEL)	+= sch_fq_codel.o
+obj-$(CONFIG_NET_SCH_CAKE)	+= sch_cake.o
 obj-$(CONFIG_NET_SCH_FQ)	+= sch_fq.o
 obj-$(CONFIG_NET_SCH_HHF)	+= sch_hhf.o
 obj-$(CONFIG_NET_SCH_PIE)	+= sch_pie.o
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
new file mode 100644
index 000000000000..8e2f2ba2ed5d
--- /dev/null
+++ b/net/sched/sch_cake.c
@@ -0,0 +1,1683 @@
+// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
+
+/* COMMON Applications Kept Enhanced (CAKE) discipline
+ *
+ * Copyright (C) 2014-2018 Jonathan Morton <chromatix99@gmail.com>
+ * Copyright (C) 2015-2018 Toke Høiland-Jørgensen <toke@toke.dk>
+ * Copyright (C) 2014-2018 Dave Täht <dave.taht@gmail.com>
+ * Copyright (C) 2015-2018 Sebastian Moeller <moeller0@gmx.de>
+ * (C) 2015-2018 Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk>
+ * Copyright (C) 2017 Ryan Mounce <ryan@mounce.com.au>
+ *
+ * The CAKE Principles:
+ *		   (or, how to have your cake and eat it too)
+ *
+ * This is a combination of several shaping, AQM and FQ techniques into one
+ * easy-to-use package:
+ *
+ * - An overall bandwidth shaper, to move the bottleneck away from dumb CPE
+ *   equipment and bloated MACs.  This operates in deficit mode (as in sch_fq),
+ *   eliminating the need for any sort of burst parameter (eg. token bucket
+ *   depth).  Burst support is limited to that necessary to overcome scheduling
+ *   latency.
+ *
+ * - A Diffserv-aware priority queue, giving more priority to certain classes,
+ *   up to a specified fraction of bandwidth.  Above that bandwidth threshold,
+ *   the priority is reduced to avoid starving other tins.
+ *
+ * - Each priority tin has a separate Flow Queue system, to isolate traffic
+ *   flows from each other.  This prevents a burst on one flow from increasing
+ *   the delay to another.  Flows are distributed to queues using a
+ *   set-associative hash function.
+ *
+ * - Each queue is actively managed by Cobalt, which is a combination of the
+ *   Codel and Blue AQM algorithms.  This serves flows fairly, and signals
+ *   congestion early via ECN (if available) and/or packet drops, to keep
+ *   latency low.  The codel parameters are auto-tuned based on the bandwidth
+ *   setting, as is necessary at low bandwidths.
+ *
+ * The configuration parameters are kept deliberately simple for ease of use.
+ * Everything has sane defaults.  Complete generality of configuration is *not*
+ * a goal.
+ *
+ * The priority queue operates according to a weighted DRR scheme, combined with
+ * a bandwidth tracker which reuses the shaper logic to detect which side of the
+ * bandwidth sharing threshold the tin is operating.  This determines whether a
+ * priority-based weight (high) or a bandwidth-based weight (low) is used for
+ * that tin in the current pass.
+ *
+ * This qdisc was inspired by Eric Dumazet's fq_codel code, which he kindly
+ * granted us permission to leverage.
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/jiffies.h>
+#include <linux/string.h>
+#include <linux/in.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/skbuff.h>
+#include <linux/jhash.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/reciprocal_div.h>
+#include <net/netlink.h>
+#include <linux/version.h>
+#include <linux/if_vlan.h>
+#include <net/pkt_sched.h>
+#include <net/tcp.h>
+#include <net/flow_dissector.h>
+
+#define CAKE_SET_WAYS (8)
+#define CAKE_MAX_TINS (8)
+#define CAKE_QUEUES (1024)
+#define CAKE_FLOW_MASK 63
+#define CAKE_FLOW_NAT_FLAG 64
+#define US2TIME(a) (a * (u64)NSEC_PER_USEC)
+
+typedef u64 cobalt_time_t;
+typedef s64 cobalt_tdiff_t;
+
+/**
+ * struct cobalt_params - contains codel and blue parameters
+ * @interval:	codel initial drop rate
+ * @target:     maximum persistent sojourn time & blue update rate
+ * @mtu_time:   serialisation delay of maximum-size packet
+ * @p_inc:      increment of blue drop probability (0.32 fxp)
+ * @p_dec:      decrement of blue drop probability (0.32 fxp)
+ */
+struct cobalt_params {
+	cobalt_time_t	interval;
+	cobalt_time_t	target;
+	cobalt_time_t	mtu_time;
+	u32		p_inc;
+	u32		p_dec;
+};
+
+/* struct cobalt_vars - contains codel and blue variables
+ * @count:	  codel dropping frequency
+ * @rec_inv_sqrt: reciprocal value of sqrt(count) >> 1
+ * @drop_next:    time to drop next packet, or when we dropped last
+ * @blue_timer:	  Blue time to next drop
+ * @p_drop:       BLUE drop probability (0.32 fxp)
+ * @dropping:     set if in dropping state
+ * @ecn_marked:   set if marked
+ */
+struct cobalt_vars {
+	u32		count;
+	u32		rec_inv_sqrt;
+	cobalt_time_t	drop_next;
+	cobalt_time_t	blue_timer;
+	u32     p_drop;
+	bool	dropping;
+	bool    ecn_marked;
+};
+
+enum {
+	CAKE_SET_NONE = 0,
+	CAKE_SET_SPARSE,
+	CAKE_SET_SPARSE_WAIT, /* counted in SPARSE, actually in BULK */
+	CAKE_SET_BULK,
+	CAKE_SET_DECAYING
+};
+
+struct cake_flow {
+	/* this stuff is all needed per-flow at dequeue time */
+	struct sk_buff	  *head;
+	struct sk_buff	  *tail;
+	struct sk_buff	  *ackcheck;
+	struct list_head  flowchain;
+	s32		  deficit;
+	struct cobalt_vars cvars;
+	u16		  srchost; /* index into cake_host table */
+	u16		  dsthost;
+	u8		  set;
+}; /* please try to keep this structure <= 64 bytes */
+
+struct cake_host {
+	u32 srchost_tag;
+	u32 dsthost_tag;
+	u16 srchost_refcnt;
+	u16 dsthost_refcnt;
+};
+
+struct cake_heap_entry {
+	u16 t:3, b:10;
+};
+
+struct cake_tin_data {
+	struct cake_flow flows[CAKE_QUEUES];
+	u32	backlogs[CAKE_QUEUES];
+	u32	tags[CAKE_QUEUES]; /* for set association */
+	u16	overflow_idx[CAKE_QUEUES];
+	struct cake_host hosts[CAKE_QUEUES]; /* for triple isolation */
+	u16	flow_quantum;
+
+	struct cobalt_params cparams;
+	u32	drop_overlimit;
+	u16	bulk_flow_count;
+	u16	sparse_flow_count;
+	u16	decaying_flow_count;
+	u16	unresponsive_flow_count;
+
+	u32	max_skblen;
+
+	struct list_head new_flows;
+	struct list_head old_flows;
+	struct list_head decaying_flows;
+
+	/* time_next = time_this + ((len * rate_ns) >> rate_shft) */
+	u64	tin_time_next_packet;
+	u32	tin_rate_ns;
+	u32	tin_rate_bps;
+	u16	tin_rate_shft;
+
+	u16	tin_quantum_prio;
+	u16	tin_quantum_band;
+	s32	tin_deficit;
+	u32	tin_backlog;
+	u32	tin_dropped;
+	u32	tin_ecn_mark;
+
+	u32	packets;
+	u64	bytes;
+
+	u32	ack_drops;
+
+	/* moving averages */
+	cobalt_time_t avge_delay;
+	cobalt_time_t peak_delay;
+	cobalt_time_t base_delay;
+
+	/* hash function stats */
+	u32	way_directs;
+	u32	way_hits;
+	u32	way_misses;
+	u32	way_collisions;
+}; /* number of tins is small, so size of this struct doesn't matter much */
+
+struct cake_sched_data {
+	struct cake_tin_data *tins;
+
+	struct cake_heap_entry overflow_heap[CAKE_QUEUES * CAKE_MAX_TINS];
+	u16		overflow_timeout;
+
+	u16		tin_cnt;
+	u8		tin_mode;
+	u8		flow_mode;
+	u8		ack_filter;
+	u8		atm_mode;
+
+	/* time_next = time_this + ((len * rate_ns) >> rate_shft) */
+	u16		rate_shft;
+	u64		time_next_packet;
+	u64		failsafe_next_packet;
+	u32		rate_ns;
+	u32		rate_bps;
+	u16		rate_flags;
+	s16		rate_overhead;
+	u16		rate_mpu;
+	u32		interval;
+	u32		target;
+
+	/* resource tracking */
+	u32		buffer_used;
+	u32		buffer_max_used;
+	u32		buffer_limit;
+	u32		buffer_config_limit;
+
+	/* indices for dequeue */
+	u16		cur_tin;
+	u16		cur_flow;
+
+	struct qdisc_watchdog watchdog;
+	const u8	*tin_index;
+	const u8	*tin_order;
+
+	/* bandwidth capacity estimate */
+	u64		last_packet_time;
+	u64		avg_packet_interval;
+	u64		avg_window_begin;
+	u32		avg_window_bytes;
+	u32		avg_peak_bandwidth;
+	u64		last_reconfig_time;
+
+	/* packet length stats */
+	u32 avg_netoff;
+	u16 max_netlen;
+	u16 max_adjlen;
+	u16 min_netlen;
+	u16 min_adjlen;
+};
+
+enum {
+	CAKE_FLAG_OVERHEAD	   = BIT(0),
+	CAKE_FLAG_AUTORATE_INGRESS = BIT(1),
+	CAKE_FLAG_INGRESS	   = BIT(2),
+	CAKE_FLAG_WASH		   = BIT(3),
+	CAKE_FLAG_SPLIT_GSO	   = BIT(4)
+};
+
+/* COBALT operates the Codel and BLUE algorithms in parallel, in order to
+ * obtain the best features of each.  Codel is excellent on flows which
+ * respond to congestion signals in a TCP-like way.  BLUE is more effective on
+ * unresponsive flows.
+ */
+
+struct cobalt_skb_cb {
+	cobalt_time_t enqueue_time;
+};
+
+static cobalt_time_t cobalt_get_time(void)
+{
+	return ktime_get_ns();
+}
+
+static u32 cobalt_time_to_us(cobalt_time_t val)
+{
+	do_div(val, NSEC_PER_USEC);
+	return (u32)val;
+}
+
+static struct cobalt_skb_cb *get_cobalt_cb(const struct sk_buff *skb)
+{
+	qdisc_cb_private_validate(skb, sizeof(struct cobalt_skb_cb));
+	return (struct cobalt_skb_cb *)qdisc_skb_cb(skb)->data;
+}
+
+static cobalt_time_t cobalt_get_enqueue_time(const struct sk_buff *skb)
+{
+	return get_cobalt_cb(skb)->enqueue_time;
+}
+
+static void cobalt_set_enqueue_time(struct sk_buff *skb,
+					   cobalt_time_t now)
+{
+	get_cobalt_cb(skb)->enqueue_time = now;
+}
+
+static u16 quantum_div[CAKE_QUEUES + 1] = {0};
+
+#define REC_INV_SQRT_CACHE (16)
+static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0};
+
+/* http://en.wikipedia.org/wiki/Methods_of_computing_square_roots
+ * new_invsqrt = (invsqrt / 2) * (3 - count * invsqrt^2)
+ *
+ * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32
+ */
+
+static void cobalt_newton_step(struct cobalt_vars *vars)
+{
+	u32 invsqrt = vars->rec_inv_sqrt;
+	u32 invsqrt2 = ((u64)invsqrt * invsqrt) >> 32;
+	u64 val = (3LL << 32) - ((u64)vars->count * invsqrt2);
+
+	val >>= 2; /* avoid overflow in following multiply */
+	val = (val * invsqrt) >> (32 - 2 + 1);
+
+	vars->rec_inv_sqrt = val;
+}
+
+static void cobalt_invsqrt(struct cobalt_vars *vars)
+{
+	if (vars->count < REC_INV_SQRT_CACHE)
+		vars->rec_inv_sqrt = cobalt_rec_inv_sqrt_cache[vars->count];
+	else
+		cobalt_newton_step(vars);
+}
+
+/* There is a big difference in timing between the accurate values placed in
+ * the cache and the approximations given by a single Newton step for small
+ * count values, particularly when stepping from count 1 to 2 or vice versa.
+ * Above 16, a single Newton step gives sufficient accuracy in either
+ * direction, given the precision stored.
+ *
+ * The magnitude of the error when stepping up to count 2 is such as to give
+ * the value that *should* have been produced at count 4.
+ */
+
+static void cobalt_cache_init(void)
+{
+	struct cobalt_vars v;
+
+	memset(&v, 0, sizeof(v));
+	v.rec_inv_sqrt = ~0U;
+	cobalt_rec_inv_sqrt_cache[0] = v.rec_inv_sqrt;
+
+	for (v.count = 1; v.count < REC_INV_SQRT_CACHE; v.count++) {
+		cobalt_newton_step(&v);
+		cobalt_newton_step(&v);
+		cobalt_newton_step(&v);
+		cobalt_newton_step(&v);
+
+		cobalt_rec_inv_sqrt_cache[v.count] = v.rec_inv_sqrt;
+	}
+}
+
+static void cobalt_vars_init(struct cobalt_vars *vars)
+{
+	memset(vars, 0, sizeof(*vars));
+
+	if (!cobalt_rec_inv_sqrt_cache[0]) {
+		cobalt_cache_init();
+		cobalt_rec_inv_sqrt_cache[0] = ~0;
+	}
+}
+
+/* CoDel control_law is t + interval/sqrt(count)
+ * We maintain in rec_inv_sqrt the reciprocal value of sqrt(count) to avoid
+ * both sqrt() and divide operation.
+ */
+static cobalt_time_t cobalt_control(cobalt_time_t t,
+				    cobalt_time_t interval,
+				    u32 rec_inv_sqrt)
+{
+	return t + reciprocal_scale(interval, rec_inv_sqrt);
+}
+
+/* Call this when a packet had to be dropped due to queue overflow.  Returns
+ * true if the BLUE state was quiescent before but active after this call.
+ */
+static bool cobalt_queue_full(struct cobalt_vars *vars,
+			      struct cobalt_params *p,
+			      cobalt_time_t now)
+{
+	bool up = false;
+
+	if ((now - vars->blue_timer) > p->target) {
+		up = !vars->p_drop;
+		vars->p_drop += p->p_inc;
+		if (vars->p_drop < p->p_inc)
+			vars->p_drop = ~0;
+		vars->blue_timer = now;
+	}
+	vars->dropping = true;
+	vars->drop_next = now;
+	if (!vars->count)
+		vars->count = 1;
+
+	return up;
+}
+
+/* Call this when the queue was serviced but turned out to be empty.  Returns
+ * true if the BLUE state was active before but quiescent after this call.
+ */
+static bool cobalt_queue_empty(struct cobalt_vars *vars,
+			       struct cobalt_params *p,
+			       cobalt_time_t now)
+{
+	bool down = false;
+
+	if (vars->p_drop && (now - vars->blue_timer) > p->target) {
+		if (vars->p_drop < p->p_dec)
+			vars->p_drop = 0;
+		else
+			vars->p_drop -= p->p_dec;
+		vars->blue_timer = now;
+		down = !vars->p_drop;
+	}
+	vars->dropping = false;
+
+	if (vars->count && (now - vars->drop_next) >= 0) {
+		vars->count--;
+		cobalt_invsqrt(vars);
+		vars->drop_next = cobalt_control(vars->drop_next,
+						 p->interval,
+						 vars->rec_inv_sqrt);
+	}
+
+	return down;
+}
+
+/* Call this with a freshly dequeued packet for possible congestion marking.
+ * Returns true as an instruction to drop the packet, false for delivery.
+ */
+static bool cobalt_should_drop(struct cobalt_vars *vars,
+			       struct cobalt_params *p,
+			       cobalt_time_t now,
+			       struct sk_buff *skb)
+{
+	bool drop = false;
+
+	/* Simplified Codel implementation */
+	cobalt_tdiff_t sojourn  = now - cobalt_get_enqueue_time(skb);
+
+/* The 'schedule' variable records, in its sign, whether 'now' is before or
+ * after 'drop_next'.  This allows 'drop_next' to be updated before the next
+ * scheduling decision is actually branched, without destroying that
+ * information.  Similarly, the first 'schedule' value calculated is preserved
+ * in the boolean 'next_due'.
+ *
+ * As for 'drop_next', we take advantage of the fact that 'interval' is both
+ * the delay between first exceeding 'target' and the first signalling event,
+ * *and* the scaling factor for the signalling frequency.  It's therefore very
+ * natural to use a single mechanism for both purposes, and eliminates a
+ * significant amount of reference Codel's spaghetti code.  To help with this,
+ * both the '0' and '1' entries in the invsqrt cache are 0xFFFFFFFF, as close
+ * as possible to 1.0 in fixed-point.
+ */
+
+	cobalt_tdiff_t schedule = now - vars->drop_next;
+
+	bool over_target = sojourn > p->target &&
+			   sojourn > p->mtu_time * 4;
+	bool next_due    = vars->count && schedule >= 0;
+
+	vars->ecn_marked = false;
+
+	if (over_target) {
+		if (!vars->dropping) {
+			vars->dropping = true;
+			vars->drop_next = cobalt_control(now,
+							 p->interval,
+							 vars->rec_inv_sqrt);
+		}
+		if (!vars->count)
+			vars->count = 1;
+	} else if (vars->dropping) {
+		vars->dropping = false;
+	}
+
+	if (next_due && vars->dropping) {
+		/* Use ECN mark if possible, otherwise drop */
+		drop = !(vars->ecn_marked = INET_ECN_set_ce(skb));
+
+		vars->count++;
+		if (!vars->count)
+			vars->count--;
+		cobalt_invsqrt(vars);
+		vars->drop_next = cobalt_control(vars->drop_next,
+						 p->interval,
+						 vars->rec_inv_sqrt);
+		schedule = now - vars->drop_next;
+	} else {
+		while (next_due) {
+			vars->count--;
+			cobalt_invsqrt(vars);
+			vars->drop_next = cobalt_control(vars->drop_next,
+							 p->interval,
+							 vars->rec_inv_sqrt);
+			schedule = now - vars->drop_next;
+			next_due = vars->count && schedule >= 0;
+		}
+	}
+
+	/* Simple BLUE implementation.  Lack of ECN is deliberate. */
+	if (vars->p_drop)
+		drop |= (prandom_u32() < vars->p_drop);
+
+	/* Overload the drop_next field as an activity timeout */
+	if (!vars->count)
+		vars->drop_next = now + p->interval;
+	else if (schedule > 0 && !drop)
+		vars->drop_next = now;
+
+	return drop;
+}
+
+/* Cake has several subtle multiple bit settings. In these cases you
+ *  would be matching triple isolate mode as well.
+ */
+
+static bool cake_dsrc(int flow_mode)
+{
+	return (flow_mode & CAKE_FLOW_DUAL_SRC) == CAKE_FLOW_DUAL_SRC;
+}
+
+static bool cake_ddst(int flow_mode)
+{
+	return (flow_mode & CAKE_FLOW_DUAL_DST) == CAKE_FLOW_DUAL_DST;
+}
+
+static u32 cake_hash(struct cake_tin_data *q, const struct sk_buff *skb,
+		     int flow_mode)
+{
+	struct flow_keys keys, host_keys;
+	u32 flow_hash = 0, srchost_hash, dsthost_hash;
+	u16 reduced_hash, srchost_idx, dsthost_idx;
+
+	if (unlikely(flow_mode == CAKE_FLOW_NONE))
+		return 0;
+
+	skb_flow_dissect_flow_keys(skb, &keys,
+				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
+
+	/* flow_hash_from_keys() sorts the addresses by value, so we have
+	 * to preserve their order in a separate data structure to treat
+	 * src and dst host addresses as independently selectable.
+	 */
+	host_keys = keys;
+	host_keys.ports.ports     = 0;
+	host_keys.basic.ip_proto  = 0;
+	host_keys.keyid.keyid     = 0;
+	host_keys.tags.flow_label = 0;
+
+	switch (host_keys.control.addr_type) {
+	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
+		host_keys.addrs.v4addrs.src = 0;
+		dsthost_hash = flow_hash_from_keys(&host_keys);
+		host_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
+		host_keys.addrs.v4addrs.dst = 0;
+		srchost_hash = flow_hash_from_keys(&host_keys);
+		break;
+
+	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
+		memset(&host_keys.addrs.v6addrs.src, 0,
+		       sizeof(host_keys.addrs.v6addrs.src));
+		dsthost_hash = flow_hash_from_keys(&host_keys);
+		host_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
+		memset(&host_keys.addrs.v6addrs.dst, 0,
+		       sizeof(host_keys.addrs.v6addrs.dst));
+		srchost_hash = flow_hash_from_keys(&host_keys);
+		break;
+
+	default:
+		dsthost_hash = 0;
+		srchost_hash = 0;
+	}
+
+	/* This *must* be after the above switch, since as a
+	 * side-effect it sorts the src and dst addresses.
+	 */
+	if (flow_mode & CAKE_FLOW_FLOWS)
+		flow_hash = flow_hash_from_keys(&keys);
+
+	if (!(flow_mode & CAKE_FLOW_FLOWS)) {
+		if (flow_mode & CAKE_FLOW_SRC_IP)
+			flow_hash ^= srchost_hash;
+
+		if (flow_mode & CAKE_FLOW_DST_IP)
+			flow_hash ^= dsthost_hash;
+	}
+
+	reduced_hash = flow_hash % CAKE_QUEUES;
+
+	/* set-associative hashing */
+	/* fast path if no hash collision (direct lookup succeeds) */
+	if (likely(q->tags[reduced_hash] == flow_hash &&
+		   q->flows[reduced_hash].set)) {
+		q->way_directs++;
+	} else {
+		u32 inner_hash = reduced_hash % CAKE_SET_WAYS;
+		u32 outer_hash = reduced_hash - inner_hash;
+		u32 i, k;
+		bool allocate_src = false;
+		bool allocate_dst = false;
+
+		/* check if any active queue in the set is reserved for
+		 * this flow.
+		 */
+		for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
+		     i++, k = (k + 1) % CAKE_SET_WAYS) {
+			if (q->tags[outer_hash + k] == flow_hash) {
+				if (i)
+					q->way_hits++;
+
+				if (!q->flows[outer_hash + k].set) {
+					/* need to increment host refcnts */
+					allocate_src = cake_dsrc(flow_mode);
+					allocate_dst = cake_ddst(flow_mode);
+				}
+
+				goto found;
+			}
+		}
+
+		/* no queue is reserved for this flow, look for an
+		 * empty one.
+		 */
+		for (i = 0; i < CAKE_SET_WAYS;
+			 i++, k = (k + 1) % CAKE_SET_WAYS) {
+			if (!q->flows[outer_hash + k].set) {
+				q->way_misses++;
+				allocate_src = cake_dsrc(flow_mode);
+				allocate_dst = cake_ddst(flow_mode);
+				goto found;
+			}
+		}
+
+		/* With no empty queues, default to the original
+		 * queue, accept the collision, update the host tags.
+		 */
+		q->way_collisions++;
+		q->hosts[q->flows[reduced_hash].srchost].srchost_refcnt--;
+		q->hosts[q->flows[reduced_hash].dsthost].dsthost_refcnt--;
+		allocate_src = cake_dsrc(flow_mode);
+		allocate_dst = cake_ddst(flow_mode);
+found:
+		/* reserve queue for future packets in same flow */
+		reduced_hash = outer_hash + k;
+		q->tags[reduced_hash] = flow_hash;
+
+		if (allocate_src) {
+			srchost_idx = srchost_hash % CAKE_QUEUES;
+			inner_hash = srchost_idx % CAKE_SET_WAYS;
+			outer_hash = srchost_idx - inner_hash;
+			for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
+				i++, k = (k + 1) % CAKE_SET_WAYS) {
+				if (q->hosts[outer_hash + k].srchost_tag ==
+				    srchost_hash)
+					goto found_src;
+			}
+			for (i = 0; i < CAKE_SET_WAYS;
+				i++, k = (k + 1) % CAKE_SET_WAYS) {
+				if (!q->hosts[outer_hash + k].srchost_refcnt)
+					break;
+			}
+			q->hosts[outer_hash + k].srchost_tag = srchost_hash;
+found_src:
+			srchost_idx = outer_hash + k;
+			q->hosts[srchost_idx].srchost_refcnt++;
+			q->flows[reduced_hash].srchost = srchost_idx;
+		}
+
+		if (allocate_dst) {
+			dsthost_idx = dsthost_hash % CAKE_QUEUES;
+			inner_hash = dsthost_idx % CAKE_SET_WAYS;
+			outer_hash = dsthost_idx - inner_hash;
+			for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
+			     i++, k = (k + 1) % CAKE_SET_WAYS) {
+				if (q->hosts[outer_hash + k].dsthost_tag ==
+				    dsthost_hash)
+					goto found_dst;
+			}
+			for (i = 0; i < CAKE_SET_WAYS;
+			     i++, k = (k + 1) % CAKE_SET_WAYS) {
+				if (!q->hosts[outer_hash + k].dsthost_refcnt)
+					break;
+			}
+			q->hosts[outer_hash + k].dsthost_tag = dsthost_hash;
+found_dst:
+			dsthost_idx = outer_hash + k;
+			q->hosts[dsthost_idx].dsthost_refcnt++;
+			q->flows[reduced_hash].dsthost = dsthost_idx;
+		}
+	}
+
+	return reduced_hash;
+}
+
+/* helper functions : might be changed when/if skb use a standard list_head */
+/* remove one skb from head of slot queue */
+
+static struct sk_buff *dequeue_head(struct cake_flow *flow)
+{
+	struct sk_buff *skb = flow->head;
+
+	if (skb) {
+		flow->head = skb->next;
+		skb->next = NULL;
+
+		if (skb == flow->ackcheck)
+			flow->ackcheck = NULL;
+	}
+
+	return skb;
+}
+
+/* add skb to flow queue (tail add) */
+
+static void flow_queue_add(struct cake_flow *flow, struct sk_buff *skb)
+{
+	if (!flow->head)
+		flow->head = skb;
+	else
+		flow->tail->next = skb;
+	flow->tail = skb;
+	skb->next = NULL;
+}
+
+static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
+				      u32 shift)
+{
+	avg -= avg >> shift;
+	avg += sample >> shift;
+	return avg;
+}
+
+static void cake_heap_swap(struct cake_sched_data *q, u16 i, u16 j)
+{
+	struct cake_heap_entry ii = q->overflow_heap[i];
+	struct cake_heap_entry jj = q->overflow_heap[j];
+
+	q->overflow_heap[i] = jj;
+	q->overflow_heap[j] = ii;
+
+	q->tins[ii.t].overflow_idx[ii.b] = j;
+	q->tins[jj.t].overflow_idx[jj.b] = i;
+}
+
+static u32 cake_heap_get_backlog(const struct cake_sched_data *q, u16 i)
+{
+	struct cake_heap_entry ii = q->overflow_heap[i];
+
+	return q->tins[ii.t].backlogs[ii.b];
+}
+
+static void cake_heapify(struct cake_sched_data *q, u16 i)
+{
+	static const u32 a = CAKE_MAX_TINS * CAKE_QUEUES;
+	u32 m = i;
+	u32 mb = cake_heap_get_backlog(q, m);
+
+	while (m < a) {
+		u32 l = m + m + 1;
+		u32 r = l + 1;
+
+		if (l < a) {
+			u32 lb = cake_heap_get_backlog(q, l);
+
+			if (lb > mb) {
+				m  = l;
+				mb = lb;
+			}
+		}
+
+		if (r < a) {
+			u32 rb = cake_heap_get_backlog(q, r);
+
+			if (rb > mb) {
+				m  = r;
+				mb = rb;
+			}
+		}
+
+		if (m != i) {
+			cake_heap_swap(q, i, m);
+			i = m;
+		} else {
+			break;
+		}
+	}
+}
+
+static void cake_heapify_up(struct cake_sched_data *q, u16 i)
+{
+	while (i > 0 && i < CAKE_MAX_TINS * CAKE_QUEUES) {
+		u16 p = (i - 1) >> 1;
+		u32 ib = cake_heap_get_backlog(q, i);
+		u32 pb = cake_heap_get_backlog(q, p);
+
+		if (ib > pb) {
+			cake_heap_swap(q, i, p);
+			i = p;
+		} else {
+			break;
+		}
+	}
+}
+
+static int cake_advance_shaper(struct cake_sched_data *q,
+			       struct cake_tin_data *b,
+			       struct sk_buff *skb,
+			       u64 now, bool drop)
+{
+	u32 len = qdisc_pkt_len(skb);
+
+	/* charge packet bandwidth to this tin
+	 * and to the global shaper.
+	 */
+	if (q->rate_ns) {
+		s64 tdiff1 = b->tin_time_next_packet - now;
+		s64 tdiff2 = (len * (u64)b->tin_rate_ns) >> b->tin_rate_shft;
+		s64 tdiff3 = (len * (u64)q->rate_ns) >> q->rate_shft;
+		s64 tdiff4 = tdiff3 + (tdiff3 >> 1);
+
+		if (tdiff1 < 0)
+			b->tin_time_next_packet += tdiff2;
+		else if (tdiff1 < tdiff2)
+			b->tin_time_next_packet = now + tdiff2;
+
+		q->time_next_packet += tdiff3;
+		if (!drop)
+			q->failsafe_next_packet += tdiff4;
+	}
+	return len;
+}
+
+static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct sk_buff *skb;
+	u32 idx = 0, tin = 0, len;
+	struct cake_tin_data *b;
+	struct cake_flow *flow;
+	struct cake_heap_entry qq;
+	u64 now = cobalt_get_time();
+
+	if (!q->overflow_timeout) {
+		int i;
+		/* Build fresh max-heap */
+		for (i = CAKE_MAX_TINS * CAKE_QUEUES / 2; i >= 0; i--)
+			cake_heapify(q, i);
+	}
+	q->overflow_timeout = 65535;
+
+	/* select longest queue for pruning */
+	qq  = q->overflow_heap[0];
+	tin = qq.t;
+	idx = qq.b;
+
+	b = &q->tins[tin];
+	flow = &b->flows[idx];
+	skb = dequeue_head(flow);
+	if (unlikely(!skb)) {
+		/* heap has gone wrong, rebuild it next time */
+		q->overflow_timeout = 0;
+		return idx + (tin << 16);
+	}
+
+	if (cobalt_queue_full(&flow->cvars, &b->cparams, now))
+		b->unresponsive_flow_count++;
+
+	len = qdisc_pkt_len(skb);
+	q->buffer_used      -= skb->truesize;
+	b->backlogs[idx]    -= len;
+	b->tin_backlog      -= len;
+	sch->qstats.backlog -= len;
+	qdisc_tree_reduce_backlog(sch, 1, len);
+
+	b->tin_dropped++;
+	sch->qstats.drops++;
+
+	__qdisc_drop(skb, to_free);
+	sch->q.qlen--;
+
+	cake_heapify(q, 0);
+
+	return idx + (tin << 16);
+}
+
+static void cake_reconfigure(struct Qdisc *sch);
+
+static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
+			struct sk_buff **to_free)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	u32 idx, tin;
+	struct cake_tin_data *b;
+	struct cake_flow *flow;
+	/* signed len to handle corner case filtered ACK larger than trigger */
+	int len = qdisc_pkt_len(skb);
+	u64 now = cobalt_get_time();
+
+	tin = 0;
+	b = &q->tins[tin];
+
+	/* choose flow to insert into */
+	idx = cake_hash(b, skb, q->flow_mode);
+	flow = &b->flows[idx];
+
+	/* ensure shaper state isn't stale */
+	if (!b->tin_backlog) {
+		if (b->tin_time_next_packet < now)
+			b->tin_time_next_packet = now;
+
+		if (!sch->q.qlen) {
+			if (q->time_next_packet < now) {
+				q->failsafe_next_packet = now;
+				q->time_next_packet = now;
+			} else if (q->time_next_packet > now &&
+				   q->failsafe_next_packet > now) {
+				u64 next = min(q->time_next_packet,
+					       q->failsafe_next_packet);
+				sch->qstats.overlimits++;
+				qdisc_watchdog_schedule_ns(&q->watchdog, next);
+			}
+		}
+	}
+
+	if (unlikely(len > b->max_skblen))
+		b->max_skblen = len;
+
+	cobalt_set_enqueue_time(skb, now);
+	flow_queue_add(flow, skb);
+
+	sch->q.qlen++;
+	q->buffer_used      += skb->truesize;
+
+	/* stats */
+	b->packets++;
+	b->bytes	    += len;
+	b->backlogs[idx]    += len;
+	b->tin_backlog      += len;
+	sch->qstats.backlog += len;
+	q->avg_window_bytes += len;
+
+	if (q->overflow_timeout)
+		cake_heapify_up(q, b->overflow_idx[idx]);
+
+	/* incoming bandwidth capacity estimate */
+	q->avg_window_bytes = 0;
+	q->last_packet_time = now;
+
+	/* flowchain */
+	if (!flow->set || flow->set == CAKE_SET_DECAYING) {
+		struct cake_host *srchost = &b->hosts[flow->srchost];
+		struct cake_host *dsthost = &b->hosts[flow->dsthost];
+		u16 host_load = 1;
+
+		if (!flow->set) {
+			list_add_tail(&flow->flowchain, &b->new_flows);
+		} else {
+			b->decaying_flow_count--;
+			list_move_tail(&flow->flowchain, &b->new_flows);
+		}
+		flow->set = CAKE_SET_SPARSE;
+		b->sparse_flow_count++;
+
+		if (cake_dsrc(q->flow_mode))
+			host_load = max(host_load, srchost->srchost_refcnt);
+
+		if (cake_ddst(q->flow_mode))
+			host_load = max(host_load, dsthost->dsthost_refcnt);
+
+		flow->deficit = (b->flow_quantum *
+				 quantum_div[host_load]) >> 16;
+	} else if (flow->set == CAKE_SET_SPARSE_WAIT) {
+		/* this flow was empty, accounted as a sparse flow, but actually
+		 * in the bulk rotation.
+		 */
+		flow->set = CAKE_SET_BULK;
+		b->sparse_flow_count--;
+		b->bulk_flow_count++;
+	}
+
+	if (q->buffer_used > q->buffer_max_used)
+		q->buffer_max_used = q->buffer_used;
+
+	if (q->buffer_used > q->buffer_limit) {
+		u32 dropped = 0;
+
+		while (q->buffer_used > q->buffer_limit) {
+			dropped++;
+			cake_drop(sch, to_free);
+		}
+		b->drop_overlimit += dropped;
+	}
+	return NET_XMIT_SUCCESS;
+}
+
+static struct sk_buff *cake_dequeue_one(struct Qdisc *sch)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct cake_tin_data *b = &q->tins[q->cur_tin];
+	struct cake_flow *flow = &b->flows[q->cur_flow];
+	struct sk_buff *skb = NULL;
+	u32 len;
+
+	if (flow->head) {
+		skb = dequeue_head(flow);
+		len = qdisc_pkt_len(skb);
+		b->backlogs[q->cur_flow] -= len;
+		b->tin_backlog		 -= len;
+		sch->qstats.backlog      -= len;
+		q->buffer_used		 -= skb->truesize;
+		sch->q.qlen--;
+
+		if (q->overflow_timeout)
+			cake_heapify(q, b->overflow_idx[q->cur_flow]);
+	}
+	return skb;
+}
+
+/* Discard leftover packets from a tin no longer in use. */
+static void cake_clear_tin(struct Qdisc *sch, u16 tin)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct sk_buff *skb;
+
+	q->cur_tin = tin;
+	for (q->cur_flow = 0; q->cur_flow < CAKE_QUEUES; q->cur_flow++)
+		while (!!(skb = cake_dequeue_one(sch)))
+			kfree_skb(skb);
+}
+
+static struct sk_buff *cake_dequeue(struct Qdisc *sch)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct sk_buff *skb;
+	struct cake_tin_data *b = &q->tins[q->cur_tin];
+	struct cake_flow *flow;
+	struct cake_host *srchost, *dsthost;
+	struct list_head *head;
+	u32 len;
+	u16 host_load;
+	cobalt_time_t now = ktime_get_ns();
+	cobalt_time_t delay;
+	bool first_flow = true;
+
+begin:
+	if (!sch->q.qlen)
+		return NULL;
+
+	/* global hard shaper */
+	if (q->time_next_packet > now && q->failsafe_next_packet > now) {
+		u64 next = min(q->time_next_packet, q->failsafe_next_packet);
+
+		sch->qstats.overlimits++;
+		qdisc_watchdog_schedule_ns(&q->watchdog, next);
+		return NULL;
+	}
+
+	/* Choose a class to work on. */
+	if (!q->rate_ns) {
+		/* In unlimited mode, can't rely on shaper timings, just balance
+		 * with DRR
+		 */
+		while (b->tin_deficit < 0 ||
+		       !(b->sparse_flow_count + b->bulk_flow_count)) {
+			if (b->tin_deficit <= 0)
+				b->tin_deficit += b->tin_quantum_band;
+
+			q->cur_tin++;
+			b++;
+			if (q->cur_tin >= q->tin_cnt) {
+				q->cur_tin = 0;
+				b = q->tins;
+			}
+		}
+	} else {
+		/* In shaped mode, choose:
+		 * - Highest-priority tin with queue and meeting schedule, or
+		 * - The earliest-scheduled tin with queue.
+		 */
+		int tin, best_tin = 0;
+		s64 best_time = 0xFFFFFFFFFFFFUL;
+
+		for (tin = 0; tin < q->tin_cnt; tin++) {
+			b = q->tins + tin;
+			if ((b->sparse_flow_count + b->bulk_flow_count) > 0) {
+				s64 tdiff = b->tin_time_next_packet - now;
+
+				if (tdiff <= 0 || tdiff <= best_time) {
+					best_time = tdiff;
+					best_tin = tin;
+				}
+			}
+		}
+
+		q->cur_tin = best_tin;
+		b = q->tins + best_tin;
+	}
+
+retry:
+	/* service this class */
+	head = &b->decaying_flows;
+	if (!first_flow || list_empty(head)) {
+		head = &b->new_flows;
+		if (list_empty(head)) {
+			head = &b->old_flows;
+			if (unlikely(list_empty(head))) {
+				head = &b->decaying_flows;
+				if (unlikely(list_empty(head)))
+					goto begin;
+			}
+		}
+	}
+	flow = list_first_entry(head, struct cake_flow, flowchain);
+	q->cur_flow = flow - b->flows;
+	first_flow = false;
+
+	/* triple isolation (modified DRR++) */
+	srchost = &b->hosts[flow->srchost];
+	dsthost = &b->hosts[flow->dsthost];
+	host_load = 1;
+
+	if (cake_dsrc(q->flow_mode))
+		host_load = max(host_load, srchost->srchost_refcnt);
+
+	if (cake_ddst(q->flow_mode))
+		host_load = max(host_load, dsthost->dsthost_refcnt);
+
+	WARN_ON(host_load > CAKE_QUEUES);
+
+	/* flow isolation (DRR++) */
+	if (flow->deficit <= 0) {
+		/* The shifted prandom_u32() is a way to apply dithering to
+		 * avoid accumulating roundoff errors
+		 */
+		flow->deficit += (b->flow_quantum * quantum_div[host_load] +
+				  (prandom_u32() >> 16)) >> 16;
+		list_move_tail(&flow->flowchain, &b->old_flows);
+
+		/* Keep all flows with deficits out of the sparse and decaying
+		 * rotations.  No non-empty flow can go into the decaying
+		 * rotation, so they can't get deficits
+		 */
+		if (flow->set == CAKE_SET_SPARSE) {
+			if (flow->head) {
+				b->sparse_flow_count--;
+				b->bulk_flow_count++;
+				flow->set = CAKE_SET_BULK;
+			} else {
+				/* we've moved it to the bulk rotation for
+				 * correct deficit accounting but we still want
+				 * to count it as a sparse flow, not a bulk one.
+				 */
+				flow->set = CAKE_SET_SPARSE_WAIT;
+			}
+		}
+		goto retry;
+	}
+
+	/* Retrieve a packet via the AQM */
+	while (1) {
+		skb = cake_dequeue_one(sch);
+		if (!skb) {
+			/* this queue was actually empty */
+			if (cobalt_queue_empty(&flow->cvars, &b->cparams, now))
+				b->unresponsive_flow_count--;
+
+			if (flow->cvars.p_drop || flow->cvars.count ||
+			    now < flow->cvars.drop_next) {
+				/* keep in the flowchain until the state has
+				 * decayed to rest
+				 */
+				list_move_tail(&flow->flowchain,
+					       &b->decaying_flows);
+				if (flow->set == CAKE_SET_BULK) {
+					b->bulk_flow_count--;
+					b->decaying_flow_count++;
+				} else if (flow->set == CAKE_SET_SPARSE ||
+					   flow->set == CAKE_SET_SPARSE_WAIT) {
+					b->sparse_flow_count--;
+					b->decaying_flow_count++;
+				}
+				flow->set = CAKE_SET_DECAYING;
+			} else {
+				/* remove empty queue from the flowchain */
+				list_del_init(&flow->flowchain);
+				if (flow->set == CAKE_SET_SPARSE ||
+				    flow->set == CAKE_SET_SPARSE_WAIT)
+					b->sparse_flow_count--;
+				else if (flow->set == CAKE_SET_BULK)
+					b->bulk_flow_count--;
+				else
+					b->decaying_flow_count--;
+
+				flow->set = CAKE_SET_NONE;
+				srchost->srchost_refcnt--;
+				dsthost->dsthost_refcnt--;
+			}
+			goto begin;
+		}
+
+		/* Last packet in queue may be marked, shouldn't be dropped */
+		if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb) ||
+		    !flow->head)
+			break;
+
+		b->tin_dropped++;
+		qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb));
+		qdisc_qstats_drop(sch);
+		kfree_skb(skb);
+	}
+
+	b->tin_ecn_mark += !!flow->cvars.ecn_marked;
+	qdisc_bstats_update(sch, skb);
+
+	/* collect delay stats */
+	delay = now - cobalt_get_enqueue_time(skb);
+	b->avge_delay = cake_ewma(b->avge_delay, delay, 8);
+	b->peak_delay = cake_ewma(b->peak_delay, delay,
+				  delay > b->peak_delay ? 2 : 8);
+	b->base_delay = cake_ewma(b->base_delay, delay,
+				  delay < b->base_delay ? 2 : 8);
+
+	len = cake_advance_shaper(q, b, skb, now, false);
+	flow->deficit -= len;
+	b->tin_deficit -= len;
+
+	if (q->time_next_packet > now && sch->q.qlen) {
+		u64 next = min(q->time_next_packet, q->failsafe_next_packet);
+
+		qdisc_watchdog_schedule_ns(&q->watchdog, next);
+	} else if (!sch->q.qlen) {
+		int i;
+
+		for (i = 0; i < q->tin_cnt; i++) {
+			if (q->tins[i].decaying_flow_count) {
+				u64 next = now + q->tins[i].cparams.target;
+
+				qdisc_watchdog_schedule_ns(&q->watchdog, next);
+				break;
+			}
+		}
+	}
+
+	if (q->overflow_timeout)
+		q->overflow_timeout--;
+
+	return skb;
+}
+
+static void cake_reset(struct Qdisc *sch)
+{
+	u32 c;
+
+	for (c = 0; c < CAKE_MAX_TINS; c++)
+		cake_clear_tin(sch, c);
+}
+
+static const struct nla_policy cake_policy[TCA_CAKE_MAX + 1] = {
+	[TCA_CAKE_BASE_RATE]     = { .type = NLA_U32 },
+	[TCA_CAKE_DIFFSERV_MODE] = { .type = NLA_U32 },
+	[TCA_CAKE_ATM]		 = { .type = NLA_U32 },
+	[TCA_CAKE_FLOW_MODE]     = { .type = NLA_U32 },
+	[TCA_CAKE_OVERHEAD]      = { .type = NLA_S32 },
+	[TCA_CAKE_RTT]		 = { .type = NLA_U32 },
+	[TCA_CAKE_TARGET]	 = { .type = NLA_U32 },
+	[TCA_CAKE_AUTORATE]      = { .type = NLA_U32 },
+	[TCA_CAKE_MEMORY]	 = { .type = NLA_U32 },
+	[TCA_CAKE_NAT]		 = { .type = NLA_U32 },
+	[TCA_CAKE_RAW]		 = { .type = NLA_U32 },
+	[TCA_CAKE_WASH]		 = { .type = NLA_U32 },
+	[TCA_CAKE_MPU]		 = { .type = NLA_U32 },
+	[TCA_CAKE_INGRESS]	 = { .type = NLA_U32 },
+	[TCA_CAKE_ACK_FILTER]	 = { .type = NLA_U32 },
+};
+
+static void cake_set_rate(struct cake_tin_data *b, u64 rate, u32 mtu,
+			  cobalt_time_t ns_target, cobalt_time_t rtt_est_ns)
+{
+	/* convert byte-rate into time-per-byte
+	 * so it will always unwedge in reasonable time.
+	 */
+	static const u64 MIN_RATE = 64;
+	u64 rate_ns = 0;
+	u8  rate_shft = 0;
+	cobalt_time_t byte_target_ns;
+	u32 byte_target = mtu;
+
+	b->flow_quantum = 1514;
+	if (rate) {
+		b->flow_quantum = max(min(rate >> 12, 1514ULL), 300ULL);
+		rate_shft = 32;
+		rate_ns = ((u64)NSEC_PER_SEC) << rate_shft;
+		do_div(rate_ns, max(MIN_RATE, rate));
+		while (!!(rate_ns >> 32)) {
+			rate_ns >>= 1;
+			rate_shft--;
+		}
+	} /* else unlimited, ie. zero delay */
+
+	b->tin_rate_bps  = rate;
+	b->tin_rate_ns   = rate_ns;
+	b->tin_rate_shft = rate_shft;
+
+	byte_target_ns = (byte_target * rate_ns) >> rate_shft;
+
+	b->cparams.target = max((byte_target_ns * 3) / 2, ns_target);
+	b->cparams.interval = max(rtt_est_ns +
+				     b->cparams.target - ns_target,
+				     b->cparams.target * 2);
+	b->cparams.mtu_time = byte_target_ns;
+	b->cparams.p_inc = 1 << 24; /* 1/256 */
+	b->cparams.p_dec = 1 << 20; /* 1/4096 */
+}
+
+static void cake_reconfigure(struct Qdisc *sch)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct cake_tin_data *b = &q->tins[0];
+	int c, ft = 0;
+
+	q->tin_cnt = 1;
+	cake_set_rate(b, q->rate_bps, psched_mtu(qdisc_dev(sch)),
+		      US2TIME(q->target), US2TIME(q->interval));
+	b->tin_quantum_band = 65535;
+	b->tin_quantum_prio = 65535;
+
+	for (c = q->tin_cnt; c < CAKE_MAX_TINS; c++) {
+		cake_clear_tin(sch, c);
+		q->tins[c].cparams.mtu_time = q->tins[ft].cparams.mtu_time;
+	}
+
+	q->rate_ns   = q->tins[ft].tin_rate_ns;
+	q->rate_shft = q->tins[ft].tin_rate_shft;
+
+	if (q->buffer_config_limit) {
+		q->buffer_limit = q->buffer_config_limit;
+	} else if (q->rate_bps) {
+		u64 t = (u64)q->rate_bps * q->interval;
+
+		do_div(t, USEC_PER_SEC / 4);
+		q->buffer_limit = max_t(u32, t, 4U << 20);
+	} else {
+		q->buffer_limit = ~0;
+	}
+
+	sch->flags &= ~TCQ_F_CAN_BYPASS;
+
+	q->buffer_limit = min(q->buffer_limit,
+			      max(sch->limit * psched_mtu(qdisc_dev(sch)),
+				  q->buffer_config_limit));
+}
+
+static int cake_change(struct Qdisc *sch, struct nlattr *opt,
+		       struct netlink_ext_ack *extack)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct nlattr *tb[TCA_CAKE_MAX + 1];
+	int err;
+
+	if (!opt)
+		return -EINVAL;
+
+	err = nla_parse_nested(tb, TCA_CAKE_MAX, opt, cake_policy, extack);
+	if (err < 0)
+		return err;
+
+	if (tb[TCA_CAKE_BASE_RATE])
+		q->rate_bps = nla_get_u32(tb[TCA_CAKE_BASE_RATE]);
+
+	if (tb[TCA_CAKE_FLOW_MODE])
+		q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
+				CAKE_FLOW_MASK);
+
+	if (tb[TCA_CAKE_RTT]) {
+		q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
+
+		if (!q->interval)
+			q->interval = 1;
+	}
+
+	if (tb[TCA_CAKE_TARGET]) {
+		q->target = nla_get_u32(tb[TCA_CAKE_TARGET]);
+
+		if (!q->target)
+			q->target = 1;
+	}
+
+	if (tb[TCA_CAKE_MEMORY])
+		q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
+
+	if (q->tins) {
+		sch_tree_lock(sch);
+		cake_reconfigure(sch);
+		sch_tree_unlock(sch);
+	}
+
+	return 0;
+}
+
+static void cake_free(void *addr)
+{
+	if (addr)
+		kvfree(addr);
+}
+
+static void cake_destroy(struct Qdisc *sch)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+
+	qdisc_watchdog_cancel(&q->watchdog);
+
+	if (q->tins)
+		cake_free(q->tins);
+}
+
+static int cake_init(struct Qdisc *sch, struct nlattr *opt,
+		     struct netlink_ext_ack *extack)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	int i, j;
+
+	sch->limit = 10240;
+	q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
+	q->flow_mode  = CAKE_FLOW_TRIPLE;
+
+	q->rate_bps = 0; /* unlimited by default */
+
+	q->interval = 100000; /* 100ms default */
+	q->target   =   5000; /* 5ms: codel RFC argues
+			       * for 5 to 10% of interval
+			       */
+
+	q->cur_tin = 0;
+	q->cur_flow  = 0;
+
+	if (opt) {
+		int err = cake_change(sch, opt, extack);
+
+		if (err)
+			return err;
+	}
+
+	qdisc_watchdog_init(&q->watchdog, sch);
+
+	quantum_div[0] = ~0;
+	for (i = 1; i <= CAKE_QUEUES; i++)
+		quantum_div[i] = 65535 / i;
+
+	q->tins = kvzalloc(CAKE_MAX_TINS * sizeof(struct cake_tin_data),
+			   GFP_KERNEL | __GFP_NOWARN);
+	if (!q->tins)
+		goto nomem;
+
+	for (i = 0; i < CAKE_MAX_TINS; i++) {
+		struct cake_tin_data *b = q->tins + i;
+
+		INIT_LIST_HEAD(&b->new_flows);
+		INIT_LIST_HEAD(&b->old_flows);
+		INIT_LIST_HEAD(&b->decaying_flows);
+		b->sparse_flow_count = 0;
+		b->bulk_flow_count = 0;
+		b->decaying_flow_count = 0;
+
+		for (j = 0; j < CAKE_QUEUES; j++) {
+			struct cake_flow *flow = b->flows + j;
+			u32 k = j * CAKE_MAX_TINS + i;
+
+			INIT_LIST_HEAD(&flow->flowchain);
+			cobalt_vars_init(&flow->cvars);
+
+			q->overflow_heap[k].t = i;
+			q->overflow_heap[k].b = j;
+			b->overflow_idx[j] = k;
+		}
+	}
+
+	cake_reconfigure(sch);
+	q->avg_peak_bandwidth = q->rate_bps;
+	q->min_netlen = ~0;
+	q->min_adjlen = ~0;
+	return 0;
+
+nomem:
+	cake_destroy(sch);
+	return -ENOMEM;
+}
+
+static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct nlattr *opts;
+
+	opts = nla_nest_start(skb, TCA_OPTIONS);
+	if (!opts)
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_BASE_RATE, q->rate_bps))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_DIFFSERV_MODE, q->tin_mode))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_ATM, q->atm_mode))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_FLOW_MODE,
+			q->flow_mode & CAKE_FLOW_MASK))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_NAT,
+			!!(q->flow_mode & CAKE_FLOW_NAT_FLAG)))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_SPLIT_GSO,
+			!!(q->rate_flags & CAKE_FLAG_SPLIT_GSO)))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_WASH,
+			!!(q->rate_flags & CAKE_FLAG_WASH)))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_OVERHEAD, q->rate_overhead))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_MPU, q->rate_mpu))
+		goto nla_put_failure;
+
+	if (!(q->rate_flags & CAKE_FLAG_OVERHEAD))
+		if (nla_put_u32(skb, TCA_CAKE_RAW, 0))
+			goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_RTT, q->interval))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_TARGET, q->target))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_AUTORATE,
+			!!(q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS)))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_INGRESS,
+			!!(q->rate_flags & CAKE_FLAG_INGRESS)))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_ACK_FILTER, q->ack_filter))
+		goto nla_put_failure;
+
+	if (nla_put_u32(skb, TCA_CAKE_MEMORY, q->buffer_config_limit))
+		goto nla_put_failure;
+
+	return nla_nest_end(skb, opts);
+
+nla_put_failure:
+	return -1;
+}
+
+static int cake_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
+{
+	struct cake_sched_data *q = qdisc_priv(sch);
+	struct nlattr *stats = nla_nest_start(d->skb, TCA_STATS_APP);
+	struct nlattr *tstats, *ts;
+	int i;
+
+	if (!stats)
+		return -1;
+
+#define PUT_STAT_U32(attr, data) do {				       \
+		if (nla_put_u32(d->skb, TCA_CAKE_STATS_ ## attr, data)) \
+			goto nla_put_failure;			       \
+	} while (0)
+
+	PUT_STAT_U32(CAPACITY_ESTIMATE, q->avg_peak_bandwidth);
+	PUT_STAT_U32(MEMORY_LIMIT, q->buffer_limit);
+	PUT_STAT_U32(MEMORY_USED, q->buffer_max_used);
+	PUT_STAT_U32(AVG_NETOFF, ((q->avg_netoff + 0x8000) >> 16));
+	PUT_STAT_U32(MAX_NETLEN, q->max_netlen);
+	PUT_STAT_U32(MAX_ADJLEN, q->max_adjlen);
+	PUT_STAT_U32(MIN_NETLEN, q->min_netlen);
+	PUT_STAT_U32(MIN_ADJLEN, q->min_adjlen);
+
+#undef PUT_STAT_U32
+
+	tstats = nla_nest_start(d->skb, TCA_CAKE_STATS_TIN_STATS);
+	if (!tstats)
+		goto nla_put_failure;
+
+#define PUT_TSTAT_U32(attr, data) do {					\
+		if (nla_put_u32(d->skb, TCA_CAKE_TIN_STATS_ ## attr, data)) \
+			goto nla_put_failure;				\
+	} while (0)
+#define PUT_TSTAT_U64(attr, data) do {					\
+		if (nla_put_u64_64bit(d->skb, TCA_CAKE_TIN_STATS_ ## attr, \
+					data, TCA_CAKE_TIN_STATS_PAD))	\
+			goto nla_put_failure;				\
+	} while (0)
+
+	for (i = 0; i < q->tin_cnt; i++) {
+		struct cake_tin_data *b = &q->tins[i];
+
+		ts = nla_nest_start(d->skb, i + 1);
+		if (!ts)
+			goto nla_put_failure;
+
+		PUT_TSTAT_U32(THRESHOLD_RATE, b->tin_rate_bps);
+		PUT_TSTAT_U32(TARGET_US, cobalt_time_to_us(b->cparams.target));
+		PUT_TSTAT_U32(INTERVAL_US,
+			      cobalt_time_to_us(b->cparams.interval));
+
+		PUT_TSTAT_U32(SENT_PACKETS, b->packets);
+		PUT_TSTAT_U64(SENT_BYTES64, b->bytes);
+		PUT_TSTAT_U32(DROPPED_PACKETS, b->tin_dropped);
+		PUT_TSTAT_U32(ECN_MARKED_PACKETS, b->tin_ecn_mark);
+		PUT_TSTAT_U64(BACKLOG_BYTES64, b->tin_backlog);
+		PUT_TSTAT_U32(ACKS_DROPPED_PACKETS, b->ack_drops);
+
+		PUT_TSTAT_U32(PEAK_DELAY_US, cobalt_time_to_us(b->peak_delay));
+		PUT_TSTAT_U32(AVG_DELAY_US, cobalt_time_to_us(b->avge_delay));
+		PUT_TSTAT_U32(BASE_DELAY_US, cobalt_time_to_us(b->base_delay));
+
+		PUT_TSTAT_U32(WAY_INDIRECT_HITS, b->way_hits);
+		PUT_TSTAT_U32(WAY_MISSES, b->way_misses);
+		PUT_TSTAT_U32(WAY_COLLISIONS, b->way_collisions);
+
+		PUT_TSTAT_U32(SPARSE_FLOWS, b->sparse_flow_count +
+					   b->decaying_flow_count);
+		PUT_TSTAT_U32(BULK_FLOWS, b->bulk_flow_count);
+		PUT_TSTAT_U32(UNRESPONSIVE_FLOWS, b->unresponsive_flow_count);
+		PUT_TSTAT_U32(MAX_SKBLEN, b->max_skblen);
+
+		PUT_TSTAT_U32(FLOW_QUANTUM, b->flow_quantum);
+		nla_nest_end(d->skb, ts);
+	}
+
+#undef PUT_TSTAT_U32
+#undef PUT_TSTAT_U64
+
+	nla_nest_end(d->skb, tstats);
+	return nla_nest_end(d->skb, stats);
+
+nla_put_failure:
+	nla_nest_cancel(d->skb, stats);
+	return -1;
+}
+
+static struct Qdisc_ops cake_qdisc_ops __read_mostly = {
+	.id		=	"cake",
+	.priv_size	=	sizeof(struct cake_sched_data),
+	.enqueue	=	cake_enqueue,
+	.dequeue	=	cake_dequeue,
+	.peek		=	qdisc_peek_dequeued,
+	.init		=	cake_init,
+	.reset		=	cake_reset,
+	.destroy	=	cake_destroy,
+	.change		=	cake_change,
+	.dump		=	cake_dump,
+	.dump_stats	=	cake_dump_stats,
+	.owner		=	THIS_MODULE,
+};
+
+static int __init cake_module_init(void)
+{
+	return register_qdisc(&cake_qdisc_ops);
+}
+
+static void __exit cake_module_exit(void)
+{
+	unregister_qdisc(&cake_qdisc_ops);
+}
+
+module_init(cake_module_init)
+module_exit(cake_module_exit)
+MODULE_AUTHOR("Jonathan Morton");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_DESCRIPTION("The CAKE shaper.");

Re: [PATCH net-next v8 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Cong Wang]

On Fri, May 4, 2018 at 7:02 AM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> +struct cake_sched_data {
> +       struct cake_tin_data *tins;
> +
> +       struct cake_heap_entry overflow_heap[CAKE_QUEUES * CAKE_MAX_TINS];
> +       u16             overflow_timeout;
> +
> +       u16             tin_cnt;
> +       u8              tin_mode;
> +       u8              flow_mode;
> +       u8              ack_filter;
> +       u8              atm_mode;
> +
> +       /* time_next = time_this + ((len * rate_ns) >> rate_shft) */
> +       u16             rate_shft;
> +       u64             time_next_packet;
> +       u64             failsafe_next_packet;
> +       u32             rate_ns;
> +       u32             rate_bps;
> +       u16             rate_flags;
> +       s16             rate_overhead;
> +       u16             rate_mpu;
> +       u32             interval;
> +       u32             target;
> +
> +       /* resource tracking */
> +       u32             buffer_used;
> +       u32             buffer_max_used;
> +       u32             buffer_limit;
> +       u32             buffer_config_limit;
> +
> +       /* indices for dequeue */
> +       u16             cur_tin;
> +       u16             cur_flow;
> +
> +       struct qdisc_watchdog watchdog;
> +       const u8        *tin_index;
> +       const u8        *tin_order;
> +
> +       /* bandwidth capacity estimate */
> +       u64             last_packet_time;
> +       u64             avg_packet_interval;
> +       u64             avg_window_begin;
> +       u32             avg_window_bytes;
> +       u32             avg_peak_bandwidth;
> +       u64             last_reconfig_time;
> +
> +       /* packet length stats */
> +       u32 avg_netoff;
> +       u16 max_netlen;
> +       u16 max_adjlen;
> +       u16 min_netlen;
> +       u16 min_adjlen;
> +};


So sch_cake doesn't accept normal tc filters? Is this intentional?
If so, why?


> +static u16 quantum_div[CAKE_QUEUES + 1] = {0};
> +
> +#define REC_INV_SQRT_CACHE (16)
> +static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0};
> +
> +/* http://en.wikipedia.org/wiki/Methods_of_computing_square_roots
> + * new_invsqrt = (invsqrt / 2) * (3 - count * invsqrt^2)
> + *
> + * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32
> + */
> +
> +static void cobalt_newton_step(struct cobalt_vars *vars)
> +{
> +       u32 invsqrt = vars->rec_inv_sqrt;
> +       u32 invsqrt2 = ((u64)invsqrt * invsqrt) >> 32;
> +       u64 val = (3LL << 32) - ((u64)vars->count * invsqrt2);
> +
> +       val >>= 2; /* avoid overflow in following multiply */
> +       val = (val * invsqrt) >> (32 - 2 + 1);
> +
> +       vars->rec_inv_sqrt = val;
> +}
> +
> +static void cobalt_invsqrt(struct cobalt_vars *vars)
> +{
> +       if (vars->count < REC_INV_SQRT_CACHE)
> +               vars->rec_inv_sqrt = cobalt_rec_inv_sqrt_cache[vars->count];
> +       else
> +               cobalt_newton_step(vars);
> +}


Looks pretty much duplicated with codel...


> +
> +/* There is a big difference in timing between the accurate values placed in
> + * the cache and the approximations given by a single Newton step for small
> + * count values, particularly when stepping from count 1 to 2 or vice versa.
> + * Above 16, a single Newton step gives sufficient accuracy in either
> + * direction, given the precision stored.
> + *
> + * The magnitude of the error when stepping up to count 2 is such as to give
> + * the value that *should* have been produced at count 4.
> + */
> +
> +static void cobalt_cache_init(void)
> +{
> +       struct cobalt_vars v;
> +
> +       memset(&v, 0, sizeof(v));
> +       v.rec_inv_sqrt = ~0U;
> +       cobalt_rec_inv_sqrt_cache[0] = v.rec_inv_sqrt;
> +
> +       for (v.count = 1; v.count < REC_INV_SQRT_CACHE; v.count++) {
> +               cobalt_newton_step(&v);
> +               cobalt_newton_step(&v);
> +               cobalt_newton_step(&v);
> +               cobalt_newton_step(&v);
> +
> +               cobalt_rec_inv_sqrt_cache[v.count] = v.rec_inv_sqrt;
> +       }
> +}
> +
> +static void cobalt_vars_init(struct cobalt_vars *vars)
> +{
> +       memset(vars, 0, sizeof(*vars));
> +
> +       if (!cobalt_rec_inv_sqrt_cache[0]) {
> +               cobalt_cache_init();
> +               cobalt_rec_inv_sqrt_cache[0] = ~0;
> +       }
> +}
> +
> +/* CoDel control_law is t + interval/sqrt(count)
> + * We maintain in rec_inv_sqrt the reciprocal value of sqrt(count) to avoid
> + * both sqrt() and divide operation.
> + */
> +static cobalt_time_t cobalt_control(cobalt_time_t t,
> +                                   cobalt_time_t interval,
> +                                   u32 rec_inv_sqrt)
> +{
> +       return t + reciprocal_scale(interval, rec_inv_sqrt);
> +}
> +
> +/* Call this when a packet had to be dropped due to queue overflow.  Returns
> + * true if the BLUE state was quiescent before but active after this call.
> + */
> +static bool cobalt_queue_full(struct cobalt_vars *vars,
> +                             struct cobalt_params *p,
> +                             cobalt_time_t now)
> +{
> +       bool up = false;
> +
> +       if ((now - vars->blue_timer) > p->target) {
> +               up = !vars->p_drop;
> +               vars->p_drop += p->p_inc;
> +               if (vars->p_drop < p->p_inc)
> +                       vars->p_drop = ~0;
> +               vars->blue_timer = now;
> +       }
> +       vars->dropping = true;
> +       vars->drop_next = now;
> +       if (!vars->count)
> +               vars->count = 1;
> +
> +       return up;
> +}
> +
> +/* Call this when the queue was serviced but turned out to be empty.  Returns
> + * true if the BLUE state was active before but quiescent after this call.
> + */
> +static bool cobalt_queue_empty(struct cobalt_vars *vars,
> +                              struct cobalt_params *p,
> +                              cobalt_time_t now)
> +{
> +       bool down = false;
> +
> +       if (vars->p_drop && (now - vars->blue_timer) > p->target) {
> +               if (vars->p_drop < p->p_dec)
> +                       vars->p_drop = 0;
> +               else
> +                       vars->p_drop -= p->p_dec;
> +               vars->blue_timer = now;
> +               down = !vars->p_drop;
> +       }
> +       vars->dropping = false;
> +
> +       if (vars->count && (now - vars->drop_next) >= 0) {
> +               vars->count--;
> +               cobalt_invsqrt(vars);
> +               vars->drop_next = cobalt_control(vars->drop_next,
> +                                                p->interval,
> +                                                vars->rec_inv_sqrt);
> +       }
> +
> +       return down;
> +}
> +
> +/* Call this with a freshly dequeued packet for possible congestion marking.
> + * Returns true as an instruction to drop the packet, false for delivery.
> + */
> +static bool cobalt_should_drop(struct cobalt_vars *vars,
> +                              struct cobalt_params *p,
> +                              cobalt_time_t now,
> +                              struct sk_buff *skb)
> +{
> +       bool drop = false;
> +
> +       /* Simplified Codel implementation */
> +       cobalt_tdiff_t sojourn  = now - cobalt_get_enqueue_time(skb);
> +
> +/* The 'schedule' variable records, in its sign, whether 'now' is before or
> + * after 'drop_next'.  This allows 'drop_next' to be updated before the next
> + * scheduling decision is actually branched, without destroying that
> + * information.  Similarly, the first 'schedule' value calculated is preserved
> + * in the boolean 'next_due'.
> + *
> + * As for 'drop_next', we take advantage of the fact that 'interval' is both
> + * the delay between first exceeding 'target' and the first signalling event,
> + * *and* the scaling factor for the signalling frequency.  It's therefore very
> + * natural to use a single mechanism for both purposes, and eliminates a
> + * significant amount of reference Codel's spaghetti code.  To help with this,
> + * both the '0' and '1' entries in the invsqrt cache are 0xFFFFFFFF, as close
> + * as possible to 1.0 in fixed-point.
> + */
> +
> +       cobalt_tdiff_t schedule = now - vars->drop_next;
> +
> +       bool over_target = sojourn > p->target &&
> +                          sojourn > p->mtu_time * 4;
> +       bool next_due    = vars->count && schedule >= 0;
> +
> +       vars->ecn_marked = false;
> +
> +       if (over_target) {
> +               if (!vars->dropping) {
> +                       vars->dropping = true;
> +                       vars->drop_next = cobalt_control(now,
> +                                                        p->interval,
> +                                                        vars->rec_inv_sqrt);
> +               }
> +               if (!vars->count)
> +                       vars->count = 1;
> +       } else if (vars->dropping) {
> +               vars->dropping = false;
> +       }
> +
> +       if (next_due && vars->dropping) {
> +               /* Use ECN mark if possible, otherwise drop */
> +               drop = !(vars->ecn_marked = INET_ECN_set_ce(skb));
> +
> +               vars->count++;
> +               if (!vars->count)
> +                       vars->count--;
> +               cobalt_invsqrt(vars);
> +               vars->drop_next = cobalt_control(vars->drop_next,
> +                                                p->interval,
> +                                                vars->rec_inv_sqrt);
> +               schedule = now - vars->drop_next;
> +       } else {
> +               while (next_due) {
> +                       vars->count--;
> +                       cobalt_invsqrt(vars);
> +                       vars->drop_next = cobalt_control(vars->drop_next,
> +                                                        p->interval,
> +                                                        vars->rec_inv_sqrt);
> +                       schedule = now - vars->drop_next;
> +                       next_due = vars->count && schedule >= 0;
> +               }
> +       }
> +
> +       /* Simple BLUE implementation.  Lack of ECN is deliberate. */
> +       if (vars->p_drop)
> +               drop |= (prandom_u32() < vars->p_drop);
> +
> +       /* Overload the drop_next field as an activity timeout */
> +       if (!vars->count)
> +               vars->drop_next = now + p->interval;
> +       else if (schedule > 0 && !drop)
> +               vars->drop_next = now;
> +
> +       return drop;
> +}
> +
> +/* Cake has several subtle multiple bit settings. In these cases you
> + *  would be matching triple isolate mode as well.
> + */
> +
> +static bool cake_dsrc(int flow_mode)
> +{
> +       return (flow_mode & CAKE_FLOW_DUAL_SRC) == CAKE_FLOW_DUAL_SRC;
> +}
> +
> +static bool cake_ddst(int flow_mode)
> +{
> +       return (flow_mode & CAKE_FLOW_DUAL_DST) == CAKE_FLOW_DUAL_DST;
> +}
> +
> +static u32 cake_hash(struct cake_tin_data *q, const struct sk_buff *skb,
> +                    int flow_mode)
> +{
> +       struct flow_keys keys, host_keys;
> +       u32 flow_hash = 0, srchost_hash, dsthost_hash;
> +       u16 reduced_hash, srchost_idx, dsthost_idx;
> +
> +       if (unlikely(flow_mode == CAKE_FLOW_NONE))
> +               return 0;
> +
> +       skb_flow_dissect_flow_keys(skb, &keys,
> +                                  FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
> +
> +       /* flow_hash_from_keys() sorts the addresses by value, so we have
> +        * to preserve their order in a separate data structure to treat
> +        * src and dst host addresses as independently selectable.
> +        */
> +       host_keys = keys;
> +       host_keys.ports.ports     = 0;
> +       host_keys.basic.ip_proto  = 0;
> +       host_keys.keyid.keyid     = 0;
> +       host_keys.tags.flow_label = 0;
> +
> +       switch (host_keys.control.addr_type) {
> +       case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
> +               host_keys.addrs.v4addrs.src = 0;
> +               dsthost_hash = flow_hash_from_keys(&host_keys);
> +               host_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
> +               host_keys.addrs.v4addrs.dst = 0;
> +               srchost_hash = flow_hash_from_keys(&host_keys);
> +               break;
> +
> +       case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
> +               memset(&host_keys.addrs.v6addrs.src, 0,
> +                      sizeof(host_keys.addrs.v6addrs.src));
> +               dsthost_hash = flow_hash_from_keys(&host_keys);
> +               host_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
> +               memset(&host_keys.addrs.v6addrs.dst, 0,
> +                      sizeof(host_keys.addrs.v6addrs.dst));
> +               srchost_hash = flow_hash_from_keys(&host_keys);
> +               break;
> +
> +       default:
> +               dsthost_hash = 0;
> +               srchost_hash = 0;
> +       }
> +
> +       /* This *must* be after the above switch, since as a
> +        * side-effect it sorts the src and dst addresses.
> +        */
> +       if (flow_mode & CAKE_FLOW_FLOWS)
> +               flow_hash = flow_hash_from_keys(&keys);
> +
> +       if (!(flow_mode & CAKE_FLOW_FLOWS)) {
> +               if (flow_mode & CAKE_FLOW_SRC_IP)
> +                       flow_hash ^= srchost_hash;
> +
> +               if (flow_mode & CAKE_FLOW_DST_IP)
> +                       flow_hash ^= dsthost_hash;
> +       }
> +
> +       reduced_hash = flow_hash % CAKE_QUEUES;
> +
> +       /* set-associative hashing */
> +       /* fast path if no hash collision (direct lookup succeeds) */
> +       if (likely(q->tags[reduced_hash] == flow_hash &&
> +                  q->flows[reduced_hash].set)) {
> +               q->way_directs++;
> +       } else {
> +               u32 inner_hash = reduced_hash % CAKE_SET_WAYS;
> +               u32 outer_hash = reduced_hash - inner_hash;
> +               u32 i, k;
> +               bool allocate_src = false;
> +               bool allocate_dst = false;
> +
> +               /* check if any active queue in the set is reserved for
> +                * this flow.
> +                */
> +               for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
> +                    i++, k = (k + 1) % CAKE_SET_WAYS) {
> +                       if (q->tags[outer_hash + k] == flow_hash) {
> +                               if (i)
> +                                       q->way_hits++;
> +
> +                               if (!q->flows[outer_hash + k].set) {
> +                                       /* need to increment host refcnts */
> +                                       allocate_src = cake_dsrc(flow_mode);
> +                                       allocate_dst = cake_ddst(flow_mode);
> +                               }
> +
> +                               goto found;
> +                       }
> +               }
> +
> +               /* no queue is reserved for this flow, look for an
> +                * empty one.
> +                */
> +               for (i = 0; i < CAKE_SET_WAYS;
> +                        i++, k = (k + 1) % CAKE_SET_WAYS) {
> +                       if (!q->flows[outer_hash + k].set) {
> +                               q->way_misses++;
> +                               allocate_src = cake_dsrc(flow_mode);
> +                               allocate_dst = cake_ddst(flow_mode);
> +                               goto found;
> +                       }
> +               }
> +
> +               /* With no empty queues, default to the original
> +                * queue, accept the collision, update the host tags.
> +                */
> +               q->way_collisions++;
> +               q->hosts[q->flows[reduced_hash].srchost].srchost_refcnt--;
> +               q->hosts[q->flows[reduced_hash].dsthost].dsthost_refcnt--;
> +               allocate_src = cake_dsrc(flow_mode);
> +               allocate_dst = cake_ddst(flow_mode);
> +found:
> +               /* reserve queue for future packets in same flow */
> +               reduced_hash = outer_hash + k;
> +               q->tags[reduced_hash] = flow_hash;
> +
> +               if (allocate_src) {
> +                       srchost_idx = srchost_hash % CAKE_QUEUES;
> +                       inner_hash = srchost_idx % CAKE_SET_WAYS;
> +                       outer_hash = srchost_idx - inner_hash;
> +                       for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
> +                               i++, k = (k + 1) % CAKE_SET_WAYS) {
> +                               if (q->hosts[outer_hash + k].srchost_tag ==
> +                                   srchost_hash)
> +                                       goto found_src;
> +                       }
> +                       for (i = 0; i < CAKE_SET_WAYS;
> +                               i++, k = (k + 1) % CAKE_SET_WAYS) {
> +                               if (!q->hosts[outer_hash + k].srchost_refcnt)
> +                                       break;
> +                       }
> +                       q->hosts[outer_hash + k].srchost_tag = srchost_hash;
> +found_src:
> +                       srchost_idx = outer_hash + k;
> +                       q->hosts[srchost_idx].srchost_refcnt++;
> +                       q->flows[reduced_hash].srchost = srchost_idx;
> +               }
> +
> +               if (allocate_dst) {
> +                       dsthost_idx = dsthost_hash % CAKE_QUEUES;
> +                       inner_hash = dsthost_idx % CAKE_SET_WAYS;
> +                       outer_hash = dsthost_idx - inner_hash;
> +                       for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
> +                            i++, k = (k + 1) % CAKE_SET_WAYS) {
> +                               if (q->hosts[outer_hash + k].dsthost_tag ==
> +                                   dsthost_hash)
> +                                       goto found_dst;
> +                       }
> +                       for (i = 0; i < CAKE_SET_WAYS;
> +                            i++, k = (k + 1) % CAKE_SET_WAYS) {
> +                               if (!q->hosts[outer_hash + k].dsthost_refcnt)
> +                                       break;
> +                       }
> +                       q->hosts[outer_hash + k].dsthost_tag = dsthost_hash;
> +found_dst:
> +                       dsthost_idx = outer_hash + k;
> +                       q->hosts[dsthost_idx].dsthost_refcnt++;
> +                       q->flows[reduced_hash].dsthost = dsthost_idx;
> +               }
> +       }
> +
> +       return reduced_hash;
> +}
> +
> +/* helper functions : might be changed when/if skb use a standard list_head */
> +/* remove one skb from head of slot queue */
> +
> +static struct sk_buff *dequeue_head(struct cake_flow *flow)
> +{
> +       struct sk_buff *skb = flow->head;
> +
> +       if (skb) {
> +               flow->head = skb->next;
> +               skb->next = NULL;
> +
> +               if (skb == flow->ackcheck)
> +                       flow->ackcheck = NULL;
> +       }
> +
> +       return skb;
> +}
> +
> +/* add skb to flow queue (tail add) */
> +
> +static void flow_queue_add(struct cake_flow *flow, struct sk_buff *skb)
> +{
> +       if (!flow->head)
> +               flow->head = skb;
> +       else
> +               flow->tail->next = skb;
> +       flow->tail = skb;
> +       skb->next = NULL;
> +}
> +
> +static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
> +                                     u32 shift)
> +{
> +       avg -= avg >> shift;
> +       avg += sample >> shift;
> +       return avg;
> +}
> +
> +static void cake_heap_swap(struct cake_sched_data *q, u16 i, u16 j)
> +{
> +       struct cake_heap_entry ii = q->overflow_heap[i];
> +       struct cake_heap_entry jj = q->overflow_heap[j];
> +
> +       q->overflow_heap[i] = jj;
> +       q->overflow_heap[j] = ii;
> +
> +       q->tins[ii.t].overflow_idx[ii.b] = j;
> +       q->tins[jj.t].overflow_idx[jj.b] = i;
> +}
> +
> +static u32 cake_heap_get_backlog(const struct cake_sched_data *q, u16 i)
> +{
> +       struct cake_heap_entry ii = q->overflow_heap[i];
> +
> +       return q->tins[ii.t].backlogs[ii.b];
> +}
> +
> +static void cake_heapify(struct cake_sched_data *q, u16 i)
> +{
> +       static const u32 a = CAKE_MAX_TINS * CAKE_QUEUES;
> +       u32 m = i;
> +       u32 mb = cake_heap_get_backlog(q, m);
> +
> +       while (m < a) {
> +               u32 l = m + m + 1;
> +               u32 r = l + 1;
> +
> +               if (l < a) {
> +                       u32 lb = cake_heap_get_backlog(q, l);
> +
> +                       if (lb > mb) {
> +                               m  = l;
> +                               mb = lb;
> +                       }
> +               }
> +
> +               if (r < a) {
> +                       u32 rb = cake_heap_get_backlog(q, r);
> +
> +                       if (rb > mb) {
> +                               m  = r;
> +                               mb = rb;
> +                       }
> +               }
> +
> +               if (m != i) {
> +                       cake_heap_swap(q, i, m);
> +                       i = m;
> +               } else {
> +                       break;
> +               }
> +       }
> +}
> +
> +static void cake_heapify_up(struct cake_sched_data *q, u16 i)
> +{
> +       while (i > 0 && i < CAKE_MAX_TINS * CAKE_QUEUES) {
> +               u16 p = (i - 1) >> 1;
> +               u32 ib = cake_heap_get_backlog(q, i);
> +               u32 pb = cake_heap_get_backlog(q, p);
> +
> +               if (ib > pb) {
> +                       cake_heap_swap(q, i, p);
> +                       i = p;
> +               } else {
> +                       break;
> +               }
> +       }
> +}
> +
> +static int cake_advance_shaper(struct cake_sched_data *q,
> +                              struct cake_tin_data *b,
> +                              struct sk_buff *skb,
> +                              u64 now, bool drop)
> +{
> +       u32 len = qdisc_pkt_len(skb);
> +
> +       /* charge packet bandwidth to this tin
> +        * and to the global shaper.
> +        */
> +       if (q->rate_ns) {
> +               s64 tdiff1 = b->tin_time_next_packet - now;
> +               s64 tdiff2 = (len * (u64)b->tin_rate_ns) >> b->tin_rate_shft;
> +               s64 tdiff3 = (len * (u64)q->rate_ns) >> q->rate_shft;
> +               s64 tdiff4 = tdiff3 + (tdiff3 >> 1);
> +
> +               if (tdiff1 < 0)
> +                       b->tin_time_next_packet += tdiff2;
> +               else if (tdiff1 < tdiff2)
> +                       b->tin_time_next_packet = now + tdiff2;
> +
> +               q->time_next_packet += tdiff3;
> +               if (!drop)
> +                       q->failsafe_next_packet += tdiff4;
> +       }
> +       return len;
> +}
> +
> +static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct sk_buff *skb;
> +       u32 idx = 0, tin = 0, len;
> +       struct cake_tin_data *b;
> +       struct cake_flow *flow;
> +       struct cake_heap_entry qq;
> +       u64 now = cobalt_get_time();
> +
> +       if (!q->overflow_timeout) {
> +               int i;
> +               /* Build fresh max-heap */
> +               for (i = CAKE_MAX_TINS * CAKE_QUEUES / 2; i >= 0; i--)
> +                       cake_heapify(q, i);
> +       }
> +       q->overflow_timeout = 65535;
> +
> +       /* select longest queue for pruning */
> +       qq  = q->overflow_heap[0];
> +       tin = qq.t;
> +       idx = qq.b;
> +
> +       b = &q->tins[tin];
> +       flow = &b->flows[idx];
> +       skb = dequeue_head(flow);
> +       if (unlikely(!skb)) {
> +               /* heap has gone wrong, rebuild it next time */
> +               q->overflow_timeout = 0;
> +               return idx + (tin << 16);
> +       }
> +
> +       if (cobalt_queue_full(&flow->cvars, &b->cparams, now))
> +               b->unresponsive_flow_count++;
> +
> +       len = qdisc_pkt_len(skb);
> +       q->buffer_used      -= skb->truesize;
> +       b->backlogs[idx]    -= len;
> +       b->tin_backlog      -= len;
> +       sch->qstats.backlog -= len;
> +       qdisc_tree_reduce_backlog(sch, 1, len);
> +
> +       b->tin_dropped++;
> +       sch->qstats.drops++;
> +
> +       __qdisc_drop(skb, to_free);
> +       sch->q.qlen--;
> +
> +       cake_heapify(q, 0);
> +
> +       return idx + (tin << 16);
> +}
> +
> +static void cake_reconfigure(struct Qdisc *sch);
> +
> +static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
> +                       struct sk_buff **to_free)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       u32 idx, tin;
> +       struct cake_tin_data *b;
> +       struct cake_flow *flow;
> +       /* signed len to handle corner case filtered ACK larger than trigger */
> +       int len = qdisc_pkt_len(skb);
> +       u64 now = cobalt_get_time();
> +
> +       tin = 0;
> +       b = &q->tins[tin];
> +
> +       /* choose flow to insert into */
> +       idx = cake_hash(b, skb, q->flow_mode);
> +       flow = &b->flows[idx];
> +
> +       /* ensure shaper state isn't stale */
> +       if (!b->tin_backlog) {
> +               if (b->tin_time_next_packet < now)
> +                       b->tin_time_next_packet = now;
> +
> +               if (!sch->q.qlen) {
> +                       if (q->time_next_packet < now) {
> +                               q->failsafe_next_packet = now;
> +                               q->time_next_packet = now;
> +                       } else if (q->time_next_packet > now &&
> +                                  q->failsafe_next_packet > now) {
> +                               u64 next = min(q->time_next_packet,
> +                                              q->failsafe_next_packet);
> +                               sch->qstats.overlimits++;
> +                               qdisc_watchdog_schedule_ns(&q->watchdog, next);
> +                       }
> +               }
> +       }
> +
> +       if (unlikely(len > b->max_skblen))
> +               b->max_skblen = len;
> +
> +       cobalt_set_enqueue_time(skb, now);
> +       flow_queue_add(flow, skb);
> +
> +       sch->q.qlen++;
> +       q->buffer_used      += skb->truesize;
> +
> +       /* stats */
> +       b->packets++;
> +       b->bytes            += len;
> +       b->backlogs[idx]    += len;
> +       b->tin_backlog      += len;
> +       sch->qstats.backlog += len;
> +       q->avg_window_bytes += len;
> +
> +       if (q->overflow_timeout)
> +               cake_heapify_up(q, b->overflow_idx[idx]);
> +
> +       /* incoming bandwidth capacity estimate */
> +       q->avg_window_bytes = 0;
> +       q->last_packet_time = now;
> +
> +       /* flowchain */
> +       if (!flow->set || flow->set == CAKE_SET_DECAYING) {
> +               struct cake_host *srchost = &b->hosts[flow->srchost];
> +               struct cake_host *dsthost = &b->hosts[flow->dsthost];
> +               u16 host_load = 1;
> +
> +               if (!flow->set) {
> +                       list_add_tail(&flow->flowchain, &b->new_flows);
> +               } else {
> +                       b->decaying_flow_count--;
> +                       list_move_tail(&flow->flowchain, &b->new_flows);
> +               }
> +               flow->set = CAKE_SET_SPARSE;
> +               b->sparse_flow_count++;
> +
> +               if (cake_dsrc(q->flow_mode))
> +                       host_load = max(host_load, srchost->srchost_refcnt);
> +
> +               if (cake_ddst(q->flow_mode))
> +                       host_load = max(host_load, dsthost->dsthost_refcnt);
> +
> +               flow->deficit = (b->flow_quantum *
> +                                quantum_div[host_load]) >> 16;
> +       } else if (flow->set == CAKE_SET_SPARSE_WAIT) {
> +               /* this flow was empty, accounted as a sparse flow, but actually
> +                * in the bulk rotation.
> +                */
> +               flow->set = CAKE_SET_BULK;
> +               b->sparse_flow_count--;
> +               b->bulk_flow_count++;
> +       }
> +
> +       if (q->buffer_used > q->buffer_max_used)
> +               q->buffer_max_used = q->buffer_used;
> +
> +       if (q->buffer_used > q->buffer_limit) {
> +               u32 dropped = 0;
> +
> +               while (q->buffer_used > q->buffer_limit) {
> +                       dropped++;
> +                       cake_drop(sch, to_free);
> +               }
> +               b->drop_overlimit += dropped;
> +       }
> +       return NET_XMIT_SUCCESS;
> +}
> +
> +static struct sk_buff *cake_dequeue_one(struct Qdisc *sch)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct cake_tin_data *b = &q->tins[q->cur_tin];
> +       struct cake_flow *flow = &b->flows[q->cur_flow];
> +       struct sk_buff *skb = NULL;
> +       u32 len;
> +
> +       if (flow->head) {
> +               skb = dequeue_head(flow);
> +               len = qdisc_pkt_len(skb);
> +               b->backlogs[q->cur_flow] -= len;
> +               b->tin_backlog           -= len;
> +               sch->qstats.backlog      -= len;
> +               q->buffer_used           -= skb->truesize;
> +               sch->q.qlen--;
> +
> +               if (q->overflow_timeout)
> +                       cake_heapify(q, b->overflow_idx[q->cur_flow]);
> +       }
> +       return skb;
> +}
> +
> +/* Discard leftover packets from a tin no longer in use. */
> +static void cake_clear_tin(struct Qdisc *sch, u16 tin)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct sk_buff *skb;
> +
> +       q->cur_tin = tin;
> +       for (q->cur_flow = 0; q->cur_flow < CAKE_QUEUES; q->cur_flow++)
> +               while (!!(skb = cake_dequeue_one(sch)))
> +                       kfree_skb(skb);
> +}
> +
> +static struct sk_buff *cake_dequeue(struct Qdisc *sch)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct sk_buff *skb;
> +       struct cake_tin_data *b = &q->tins[q->cur_tin];
> +       struct cake_flow *flow;
> +       struct cake_host *srchost, *dsthost;
> +       struct list_head *head;
> +       u32 len;
> +       u16 host_load;
> +       cobalt_time_t now = ktime_get_ns();
> +       cobalt_time_t delay;
> +       bool first_flow = true;
> +
> +begin:
> +       if (!sch->q.qlen)
> +               return NULL;
> +
> +       /* global hard shaper */
> +       if (q->time_next_packet > now && q->failsafe_next_packet > now) {
> +               u64 next = min(q->time_next_packet, q->failsafe_next_packet);
> +
> +               sch->qstats.overlimits++;
> +               qdisc_watchdog_schedule_ns(&q->watchdog, next);
> +               return NULL;
> +       }
> +
> +       /* Choose a class to work on. */
> +       if (!q->rate_ns) {
> +               /* In unlimited mode, can't rely on shaper timings, just balance
> +                * with DRR
> +                */
> +               while (b->tin_deficit < 0 ||
> +                      !(b->sparse_flow_count + b->bulk_flow_count)) {
> +                       if (b->tin_deficit <= 0)
> +                               b->tin_deficit += b->tin_quantum_band;
> +
> +                       q->cur_tin++;
> +                       b++;
> +                       if (q->cur_tin >= q->tin_cnt) {
> +                               q->cur_tin = 0;
> +                               b = q->tins;
> +                       }
> +               }
> +       } else {
> +               /* In shaped mode, choose:
> +                * - Highest-priority tin with queue and meeting schedule, or
> +                * - The earliest-scheduled tin with queue.
> +                */
> +               int tin, best_tin = 0;
> +               s64 best_time = 0xFFFFFFFFFFFFUL;
> +
> +               for (tin = 0; tin < q->tin_cnt; tin++) {
> +                       b = q->tins + tin;
> +                       if ((b->sparse_flow_count + b->bulk_flow_count) > 0) {
> +                               s64 tdiff = b->tin_time_next_packet - now;
> +
> +                               if (tdiff <= 0 || tdiff <= best_time) {
> +                                       best_time = tdiff;
> +                                       best_tin = tin;
> +                               }
> +                       }
> +               }
> +
> +               q->cur_tin = best_tin;
> +               b = q->tins + best_tin;
> +       }
> +
> +retry:
> +       /* service this class */
> +       head = &b->decaying_flows;
> +       if (!first_flow || list_empty(head)) {
> +               head = &b->new_flows;
> +               if (list_empty(head)) {
> +                       head = &b->old_flows;
> +                       if (unlikely(list_empty(head))) {
> +                               head = &b->decaying_flows;
> +                               if (unlikely(list_empty(head)))
> +                                       goto begin;
> +                       }
> +               }
> +       }
> +       flow = list_first_entry(head, struct cake_flow, flowchain);
> +       q->cur_flow = flow - b->flows;
> +       first_flow = false;
> +
> +       /* triple isolation (modified DRR++) */
> +       srchost = &b->hosts[flow->srchost];
> +       dsthost = &b->hosts[flow->dsthost];
> +       host_load = 1;
> +
> +       if (cake_dsrc(q->flow_mode))
> +               host_load = max(host_load, srchost->srchost_refcnt);
> +
> +       if (cake_ddst(q->flow_mode))
> +               host_load = max(host_load, dsthost->dsthost_refcnt);
> +
> +       WARN_ON(host_load > CAKE_QUEUES);
> +
> +       /* flow isolation (DRR++) */
> +       if (flow->deficit <= 0) {
> +               /* The shifted prandom_u32() is a way to apply dithering to
> +                * avoid accumulating roundoff errors
> +                */
> +               flow->deficit += (b->flow_quantum * quantum_div[host_load] +
> +                                 (prandom_u32() >> 16)) >> 16;
> +               list_move_tail(&flow->flowchain, &b->old_flows);
> +
> +               /* Keep all flows with deficits out of the sparse and decaying
> +                * rotations.  No non-empty flow can go into the decaying
> +                * rotation, so they can't get deficits
> +                */
> +               if (flow->set == CAKE_SET_SPARSE) {
> +                       if (flow->head) {
> +                               b->sparse_flow_count--;
> +                               b->bulk_flow_count++;
> +                               flow->set = CAKE_SET_BULK;
> +                       } else {
> +                               /* we've moved it to the bulk rotation for
> +                                * correct deficit accounting but we still want
> +                                * to count it as a sparse flow, not a bulk one.
> +                                */
> +                               flow->set = CAKE_SET_SPARSE_WAIT;
> +                       }
> +               }
> +               goto retry;
> +       }
> +
> +       /* Retrieve a packet via the AQM */
> +       while (1) {
> +               skb = cake_dequeue_one(sch);
> +               if (!skb) {
> +                       /* this queue was actually empty */
> +                       if (cobalt_queue_empty(&flow->cvars, &b->cparams, now))
> +                               b->unresponsive_flow_count--;
> +
> +                       if (flow->cvars.p_drop || flow->cvars.count ||
> +                           now < flow->cvars.drop_next) {
> +                               /* keep in the flowchain until the state has
> +                                * decayed to rest
> +                                */
> +                               list_move_tail(&flow->flowchain,
> +                                              &b->decaying_flows);
> +                               if (flow->set == CAKE_SET_BULK) {
> +                                       b->bulk_flow_count--;
> +                                       b->decaying_flow_count++;
> +                               } else if (flow->set == CAKE_SET_SPARSE ||
> +                                          flow->set == CAKE_SET_SPARSE_WAIT) {
> +                                       b->sparse_flow_count--;
> +                                       b->decaying_flow_count++;
> +                               }
> +                               flow->set = CAKE_SET_DECAYING;
> +                       } else {
> +                               /* remove empty queue from the flowchain */
> +                               list_del_init(&flow->flowchain);
> +                               if (flow->set == CAKE_SET_SPARSE ||
> +                                   flow->set == CAKE_SET_SPARSE_WAIT)
> +                                       b->sparse_flow_count--;
> +                               else if (flow->set == CAKE_SET_BULK)
> +                                       b->bulk_flow_count--;
> +                               else
> +                                       b->decaying_flow_count--;
> +
> +                               flow->set = CAKE_SET_NONE;
> +                               srchost->srchost_refcnt--;
> +                               dsthost->dsthost_refcnt--;
> +                       }
> +                       goto begin;
> +               }
> +
> +               /* Last packet in queue may be marked, shouldn't be dropped */
> +               if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb) ||
> +                   !flow->head)
> +                       break;
> +
> +               b->tin_dropped++;
> +               qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb));
> +               qdisc_qstats_drop(sch);
> +               kfree_skb(skb);
> +       }
> +
> +       b->tin_ecn_mark += !!flow->cvars.ecn_marked;
> +       qdisc_bstats_update(sch, skb);
> +
> +       /* collect delay stats */
> +       delay = now - cobalt_get_enqueue_time(skb);
> +       b->avge_delay = cake_ewma(b->avge_delay, delay, 8);
> +       b->peak_delay = cake_ewma(b->peak_delay, delay,
> +                                 delay > b->peak_delay ? 2 : 8);
> +       b->base_delay = cake_ewma(b->base_delay, delay,
> +                                 delay < b->base_delay ? 2 : 8);
> +
> +       len = cake_advance_shaper(q, b, skb, now, false);
> +       flow->deficit -= len;
> +       b->tin_deficit -= len;
> +
> +       if (q->time_next_packet > now && sch->q.qlen) {
> +               u64 next = min(q->time_next_packet, q->failsafe_next_packet);
> +
> +               qdisc_watchdog_schedule_ns(&q->watchdog, next);
> +       } else if (!sch->q.qlen) {
> +               int i;
> +
> +               for (i = 0; i < q->tin_cnt; i++) {
> +                       if (q->tins[i].decaying_flow_count) {
> +                               u64 next = now + q->tins[i].cparams.target;
> +
> +                               qdisc_watchdog_schedule_ns(&q->watchdog, next);
> +                               break;
> +                       }
> +               }
> +       }
> +
> +       if (q->overflow_timeout)
> +               q->overflow_timeout--;
> +
> +       return skb;
> +}
> +
> +static void cake_reset(struct Qdisc *sch)
> +{
> +       u32 c;
> +
> +       for (c = 0; c < CAKE_MAX_TINS; c++)
> +               cake_clear_tin(sch, c);
> +}
> +
> +static const struct nla_policy cake_policy[TCA_CAKE_MAX + 1] = {
> +       [TCA_CAKE_BASE_RATE]     = { .type = NLA_U32 },
> +       [TCA_CAKE_DIFFSERV_MODE] = { .type = NLA_U32 },
> +       [TCA_CAKE_ATM]           = { .type = NLA_U32 },
> +       [TCA_CAKE_FLOW_MODE]     = { .type = NLA_U32 },
> +       [TCA_CAKE_OVERHEAD]      = { .type = NLA_S32 },
> +       [TCA_CAKE_RTT]           = { .type = NLA_U32 },
> +       [TCA_CAKE_TARGET]        = { .type = NLA_U32 },
> +       [TCA_CAKE_AUTORATE]      = { .type = NLA_U32 },
> +       [TCA_CAKE_MEMORY]        = { .type = NLA_U32 },
> +       [TCA_CAKE_NAT]           = { .type = NLA_U32 },
> +       [TCA_CAKE_RAW]           = { .type = NLA_U32 },
> +       [TCA_CAKE_WASH]          = { .type = NLA_U32 },
> +       [TCA_CAKE_MPU]           = { .type = NLA_U32 },
> +       [TCA_CAKE_INGRESS]       = { .type = NLA_U32 },
> +       [TCA_CAKE_ACK_FILTER]    = { .type = NLA_U32 },
> +};
> +
> +static void cake_set_rate(struct cake_tin_data *b, u64 rate, u32 mtu,
> +                         cobalt_time_t ns_target, cobalt_time_t rtt_est_ns)
> +{
> +       /* convert byte-rate into time-per-byte
> +        * so it will always unwedge in reasonable time.
> +        */
> +       static const u64 MIN_RATE = 64;
> +       u64 rate_ns = 0;
> +       u8  rate_shft = 0;
> +       cobalt_time_t byte_target_ns;
> +       u32 byte_target = mtu;
> +
> +       b->flow_quantum = 1514;
> +       if (rate) {
> +               b->flow_quantum = max(min(rate >> 12, 1514ULL), 300ULL);
> +               rate_shft = 32;
> +               rate_ns = ((u64)NSEC_PER_SEC) << rate_shft;
> +               do_div(rate_ns, max(MIN_RATE, rate));
> +               while (!!(rate_ns >> 32)) {
> +                       rate_ns >>= 1;
> +                       rate_shft--;
> +               }
> +       } /* else unlimited, ie. zero delay */
> +
> +       b->tin_rate_bps  = rate;
> +       b->tin_rate_ns   = rate_ns;
> +       b->tin_rate_shft = rate_shft;
> +
> +       byte_target_ns = (byte_target * rate_ns) >> rate_shft;
> +
> +       b->cparams.target = max((byte_target_ns * 3) / 2, ns_target);
> +       b->cparams.interval = max(rtt_est_ns +
> +                                    b->cparams.target - ns_target,
> +                                    b->cparams.target * 2);
> +       b->cparams.mtu_time = byte_target_ns;
> +       b->cparams.p_inc = 1 << 24; /* 1/256 */
> +       b->cparams.p_dec = 1 << 20; /* 1/4096 */
> +}
> +
> +static void cake_reconfigure(struct Qdisc *sch)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct cake_tin_data *b = &q->tins[0];
> +       int c, ft = 0;
> +
> +       q->tin_cnt = 1;
> +       cake_set_rate(b, q->rate_bps, psched_mtu(qdisc_dev(sch)),
> +                     US2TIME(q->target), US2TIME(q->interval));
> +       b->tin_quantum_band = 65535;
> +       b->tin_quantum_prio = 65535;
> +
> +       for (c = q->tin_cnt; c < CAKE_MAX_TINS; c++) {
> +               cake_clear_tin(sch, c);
> +               q->tins[c].cparams.mtu_time = q->tins[ft].cparams.mtu_time;
> +       }
> +
> +       q->rate_ns   = q->tins[ft].tin_rate_ns;
> +       q->rate_shft = q->tins[ft].tin_rate_shft;
> +
> +       if (q->buffer_config_limit) {
> +               q->buffer_limit = q->buffer_config_limit;
> +       } else if (q->rate_bps) {
> +               u64 t = (u64)q->rate_bps * q->interval;
> +
> +               do_div(t, USEC_PER_SEC / 4);
> +               q->buffer_limit = max_t(u32, t, 4U << 20);
> +       } else {
> +               q->buffer_limit = ~0;
> +       }
> +
> +       sch->flags &= ~TCQ_F_CAN_BYPASS;
> +
> +       q->buffer_limit = min(q->buffer_limit,
> +                             max(sch->limit * psched_mtu(qdisc_dev(sch)),
> +                                 q->buffer_config_limit));
> +}
> +
> +static int cake_change(struct Qdisc *sch, struct nlattr *opt,
> +                      struct netlink_ext_ack *extack)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct nlattr *tb[TCA_CAKE_MAX + 1];
> +       int err;
> +
> +       if (!opt)
> +               return -EINVAL;
> +
> +       err = nla_parse_nested(tb, TCA_CAKE_MAX, opt, cake_policy, extack);
> +       if (err < 0)
> +               return err;
> +
> +       if (tb[TCA_CAKE_BASE_RATE])
> +               q->rate_bps = nla_get_u32(tb[TCA_CAKE_BASE_RATE]);
> +
> +       if (tb[TCA_CAKE_FLOW_MODE])
> +               q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
> +                               CAKE_FLOW_MASK);
> +
> +       if (tb[TCA_CAKE_RTT]) {
> +               q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
> +
> +               if (!q->interval)
> +                       q->interval = 1;
> +       }
> +
> +       if (tb[TCA_CAKE_TARGET]) {
> +               q->target = nla_get_u32(tb[TCA_CAKE_TARGET]);
> +
> +               if (!q->target)
> +                       q->target = 1;
> +       }
> +
> +       if (tb[TCA_CAKE_MEMORY])
> +               q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
> +
> +       if (q->tins) {
> +               sch_tree_lock(sch);
> +               cake_reconfigure(sch);
> +               sch_tree_unlock(sch);
> +       }
> +
> +       return 0;
> +}
> +
> +static void cake_free(void *addr)
> +{
> +       if (addr)
> +               kvfree(addr);
> +}


Are you sure you have to check NULL for kvfree()?



> +
> +static void cake_destroy(struct Qdisc *sch)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +
> +       qdisc_watchdog_cancel(&q->watchdog);
> +
> +       if (q->tins)
> +               cake_free(q->tins);


Duplicated NULL check.


> +}
> +
> +static int cake_init(struct Qdisc *sch, struct nlattr *opt,
> +                    struct netlink_ext_ack *extack)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       int i, j;
> +
> +       sch->limit = 10240;
> +       q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
> +       q->flow_mode  = CAKE_FLOW_TRIPLE;
> +
> +       q->rate_bps = 0; /* unlimited by default */
> +
> +       q->interval = 100000; /* 100ms default */
> +       q->target   =   5000; /* 5ms: codel RFC argues
> +                              * for 5 to 10% of interval
> +                              */
> +
> +       q->cur_tin = 0;
> +       q->cur_flow  = 0;
> +
> +       if (opt) {
> +               int err = cake_change(sch, opt, extack);
> +
> +               if (err)
> +                       return err;


Not sure if you really want to reallocate q->tines below for this
case.


> +       }
> +
> +       qdisc_watchdog_init(&q->watchdog, sch);
> +
> +       quantum_div[0] = ~0;
> +       for (i = 1; i <= CAKE_QUEUES; i++)
> +               quantum_div[i] = 65535 / i;
> +
> +       q->tins = kvzalloc(CAKE_MAX_TINS * sizeof(struct cake_tin_data),
> +                          GFP_KERNEL | __GFP_NOWARN);



Why __GFP_NOWARN?



> +       if (!q->tins)
> +               goto nomem;
> +
> +       for (i = 0; i < CAKE_MAX_TINS; i++) {
> +               struct cake_tin_data *b = q->tins + i;
> +
> +               INIT_LIST_HEAD(&b->new_flows);
> +               INIT_LIST_HEAD(&b->old_flows);
> +               INIT_LIST_HEAD(&b->decaying_flows);
> +               b->sparse_flow_count = 0;
> +               b->bulk_flow_count = 0;
> +               b->decaying_flow_count = 0;
> +
> +               for (j = 0; j < CAKE_QUEUES; j++) {
> +                       struct cake_flow *flow = b->flows + j;
> +                       u32 k = j * CAKE_MAX_TINS + i;
> +
> +                       INIT_LIST_HEAD(&flow->flowchain);
> +                       cobalt_vars_init(&flow->cvars);
> +
> +                       q->overflow_heap[k].t = i;
> +                       q->overflow_heap[k].b = j;
> +                       b->overflow_idx[j] = k;
> +               }
> +       }
> +
> +       cake_reconfigure(sch);
> +       q->avg_peak_bandwidth = q->rate_bps;
> +       q->min_netlen = ~0;
> +       q->min_adjlen = ~0;
> +       return 0;
> +
> +nomem:
> +       cake_destroy(sch);
> +       return -ENOMEM;
> +}
> +
> +static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct nlattr *opts;
> +
> +       opts = nla_nest_start(skb, TCA_OPTIONS);
> +       if (!opts)
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_BASE_RATE, q->rate_bps))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_DIFFSERV_MODE, q->tin_mode))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_ATM, q->atm_mode))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_FLOW_MODE,
> +                       q->flow_mode & CAKE_FLOW_MASK))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_NAT,
> +                       !!(q->flow_mode & CAKE_FLOW_NAT_FLAG)))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_SPLIT_GSO,
> +                       !!(q->rate_flags & CAKE_FLAG_SPLIT_GSO)))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_WASH,
> +                       !!(q->rate_flags & CAKE_FLAG_WASH)))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_OVERHEAD, q->rate_overhead))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_MPU, q->rate_mpu))
> +               goto nla_put_failure;
> +
> +       if (!(q->rate_flags & CAKE_FLAG_OVERHEAD))
> +               if (nla_put_u32(skb, TCA_CAKE_RAW, 0))
> +                       goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_RTT, q->interval))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_TARGET, q->target))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_AUTORATE,
> +                       !!(q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS)))
> +               goto nla_put_failure;
> +
> +       if (nla_put_u32(skb, TCA_CAKE_INGRESS,
> +                       !!(q->rate_flags & CAKE_FLAG_INGRESS)))
> +               goto nla_put_failure;


Is this used by a following patch? If so, please move it there.



> +
> +       if (nla_put_u32(skb, TCA_CAKE_ACK_FILTER, q->ack_filter))
> +               goto nla_put_failure;


Ditto.



> +
> +       if (nla_put_u32(skb, TCA_CAKE_MEMORY, q->buffer_config_limit))
> +               goto nla_put_failure;
> +
> +       return nla_nest_end(skb, opts);
> +
> +nla_put_failure:
> +       return -1;
> +}
> +
> +static int cake_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
> +{
> +       struct cake_sched_data *q = qdisc_priv(sch);
> +       struct nlattr *stats = nla_nest_start(d->skb, TCA_STATS_APP);
> +       struct nlattr *tstats, *ts;
> +       int i;
> +
> +       if (!stats)
> +               return -1;
> +
> +#define PUT_STAT_U32(attr, data) do {                                 \
> +               if (nla_put_u32(d->skb, TCA_CAKE_STATS_ ## attr, data)) \
> +                       goto nla_put_failure;                          \
> +       } while (0)
> +
> +       PUT_STAT_U32(CAPACITY_ESTIMATE, q->avg_peak_bandwidth);
> +       PUT_STAT_U32(MEMORY_LIMIT, q->buffer_limit);
> +       PUT_STAT_U32(MEMORY_USED, q->buffer_max_used);
> +       PUT_STAT_U32(AVG_NETOFF, ((q->avg_netoff + 0x8000) >> 16));
> +       PUT_STAT_U32(MAX_NETLEN, q->max_netlen);
> +       PUT_STAT_U32(MAX_ADJLEN, q->max_adjlen);
> +       PUT_STAT_U32(MIN_NETLEN, q->min_netlen);
> +       PUT_STAT_U32(MIN_ADJLEN, q->min_adjlen);
> +
> +#undef PUT_STAT_U32
> +
> +       tstats = nla_nest_start(d->skb, TCA_CAKE_STATS_TIN_STATS);
> +       if (!tstats)
> +               goto nla_put_failure;
> +
> +#define PUT_TSTAT_U32(attr, data) do {                                 \
> +               if (nla_put_u32(d->skb, TCA_CAKE_TIN_STATS_ ## attr, data)) \
> +                       goto nla_put_failure;                           \
> +       } while (0)
> +#define PUT_TSTAT_U64(attr, data) do {                                 \
> +               if (nla_put_u64_64bit(d->skb, TCA_CAKE_TIN_STATS_ ## attr, \
> +                                       data, TCA_CAKE_TIN_STATS_PAD))  \
> +                       goto nla_put_failure;                           \
> +       } while (0)
> +
> +       for (i = 0; i < q->tin_cnt; i++) {
> +               struct cake_tin_data *b = &q->tins[i];
> +
> +               ts = nla_nest_start(d->skb, i + 1);
> +               if (!ts)
> +                       goto nla_put_failure;
> +
> +               PUT_TSTAT_U32(THRESHOLD_RATE, b->tin_rate_bps);
> +               PUT_TSTAT_U32(TARGET_US, cobalt_time_to_us(b->cparams.target));
> +               PUT_TSTAT_U32(INTERVAL_US,
> +                             cobalt_time_to_us(b->cparams.interval));
> +
> +               PUT_TSTAT_U32(SENT_PACKETS, b->packets);
> +               PUT_TSTAT_U64(SENT_BYTES64, b->bytes);
> +               PUT_TSTAT_U32(DROPPED_PACKETS, b->tin_dropped);
> +               PUT_TSTAT_U32(ECN_MARKED_PACKETS, b->tin_ecn_mark);
> +               PUT_TSTAT_U64(BACKLOG_BYTES64, b->tin_backlog);
> +               PUT_TSTAT_U32(ACKS_DROPPED_PACKETS, b->ack_drops);
> +
> +               PUT_TSTAT_U32(PEAK_DELAY_US, cobalt_time_to_us(b->peak_delay));
> +               PUT_TSTAT_U32(AVG_DELAY_US, cobalt_time_to_us(b->avge_delay));
> +               PUT_TSTAT_U32(BASE_DELAY_US, cobalt_time_to_us(b->base_delay));
> +
> +               PUT_TSTAT_U32(WAY_INDIRECT_HITS, b->way_hits);
> +               PUT_TSTAT_U32(WAY_MISSES, b->way_misses);
> +               PUT_TSTAT_U32(WAY_COLLISIONS, b->way_collisions);
> +
> +               PUT_TSTAT_U32(SPARSE_FLOWS, b->sparse_flow_count +
> +                                          b->decaying_flow_count);
> +               PUT_TSTAT_U32(BULK_FLOWS, b->bulk_flow_count);
> +               PUT_TSTAT_U32(UNRESPONSIVE_FLOWS, b->unresponsive_flow_count);
> +               PUT_TSTAT_U32(MAX_SKBLEN, b->max_skblen);
> +
> +               PUT_TSTAT_U32(FLOW_QUANTUM, b->flow_quantum);
> +               nla_nest_end(d->skb, ts);
> +       }
> +
> +#undef PUT_TSTAT_U32
> +#undef PUT_TSTAT_U64
> +
> +       nla_nest_end(d->skb, tstats);
> +       return nla_nest_end(d->skb, stats);
> +
> +nla_put_failure:
> +       nla_nest_cancel(d->skb, stats);
> +       return -1;
> +}
> +
> +static struct Qdisc_ops cake_qdisc_ops __read_mostly = {
> +       .id             =       "cake",
> +       .priv_size      =       sizeof(struct cake_sched_data),
> +       .enqueue        =       cake_enqueue,
> +       .dequeue        =       cake_dequeue,
> +       .peek           =       qdisc_peek_dequeued,
> +       .init           =       cake_init,
> +       .reset          =       cake_reset,
> +       .destroy        =       cake_destroy,
> +       .change         =       cake_change,
> +       .dump           =       cake_dump,
> +       .dump_stats     =       cake_dump_stats,
> +       .owner          =       THIS_MODULE,
> +};
> +
> +static int __init cake_module_init(void)
> +{
> +       return register_qdisc(&cake_qdisc_ops);
> +}
> +
> +static void __exit cake_module_exit(void)
> +{
> +       unregister_qdisc(&cake_qdisc_ops);
> +}
> +
> +module_init(cake_module_init)
> +module_exit(cake_module_exit)
> +MODULE_AUTHOR("Jonathan Morton");
> +MODULE_LICENSE("Dual BSD/GPL");
> +MODULE_DESCRIPTION("The CAKE shaper.");
>

Re: [PATCH net-next v8 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Toke Høiland-Jørgensen]

Thank you for the review! A few comments below, I'll fix the rest.

> [...]
> 
> So sch_cake doesn't accept normal tc filters? Is this intentional?
> If so, why?

For two reasons:

- The two-level scheduling used in CAKE (tins / diffserv classes, and
  flow hashing) does not map in an obvious way to the classification
  index of tc filters.

- No one has asked for it. We have done our best to accommodate the
  features people want in a home router qdisc directly in CAKE, and the
  ability to integrate tc filters has never been requested.

>> +static u16 quantum_div[CAKE_QUEUES + 1] = {0};
>> +
>> +#define REC_INV_SQRT_CACHE (16)
>> +static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0};
>> +
>> +/* http://en.wikipedia.org/wiki/Methods_of_computing_square_roots
>> + * new_invsqrt = (invsqrt / 2) * (3 - count * invsqrt^2)
>> + *
>> + * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32
>> + */
>> +
>> +static void cobalt_newton_step(struct cobalt_vars *vars)
>> +{
>> +       u32 invsqrt = vars->rec_inv_sqrt;
>> +       u32 invsqrt2 = ((u64)invsqrt * invsqrt) >> 32;
>> +       u64 val = (3LL << 32) - ((u64)vars->count * invsqrt2);
>> +
>> +       val >>= 2; /* avoid overflow in following multiply */
>> +       val = (val * invsqrt) >> (32 - 2 + 1);
>> +
>> +       vars->rec_inv_sqrt = val;
>> +}
>> +
>> +static void cobalt_invsqrt(struct cobalt_vars *vars)
>> +{
>> +       if (vars->count < REC_INV_SQRT_CACHE)
>> +               vars->rec_inv_sqrt = cobalt_rec_inv_sqrt_cache[vars->count];
>> +       else
>> +               cobalt_newton_step(vars);
>> +}
>
> Looks pretty much duplicated with codel...

Cobalt is derived from CoDel, and so naturally shares some features with
it. However, it is quite different in other respects, so we can't just
use the existing CoDel code for the parts that are similar. We don't
feel quite confident enough in Cobalt (yet) to propose it replace CoDel
everywhere else in the kernel; so we have elected to keep it internal to
CAKE instead.

>> [...]
>>
>> +static int cake_init(struct Qdisc *sch, struct nlattr *opt,
>> +                    struct netlink_ext_ack *extack)
>> +{
>> +       struct cake_sched_data *q = qdisc_priv(sch);
>> +       int i, j;
>> +
>> +       sch->limit = 10240;
>> +       q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
>> +       q->flow_mode  = CAKE_FLOW_TRIPLE;
>> +
>> +       q->rate_bps = 0; /* unlimited by default */
>> +
>> +       q->interval = 100000; /* 100ms default */
>> +       q->target   =   5000; /* 5ms: codel RFC argues
>> +                              * for 5 to 10% of interval
>> +                              */
>> +
>> +       q->cur_tin = 0;
>> +       q->cur_flow  = 0;
>> +
>> +       if (opt) {
>> +               int err = cake_change(sch, opt, extack);
>> +
>> +               if (err)
>> +                       return err;
>
>
> Not sure if you really want to reallocate q->tines below for this
> case.

I'm not sure what you mean here? If there's an error we return it and
the qdisc is not created. If there's not, we allocate and on subsequent
changes cake_change() will be called directly, or? Can the init function
ever be called again during the lifetime of the qdisc?

-Toke

Re: [PATCH net-next v8 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Cong Wang]

On Fri, May 4, 2018 at 12:10 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Thank you for the review! A few comments below, I'll fix the rest.
>
>> [...]
>>
>> So sch_cake doesn't accept normal tc filters? Is this intentional?
>> If so, why?
>
> For two reasons:
>
> - The two-level scheduling used in CAKE (tins / diffserv classes, and
>   flow hashing) does not map in an obvious way to the classification
>   index of tc filters.

Sounds like you need to extend struct tcf_result?

>
> - No one has asked for it. We have done our best to accommodate the
>   features people want in a home router qdisc directly in CAKE, and the
>   ability to integrate tc filters has never been requested.

It is not hard to integrate, basically you need to call tcf_classify().
Although it is not mandatory, it is odd to merge a qdisc doesn't work
with existing tc filters (and actions too).


>>> +static int cake_init(struct Qdisc *sch, struct nlattr *opt,
>>> +                    struct netlink_ext_ack *extack)
>>> +{
>>> +       struct cake_sched_data *q = qdisc_priv(sch);
>>> +       int i, j;
>>> +
>>> +       sch->limit = 10240;
>>> +       q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
>>> +       q->flow_mode  = CAKE_FLOW_TRIPLE;
>>> +
>>> +       q->rate_bps = 0; /* unlimited by default */
>>> +
>>> +       q->interval = 100000; /* 100ms default */
>>> +       q->target   =   5000; /* 5ms: codel RFC argues
>>> +                              * for 5 to 10% of interval
>>> +                              */
>>> +
>>> +       q->cur_tin = 0;
>>> +       q->cur_flow  = 0;
>>> +
>>> +       if (opt) {
>>> +               int err = cake_change(sch, opt, extack);
>>> +
>>> +               if (err)
>>> +                       return err;
>>
>>
>> Not sure if you really want to reallocate q->tines below for this
>> case.
>
> I'm not sure what you mean here? If there's an error we return it and
> the qdisc is not created. If there's not, we allocate and on subsequent
> changes cake_change() will be called directly, or? Can the init function
> ever be called again during the lifetime of the qdisc?
>

In non-error case, you call cake_change() first and then allocate ->tins
with kvzalloc() below. For me it looks like you don't need to allocate it
again when ->tins!=NULL.

Re: [PATCH net-next v8 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Toke Høiland-Jørgensen]

Cong Wang <xiyou.wangcong@gmail.com> writes:

> On Fri, May 4, 2018 at 12:10 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>> Thank you for the review! A few comments below, I'll fix the rest.
>>
>>> [...]
>>>
>>> So sch_cake doesn't accept normal tc filters? Is this intentional?
>>> If so, why?
>>
>> For two reasons:
>>
>> - The two-level scheduling used in CAKE (tins / diffserv classes, and
>>   flow hashing) does not map in an obvious way to the classification
>>   index of tc filters.
>
> Sounds like you need to extend struct tcf_result?

Well, the obvious way to support filters would be to have skb->priority
override the diffserv mapping if set, and have the filter classification
result select the queue within that tier. That would probably be doable,
but see below.

>> - No one has asked for it. We have done our best to accommodate the
>>   features people want in a home router qdisc directly in CAKE, and the
>>   ability to integrate tc filters has never been requested.
>
> It is not hard to integrate, basically you need to call
> tcf_classify(). Although it is not mandatory, it is odd to merge a
> qdisc doesn't work with existing tc filters (and actions too).

I looked at the fq_codel code to do this. Is it possible to support
filtering without implementing Qdisc_class_ops? If so, I'll give it a
shot; but implementing the class ops is more than I can commit to...

>>>> +static int cake_init(struct Qdisc *sch, struct nlattr *opt,
>>>> +                    struct netlink_ext_ack *extack)
>>>> +{
>>>> +       struct cake_sched_data *q = qdisc_priv(sch);
>>>> +       int i, j;
>>>> +
>>>> +       sch->limit = 10240;
>>>> +       q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
>>>> +       q->flow_mode  = CAKE_FLOW_TRIPLE;
>>>> +
>>>> +       q->rate_bps = 0; /* unlimited by default */
>>>> +
>>>> +       q->interval = 100000; /* 100ms default */
>>>> +       q->target   =   5000; /* 5ms: codel RFC argues
>>>> +                              * for 5 to 10% of interval
>>>> +                              */
>>>> +
>>>> +       q->cur_tin = 0;
>>>> +       q->cur_flow  = 0;
>>>> +
>>>> +       if (opt) {
>>>> +               int err = cake_change(sch, opt, extack);
>>>> +
>>>> +               if (err)
>>>> +                       return err;
>>>
>>>
>>> Not sure if you really want to reallocate q->tines below for this
>>> case.
>>
>> I'm not sure what you mean here? If there's an error we return it and
>> the qdisc is not created. If there's not, we allocate and on subsequent
>> changes cake_change() will be called directly, or? Can the init function
>> ever be called again during the lifetime of the qdisc?
>>
>
> In non-error case, you call cake_change() first and then allocate
> ->tins with kvzalloc() below. For me it looks like you don't need to
> allocate it again when ->tins!=NULL.

No, we definitely don't. It's just not clear to me how cake_init() could
ever be called with q->tins already allocated?

I can add a check in any case, though, I see that there is one in
fq_codel as well...

-Toke

Re: [PATCH net-next v8 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc

[Cong Wang]

On Mon, May 7, 2018 at 11:37 AM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Cong Wang <xiyou.wangcong@gmail.com> writes:
>
>> On Fri, May 4, 2018 at 12:10 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>>> Thank you for the review! A few comments below, I'll fix the rest.
>>>
>>>> [...]
>>>>
>>>> So sch_cake doesn't accept normal tc filters? Is this intentional?
>>>> If so, why?
>>>
>>> For two reasons:
>>>
>>> - The two-level scheduling used in CAKE (tins / diffserv classes, and
>>>   flow hashing) does not map in an obvious way to the classification
>>>   index of tc filters.
>>
>> Sounds like you need to extend struct tcf_result?
>
> Well, the obvious way to support filters would be to have skb->priority
> override the diffserv mapping if set, and have the filter classification
> result select the queue within that tier. That would probably be doable,
> but see below.
>
>>> - No one has asked for it. We have done our best to accommodate the
>>>   features people want in a home router qdisc directly in CAKE, and the
>>>   ability to integrate tc filters has never been requested.
>>
>> It is not hard to integrate, basically you need to call
>> tcf_classify(). Although it is not mandatory, it is odd to merge a
>> qdisc doesn't work with existing tc filters (and actions too).
>
> I looked at the fq_codel code to do this. Is it possible to support
> filtering without implementing Qdisc_class_ops? If so, I'll give it a
> shot; but implementing the class ops is more than I can commit to...

Good question. The tc classes in flow-based qdisc's are actually
used as flows rather than a normal tc class in a hierarchy qdisc.
Like in fq_code, the classes are mapped to each flow and because
of that we can dump stats of each flow.

I am not sure if you can totally bypass class_ops, you need to look
into these API's. Most of them are easy to implement, probably
only except the ->dump_stats(), so I don't think it is a barrier here.


>
>>>>> +static int cake_init(struct Qdisc *sch, struct nlattr *opt,
>>>>> +                    struct netlink_ext_ack *extack)
>>>>> +{
>>>>> +       struct cake_sched_data *q = qdisc_priv(sch);
>>>>> +       int i, j;
>>>>> +
>>>>> +       sch->limit = 10240;
>>>>> +       q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
>>>>> +       q->flow_mode  = CAKE_FLOW_TRIPLE;
>>>>> +
>>>>> +       q->rate_bps = 0; /* unlimited by default */
>>>>> +
>>>>> +       q->interval = 100000; /* 100ms default */
>>>>> +       q->target   =   5000; /* 5ms: codel RFC argues
>>>>> +                              * for 5 to 10% of interval
>>>>> +                              */
>>>>> +
>>>>> +       q->cur_tin = 0;
>>>>> +       q->cur_flow  = 0;
>>>>> +
>>>>> +       if (opt) {
>>>>> +               int err = cake_change(sch, opt, extack);
>>>>> +
>>>>> +               if (err)
>>>>> +                       return err;
>>>>
>>>>
>>>> Not sure if you really want to reallocate q->tines below for this
>>>> case.
>>>
>>> I'm not sure what you mean here? If there's an error we return it and
>>> the qdisc is not created. If there's not, we allocate and on subsequent
>>> changes cake_change() will be called directly, or? Can the init function
>>> ever be called again during the lifetime of the qdisc?
>>>
>>
>> In non-error case, you call cake_change() first and then allocate
>> ->tins with kvzalloc() below. For me it looks like you don't need to
>> allocate it again when ->tins!=NULL.
>
> No, we definitely don't. It's just not clear to me how cake_init() could
> ever be called with q->tins already allocated?
>
> I can add a check in any case, though, I see that there is one in
> fq_codel as well...

Ah, that's right, you have a check in cake_change() before
cake_reconfigure().