diff options
author | David S. Miller <davem@davemloft.net> | 2020-01-23 11:38:31 +0100 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2020-01-23 11:38:31 +0100 |
commit | 6d9f6e6790e794461ff3a16e4f3778f01ff6cca2 (patch) | |
tree | 2e5491ce3a3ff1b637edf2e24dce419d4c1da678 | |
parent | fd786fb1d2cad70b9aaba8c73872cbf63262bd58 (diff) | |
parent | ec97ecf1ebe485a17cd8395a5f35e6b80b57665a (diff) |
Merge branch 'net-sched-add-Flow-Queue-PIE-packet-scheduler'
Gautam Ramakrishnan says:
====================
net: sched: add Flow Queue PIE packet scheduler
Flow Queue PIE packet scheduler
This patch series implements the Flow Queue Proportional
Integral controller Enhanced (FQ-PIE) active queue
Management algorithm. It is an enhancement over the PIE
algorithm. It integrates the PIE aqm with a deficit round robin
scheme.
FQ-PIE is implemented over the latest version of PIE which
uses timestamps to calculate queue delay with an additional
option of using average dequeue rate to calculate the queue
delay. This patch also adds a memory limit of all the packets
across all queues to a default value of 32Mb.
- Patch #1
- Creates pie.h and moves all small functions and structures
common to PIE and FQ-PIE here. The functions are all made
inline.
- Patch #2 - #8
- Addresses code formatting, indentation, comment changes
and rearrangement of structure members.
- Patch #9
- Refactors sch_pie.c by changing arguments to
calculate_probability(), [pie_]drop_early() and
pie_process_dequeue() to make it generic enough to
be used by sch_fq_pie.c. These functions are exported
to be used by sch_fq_pie.c.
- Patch #10
- Adds the FQ-PIE Qdisc.
For more information:
https://tools.ietf.org/html/rfc8033
Changes from v6 to v7
- Call tcf_block_put() when destroying the Qdisc as suggested
by Jakub Kicinski.
Changes from v5 to v6
- Rearranged struct members according to their access pattern
and to remove holes.
Changes from v4 to v5
- This patch series breaks down patch 1 of v4 into
separate logical commits as suggested by David Miller.
Changes from v3 to v4
- Used non deprecated version of nla_parse_nested
- Used SZ_32M macro
- Removed an unused variable
- Code cleanup
All suggested by Jakub and Toke.
Changes from v2 to v3
- Exported drop_early, pie_process_dequeue and
calculate_probability functions from sch_pie as
suggested by Stephen Hemminger.
Changes from v1 ( and RFC patch) to v2
- Added timestamp to calculate queue delay as recommended
by Dave Taht
- Packet memory limit implemented as recommended by Toke.
- Added external classifier as recommended by Toke.
- Used NET_XMIT_CN instead of NET_XMIT_DROP as the return
value in the fq_pie_qdisc_enqueue function.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
-rw-r--r-- | include/net/pie.h | 138 | ||||
-rw-r--r-- | include/uapi/linux/pkt_sched.h | 31 | ||||
-rw-r--r-- | net/sched/Kconfig | 13 | ||||
-rw-r--r-- | net/sched/Makefile | 1 | ||||
-rw-r--r-- | net/sched/sch_fq_pie.c | 562 | ||||
-rw-r--r-- | net/sched/sch_pie.c | 289 |
6 files changed, 849 insertions, 185 deletions
diff --git a/include/net/pie.h b/include/net/pie.h new file mode 100644 index 000000000000..fd5a37cb7993 --- /dev/null +++ b/include/net/pie.h @@ -0,0 +1,138 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +#ifndef __NET_SCHED_PIE_H +#define __NET_SCHED_PIE_H + +#include <linux/ktime.h> +#include <linux/skbuff.h> +#include <linux/types.h> +#include <net/inet_ecn.h> +#include <net/pkt_sched.h> + +#define MAX_PROB U64_MAX +#define DTIME_INVALID U64_MAX +#define QUEUE_THRESHOLD 16384 +#define DQCOUNT_INVALID -1 +#define PIE_SCALE 8 + +/** + * struct pie_params - contains pie parameters + * @target: target delay in pschedtime + * @tudpate: interval at which drop probability is calculated + * @limit: total number of packets that can be in the queue + * @alpha: parameter to control drop probability + * @beta: parameter to control drop probability + * @ecn: is ECN marking of packets enabled + * @bytemode: is drop probability scaled based on pkt size + * @dq_rate_estimator: is Little's law used for qdelay calculation + */ +struct pie_params { + psched_time_t target; + u32 tupdate; + u32 limit; + u32 alpha; + u32 beta; + u8 ecn; + u8 bytemode; + u8 dq_rate_estimator; +}; + +/** + * struct pie_vars - contains pie variables + * @qdelay: current queue delay + * @qdelay_old: queue delay in previous qdelay calculation + * @burst_time: burst time allowance + * @dq_tstamp: timestamp at which dq rate was last calculated + * @prob: drop probability + * @accu_prob: accumulated drop probability + * @dq_count: number of bytes dequeued in a measurement cycle + * @avg_dq_rate: calculated average dq rate + * @qlen_old: queue length during previous qdelay calculation + * @accu_prob_overflows: number of times accu_prob overflows + */ +struct pie_vars { + psched_time_t qdelay; + psched_time_t qdelay_old; + psched_time_t burst_time; + psched_time_t dq_tstamp; + u64 prob; + u64 accu_prob; + u64 dq_count; + u32 avg_dq_rate; + u32 qlen_old; + u8 accu_prob_overflows; +}; + +/** + * struct pie_stats - contains pie stats + * @packets_in: total number of packets enqueued + * @dropped: packets dropped due to pie action + * @overlimit: packets dropped due to lack of space in queue + * @ecn_mark: packets marked with ECN + * @maxq: maximum queue size + */ +struct pie_stats { + u32 packets_in; + u32 dropped; + u32 overlimit; + u32 ecn_mark; + u32 maxq; +}; + +/** + * struct pie_skb_cb - contains private skb vars + * @enqueue_time: timestamp when the packet is enqueued + * @mem_usage: size of the skb during enqueue + */ +struct pie_skb_cb { + psched_time_t enqueue_time; + u32 mem_usage; +}; + +static inline void pie_params_init(struct pie_params *params) +{ + params->target = PSCHED_NS2TICKS(15 * NSEC_PER_MSEC); /* 15 ms */ + params->tupdate = usecs_to_jiffies(15 * USEC_PER_MSEC); /* 15 ms */ + params->limit = 1000; + params->alpha = 2; + params->beta = 20; + params->ecn = false; + params->bytemode = false; + params->dq_rate_estimator = false; +} + +static inline void pie_vars_init(struct pie_vars *vars) +{ + vars->burst_time = PSCHED_NS2TICKS(150 * NSEC_PER_MSEC); /* 150 ms */ + vars->dq_tstamp = DTIME_INVALID; + vars->accu_prob = 0; + vars->dq_count = DQCOUNT_INVALID; + vars->avg_dq_rate = 0; + vars->accu_prob_overflows = 0; +} + +static inline struct pie_skb_cb *get_pie_cb(const struct sk_buff *skb) +{ + qdisc_cb_private_validate(skb, sizeof(struct pie_skb_cb)); + return (struct pie_skb_cb *)qdisc_skb_cb(skb)->data; +} + +static inline psched_time_t pie_get_enqueue_time(const struct sk_buff *skb) +{ + return get_pie_cb(skb)->enqueue_time; +} + +static inline void pie_set_enqueue_time(struct sk_buff *skb) +{ + get_pie_cb(skb)->enqueue_time = psched_get_time(); +} + +bool pie_drop_early(struct Qdisc *sch, struct pie_params *params, + struct pie_vars *vars, u32 qlen, u32 packet_size); + +void pie_process_dequeue(struct sk_buff *skb, struct pie_params *params, + struct pie_vars *vars, u32 qlen); + +void pie_calculate_probability(struct pie_params *params, struct pie_vars *vars, + u32 qlen); + +#endif diff --git a/include/uapi/linux/pkt_sched.h b/include/uapi/linux/pkt_sched.h index bf5a5b1dfb0b..bbe791b24168 100644 --- a/include/uapi/linux/pkt_sched.h +++ b/include/uapi/linux/pkt_sched.h @@ -971,6 +971,37 @@ struct tc_pie_xstats { __u32 ecn_mark; /* packets marked with ecn*/ }; +/* FQ PIE */ +enum { + TCA_FQ_PIE_UNSPEC, + TCA_FQ_PIE_LIMIT, + TCA_FQ_PIE_FLOWS, + TCA_FQ_PIE_TARGET, + TCA_FQ_PIE_TUPDATE, + TCA_FQ_PIE_ALPHA, + TCA_FQ_PIE_BETA, + TCA_FQ_PIE_QUANTUM, + TCA_FQ_PIE_MEMORY_LIMIT, + TCA_FQ_PIE_ECN_PROB, + TCA_FQ_PIE_ECN, + TCA_FQ_PIE_BYTEMODE, + TCA_FQ_PIE_DQ_RATE_ESTIMATOR, + __TCA_FQ_PIE_MAX +}; +#define TCA_FQ_PIE_MAX (__TCA_FQ_PIE_MAX - 1) + +struct tc_fq_pie_xstats { + __u32 packets_in; /* total number of packets enqueued */ + __u32 dropped; /* packets dropped due to fq_pie_action */ + __u32 overlimit; /* dropped due to lack of space in queue */ + __u32 overmemory; /* dropped due to lack of memory in queue */ + __u32 ecn_mark; /* packets marked with ecn */ + __u32 new_flow_count; /* count of new flows created by packets */ + __u32 new_flows_len; /* count of flows in new list */ + __u32 old_flows_len; /* count of flows in old list */ + __u32 memory_usage; /* total memory across all queues */ +}; + /* CBS */ struct tc_cbs_qopt { __u8 offload; diff --git a/net/sched/Kconfig b/net/sched/Kconfig index b1e7ec726958..edde0e519438 100644 --- a/net/sched/Kconfig +++ b/net/sched/Kconfig @@ -366,6 +366,19 @@ config NET_SCH_PIE If unsure, say N. +config NET_SCH_FQ_PIE + depends on NET_SCH_PIE + tristate "Flow Queue Proportional Integral controller Enhanced (FQ-PIE)" + help + Say Y here if you want to use the Flow Queue Proportional Integral + controller Enhanced (FQ-PIE) packet scheduling algorithm. + For more information, please see https://tools.ietf.org/html/rfc8033 + + To compile this driver as a module, choose M here: the module + will be called sch_fq_pie. + + If unsure, say N. + config NET_SCH_INGRESS tristate "Ingress/classifier-action Qdisc" depends on NET_CLS_ACT diff --git a/net/sched/Makefile b/net/sched/Makefile index bc8856b865ff..31c367a6cd09 100644 --- a/net/sched/Makefile +++ b/net/sched/Makefile @@ -59,6 +59,7 @@ 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 +obj-$(CONFIG_NET_SCH_FQ_PIE) += sch_fq_pie.o obj-$(CONFIG_NET_SCH_CBS) += sch_cbs.o obj-$(CONFIG_NET_SCH_ETF) += sch_etf.o obj-$(CONFIG_NET_SCH_TAPRIO) += sch_taprio.o diff --git a/net/sched/sch_fq_pie.c b/net/sched/sch_fq_pie.c new file mode 100644 index 000000000000..bbd0dea6b6b9 --- /dev/null +++ b/net/sched/sch_fq_pie.c @@ -0,0 +1,562 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Flow Queue PIE discipline + * + * Copyright (C) 2019 Mohit P. Tahiliani <tahiliani@nitk.edu.in> + * Copyright (C) 2019 Sachin D. Patil <sdp.sachin@gmail.com> + * Copyright (C) 2019 V. Saicharan <vsaicharan1998@gmail.com> + * Copyright (C) 2019 Mohit Bhasi <mohitbhasi1998@gmail.com> + * Copyright (C) 2019 Leslie Monis <lesliemonis@gmail.com> + * Copyright (C) 2019 Gautam Ramakrishnan <gautamramk@gmail.com> + */ + +#include <linux/jhash.h> +#include <linux/sizes.h> +#include <linux/vmalloc.h> +#include <net/pkt_cls.h> +#include <net/pie.h> + +/* Flow Queue PIE + * + * Principles: + * - Packets are classified on flows. + * - This is a Stochastic model (as we use a hash, several flows might + * be hashed to the same slot) + * - Each flow has a PIE managed queue. + * - Flows are linked onto two (Round Robin) lists, + * so that new flows have priority on old ones. + * - For a given flow, packets are not reordered. + * - Drops during enqueue only. + * - ECN capability is off by default. + * - ECN threshold (if ECN is enabled) is at 10% by default. + * - Uses timestamps to calculate queue delay by default. + */ + +/** + * struct fq_pie_flow - contains data for each flow + * @vars: pie vars associated with the flow + * @deficit: number of remaining byte credits + * @backlog: size of data in the flow + * @qlen: number of packets in the flow + * @flowchain: flowchain for the flow + * @head: first packet in the flow + * @tail: last packet in the flow + */ +struct fq_pie_flow { + struct pie_vars vars; + s32 deficit; + u32 backlog; + u32 qlen; + struct list_head flowchain; + struct sk_buff *head; + struct sk_buff *tail; +}; + +struct fq_pie_sched_data { + struct tcf_proto __rcu *filter_list; /* optional external classifier */ + struct tcf_block *block; + struct fq_pie_flow *flows; + struct Qdisc *sch; + struct list_head old_flows; + struct list_head new_flows; + struct pie_params p_params; + u32 ecn_prob; + u32 flows_cnt; + u32 quantum; + u32 memory_limit; + u32 new_flow_count; + u32 memory_usage; + u32 overmemory; + struct pie_stats stats; + struct timer_list adapt_timer; +}; + +static unsigned int fq_pie_hash(const struct fq_pie_sched_data *q, + struct sk_buff *skb) +{ + return reciprocal_scale(skb_get_hash(skb), q->flows_cnt); +} + +static unsigned int fq_pie_classify(struct sk_buff *skb, struct Qdisc *sch, + int *qerr) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + struct tcf_proto *filter; + struct tcf_result res; + int result; + + if (TC_H_MAJ(skb->priority) == sch->handle && + TC_H_MIN(skb->priority) > 0 && + TC_H_MIN(skb->priority) <= q->flows_cnt) + return TC_H_MIN(skb->priority); + + filter = rcu_dereference_bh(q->filter_list); + if (!filter) + return fq_pie_hash(q, skb) + 1; + + *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; + result = tcf_classify(skb, filter, &res, false); + if (result >= 0) { +#ifdef CONFIG_NET_CLS_ACT + switch (result) { + case TC_ACT_STOLEN: + case TC_ACT_QUEUED: + case TC_ACT_TRAP: + *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN; + /* fall through */ + case TC_ACT_SHOT: + return 0; + } +#endif + if (TC_H_MIN(res.classid) <= q->flows_cnt) + return TC_H_MIN(res.classid); + } + return 0; +} + +/* add skb to flow queue (tail add) */ +static inline void flow_queue_add(struct fq_pie_flow *flow, + struct sk_buff *skb) +{ + if (!flow->head) + flow->head = skb; + else + flow->tail->next = skb; + flow->tail = skb; + skb->next = NULL; +} + +static int fq_pie_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch, + struct sk_buff **to_free) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + struct fq_pie_flow *sel_flow; + int uninitialized_var(ret); + u8 memory_limited = false; + u8 enqueue = false; + u32 pkt_len; + u32 idx; + + /* Classifies packet into corresponding flow */ + idx = fq_pie_classify(skb, sch, &ret); + sel_flow = &q->flows[idx]; + + /* Checks whether adding a new packet would exceed memory limit */ + get_pie_cb(skb)->mem_usage = skb->truesize; + memory_limited = q->memory_usage > q->memory_limit + skb->truesize; + + /* Checks if the qdisc is full */ + if (unlikely(qdisc_qlen(sch) >= sch->limit)) { + q->stats.overlimit++; + goto out; + } else if (unlikely(memory_limited)) { + q->overmemory++; + } + + if (!pie_drop_early(sch, &q->p_params, &sel_flow->vars, + sel_flow->backlog, skb->len)) { + enqueue = true; + } else if (q->p_params.ecn && + sel_flow->vars.prob <= (MAX_PROB / 100) * q->ecn_prob && + INET_ECN_set_ce(skb)) { + /* If packet is ecn capable, mark it if drop probability + * is lower than the parameter ecn_prob, else drop it. + */ + q->stats.ecn_mark++; + enqueue = true; + } + if (enqueue) { + /* Set enqueue time only when dq_rate_estimator is disabled. */ + if (!q->p_params.dq_rate_estimator) + pie_set_enqueue_time(skb); + + pkt_len = qdisc_pkt_len(skb); + q->stats.packets_in++; + q->memory_usage += skb->truesize; + sch->qstats.backlog += pkt_len; + sch->q.qlen++; + flow_queue_add(sel_flow, skb); + if (list_empty(&sel_flow->flowchain)) { + list_add_tail(&sel_flow->flowchain, &q->new_flows); + q->new_flow_count++; + sel_flow->deficit = q->quantum; + sel_flow->qlen = 0; + sel_flow->backlog = 0; + } + sel_flow->qlen++; + sel_flow->backlog += pkt_len; + return NET_XMIT_SUCCESS; + } +out: + q->stats.dropped++; + sel_flow->vars.accu_prob = 0; + sel_flow->vars.accu_prob_overflows = 0; + __qdisc_drop(skb, to_free); + qdisc_qstats_drop(sch); + return NET_XMIT_CN; +} + +static const struct nla_policy fq_pie_policy[TCA_FQ_PIE_MAX + 1] = { + [TCA_FQ_PIE_LIMIT] = {.type = NLA_U32}, + [TCA_FQ_PIE_FLOWS] = {.type = NLA_U32}, + [TCA_FQ_PIE_TARGET] = {.type = NLA_U32}, + [TCA_FQ_PIE_TUPDATE] = {.type = NLA_U32}, + [TCA_FQ_PIE_ALPHA] = {.type = NLA_U32}, + [TCA_FQ_PIE_BETA] = {.type = NLA_U32}, + [TCA_FQ_PIE_QUANTUM] = {.type = NLA_U32}, + [TCA_FQ_PIE_MEMORY_LIMIT] = {.type = NLA_U32}, + [TCA_FQ_PIE_ECN_PROB] = {.type = NLA_U32}, + [TCA_FQ_PIE_ECN] = {.type = NLA_U32}, + [TCA_FQ_PIE_BYTEMODE] = {.type = NLA_U32}, + [TCA_FQ_PIE_DQ_RATE_ESTIMATOR] = {.type = NLA_U32}, +}; + +static inline struct sk_buff *dequeue_head(struct fq_pie_flow *flow) +{ + struct sk_buff *skb = flow->head; + + flow->head = skb->next; + skb->next = NULL; + return skb; +} + +static struct sk_buff *fq_pie_qdisc_dequeue(struct Qdisc *sch) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + struct sk_buff *skb = NULL; + struct fq_pie_flow *flow; + struct list_head *head; + u32 pkt_len; + +begin: + head = &q->new_flows; + if (list_empty(head)) { + head = &q->old_flows; + if (list_empty(head)) + return NULL; + } + + flow = list_first_entry(head, struct fq_pie_flow, flowchain); + /* Flow has exhausted all its credits */ + if (flow->deficit <= 0) { + flow->deficit += q->quantum; + list_move_tail(&flow->flowchain, &q->old_flows); + goto begin; + } + + if (flow->head) { + skb = dequeue_head(flow); + pkt_len = qdisc_pkt_len(skb); + sch->qstats.backlog -= pkt_len; + sch->q.qlen--; + qdisc_bstats_update(sch, skb); + } + + if (!skb) { + /* force a pass through old_flows to prevent starvation */ + if (head == &q->new_flows && !list_empty(&q->old_flows)) + list_move_tail(&flow->flowchain, &q->old_flows); + else + list_del_init(&flow->flowchain); + goto begin; + } + + flow->qlen--; + flow->deficit -= pkt_len; + flow->backlog -= pkt_len; + q->memory_usage -= get_pie_cb(skb)->mem_usage; + pie_process_dequeue(skb, &q->p_params, &flow->vars, flow->backlog); + return skb; +} + +static int fq_pie_change(struct Qdisc *sch, struct nlattr *opt, + struct netlink_ext_ack *extack) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + struct nlattr *tb[TCA_FQ_PIE_MAX + 1]; + unsigned int len_dropped = 0; + unsigned int num_dropped = 0; + int err; + + if (!opt) + return -EINVAL; + + err = nla_parse_nested(tb, TCA_FQ_PIE_MAX, opt, fq_pie_policy, extack); + if (err < 0) + return err; + + sch_tree_lock(sch); + if (tb[TCA_FQ_PIE_LIMIT]) { + u32 limit = nla_get_u32(tb[TCA_FQ_PIE_LIMIT]); + + q->p_params.limit = limit; + sch->limit = limit; + } + if (tb[TCA_FQ_PIE_FLOWS]) { + if (q->flows) { + NL_SET_ERR_MSG_MOD(extack, + "Number of flows cannot be changed"); + goto flow_error; + } + q->flows_cnt = nla_get_u32(tb[TCA_FQ_PIE_FLOWS]); + if (!q->flows_cnt || q->flows_cnt > 65536) { + NL_SET_ERR_MSG_MOD(extack, + "Number of flows must be < 65536"); + goto flow_error; + } + } + + /* convert from microseconds to pschedtime */ + if (tb[TCA_FQ_PIE_TARGET]) { + /* target is in us */ + u32 target = nla_get_u32(tb[TCA_FQ_PIE_TARGET]); + + /* convert to pschedtime */ + q->p_params.target = + PSCHED_NS2TICKS((u64)target * NSEC_PER_USEC); + } + + /* tupdate is in jiffies */ + if (tb[TCA_FQ_PIE_TUPDATE]) + q->p_params.tupdate = + usecs_to_jiffies(nla_get_u32(tb[TCA_FQ_PIE_TUPDATE])); + + if (tb[TCA_FQ_PIE_ALPHA]) + q->p_params.alpha = nla_get_u32(tb[TCA_FQ_PIE_ALPHA]); + + if (tb[TCA_FQ_PIE_BETA]) + q->p_params.beta = nla_get_u32(tb[TCA_FQ_PIE_BETA]); + + if (tb[TCA_FQ_PIE_QUANTUM]) + q->quantum = nla_get_u32(tb[TCA_FQ_PIE_QUANTUM]); + + if (tb[TCA_FQ_PIE_MEMORY_LIMIT]) + q->memory_limit = nla_get_u32(tb[TCA_FQ_PIE_MEMORY_LIMIT]); + + if (tb[TCA_FQ_PIE_ECN_PROB]) + q->ecn_prob = nla_get_u32(tb[TCA_FQ_PIE_ECN_PROB]); + + if (tb[TCA_FQ_PIE_ECN]) + q->p_params.ecn = nla_get_u32(tb[TCA_FQ_PIE_ECN]); + + if (tb[TCA_FQ_PIE_BYTEMODE]) + q->p_params.bytemode = nla_get_u32(tb[TCA_FQ_PIE_BYTEMODE]); + + if (tb[TCA_FQ_PIE_DQ_RATE_ESTIMATOR]) + q->p_params.dq_rate_estimator = + nla_get_u32(tb[TCA_FQ_PIE_DQ_RATE_ESTIMATOR]); + + /* Drop excess packets if new limit is lower */ + while (sch->q.qlen > sch->limit) { + struct sk_buff *skb = fq_pie_qdisc_dequeue(sch); + + kfree_skb(skb); + len_dropped += qdisc_pkt_len(skb); + num_dropped += 1; + } + qdisc_tree_reduce_backlog(sch, num_dropped, len_dropped); + + sch_tree_unlock(sch); + return 0; + +flow_error: + sch_tree_unlock(sch); + return -EINVAL; +} + +static void fq_pie_timer(struct timer_list *t) +{ + struct fq_pie_sched_data *q = from_timer(q, t, adapt_timer); + struct Qdisc *sch = q->sch; + spinlock_t *root_lock; /* to lock qdisc for probability calculations */ + u16 idx; + + root_lock = qdisc_lock(qdisc_root_sleeping(sch)); + spin_lock(root_lock); + + for (idx = 0; idx < q->flows_cnt; idx++) + pie_calculate_probability(&q->p_params, &q->flows[idx].vars, + q->flows[idx].backlog); + + /* reset the timer to fire after 'tupdate' jiffies. */ + if (q->p_params.tupdate) + mod_timer(&q->adapt_timer, jiffies + q->p_params.tupdate); + + spin_unlock(root_lock); +} + +static int fq_pie_init(struct Qdisc *sch, struct nlattr *opt, + struct netlink_ext_ack *extack) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + int err; + u16 idx; + + pie_params_init(&q->p_params); + sch->limit = 10 * 1024; + q->p_params.limit = sch->limit; + q->quantum = psched_mtu(qdisc_dev(sch)); + q->sch = sch; + q->ecn_prob = 10; + q->flows_cnt = 1024; + q->memory_limit = SZ_32M; + + INIT_LIST_HEAD(&q->new_flows); + INIT_LIST_HEAD(&q->old_flows); + + if (opt) { + err = fq_pie_change(sch, opt, extack); + + if (err) + return err; + } + + err = tcf_block_get(&q->block, &q->filter_list, sch, extack); + if (err) + goto init_failure; + + q->flows = kvcalloc(q->flows_cnt, sizeof(struct fq_pie_flow), + GFP_KERNEL); + if (!q->flows) { + err = -ENOMEM; + goto init_failure; + } + for (idx = 0; idx < q->flows_cnt; idx++) { + struct fq_pie_flow *flow = q->flows + idx; + + INIT_LIST_HEAD(&flow->flowchain); + pie_vars_init(&flow->vars); + } + + timer_setup(&q->adapt_timer, fq_pie_timer, 0); + mod_timer(&q->adapt_timer, jiffies + HZ / 2); + + return 0; + +init_failure: + q->flows_cnt = 0; + + return err; +} + +static int fq_pie_dump(struct Qdisc *sch, struct sk_buff *skb) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + struct nlattr *opts; + + opts = nla_nest_start(skb, TCA_OPTIONS); + if (!opts) + return -EMSGSIZE; + + /* convert target from pschedtime to us */ + if (nla_put_u32(skb, TCA_FQ_PIE_LIMIT, sch->limit) || + nla_put_u32(skb, TCA_FQ_PIE_FLOWS, q->flows_cnt) || + nla_put_u32(skb, TCA_FQ_PIE_TARGET, + ((u32)PSCHED_TICKS2NS(q->p_params.target)) / + NSEC_PER_USEC) || + nla_put_u32(skb, TCA_FQ_PIE_TUPDATE, + jiffies_to_usecs(q->p_params.tupdate)) || + nla_put_u32(skb, TCA_FQ_PIE_ALPHA, q->p_params.alpha) || + nla_put_u32(skb, TCA_FQ_PIE_BETA, q->p_params.beta) || + nla_put_u32(skb, TCA_FQ_PIE_QUANTUM, q->quantum) || + nla_put_u32(skb, TCA_FQ_PIE_MEMORY_LIMIT, q->memory_limit) || + nla_put_u32(skb, TCA_FQ_PIE_ECN_PROB, q->ecn_prob) || + nla_put_u32(skb, TCA_FQ_PIE_ECN, q->p_params.ecn) || + nla_put_u32(skb, TCA_FQ_PIE_BYTEMODE, q->p_params.bytemode) || + nla_put_u32(skb, TCA_FQ_PIE_DQ_RATE_ESTIMATOR, + q->p_params.dq_rate_estimator)) + goto nla_put_failure; + + return nla_nest_end(skb, opts); + +nla_put_failure: + nla_nest_cancel(skb, opts); + return -EMSGSIZE; +} + +static int fq_pie_dump_stats(struct Qdisc *sch, struct gnet_dump *d) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + struct tc_fq_pie_xstats st = { + .packets_in = q->stats.packets_in, + .overlimit = q->stats.overlimit, + .overmemory = q->overmemory, + .dropped = q->stats.dropped, + .ecn_mark = q->stats.ecn_mark, + .new_flow_count = q->new_flow_count, + .memory_usage = q->memory_usage, + }; + struct list_head *pos; + + sch_tree_lock(sch); + list_for_each(pos, &q->new_flows) + st.new_flows_len++; + + list_for_each(pos, &q->old_flows) + st.old_flows_len++; + sch_tree_unlock(sch); + + return gnet_stats_copy_app(d, &st, sizeof(st)); +} + +static void fq_pie_reset(struct Qdisc *sch) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + u16 idx; + + INIT_LIST_HEAD(&q->new_flows); + INIT_LIST_HEAD(&q->old_flows); + for (idx = 0; idx < q->flows_cnt; idx++) { + struct fq_pie_flow *flow = q->flows + idx; + + /* Removes all packets from flow */ + rtnl_kfree_skbs(flow->head, flow->tail); + flow->head = NULL; + + INIT_LIST_HEAD(&flow->flowchain); + pie_vars_init(&flow->vars); + } + + sch->q.qlen = 0; + sch->qstats.backlog = 0; +} + +static void fq_pie_destroy(struct Qdisc *sch) +{ + struct fq_pie_sched_data *q = qdisc_priv(sch); + + tcf_block_put(q->block); + del_timer_sync(&q->adapt_timer); + kvfree(q->flows); +} + +static struct Qdisc_ops fq_pie_qdisc_ops __read_mostly = { + .id = "fq_pie", + .priv_size = sizeof(struct fq_pie_sched_data), + .enqueue = fq_pie_qdisc_enqueue, + .dequeue = fq_pie_qdisc_dequeue, + .peek = qdisc_peek_dequeued, + .init = fq_pie_init, + .destroy = fq_pie_destroy, + .reset = fq_pie_reset, + .change = fq_pie_change, + .dump = fq_pie_dump, + .dump_stats = fq_pie_dump_stats, + .owner = THIS_MODULE, +}; + +static int __init fq_pie_module_init(void) +{ + return register_qdisc(&fq_pie_qdisc_ops); +} + +static void __exit fq_pie_module_exit(void) +{ + unregister_qdisc(&fq_pie_qdisc_ops); +} + +module_init(fq_pie_module_init); +module_exit(fq_pie_module_exit); + +MODULE_DESCRIPTION("Flow Queue Proportional Integral controller Enhanced (FQ-PIE)"); +MODULE_AUTHOR("Mohit P. Tahiliani"); +MODULE_LICENSE("GPL"); diff --git a/net/sched/sch_pie.c b/net/sched/sch_pie.c index b0b0dc46af61..915bcdb59a9f 100644 --- a/net/sched/sch_pie.c +++ b/net/sched/sch_pie.c @@ -19,159 +19,76 @@ #include <linux/skbuff.h> #include <net/pkt_sched.h> #include <net/inet_ecn.h> - -#define QUEUE_THRESHOLD 16384 -#define DQCOUNT_INVALID -1 -#define DTIME_INVALID 0xffffffffffffffff -#define MAX_PROB 0xffffffffffffffff -#define PIE_SCALE 8 - -/* parameters used */ -struct pie_params { - psched_time_t target; /* user specified target delay in pschedtime */ - u32 tupdate; /* timer frequency (in jiffies) */ - u32 limit; /* number of packets that can be enqueued */ - u32 alpha; /* alpha and beta are between 0 and 32 */ - u32 beta; /* and are used for shift relative to 1 */ - bool ecn; /* true if ecn is enabled */ - bool bytemode; /* to scale drop early prob based on pkt size */ - u8 dq_rate_estimator; /* to calculate delay using Little's law */ -}; - -/* variables used */ -struct pie_vars { - u64 prob; /* probability but scaled by u64 limit. */ - psched_time_t burst_time; - psched_time_t qdelay; - psched_time_t qdelay_old; - u64 dq_count; /* measured in bytes */ - psched_time_t dq_tstamp; /* drain rate */ - u64 accu_prob; /* accumulated drop probability */ - u32 avg_dq_rate; /* bytes per pschedtime tick,scaled */ - u32 qlen_old; /* in bytes */ - u8 accu_prob_overflows; /* overflows of accu_prob */ -}; - -/* statistics gathering */ -struct pie_stats { - u32 packets_in; /* total number of packets enqueued */ - u32 dropped; /* packets dropped due to pie_action */ - u32 overlimit; /* dropped due to lack of space in queue */ - u32 maxq; /* maximum queue size */ - u32 ecn_mark; /* packets marked with ECN */ -}; +#include <net/pie.h> /* private data for the Qdisc */ struct pie_sched_data { - struct pie_params params; struct pie_vars vars; + struct pie_params params; struct pie_stats stats; struct timer_list adapt_timer; struct Qdisc *sch; }; -static void pie_params_init(struct pie_params *params) +bool pie_drop_early(struct Qdisc *sch, struct pie_params *params, + struct pie_vars *vars, u32 qlen, u32 packet_size) { - params->alpha = 2; - params->beta = 20; - params->tupdate = usecs_to_jiffies(15 * USEC_PER_MSEC); /* 15 ms */ - params->limit = 1000; /* default of 1000 packets */ - params->target = PSCHED_NS2TICKS(15 * NSEC_PER_MSEC); /* 15 ms */ - params->ecn = false; - params->bytemode = false; - params->dq_rate_estimator = false; -} - -/* private skb vars */ -struct pie_skb_cb { - psched_time_t enqueue_time; -}; - -static struct pie_skb_cb *get_pie_cb(const struct sk_buff *skb) -{ - qdisc_cb_private_validate(skb, sizeof(struct pie_skb_cb)); - return (struct pie_skb_cb *)qdisc_skb_cb(skb)->data; -} - -static psched_time_t pie_get_enqueue_time(const struct sk_buff *skb) -{ - return get_pie_cb(skb)->enqueue_time; -} - -static void pie_set_enqueue_time(struct sk_buff *skb) -{ - get_pie_cb(skb)->enqueue_time = psched_get_time(); -} - -static void pie_vars_init(struct pie_vars *vars) -{ - vars->dq_count = DQCOUNT_INVALID; - vars->dq_tstamp = DTIME_INVALID; - vars->accu_prob = 0; - vars->avg_dq_rate = 0; - /* default of 150 ms in pschedtime */ - vars->burst_time = PSCHED_NS2TICKS(150 * NSEC_PER_MSEC); - vars->accu_prob_overflows = 0; -} - -static bool drop_early(struct Qdisc *sch, u32 packet_size) -{ - struct pie_sched_data *q = qdisc_priv(sch); u64 rnd; - u64 local_prob = q->vars.prob; + u64 local_prob = vars->prob; u32 mtu = psched_mtu(qdisc_dev(sch)); /* If there is still burst allowance left skip random early drop */ - if (q->vars.burst_time > 0) + if (vars->burst_time > 0) return false; /* If current delay is less than half of target, and * if drop prob is low already, disable early_drop */ - if ((q->vars.qdelay < q->params.target / 2) && - (q->vars.prob < MAX_PROB / 5)) + if ((vars->qdelay < params->target / 2) && + (vars->prob < MAX_PROB / 5)) return false; - /* If we have fewer than 2 mtu-sized packets, disable drop_early, + /* If we have fewer than 2 mtu-sized packets, disable pie_drop_early, * similar to min_th in RED */ - if (sch->qstats.backlog < 2 * mtu) + if (qlen < 2 * mtu) return false; /* If bytemode is turned on, use packet size to compute new * probablity. Smaller packets will have lower drop prob in this case */ - if (q->params.bytemode && packet_size <= mtu) + if (params->bytemode && packet_size <= mtu) local_prob = (u64)packet_size * div_u64(local_prob, mtu); else - local_prob = q->vars.prob; + local_prob = vars->prob; if (local_prob == 0) { - q->vars.accu_prob = 0; - q->vars.accu_prob_overflows = 0; + vars->accu_prob = 0; + vars->accu_prob_overflows = 0; } - if (local_prob > MAX_PROB - q->vars.accu_prob) - q->vars.accu_prob_overflows++; + if (local_prob > MAX_PROB - vars->accu_prob) + vars->accu_prob_overflows++; - q->vars.accu_prob += local_prob; + vars->accu_prob += local_prob; - if (q->vars.accu_prob_overflows == 0 && - q->vars.accu_prob < (MAX_PROB / 100) * 85) + if (vars->accu_prob_overflows == 0 && + vars->accu_prob < (MAX_PROB / 100) * 85) return false; - if (q->vars.accu_prob_overflows == 8 && - q->vars.accu_prob >= MAX_PROB / 2) + if (vars->accu_prob_overflows == 8 && + vars->accu_prob >= MAX_PROB / 2) return true; prandom_bytes(&rnd, 8); if (rnd < local_prob) { - q->vars.accu_prob = 0; - q->vars.accu_prob_overflows = 0; + vars->accu_prob = 0; + vars->accu_prob_overflows = 0; return true; } return false; } +EXPORT_SYMBOL_GPL(pie_drop_early); static int pie_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) @@ -184,7 +101,8 @@ static int pie_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch, goto out; } - if (!drop_early(sch, skb->len)) { + if (!pie_drop_early(sch, &q->params, &q->vars, sch->qstats.backlog, + skb->len)) { enqueue = true; } else if (q->params.ecn && (q->vars.prob <= MAX_PROB / 10) && INET_ECN_set_ce(skb)) { @@ -216,14 +134,14 @@ out: } static const struct nla_policy pie_policy[TCA_PIE_MAX + 1] = { - [TCA_PIE_TARGET] = {.type = NLA_U32}, - [TCA_PIE_LIMIT] = {.type = NLA_U32}, - [TCA_PIE_TUPDATE] = {.type = NLA_U32}, - [TCA_PIE_ALPHA] = {.type = NLA_U32}, - [TCA_PIE_BETA] = {.type = NLA_U32}, - [TCA_PIE_ECN] = {.type = NLA_U32}, - [TCA_PIE_BYTEMODE] = {.type = NLA_U32}, - [TCA_PIE_DQ_RATE_ESTIMATOR] = {.type = NLA_U32}, + [TCA_PIE_TARGET] = {.type = NLA_U32}, + [TCA_PIE_LIMIT] = {.type = NLA_U32}, + [TCA_PIE_TUPDATE] = {.type = NLA_U32}, + [TCA_PIE_ALPHA] = {.type = NLA_U32}, + [TCA_PIE_BETA] = {.type = NLA_U32}, + [TCA_PIE_ECN] = {.type = NLA_U32}, + [TCA_PIE_BYTEMODE] = {.type = NLA_U32}, + [TCA_PIE_DQ_RATE_ESTIMATOR] = {.type = NLA_U32}, }; static int pie_change(struct Qdisc *sch, struct nlattr *opt, @@ -296,26 +214,25 @@ static int pie_change(struct Qdisc *sch, struct nlattr *opt, return 0; } -static void pie_process_dequeue(struct Qdisc *sch, struct sk_buff *skb) +void pie_process_dequeue(struct sk_buff *skb, struct pie_params *params, + struct pie_vars *vars, u32 qlen) { - struct pie_sched_data *q = qdisc_priv(sch); - int qlen = sch->qstats.backlog; /* current queue size in bytes */ psched_time_t now = psched_get_time(); u32 dtime = 0; /* If dq_rate_estimator is disabled, calculate qdelay using the * packet timestamp. */ - if (!q->params.dq_rate_estimator) { - q->vars.qdelay = now - pie_get_enqueue_time(skb); + if (!params->dq_rate_estimator) { + vars->qdelay = now - pie_get_enqueue_time(skb); - if (q->vars.dq_tstamp != DTIME_INVALID) - dtime = now - q->vars.dq_tstamp; + if (vars->dq_tstamp != DTIME_INVALID) + dtime = now - vars->dq_tstamp; - q->vars.dq_tstamp = now; + vars->dq_tstamp = now; if (qlen == 0) - q->vars.qdelay = 0; + vars->qdelay = 0; if (dtime == 0) return; @@ -327,39 +244,39 @@ static void pie_process_dequeue(struct Qdisc *sch, struct sk_buff *skb) * we have enough packets to calculate the drain rate. Save |