Patchwork [RFC] sched: QFQ - quick fair queue scheduler

login
register
mail settings
Submitter stephen hemminger
Date Jan. 7, 2011, 3:56 a.m.
Message ID <20110106195614.20dbc402@nehalam>
Download mbox | patch
Permalink /patch/77845/
State RFC
Delegated to: David Miller
Headers show

Comments

stephen hemminger - Jan. 7, 2011, 3:56 a.m.
This is an implementation of the Quick Fair Queue scheduler developed
by Fabio Checconi and Luigi Rizzo. The same algorithm is already implemented in ipfw
in FreeBSD. Fabio had an earlier version developed on Linux, I just
did some cleanup, and backporting of FreeBSD version.

For more information see web page: http://info.iet.unipi.it/~luigi/qfq/
and Google tech talk: http://www.youtube.com/watch?v=r8vBmybeKlE

This is for inspection at this point, barely tested. 

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>

---
Patch against net-next-2.6.
Configuration may get patch fuzz because of testing CHOKe in
same tree.

 include/linux/pkt_sched.h |   14 
 net/sched/Kconfig         |   11 
 net/sched/Makefile        |    1 
 net/sched/sch_qfq.c       | 1012 ++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 1038 insertions(+)

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Changli Gao - Jan. 8, 2011, 2:56 a.m.
On Fri, Jan 7, 2011 at 11:56 AM, Stephen Hemminger
<shemminger@vyatta.com> wrote:
> This is an implementation of the Quick Fair Queue scheduler developed
> by Fabio Checconi and Luigi Rizzo. The same algorithm is already implemented in ipfw
> in FreeBSD. Fabio had an earlier version developed on Linux, I just
> did some cleanup, and backporting of FreeBSD version.
>
> For more information see web page: http://info.iet.unipi.it/~luigi/qfq/
> and Google tech talk: http://www.youtube.com/watch?v=r8vBmybeKlE
>
> This is for inspection at this point, barely tested.
>
> Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
>
> ---
> Patch against net-next-2.6.
> Configuration may get patch fuzz because of testing CHOKe in
> same tree.
>
>  include/linux/pkt_sched.h |   14
>  net/sched/Kconfig         |   11
>  net/sched/Makefile        |    1
>  net/sched/sch_qfq.c       | 1012 ++++++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 1038 insertions(+)
>
> --- a/include/linux/pkt_sched.h 2011-01-05 09:01:33.268032043 -0800
> +++ b/include/linux/pkt_sched.h 2011-01-05 23:17:20.637390255 -0800
> @@ -481,4 +481,18 @@ struct tc_drr_stats {
>        __u32   deficit;
>  };
>
> +/* QFQ */
> +enum {
> +       TCA_QFQ_WEIGHT,
> +       TCA_QFQ_LMAX,
> +       __TCA_QFQ_MAX
> +};
> +
> +#define TCA_QFQ_MAX    (__TCA_QFQ_MAX - 1)
> +
> +struct tc_qfq_stats {
> +       __u32 weight;
> +       __u32 lmax;
> +};
> +
>  #endif
> --- a/net/sched/Kconfig 2011-01-05 09:01:33.280032462 -0800
> +++ b/net/sched/Kconfig 2011-01-05 23:17:20.637390255 -0800
> @@ -216,6 +216,17 @@ config NET_SCH_CHOKE
>          To compile this code as a module, choose M here: the
>          module will be called sch_choke.
>
> +config NET_SCH_QFQ
> +        tristate "Quick Fair Queueing Scheduler (QFQ)"
> +       help
> +         Say Y here if you want to use the Quick Fair Queueing Scheduler (QFQ)
> +         packet scheduling algorithm.
> +
> +         To compile this driver as a module, choose M here: the module
> +         will be called sch_qfq.
> +
> +         If unsure, say N.
> +
>  config NET_SCH_INGRESS
>        tristate "Ingress Qdisc"
>        depends on NET_CLS_ACT
> --- a/net/sched/Makefile        2011-01-05 09:01:33.284032598 -0800
> +++ b/net/sched/Makefile        2011-01-05 23:17:20.645389829 -0800
> @@ -32,6 +32,7 @@ obj-$(CONFIG_NET_SCH_MULTIQ)  += sch_mult
>  obj-$(CONFIG_NET_SCH_ATM)      += sch_atm.o
>  obj-$(CONFIG_NET_SCH_NETEM)    += sch_netem.o
>  obj-$(CONFIG_NET_SCH_DRR)      += sch_drr.o
> +obj-$(CONFIG_NET_SCH_QFQ)      += sch_qfq.o
>  obj-$(CONFIG_NET_SCH_CHOKE)    += sch_choke.o
>  obj-$(CONFIG_NET_CLS_U32)      += cls_u32.o
>  obj-$(CONFIG_NET_CLS_ROUTE4)   += cls_route.o
> --- /dev/null   1970-01-01 00:00:00.000000000 +0000
> +++ b/net/sched/sch_qfq.c       2011-01-06 12:51:28.498280327 -0800
> @@ -0,0 +1,1125 @@
> +/*
> + * net/sched/sch_qfq.c         Quick Fair Queueing Scheduler.
> + *
> + * Copyright (c) 2009 Fabio Checconi, Luigi Rizzo, and Paolo Valente.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * version 2 as published by the Free Software Foundation.
> + */
> +
> +#include <linux/module.h>
> +#include <linux/init.h>
> +#include <linux/bitops.h>
> +#include <linux/errno.h>
> +#include <linux/netdevice.h>
> +#include <linux/pkt_sched.h>
> +#include <net/sch_generic.h>
> +#include <net/pkt_sched.h>
> +#include <net/pkt_cls.h>
> +
> +/*  Quick Fair Queueing
> +    ===================
> +
> +    Sources:
> +    Fabio Checconi and Scuola Superiore and S. Anna
> +    and Paolo Valente and Luigi Riz "QFQ: Efficient Packet Scheduling
> +    with Tight Bandwidth Distribution Guarantees", SIGCOMM 2010
> +
> +    See also:
> +    http://retis.sssup.it/~fabio/linux/qfq/
> + */
> +
> +/*
> +
> +  Virtual time computations.
> +
> +  S, F and V are all computed in fixed point arithmetic with
> +  FRAC_BITS decimal bits.
> +
> +  QFQ_MAX_INDEX is the maximum index allowed for a group. We need
> +       one bit per index.
> +  QFQ_MAX_WSHIFT is the maximum power of two supported as a weight.
> +
> +  The layout of the bits is as below:
> +
> +                   [ MTU_SHIFT ][      FRAC_BITS    ]
> +                   [ MAX_INDEX    ][ MIN_SLOT_SHIFT ]
> +                                ^.__grp->index = 0
> +                                *.__grp->slot_shift
> +
> +  where MIN_SLOT_SHIFT is derived by difference from the others.
> +
> +  The max group index corresponds to Lmax/w_min, where
> +  Lmax=1<<MTU_SHIFT, w_min = 1 .
> +  From this, and knowing how many groups (MAX_INDEX) we want,
> +  we can derive the shift corresponding to each group.
> +
> +  Because we often need to compute
> +       F = S + len/w_i  and V = V + len/wsum
> +  instead of storing w_i store the value
> +       inv_w = (1<<FRAC_BITS)/w_i
> +  so we can do F = S + len * inv_w * wsum.
> +  We use W_TOT in the formulas so we can easily move between
> +  static and adaptive weight sum.
> +
> +  The per-scheduler-instance data contain all the data structures
> +  for the scheduler: bitmaps and bucket lists.
> +
> + */
> +
> +/*
> + * Maximum number of consecutive slots occupied by backlogged classes
> + * inside a group.
> + */
> +#define QFQ_MAX_SLOTS  32
> +
> +/*
> + * Shifts used for class<->group mapping.  We allow class weights that are
> + * in the range [1, 2^MAX_WSHIFT], and we try to map each class i to the
> + * group with the smallest index that can support the L_i / r_i configured
> + * for the class.
> + *
> + * grp->index is the index of the group; and grp->slot_shift
> + * is the shift for the corresponding (scaled) sigma_i.
> + */
> +#define QFQ_MAX_INDEX          19
> +#define QFQ_MAX_WSHIFT         16
> +
> +#define        QFQ_MAX_WEIGHT          (1<<QFQ_MAX_WSHIFT)
> +#define QFQ_MAX_WSUM           (2*QFQ_MAX_WEIGHT)
> +
> +#define FRAC_BITS              30      /* fixed point arithmetic */
> +#define ONE_FP                 (1UL << FRAC_BITS)
> +#define IWSUM                  (ONE_FP/QFQ_MAX_WSUM)
> +
> +#define QFQ_MTU_SHIFT          11
> +#define QFQ_MIN_SLOT_SHIFT     (FRAC_BITS + QFQ_MTU_SHIFT - QFQ_MAX_INDEX)
> +
> +/*
> + * Possible group states.  These values are used as indexes for the bitmaps
> + * array of struct qfq_queue.
> + */
> +enum qfq_state { ER, IR, EB, IB, QFQ_MAX_STATE };
> +
> +struct qfq_group;
> +
> +struct qfq_class {
> +       struct Qdisc_class_common common;
> +
> +       unsigned int refcnt;
> +       unsigned int filter_cnt;
> +
> +       struct gnet_stats_basic_packed bstats;
> +       struct gnet_stats_queue qstats;
> +       struct gnet_stats_rate_est rate_est;
> +       struct Qdisc *qdisc;
> +
> +       struct qfq_class *next; /* Link for the slot list. */
> +       u64 S, F;               /* flow timestamps (exact) */
> +
> +       /* group we belong to. In principle we would need the index,
> +        * which is log_2(lmax/weight), but we never reference it
> +        * directly, only the group.
> +        */
> +       struct qfq_group *grp;
> +
> +       /* these are copied from the flowset. */
> +       u32     inv_w;          /* ONE_FP/weight */
> +       u32     lmax;           /* Max packet size for this flow. */
> +};
> +
> +struct qfq_group {
> +       uint64_t S, F;                  /* group timestamps (approx). */
> +       unsigned int slot_shift;        /* Slot shift. */
> +       unsigned int index;             /* Group index. */
> +       unsigned int front;             /* Index of the front slot. */
> +       unsigned long full_slots;       /* non-empty slots */
> +
> +       /* Array of RR lists of active classes. */
> +       struct qfq_class *slots[QFQ_MAX_SLOTS];
> +};
> +
> +struct qfq_sched {
> +       struct tcf_proto *filter_list;
> +       struct Qdisc_class_hash clhash;
> +
> +       uint64_t        V;              /* Precise virtual time. */
> +       u32 wsum;                       /* weight sum */
> +
> +       unsigned long bitmaps[QFQ_MAX_STATE];       /* Group bitmaps. */
> +       struct qfq_group groups[QFQ_MAX_INDEX + 1]; /* The groups. */
> +};
> +
> +static struct qfq_class *qfq_find_class(struct Qdisc *sch, u32 classid)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct Qdisc_class_common *clc;
> +
> +       clc = qdisc_class_find(&q->clhash, classid);
> +       if (clc == NULL)
> +               return NULL;
> +       return container_of(clc, struct qfq_class, common);
> +}
> +
> +static void qfq_purge_queue(struct qfq_class *cl)
> +{
> +       unsigned int len = cl->qdisc->q.qlen;
> +
> +       qdisc_reset(cl->qdisc);
> +       qdisc_tree_decrease_qlen(cl->qdisc, len);
> +}
> +
> +static const struct nla_policy qfq_policy[TCA_QFQ_MAX + 1] = {
> +       [TCA_QFQ_WEIGHT] = { .type = NLA_U32 },
> +       [TCA_QFQ_LMAX] = { .type = NLA_U32 },
> +};
> +
> +/*
> + * Calculate a flow index, given its weight and maximum packet length.
> + * index = log_2(maxlen/weight) but we need to apply the scaling.
> + * This is used only once at flow creation.
> + */
> +static int qfq_calc_index(u32 inv_w, unsigned int maxlen)
> +{
> +       u64 slot_size = (u64)maxlen *inv_w;
> +       unsigned long size_map;
> +       int index = 0;
> +
> +       size_map = slot_size >> QFQ_MIN_SLOT_SHIFT;
> +       if (!size_map)
> +               goto out;
> +
> +       index = __fls(size_map) + 1;    /* basically a log_2 */
> +       index -= !(slot_size - (1ULL << (index + QFQ_MIN_SLOT_SHIFT - 1)));
> +
> +       if (index < 0)
> +               index = 0;
> +out:
> +       pr_debug("qfq calc_index: W = %lu, L = %u, I = %d\n",
> +                (unsigned long) ONE_FP/inv_w, maxlen, index);
> +
> +       return index;
> +}
> +
> +static int qfq_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
> +                           struct nlattr **tca, unsigned long *arg)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_class *cl = (struct qfq_class *)*arg;
> +       struct nlattr *tb[TCA_QFQ_MAX + 1];
> +       u32 weight, lmax, inv_w;
> +       int i, err;
> +
> +       if (tca[TCA_OPTIONS] == NULL)
> +               return -EINVAL;
> +
> +       err = nla_parse_nested(tb, TCA_QFQ_MAX, tca[TCA_OPTIONS], qfq_policy);
> +       if (err < 0)
> +               return err;
> +
> +       if (tb[TCA_QFQ_WEIGHT]) {
> +               weight = nla_get_u32(tb[TCA_QFQ_WEIGHT]);
> +               if (!weight || weight > (1UL << QFQ_MAX_WSHIFT)) {
> +                       pr_notice("qfq: invalid weight %u\n", weight);
> +                       return -EINVAL;
> +               }
> +       } else
> +               weight = 1;
> +
> +       inv_w = ONE_FP / weight;
> +       weight = ONE_FP / inv_w;
> +       if (q->wsum + weight > QFQ_MAX_WSUM) {
> +               pr_notice("qfq: total weight out of range (%u + %u)\n",
> +                         weight, q->wsum);
> +               return -EINVAL;
> +       }
> +
> +       if (tb[TCA_QFQ_LMAX]) {
> +               lmax = nla_get_u32(tb[TCA_QFQ_LMAX]);
> +               if (!lmax || lmax > (1UL << QFQ_MTU_SHIFT)) {
> +                       pr_notice("qfq: invalid max length %u\n", lmax);
> +                       return -EINVAL;
> +               }
> +       } else
> +               lmax = 1UL << QFQ_MTU_SHIFT;
> +
> +       if (cl != NULL) {
> +               if (tca[TCA_RATE]) {
> +                       err = gen_replace_estimator(&cl->bstats, &cl->rate_est,
> +                                                   qdisc_root_sleeping_lock(sch),
> +                                                   tca[TCA_RATE]);
> +                       if (err)
> +                               return err;
> +               }
> +
> +               sch_tree_lock(sch);
> +               if (tb[TCA_QFQ_WEIGHT]) {
> +                       q->wsum = weight - ONE_FP / cl->inv_w;
> +                       cl->inv_w = inv_w;
> +               }
> +               sch_tree_unlock(sch);
> +
> +               return 0;
> +       }
> +
> +       cl = kzalloc(sizeof(struct qfq_class), GFP_KERNEL);
> +       if (cl == NULL)
> +               return -ENOBUFS;
> +
> +       cl->refcnt = 1;
> +       cl->common.classid = classid;
> +       cl->lmax = lmax;
> +       cl->inv_w = inv_w;
> +       i = qfq_calc_index(cl->inv_w, cl->lmax);
> +
> +       cl->grp = &q->groups[i];
> +       q->wsum += weight;
> +
> +       cl->qdisc = qdisc_create_dflt(sch->dev_queue,
> +                                     &pfifo_qdisc_ops, classid);
> +       if (cl->qdisc == NULL)
> +               cl->qdisc = &noop_qdisc;
> +
> +       if (tca[TCA_RATE]) {
> +               err = gen_new_estimator(&cl->bstats, &cl->rate_est,
> +                                       qdisc_root_sleeping_lock(sch),
> +                                       tca[TCA_RATE]);
> +               if (err) {
> +                       qdisc_destroy(cl->qdisc);
> +                       kfree(cl);
> +                       return err;
> +               }
> +       }
> +
> +       sch_tree_lock(sch);
> +       qdisc_class_hash_insert(&q->clhash, &cl->common);
> +       sch_tree_unlock(sch);
> +
> +       qdisc_class_hash_grow(sch, &q->clhash);
> +
> +       *arg = (unsigned long)cl;
> +       return 0;
> +}
> +
> +static void qfq_destroy_class(struct Qdisc *sch, struct qfq_class *cl)
> +{
> +       struct qfq_sched *q = (struct qfq_sched *)sch;
> +
> +       if (cl->inv_w) {
> +               q->wsum -= ONE_FP / cl->inv_w;
> +               cl->inv_w = 0;
> +       }
> +
> +       gen_kill_estimator(&cl->bstats, &cl->rate_est);
> +       qdisc_destroy(cl->qdisc);
> +       kfree(cl);
> +}
> +
> +static int qfq_delete_class(struct Qdisc *sch, unsigned long arg)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +
> +       if (cl->filter_cnt > 0)
> +               return -EBUSY;
> +
> +       sch_tree_lock(sch);
> +
> +       qfq_purge_queue(cl);
> +       qdisc_class_hash_remove(&q->clhash, &cl->common);
> +
> +       if (--cl->refcnt == 0)
> +               qfq_destroy_class(sch, cl);
> +
> +       sch_tree_unlock(sch);
> +       return 0;
> +}
> +
> +static unsigned long qfq_get_class(struct Qdisc *sch, u32 classid)
> +{
> +       struct qfq_class *cl = qfq_find_class(sch, classid);
> +
> +       if (cl != NULL)
> +               cl->refcnt++;
> +
> +       return (unsigned long)cl;
> +}
> +
> +static void qfq_put_class(struct Qdisc *sch, unsigned long arg)
> +{
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +
> +       if (--cl->refcnt == 0)
> +               qfq_destroy_class(sch, cl);
> +}
> +
> +static struct tcf_proto **qfq_tcf_chain(struct Qdisc *sch, unsigned long cl)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +
> +       if (cl)
> +               return NULL;
> +
> +       return &q->filter_list;
> +}
> +
> +static unsigned long qfq_bind_tcf(struct Qdisc *sch, unsigned long parent,
> +                                 u32 classid)
> +{
> +       struct qfq_class *cl = qfq_find_class(sch, classid);
> +
> +       if (cl != NULL)
> +               cl->filter_cnt++;
> +
> +       return (unsigned long)cl;
> +}
> +
> +static void qfq_unbind_tcf(struct Qdisc *sch, unsigned long arg)
> +{
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +
> +       cl->filter_cnt--;
> +}
> +
> +static int qfq_graft_class(struct Qdisc *sch, unsigned long arg,
> +                          struct Qdisc *new, struct Qdisc **old)
> +{
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +
> +       if (new == NULL) {
> +               new = qdisc_create_dflt(sch->dev_queue,
> +                                       &pfifo_qdisc_ops, cl->common.classid);
> +               if (new == NULL)
> +                       new = &noop_qdisc;
> +       }
> +
> +       sch_tree_lock(sch);
> +       qfq_purge_queue(cl);
> +       *old = cl->qdisc;
> +       cl->qdisc = new;
> +       sch_tree_unlock(sch);
> +       return 0;
> +}
> +
> +static struct Qdisc *qfq_class_leaf(struct Qdisc *sch, unsigned long arg)
> +{
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +
> +       return cl->qdisc;
> +}
> +
> +static int qfq_dump_class(struct Qdisc *sch, unsigned long arg,
> +                         struct sk_buff *skb, struct tcmsg *tcm)
> +{
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +       struct nlattr *nest;
> +
> +       tcm->tcm_parent = TC_H_ROOT;
> +       tcm->tcm_handle = cl->common.classid;
> +       tcm->tcm_info   = cl->qdisc->handle;
> +
> +       nest = nla_nest_start(skb, TCA_OPTIONS);
> +       if (nest == NULL)
> +               goto nla_put_failure;
> +       NLA_PUT_U32(skb, TCA_QFQ_WEIGHT, ONE_FP/cl->inv_w);
> +       NLA_PUT_U32(skb, TCA_QFQ_LMAX, cl->lmax);
> +       return nla_nest_end(skb, nest);
> +
> +nla_put_failure:
> +       nla_nest_cancel(skb, nest);
> +       return -EMSGSIZE;
> +}
> +
> +static int qfq_dump_class_stats(struct Qdisc *sch, unsigned long arg,
> +                               struct gnet_dump *d)
> +{
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +       struct tc_qfq_stats xstats;
> +
> +       memset(&xstats, 0, sizeof(xstats));
> +
> +       xstats.weight = ONE_FP/cl->inv_w;
> +       xstats.lmax = cl->lmax;
> +
> +       if (gnet_stats_copy_basic(d, &cl->bstats) < 0 ||
> +           gnet_stats_copy_rate_est(d, NULL, &cl->rate_est) < 0 ||
> +           gnet_stats_copy_queue(d, &cl->qdisc->qstats) < 0)
> +               return -1;
> +
> +       return gnet_stats_copy_app(d, &xstats, sizeof(xstats));
> +}
> +
> +static void qfq_walk(struct Qdisc *sch, struct qdisc_walker *arg)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_class *cl;
> +       struct hlist_node *n;
> +       unsigned int i;
> +
> +       if (arg->stop)
> +               return;
> +
> +       for (i = 0; i < q->clhash.hashsize; i++) {
> +               hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) {
> +                       if (arg->count < arg->skip) {
> +                               arg->count++;
> +                               continue;
> +                       }
> +                       if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
> +                               arg->stop = 1;
> +                               return;
> +                       }
> +                       arg->count++;
> +               }
> +       }
> +}
> +
> +static struct qfq_class *qfq_classify(struct sk_buff *skb, struct Qdisc *sch,
> +                                     int *qerr)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_class *cl;
> +       struct tcf_result res;
> +       int result;
> +
> +       if (TC_H_MAJ(skb->priority ^ sch->handle) == 0) {
> +               cl = qfq_find_class(sch, skb->priority);
> +               if (cl != NULL)
> +                       return cl;
> +       }
> +
> +       *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
> +       result = tc_classify(skb, q->filter_list, &res);
> +       if (result >= 0) {
> +#ifdef CONFIG_NET_CLS_ACT
> +               switch (result) {
> +               case TC_ACT_QUEUED:
> +               case TC_ACT_STOLEN:
> +                       *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
> +               case TC_ACT_SHOT:
> +                       return NULL;
> +               }
> +#endif
> +               cl = (struct qfq_class *)res.class;
> +               if (cl == NULL)
> +                       cl = qfq_find_class(sch, res.classid);
> +               return cl;
> +       }
> +
> +       return NULL;
> +}
> +
> +/* Generic comparison function, handling wraparound. */
> +static inline int qfq_gt(u64 a, u64 b)
> +{
> +       return (s64)(a - b) > 0;
> +}
> +
> +/* Round a precise timestamp to its slotted value. */
> +static inline u64 qfq_round_down(u64 ts, unsigned int shift)
> +{
> +       return ts & ~((1ULL << shift) - 1);
> +}
> +
> +/* return the pointer to the group with lowest index in the bitmap */
> +static inline struct qfq_group *qfq_ffs(struct qfq_sched *q,
> +                                       unsigned long bitmap)
> +{
> +       int index = __ffs(bitmap); // zero-based
> +       return &q->groups[index];
> +}
> +/* Calculate a mask to mimic what would be ffs_from(). */
> +static inline unsigned long mask_from(unsigned long bitmap, int from)
> +{
> +       return bitmap & ~((1UL << from) - 1);
> +}
> +
> +/*
> + * The state computation relies on ER=0, IR=1, EB=2, IB=3
> + * First compute eligibility comparing grp->S, q->V,
> + * then check if someone is blocking us and possibly add EB
> + */
> +static int qfq_calc_state(struct qfq_sched *q, const struct qfq_group *grp)
> +{
> +       /* if S > V we are not eligible */
> +       unsigned int state = qfq_gt(grp->S, q->V);
> +       unsigned long mask = mask_from(q->bitmaps[ER], grp->index);
> +       struct qfq_group *next;
> +
> +       if (mask) {
> +               next = qfq_ffs(q, mask);
> +               if (qfq_gt(grp->F, next->F))
> +                       state |= EB;
> +       }
> +
> +       return state;
> +}
> +
> +
> +/*
> + * In principle
> + *     q->bitmaps[dst] |= q->bitmaps[src] & mask;
> + *     q->bitmaps[src] &= ~mask;
> + * but we should make sure that src != dst
> + */
> +static inline void qfq_move_groups(struct qfq_sched *q, unsigned long mask,
> +                                  int src, int dst)
> +{
> +       q->bitmaps[dst] |= q->bitmaps[src] & mask;
> +       q->bitmaps[src] &= ~mask;
> +}
> +
> +static void qfq_unblock_groups(struct qfq_sched *q, int index, u64 old_F)
> +{
> +       unsigned long mask = mask_from(q->bitmaps[ER], index + 1);
> +       struct qfq_group *next;
> +
> +       if (mask) {
> +               next = qfq_ffs(q, mask);
> +               if (!qfq_gt(next->F, old_F))
> +                       return;
> +       }
> +
> +       mask = (1UL << index) - 1;
> +       qfq_move_groups(q, mask, EB, ER);
> +       qfq_move_groups(q, mask, IB, IR);
> +}
> +
> +/*
> + * perhaps
> + *
> +       old_V ^= q->V;
> +       old_V >>= QFQ_MIN_SLOT_SHIFT;
> +       if (old_V) {
> +               ...
> +       }
> + *
> + */
> +static void qfq_make_eligible(struct qfq_sched *q, u64 old_V)
> +{
> +       unsigned long vslot = q->V >> QFQ_MIN_SLOT_SHIFT;
> +       unsigned long old_vslot = old_V >> QFQ_MIN_SLOT_SHIFT;
> +
> +       if (vslot != old_vslot) {
> +               unsigned long mask = (1UL << fls(vslot ^ old_vslot)) - 1;
> +               qfq_move_groups(q, mask, IR, ER);
> +               qfq_move_groups(q, mask, IB, EB);
> +       }
> +}
> +
> +/*
> + * XXX we should make sure that slot becomes less than 32.
> + * This is guaranteed by the input values.
> + * roundedS is always cl->S rounded on grp->slot_shift bits.
> + */
> +static void qfq_slot_insert(struct qfq_group *grp, struct qfq_class *cl,
> +                                  u64 roundedS)
> +{
> +       u64 slot = (roundedS - grp->S) >> grp->slot_shift;
> +       unsigned int i = (grp->front + slot) % QFQ_MAX_SLOTS;
> +
> +       cl->next = grp->slots[i];
> +       grp->slots[i] = cl;
> +       __set_bit(slot, &grp->full_slots);
> +}
> +
> +/*
> + * remove the entry from the slot
> + */
> +static void qfq_front_slot_remove(struct qfq_group *grp)
> +{
> +       struct qfq_class **h = &grp->slots[grp->front];
> +
> +       *h = (*h)->next;
> +       if (!*h)
> +               __clear_bit(0, &grp->full_slots);
> +}
> +
> +/*
> + * Returns the first full queue in a group. As a side effect,
> + * adjust the bucket list so the first non-empty bucket is at
> + * position 0 in full_slots.
> + */
> +static struct qfq_class *qfq_slot_scan(struct qfq_group *grp)
> +{
> +       unsigned int i;
> +
> +       pr_debug("qfq slot_scan: grp %u full %#lx\n",
> +                grp->index, grp->full_slots);
> +
> +       if (!grp->full_slots)
> +               return NULL;
> +
> +       i = __ffs(grp->full_slots);  /* zero based */
> +       if (i > 0) {
> +               grp->front = (grp->front + i) % QFQ_MAX_SLOTS;
> +               grp->full_slots >>= i;
> +       }
> +
> +       return grp->slots[grp->front];
> +}
> +
> +/*
> + * adjust the bucket list. When the start time of a group decreases,
> + * we move the index down (modulo QFQ_MAX_SLOTS) so we don't need to
> + * move the objects. The mask of occupied slots must be shifted
> + * because we use ffs() to find the first non-empty slot.
> + * This covers decreases in the group's start time, but what about
> + * increases of the start time ?
> + * Here too we should make sure that i is less than 32
> + */
> +static void qfq_slot_rotate(struct qfq_group *grp, u64 roundedS)
> +{
> +       unsigned int i = (grp->S - roundedS) >> grp->slot_shift;
> +
> +       grp->full_slots <<= i;
> +       grp->front = (grp->front - i) % QFQ_MAX_SLOTS;
> +}
> +
> +static void qfq_update_eligible(struct qfq_sched *q, u64 old_V)
> +{
> +       struct qfq_group *grp;
> +       unsigned long ineligible;
> +
> +       ineligible = q->bitmaps[IR] | q->bitmaps[IB];
> +       if (ineligible) {
> +               if (!q->bitmaps[ER]) {
> +                       grp = qfq_ffs(q, ineligible);
> +                       if (qfq_gt(grp->S, q->V))
> +                               q->V = grp->S;
> +               }
> +               qfq_make_eligible(q, old_V);
> +       }
> +}
> +
> +/* What is length of next packet in queue (0 if queue is empty) */
> +static unsigned int qdisc_peek_len(struct Qdisc *sch)
> +{
> +       struct sk_buff *skb;
> +
> +       skb = sch->ops->peek(sch);
> +       return skb ? qdisc_pkt_len(skb) : 0;
> +}
> +
> +/*
> + * Updates the class, returns true if also the group needs to be updated.
> + */
> +static bool qfq_update_class(struct qfq_group *grp, struct qfq_class *cl)
> +{
> +       unsigned int len = qdisc_peek_len(cl->qdisc);
> +
> +       cl->S = cl->F;
> +       if (!len)
> +               qfq_front_slot_remove(grp);     /* queue is empty */
> +       else {
> +               u64 roundedS;
> +
> +               cl->F = cl->S + (u64)len * cl->inv_w;
> +               roundedS = qfq_round_down(cl->S, grp->slot_shift);
> +               if (roundedS == grp->S)
> +                       return false;
> +
> +               qfq_front_slot_remove(grp);
> +               qfq_slot_insert(grp, cl, roundedS);
> +       }
> +
> +       return true;
> +}
> +
> +static struct sk_buff *qfq_dequeue(struct Qdisc *sch)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_group *grp;
> +       struct qfq_class *cl;
> +       struct sk_buff *skb;
> +       unsigned int len;
> +       u64 old_V;
> +
> +       if (!q->bitmaps[ER])
> +               return NULL;
> +
> +       grp = qfq_ffs(q, q->bitmaps[ER]);
> +
> +       cl = grp->slots[grp->front];
> +       skb = qdisc_dequeue_peeked(cl->qdisc);
> +       if (!skb) {
> +               WARN_ONCE(1, "qfq_dequeue: non-workconserving leaf\n");
> +               return NULL;
> +       }
> +
> +       sch->q.qlen--;
> +
> +       old_V = q->V;
> +       len = qdisc_pkt_len(skb);
> +       q->V += (u64)len * IWSUM;
> +       pr_debug("qfq enqueue: len %u F %lld now %lld\n",
> +                len, (unsigned long long) cl->F, (unsigned long long) q->V);
> +
> +       if (qfq_update_class(grp, cl)) {
> +               u64 old_F = grp->F;
> +
> +               cl = qfq_slot_scan(grp);
> +               if (!cl)
> +                       __clear_bit(grp->index, &q->bitmaps[ER]);
> +               else {
> +                       u64 roundedS = qfq_round_down(cl->S, grp->slot_shift);
> +                       unsigned int s;
> +
> +                       if (grp->S == roundedS)
> +                               goto skip_unblock;
> +                       grp->S = roundedS;
> +                       grp->F = roundedS + (2ULL << grp->slot_shift);
> +                       __clear_bit(grp->index, &q->bitmaps[ER]);
> +                       s = qfq_calc_state(q, grp);
> +                       __set_bit(grp->index, &q->bitmaps[s]);
> +               }
> +
> +               qfq_unblock_groups(q, grp->index, old_F);
> +       }
> +
> +skip_unblock:
> +       qfq_update_eligible(q, old_V);
> +
> +       return skb;
> +}
> +
> +/*
> + * Assign a reasonable start time for a new flow k in group i.
> + * Admissible values for \hat(F) are multiples of \sigma_i
> + * no greater than V+\sigma_i . Larger values mean that
> + * we had a wraparound so we consider the timestamp to be stale.
> + *
> + * If F is not stale and F >= V then we set S = F.
> + * Otherwise we should assign S = V, but this may violate
> + * the ordering in ER. So, if we have groups in ER, set S to
> + * the F_j of the first group j which would be blocking us.
> + * We are guaranteed not to move S backward because
> + * otherwise our group i would still be blocked.
> + */
> +static void qfq_update_start(struct qfq_sched *q, struct qfq_class *cl)
> +{
> +       unsigned long mask;
> +       uint32_t limit, roundedF;
> +       int slot_shift = cl->grp->slot_shift;
> +
> +       roundedF = qfq_round_down(cl->F, slot_shift);
> +       limit = qfq_round_down(q->V, slot_shift) + (1UL << slot_shift);
> +
> +       if (!qfq_gt(cl->F, q->V) || qfq_gt(roundedF, limit)) {
> +               /* timestamp was stale */
> +               mask = mask_from(q->bitmaps[ER], cl->grp->index);
> +               if (mask) {
> +                       struct qfq_group *next = qfq_ffs(q, mask);
> +                       if (qfq_gt(roundedF, next->F)) {
> +                               cl->S = next->F;
> +                               return;
> +                       }
> +               }
> +               cl->S = q->V;
> +       } else { /* timestamp is not stale */
> +               cl->S = cl->F;
> +       }
> +}
> +
> +static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_group *grp;
> +       struct qfq_class *cl;
> +       unsigned int len;
> +       int err;
> +       u64 roundedS;
> +       int s;
> +
> +       cl = qfq_classify(skb, sch, &err);
> +       if (cl == NULL || cl->qdisc->q.qlen > 80) {
> +               if (err & __NET_XMIT_BYPASS)
> +                       sch->qstats.drops++;
> +               kfree_skb(skb);
> +               return err;
> +       }
> +
> +       len = qdisc_pkt_len(skb);
> +       err = qdisc_enqueue(skb, cl->qdisc);
> +       if (unlikely(err != NET_XMIT_SUCCESS)) {
> +               if (net_xmit_drop_count(err)) {
> +                       cl->qstats.drops++;
> +                       sch->qstats.drops++;
> +               }
> +               return err;
> +       }
> +
> +       cl->bstats.packets += skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1;

Hmm, there is no other packets schedulers which account packets in
this way. Which one is better? I am not sure. And in this patch,
qstats.drops isn't maintained in the same way. Would these two be
consistent.

> +       cl->bstats.bytes += qdisc_pkt_len(skb);
> +
> +       sch->q.qlen++;
> +       sch->bstats.packets += skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1;
> +       sch->bstats.bytes += qdisc_pkt_len(skb);
> +
> +       if (qdisc_peek_head(sch) != skb)
> +               return err;

I suspect that it is wrong.

Here is the fake code from the paper:


5 i f ( f low . queue . head != pkt )
6 return ; // Flow already backlogged, we are don

So the correct code should be:
    if (qdisc_peek_head(cl->qdisc) != skb)
           return err;

However, we can't assume the cl->qdisc is work conserving, so the code
should be:
   if (cl->qdisc->q.qlen > 1)
          return err;

> +
> +       /* If reach this point, queue q was idle */
> +       grp = cl->grp;
> +       qfq_update_start(q, cl);
> +
> +       /* compute new finish time and rounded start. */
> +       cl->F = cl->S + (u64)qdisc_pkt_len(skb) * cl->inv_w;
> +       roundedS = qfq_round_down(cl->S, grp->slot_shift);
> +
> +       /*
> +        * insert cl in the correct bucket.
> +        * If cl->S >= grp->S we don't need to adjust the
> +        * bucket list and simply go to the insertion phase.
> +        * Otherwise grp->S is decreasing, we must make room
> +        * in the bucket list, and also recompute the group state.
> +        * Finally, if there were no flows in this group and nobody
> +        * was in ER make sure to adjust V.
> +        */
> +       if (grp->full_slots) {
> +               if (!qfq_gt(grp->S, cl->S))
> +                       goto skip_update;
> +
> +               /* create a slot for this cl->S */
> +               qfq_slot_rotate(grp, roundedS);
> +               /* group was surely ineligible, remove */
> +               __clear_bit(grp->index, &q->bitmaps[IR]);
> +               __clear_bit(grp->index, &q->bitmaps[IB]);
> +       } else if (!q->bitmaps[ER] && qfq_gt(roundedS, q->V))
> +               q->V = roundedS;
> +
> +       grp->S = roundedS;
> +       grp->F = roundedS + (2ULL << grp->slot_shift);
> +       s = qfq_calc_state(q, grp);
> +       __set_bit(grp->index, &q->bitmaps[s]);
> +
> +       pr_debug("qfq enqueue: new state %d %#lx S %lld F %lld V %lld\n",
> +                s, q->bitmaps[s],
> +                (unsigned long long) cl->S,
> +                (unsigned long long) cl->F,
> +                (unsigned long long) q->V);
> +
> +skip_update:
> +       qfq_slot_insert(grp, cl, roundedS);
> +
> +       return err;
> +}
> +
> +
> +static void qfq_slot_remove(struct qfq_sched *q, struct qfq_group *grp,
> +                           struct qfq_class *cl, struct qfq_class **pprev)
> +{
> +       unsigned int i, offset;
> +       u64 roundedS;
> +
> +       roundedS = qfq_round_down(cl->S, grp->slot_shift);
> +       offset = (roundedS - grp->S) >> grp->slot_shift;
> +       i = (grp->front + offset) % QFQ_MAX_SLOTS;
> +
> +       if (!pprev) {
> +               pprev = &grp->slots[i];
> +               while (*pprev && *pprev != cl)
> +                       pprev = &(*pprev)->next;
> +       }
> +
> +       *pprev = cl->next;
> +       if (!grp->slots[i])
> +               __clear_bit(offset, &grp->full_slots);
> +}
> +
> +/*
> + * called to forcibly destroy a queue.
> + * If the queue is not in the front bucket, or if it has
> + * other queues in the front bucket, we can simply remove
> + * the queue with no other side effects.
> + * Otherwise we must propagate the event up.
> + */
> +static void qfq_deactivate_class(struct qfq_sched *q, struct qfq_class *cl,
> +                                struct qfq_class **pprev)
> +{
> +       struct qfq_group *grp = cl->grp;
> +       unsigned long mask;
> +       u64 roundedS;
> +       int s;
> +
> +       cl->F = cl->S;
> +       qfq_slot_remove(q, grp, cl, pprev);
> +
> +       if (!grp->full_slots) {
> +               __clear_bit(grp->index, &q->bitmaps[IR]);
> +               __clear_bit(grp->index, &q->bitmaps[EB]);
> +               __clear_bit(grp->index, &q->bitmaps[IB]);
> +
> +               if (test_bit(grp->index, &q->bitmaps[ER]) &&
> +                   !(q->bitmaps[ER] & ~((1UL << grp->index) - 1))) {
> +                       mask = q->bitmaps[ER] & ((1UL << grp->index) - 1);
> +                       if (mask)
> +                               mask = ~((1UL << __fls(mask)) - 1);
> +                       else
> +                               mask = ~0UL;
> +                       qfq_move_groups(q, mask, EB, ER);
> +                       qfq_move_groups(q, mask, IB, IR);
> +               }
> +               __clear_bit(grp->index, &q->bitmaps[ER]);
> +       } else if (!grp->slots[grp->front]) {
> +               cl = qfq_slot_scan(grp);
> +               roundedS = qfq_round_down(cl->S, grp->slot_shift);
> +               if (grp->S != roundedS) {
> +                       __clear_bit(grp->index, &q->bitmaps[ER]);
> +                       __clear_bit(grp->index, &q->bitmaps[IR]);
> +                       __clear_bit(grp->index, &q->bitmaps[EB]);
> +                       __clear_bit(grp->index, &q->bitmaps[IB]);
> +                       grp->S = roundedS;
> +                       grp->F = roundedS + (2ULL << grp->slot_shift);
> +                       s = qfq_calc_state(q, grp);
> +                       __set_bit(grp->index, &q->bitmaps[s]);
> +               }
> +       }
> +
> +       qfq_update_eligible(q, q->V);
> +}
> +
> +static void qfq_qlen_notify(struct Qdisc *sch, unsigned long arg)
> +{
> +       struct qfq_sched *q = (struct qfq_sched *)sch;
> +       struct qfq_class *cl = (struct qfq_class *)arg;
> +
> +       if (cl->qdisc->q.qlen == 0)
> +               qfq_deactivate_class(q, cl, NULL);
> +}
> +
> +static unsigned int qfq_drop(struct Qdisc *sch)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_group *grp;
> +       struct qfq_class *cl, **pp;
> +       unsigned int i, j, len;
> +
> +       for (i = 0; i <= QFQ_MAX_INDEX; i++) {
> +               grp = &q->groups[i];
> +               for (j = 0; j < QFQ_MAX_SLOTS; j++) {
> +                       for (pp = &grp->slots[j]; *pp; pp = &(*pp)->next) {
> +                               cl = *pp;
> +                               if (!cl->qdisc->ops->drop)
> +                                       continue;
> +
> +                               len = cl->qdisc->ops->drop(cl->qdisc);
> +                               if (len > 0) {
> +                                       sch->q.qlen--;
> +                                       if (!cl->qdisc->q.qlen)
> +                                               qfq_deactivate_class(q, cl, pp);
> +
> +                                       return len;
> +                               }
> +                       }
> +               }
> +       }
> +
> +       return 0;
> +}
> +
> +static int qfq_init_qdisc(struct Qdisc *sch, struct nlattr *opt)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_group *grp;
> +       int i, err;
> +
> +       err = qdisc_class_hash_init(&q->clhash);
> +       if (err < 0)
> +               return err;
> +
> +       for (i = 0; i <= QFQ_MAX_INDEX; i++) {
> +               grp = &q->groups[i];
> +               grp->index = i;
> +       }
> +
> +       return 0;
> +}
> +
> +static void qfq_reset_qdisc(struct Qdisc *sch)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_group *grp;
> +       struct qfq_class *cl, **pp;
> +       struct hlist_node *n;
> +       unsigned int i, j;
> +
> +       for (i = 0; i <= QFQ_MAX_INDEX; i++) {
> +               grp = &q->groups[i];
> +               for (j = 0; j < QFQ_MAX_SLOTS; j++) {
> +                       for (pp = &grp->slots[j]; *pp; pp = &(*pp)->next) {
> +                               cl = *pp;
> +                               if (cl->qdisc->q.qlen)
> +                                       qfq_deactivate_class(q, cl, pp);
> +                       }
> +               }
> +       }
> +
> +       for (i = 0; i < q->clhash.hashsize; i++) {
> +               hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode)
> +                       qdisc_reset(cl->qdisc);
> +       }
> +       sch->q.qlen = 0;
> +}
> +
> +static void qfq_destroy_qdisc(struct Qdisc *sch)
> +{
> +       struct qfq_sched *q = qdisc_priv(sch);
> +       struct qfq_class *cl;
> +       struct hlist_node *n, *next;
> +       unsigned int i;
> +
> +       tcf_destroy_chain(&q->filter_list);
> +
> +       for (i = 0; i < q->clhash.hashsize; i++) {
> +               hlist_for_each_entry_safe(cl, n, next, &q->clhash.hash[i],
> +                                         common.hnode)
> +                       qfq_destroy_class(sch, cl);
> +       }
> +       qdisc_class_hash_destroy(&q->clhash);
> +}
> +
> +static const struct Qdisc_class_ops qfq_class_ops = {
> +       .change         = qfq_change_class,
> +       .delete         = qfq_delete_class,
> +       .get            = qfq_get_class,
> +       .put            = qfq_put_class,
> +       .tcf_chain      = qfq_tcf_chain,
> +       .bind_tcf       = qfq_bind_tcf,
> +       .unbind_tcf     = qfq_unbind_tcf,
> +       .graft          = qfq_graft_class,
> +       .leaf           = qfq_class_leaf,
> +       .qlen_notify    = qfq_qlen_notify,
> +       .dump           = qfq_dump_class,
> +       .dump_stats     = qfq_dump_class_stats,
> +       .walk           = qfq_walk,
> +};
> +
> +static struct Qdisc_ops qfq_qdisc_ops __read_mostly = {
> +       .cl_ops         = &qfq_class_ops,
> +       .id             = "qfq",
> +       .priv_size      = sizeof(struct qfq_sched),
> +       .enqueue        = qfq_enqueue,
> +       .dequeue        = qfq_dequeue,
> +       .peek           = qdisc_peek_dequeued,
> +       .drop           = qfq_drop,
> +       .init           = qfq_init_qdisc,
> +       .reset          = qfq_reset_qdisc,
> +       .destroy        = qfq_destroy_qdisc,
> +       .owner          = THIS_MODULE,
> +};
> +
> +static int __init qfq_init(void)
> +{
> +       return register_qdisc(&qfq_qdisc_ops);
> +}
> +
> +static void __exit qfq_exit(void)
> +{
> +       unregister_qdisc(&qfq_qdisc_ops);
> +}
> +
> +module_init(qfq_init);
> +module_exit(qfq_exit);
> +MODULE_LICENSE("GPL");
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
stephen hemminger - Jan. 8, 2011, 4:02 a.m.
On Sat, 8 Jan 2011 10:56:33 +0800
Changli Gao <xiaosuo@gmail.com> wrote:

> > +       cl->bstats.packets += skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1;  
> 
> Hmm, there is no other packets schedulers which account packets in
> this way. Which one is better? I am not sure. And in this patch,
> qstats.drops isn't maintained in the same way. Would these two be
> consistent.

HTB uses this accounting.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller - Jan. 8, 2011, 4:34 a.m.
Changli, please do not quote an entire patch just to comment upon
one specific portion of that patch.  Quote only the hunks of the
patch you actually want to talk about.

When you quote the entire patch, it wastes bandwith, and makes it
harder for people to scan around to see your feedback.

I think people who do this have no idea how much pain they cause
for every single person reading their postings.  Nor do they
realize that this makes their feedback get unread completely by
many people.

Please, never do this again.  You contribute far too much for this
to become a habit.

Thank you.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

--- a/include/linux/pkt_sched.h	2011-01-05 09:01:33.268032043 -0800
+++ b/include/linux/pkt_sched.h	2011-01-05 23:17:20.637390255 -0800
@@ -481,4 +481,18 @@  struct tc_drr_stats {
 	__u32	deficit;
 };
 
+/* QFQ */
+enum {
+	TCA_QFQ_WEIGHT,
+	TCA_QFQ_LMAX,
+	__TCA_QFQ_MAX
+};
+
+#define TCA_QFQ_MAX	(__TCA_QFQ_MAX - 1)
+
+struct tc_qfq_stats {
+	__u32 weight;
+	__u32 lmax;
+};
+
 #endif
--- a/net/sched/Kconfig	2011-01-05 09:01:33.280032462 -0800
+++ b/net/sched/Kconfig	2011-01-05 23:17:20.637390255 -0800
@@ -216,6 +216,17 @@  config NET_SCH_CHOKE
 	  To compile this code as a module, choose M here: the
 	  module will be called sch_choke.
 
+config NET_SCH_QFQ
+        tristate "Quick Fair Queueing Scheduler (QFQ)"
+	help
+	  Say Y here if you want to use the Quick Fair Queueing Scheduler (QFQ)
+	  packet scheduling algorithm.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called sch_qfq.
+
+	  If unsure, say N.
+
 config NET_SCH_INGRESS
 	tristate "Ingress Qdisc"
 	depends on NET_CLS_ACT
--- a/net/sched/Makefile	2011-01-05 09:01:33.284032598 -0800
+++ b/net/sched/Makefile	2011-01-05 23:17:20.645389829 -0800
@@ -32,6 +32,7 @@  obj-$(CONFIG_NET_SCH_MULTIQ)	+= sch_mult
 obj-$(CONFIG_NET_SCH_ATM)	+= sch_atm.o
 obj-$(CONFIG_NET_SCH_NETEM)	+= sch_netem.o
 obj-$(CONFIG_NET_SCH_DRR)	+= sch_drr.o
+obj-$(CONFIG_NET_SCH_QFQ)	+= sch_qfq.o
 obj-$(CONFIG_NET_SCH_CHOKE)	+= sch_choke.o
 obj-$(CONFIG_NET_CLS_U32)	+= cls_u32.o
 obj-$(CONFIG_NET_CLS_ROUTE4)	+= cls_route.o
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ b/net/sched/sch_qfq.c	2011-01-06 12:51:28.498280327 -0800
@@ -0,0 +1,1125 @@ 
+/*
+ * net/sched/sch_qfq.c         Quick Fair Queueing Scheduler.
+ *
+ * Copyright (c) 2009 Fabio Checconi, Luigi Rizzo, and Paolo Valente.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * version 2 as published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/bitops.h>
+#include <linux/errno.h>
+#include <linux/netdevice.h>
+#include <linux/pkt_sched.h>
+#include <net/sch_generic.h>
+#include <net/pkt_sched.h>
+#include <net/pkt_cls.h>
+
+/*  Quick Fair Queueing
+    ===================
+
+    Sources:
+    Fabio Checconi and Scuola Superiore and S. Anna
+    and Paolo Valente and Luigi Riz "QFQ: Efficient Packet Scheduling
+    with Tight Bandwidth Distribution Guarantees", SIGCOMM 2010
+
+    See also:
+    http://retis.sssup.it/~fabio/linux/qfq/
+ */
+
+/*
+
+  Virtual time computations.
+
+  S, F and V are all computed in fixed point arithmetic with
+  FRAC_BITS decimal bits.
+
+  QFQ_MAX_INDEX is the maximum index allowed for a group. We need
+  	one bit per index.
+  QFQ_MAX_WSHIFT is the maximum power of two supported as a weight.
+
+  The layout of the bits is as below:
+
+                   [ MTU_SHIFT ][      FRAC_BITS    ]
+                   [ MAX_INDEX    ][ MIN_SLOT_SHIFT ]
+  				 ^.__grp->index = 0
+  				 *.__grp->slot_shift
+
+  where MIN_SLOT_SHIFT is derived by difference from the others.
+
+  The max group index corresponds to Lmax/w_min, where
+  Lmax=1<<MTU_SHIFT, w_min = 1 .
+  From this, and knowing how many groups (MAX_INDEX) we want,
+  we can derive the shift corresponding to each group.
+
+  Because we often need to compute
+	F = S + len/w_i  and V = V + len/wsum
+  instead of storing w_i store the value
+	inv_w = (1<<FRAC_BITS)/w_i
+  so we can do F = S + len * inv_w * wsum.
+  We use W_TOT in the formulas so we can easily move between
+  static and adaptive weight sum.
+
+  The per-scheduler-instance data contain all the data structures
+  for the scheduler: bitmaps and bucket lists.
+
+ */
+
+/*
+ * Maximum number of consecutive slots occupied by backlogged classes
+ * inside a group.
+ */
+#define QFQ_MAX_SLOTS	32
+
+/*
+ * Shifts used for class<->group mapping.  We allow class weights that are
+ * in the range [1, 2^MAX_WSHIFT], and we try to map each class i to the
+ * group with the smallest index that can support the L_i / r_i configured
+ * for the class.
+ *
+ * grp->index is the index of the group; and grp->slot_shift
+ * is the shift for the corresponding (scaled) sigma_i.
+ */
+#define QFQ_MAX_INDEX		19
+#define QFQ_MAX_WSHIFT		16
+
+#define	QFQ_MAX_WEIGHT		(1<<QFQ_MAX_WSHIFT)
+#define QFQ_MAX_WSUM		(2*QFQ_MAX_WEIGHT)
+
+#define FRAC_BITS		30	/* fixed point arithmetic */
+#define ONE_FP			(1UL << FRAC_BITS)
+#define IWSUM			(ONE_FP/QFQ_MAX_WSUM)
+
+#define QFQ_MTU_SHIFT		11
+#define QFQ_MIN_SLOT_SHIFT	(FRAC_BITS + QFQ_MTU_SHIFT - QFQ_MAX_INDEX)
+
+/*
+ * Possible group states.  These values are used as indexes for the bitmaps
+ * array of struct qfq_queue.
+ */
+enum qfq_state { ER, IR, EB, IB, QFQ_MAX_STATE };
+
+struct qfq_group;
+
+struct qfq_class {
+	struct Qdisc_class_common common;
+
+	unsigned int refcnt;
+	unsigned int filter_cnt;
+
+	struct gnet_stats_basic_packed bstats;
+	struct gnet_stats_queue qstats;
+	struct gnet_stats_rate_est rate_est;
+	struct Qdisc *qdisc;
+
+	struct qfq_class *next;	/* Link for the slot list. */
+	u64 S, F;		/* flow timestamps (exact) */
+
+	/* group we belong to. In principle we would need the index,
+	 * which is log_2(lmax/weight), but we never reference it
+	 * directly, only the group.
+	 */
+	struct qfq_group *grp;
+
+	/* these are copied from the flowset. */
+	u32	inv_w;		/* ONE_FP/weight */
+	u32 	lmax;		/* Max packet size for this flow. */
+};
+
+struct qfq_group {
+	uint64_t S, F;			/* group timestamps (approx). */
+	unsigned int slot_shift;	/* Slot shift. */
+	unsigned int index;		/* Group index. */
+	unsigned int front;		/* Index of the front slot. */
+	unsigned long full_slots;	/* non-empty slots */
+
+	/* Array of RR lists of active classes. */
+	struct qfq_class *slots[QFQ_MAX_SLOTS];
+};
+
+struct qfq_sched {
+	struct tcf_proto *filter_list;
+	struct Qdisc_class_hash clhash;
+
+	uint64_t	V;		/* Precise virtual time. */
+	u32 wsum;			/* weight sum */
+
+	unsigned long bitmaps[QFQ_MAX_STATE];	    /* Group bitmaps. */
+	struct qfq_group groups[QFQ_MAX_INDEX + 1]; /* The groups. */
+};
+
+static struct qfq_class *qfq_find_class(struct Qdisc *sch, u32 classid)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct Qdisc_class_common *clc;
+
+	clc = qdisc_class_find(&q->clhash, classid);
+	if (clc == NULL)
+		return NULL;
+	return container_of(clc, struct qfq_class, common);
+}
+
+static void qfq_purge_queue(struct qfq_class *cl)
+{
+	unsigned int len = cl->qdisc->q.qlen;
+
+	qdisc_reset(cl->qdisc);
+	qdisc_tree_decrease_qlen(cl->qdisc, len);
+}
+
+static const struct nla_policy qfq_policy[TCA_QFQ_MAX + 1] = {
+	[TCA_QFQ_WEIGHT] = { .type = NLA_U32 },
+	[TCA_QFQ_LMAX] = { .type = NLA_U32 },
+};
+
+/*
+ * Calculate a flow index, given its weight and maximum packet length.
+ * index = log_2(maxlen/weight) but we need to apply the scaling.
+ * This is used only once at flow creation.
+ */
+static int qfq_calc_index(u32 inv_w, unsigned int maxlen)
+{
+	u64 slot_size = (u64)maxlen *inv_w;
+	unsigned long size_map;
+	int index = 0;
+
+	size_map = slot_size >> QFQ_MIN_SLOT_SHIFT;
+	if (!size_map)
+		goto out;
+
+	index = __fls(size_map) + 1;	/* basically a log_2 */
+	index -= !(slot_size - (1ULL << (index + QFQ_MIN_SLOT_SHIFT - 1)));
+
+	if (index < 0)
+		index = 0;
+out:
+	pr_debug("qfq calc_index: W = %lu, L = %u, I = %d\n",
+		 (unsigned long) ONE_FP/inv_w, maxlen, index);
+
+	return index;
+}
+
+static int qfq_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
+			    struct nlattr **tca, unsigned long *arg)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_class *cl = (struct qfq_class *)*arg;
+	struct nlattr *tb[TCA_QFQ_MAX + 1];
+	u32 weight, lmax, inv_w;
+	int i, err;
+
+	if (tca[TCA_OPTIONS] == NULL)
+		return -EINVAL;
+
+	err = nla_parse_nested(tb, TCA_QFQ_MAX, tca[TCA_OPTIONS], qfq_policy);
+	if (err < 0)
+		return err;
+
+	if (tb[TCA_QFQ_WEIGHT]) {
+		weight = nla_get_u32(tb[TCA_QFQ_WEIGHT]);
+		if (!weight || weight > (1UL << QFQ_MAX_WSHIFT)) {
+			pr_notice("qfq: invalid weight %u\n", weight);
+			return -EINVAL;
+		}
+	} else
+		weight = 1;
+
+	inv_w = ONE_FP / weight;
+	weight = ONE_FP / inv_w;
+	if (q->wsum + weight > QFQ_MAX_WSUM) {
+		pr_notice("qfq: total weight out of range (%u + %u)\n",
+			  weight, q->wsum);
+		return -EINVAL;
+	}
+
+	if (tb[TCA_QFQ_LMAX]) {
+		lmax = nla_get_u32(tb[TCA_QFQ_LMAX]);
+		if (!lmax || lmax > (1UL << QFQ_MTU_SHIFT)) {
+			pr_notice("qfq: invalid max length %u\n", lmax);
+			return -EINVAL;
+		}
+	} else
+		lmax = 1UL << QFQ_MTU_SHIFT;
+
+	if (cl != NULL) {
+		if (tca[TCA_RATE]) {
+			err = gen_replace_estimator(&cl->bstats, &cl->rate_est,
+						    qdisc_root_sleeping_lock(sch),
+						    tca[TCA_RATE]);
+			if (err)
+				return err;
+		}
+
+		sch_tree_lock(sch);
+		if (tb[TCA_QFQ_WEIGHT]) {
+			q->wsum = weight - ONE_FP / cl->inv_w;
+			cl->inv_w = inv_w;
+		}
+		sch_tree_unlock(sch);
+
+		return 0;
+	}
+
+	cl = kzalloc(sizeof(struct qfq_class), GFP_KERNEL);
+	if (cl == NULL)
+		return -ENOBUFS;
+
+	cl->refcnt = 1;
+	cl->common.classid = classid;
+	cl->lmax = lmax;
+	cl->inv_w = inv_w;
+	i = qfq_calc_index(cl->inv_w, cl->lmax);
+
+	cl->grp = &q->groups[i];
+	q->wsum += weight;
+
+	cl->qdisc = qdisc_create_dflt(sch->dev_queue,
+				      &pfifo_qdisc_ops, classid);
+	if (cl->qdisc == NULL)
+		cl->qdisc = &noop_qdisc;
+
+	if (tca[TCA_RATE]) {
+		err = gen_new_estimator(&cl->bstats, &cl->rate_est,
+					qdisc_root_sleeping_lock(sch),
+					tca[TCA_RATE]);
+		if (err) {
+			qdisc_destroy(cl->qdisc);
+			kfree(cl);
+			return err;
+		}
+	}
+
+	sch_tree_lock(sch);
+	qdisc_class_hash_insert(&q->clhash, &cl->common);
+	sch_tree_unlock(sch);
+
+	qdisc_class_hash_grow(sch, &q->clhash);
+
+	*arg = (unsigned long)cl;
+	return 0;
+}
+
+static void qfq_destroy_class(struct Qdisc *sch, struct qfq_class *cl)
+{
+	struct qfq_sched *q = (struct qfq_sched *)sch;
+
+	if (cl->inv_w) {
+		q->wsum -= ONE_FP / cl->inv_w;
+		cl->inv_w = 0;
+	}
+
+	gen_kill_estimator(&cl->bstats, &cl->rate_est);
+	qdisc_destroy(cl->qdisc);
+	kfree(cl);
+}
+
+static int qfq_delete_class(struct Qdisc *sch, unsigned long arg)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_class *cl = (struct qfq_class *)arg;
+
+	if (cl->filter_cnt > 0)
+		return -EBUSY;
+
+	sch_tree_lock(sch);
+
+	qfq_purge_queue(cl);
+	qdisc_class_hash_remove(&q->clhash, &cl->common);
+
+	if (--cl->refcnt == 0)
+		qfq_destroy_class(sch, cl);
+
+	sch_tree_unlock(sch);
+	return 0;
+}
+
+static unsigned long qfq_get_class(struct Qdisc *sch, u32 classid)
+{
+	struct qfq_class *cl = qfq_find_class(sch, classid);
+
+	if (cl != NULL)
+		cl->refcnt++;
+
+	return (unsigned long)cl;
+}
+
+static void qfq_put_class(struct Qdisc *sch, unsigned long arg)
+{
+	struct qfq_class *cl = (struct qfq_class *)arg;
+
+	if (--cl->refcnt == 0)
+		qfq_destroy_class(sch, cl);
+}
+
+static struct tcf_proto **qfq_tcf_chain(struct Qdisc *sch, unsigned long cl)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+
+	if (cl)
+		return NULL;
+
+	return &q->filter_list;
+}
+
+static unsigned long qfq_bind_tcf(struct Qdisc *sch, unsigned long parent,
+				  u32 classid)
+{
+	struct qfq_class *cl = qfq_find_class(sch, classid);
+
+	if (cl != NULL)
+		cl->filter_cnt++;
+
+	return (unsigned long)cl;
+}
+
+static void qfq_unbind_tcf(struct Qdisc *sch, unsigned long arg)
+{
+	struct qfq_class *cl = (struct qfq_class *)arg;
+
+	cl->filter_cnt--;
+}
+
+static int qfq_graft_class(struct Qdisc *sch, unsigned long arg,
+			   struct Qdisc *new, struct Qdisc **old)
+{
+	struct qfq_class *cl = (struct qfq_class *)arg;
+
+	if (new == NULL) {
+		new = qdisc_create_dflt(sch->dev_queue,
+					&pfifo_qdisc_ops, cl->common.classid);
+		if (new == NULL)
+			new = &noop_qdisc;
+	}
+
+	sch_tree_lock(sch);
+	qfq_purge_queue(cl);
+	*old = cl->qdisc;
+	cl->qdisc = new;
+	sch_tree_unlock(sch);
+	return 0;
+}
+
+static struct Qdisc *qfq_class_leaf(struct Qdisc *sch, unsigned long arg)
+{
+	struct qfq_class *cl = (struct qfq_class *)arg;
+
+	return cl->qdisc;
+}
+
+static int qfq_dump_class(struct Qdisc *sch, unsigned long arg,
+			  struct sk_buff *skb, struct tcmsg *tcm)
+{
+	struct qfq_class *cl = (struct qfq_class *)arg;
+	struct nlattr *nest;
+
+	tcm->tcm_parent	= TC_H_ROOT;
+	tcm->tcm_handle	= cl->common.classid;
+	tcm->tcm_info	= cl->qdisc->handle;
+
+	nest = nla_nest_start(skb, TCA_OPTIONS);
+	if (nest == NULL)
+		goto nla_put_failure;
+	NLA_PUT_U32(skb, TCA_QFQ_WEIGHT, ONE_FP/cl->inv_w);
+	NLA_PUT_U32(skb, TCA_QFQ_LMAX, cl->lmax);
+	return nla_nest_end(skb, nest);
+
+nla_put_failure:
+	nla_nest_cancel(skb, nest);
+	return -EMSGSIZE;
+}
+
+static int qfq_dump_class_stats(struct Qdisc *sch, unsigned long arg,
+				struct gnet_dump *d)
+{
+	struct qfq_class *cl = (struct qfq_class *)arg;
+	struct tc_qfq_stats xstats;
+
+	memset(&xstats, 0, sizeof(xstats));
+
+	xstats.weight = ONE_FP/cl->inv_w;
+	xstats.lmax = cl->lmax;
+
+	if (gnet_stats_copy_basic(d, &cl->bstats) < 0 ||
+	    gnet_stats_copy_rate_est(d, NULL, &cl->rate_est) < 0 ||
+	    gnet_stats_copy_queue(d, &cl->qdisc->qstats) < 0)
+		return -1;
+
+	return gnet_stats_copy_app(d, &xstats, sizeof(xstats));
+}
+
+static void qfq_walk(struct Qdisc *sch, struct qdisc_walker *arg)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_class *cl;
+	struct hlist_node *n;
+	unsigned int i;
+
+	if (arg->stop)
+		return;
+
+	for (i = 0; i < q->clhash.hashsize; i++) {
+		hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) {
+			if (arg->count < arg->skip) {
+				arg->count++;
+				continue;
+			}
+			if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
+				arg->stop = 1;
+				return;
+			}
+			arg->count++;
+		}
+	}
+}
+
+static struct qfq_class *qfq_classify(struct sk_buff *skb, struct Qdisc *sch,
+				      int *qerr)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_class *cl;
+	struct tcf_result res;
+	int result;
+
+	if (TC_H_MAJ(skb->priority ^ sch->handle) == 0) {
+		cl = qfq_find_class(sch, skb->priority);
+		if (cl != NULL)
+			return cl;
+	}
+
+	*qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
+	result = tc_classify(skb, q->filter_list, &res);
+	if (result >= 0) {
+#ifdef CONFIG_NET_CLS_ACT
+		switch (result) {
+		case TC_ACT_QUEUED:
+		case TC_ACT_STOLEN:
+			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
+		case TC_ACT_SHOT:
+			return NULL;
+		}
+#endif
+		cl = (struct qfq_class *)res.class;
+		if (cl == NULL)
+			cl = qfq_find_class(sch, res.classid);
+		return cl;
+	}
+
+	return NULL;
+}
+
+/* Generic comparison function, handling wraparound. */
+static inline int qfq_gt(u64 a, u64 b)
+{
+	return (s64)(a - b) > 0;
+}
+
+/* Round a precise timestamp to its slotted value. */
+static inline u64 qfq_round_down(u64 ts, unsigned int shift)
+{
+	return ts & ~((1ULL << shift) - 1);
+}
+
+/* return the pointer to the group with lowest index in the bitmap */
+static inline struct qfq_group *qfq_ffs(struct qfq_sched *q,
+					unsigned long bitmap)
+{
+	int index = __ffs(bitmap); // zero-based
+	return &q->groups[index];
+}
+/* Calculate a mask to mimic what would be ffs_from(). */
+static inline unsigned long mask_from(unsigned long bitmap, int from)
+{
+	return bitmap & ~((1UL << from) - 1);
+}
+
+/*
+ * The state computation relies on ER=0, IR=1, EB=2, IB=3
+ * First compute eligibility comparing grp->S, q->V,
+ * then check if someone is blocking us and possibly add EB
+ */
+static int qfq_calc_state(struct qfq_sched *q, const struct qfq_group *grp)
+{
+	/* if S > V we are not eligible */
+	unsigned int state = qfq_gt(grp->S, q->V);
+	unsigned long mask = mask_from(q->bitmaps[ER], grp->index);
+	struct qfq_group *next;
+
+	if (mask) {
+		next = qfq_ffs(q, mask);
+		if (qfq_gt(grp->F, next->F))
+			state |= EB;
+	}
+
+	return state;
+}
+
+
+/*
+ * In principle
+ *	q->bitmaps[dst] |= q->bitmaps[src] & mask;
+ *	q->bitmaps[src] &= ~mask;
+ * but we should make sure that src != dst
+ */
+static inline void qfq_move_groups(struct qfq_sched *q, unsigned long mask,
+				   int src, int dst)
+{
+	q->bitmaps[dst] |= q->bitmaps[src] & mask;
+	q->bitmaps[src] &= ~mask;
+}
+
+static void qfq_unblock_groups(struct qfq_sched *q, int index, u64 old_F)
+{
+	unsigned long mask = mask_from(q->bitmaps[ER], index + 1);
+	struct qfq_group *next;
+
+	if (mask) {
+		next = qfq_ffs(q, mask);
+		if (!qfq_gt(next->F, old_F))
+			return;
+	}
+
+	mask = (1UL << index) - 1;
+	qfq_move_groups(q, mask, EB, ER);
+	qfq_move_groups(q, mask, IB, IR);
+}
+
+/*
+ * perhaps
+ *
+	old_V ^= q->V;
+	old_V >>= QFQ_MIN_SLOT_SHIFT;
+	if (old_V) {
+		...
+	}
+ *
+ */
+static void qfq_make_eligible(struct qfq_sched *q, u64 old_V)
+{
+	unsigned long vslot = q->V >> QFQ_MIN_SLOT_SHIFT;
+	unsigned long old_vslot = old_V >> QFQ_MIN_SLOT_SHIFT;
+
+	if (vslot != old_vslot) {
+		unsigned long mask = (1UL << fls(vslot ^ old_vslot)) - 1;
+		qfq_move_groups(q, mask, IR, ER);
+		qfq_move_groups(q, mask, IB, EB);
+	}
+}
+
+/*
+ * XXX we should make sure that slot becomes less than 32.
+ * This is guaranteed by the input values.
+ * roundedS is always cl->S rounded on grp->slot_shift bits.
+ */
+static void qfq_slot_insert(struct qfq_group *grp, struct qfq_class *cl,
+				   u64 roundedS)
+{
+	u64 slot = (roundedS - grp->S) >> grp->slot_shift;
+	unsigned int i = (grp->front + slot) % QFQ_MAX_SLOTS;
+
+	cl->next = grp->slots[i];
+	grp->slots[i] = cl;
+	__set_bit(slot, &grp->full_slots);
+}
+
+/*
+ * remove the entry from the slot
+ */
+static void qfq_front_slot_remove(struct qfq_group *grp)
+{
+	struct qfq_class **h = &grp->slots[grp->front];
+
+	*h = (*h)->next;
+	if (!*h)
+		__clear_bit(0, &grp->full_slots);
+}
+
+/*
+ * Returns the first full queue in a group. As a side effect,
+ * adjust the bucket list so the first non-empty bucket is at
+ * position 0 in full_slots.
+ */
+static struct qfq_class *qfq_slot_scan(struct qfq_group *grp)
+{
+	unsigned int i;
+
+	pr_debug("qfq slot_scan: grp %u full %#lx\n",
+		 grp->index, grp->full_slots);
+
+	if (!grp->full_slots)
+		return NULL;
+
+	i = __ffs(grp->full_slots);  /* zero based */
+	if (i > 0) {
+		grp->front = (grp->front + i) % QFQ_MAX_SLOTS;
+		grp->full_slots >>= i;
+	}
+
+	return grp->slots[grp->front];
+}
+
+/*
+ * adjust the bucket list. When the start time of a group decreases,
+ * we move the index down (modulo QFQ_MAX_SLOTS) so we don't need to
+ * move the objects. The mask of occupied slots must be shifted
+ * because we use ffs() to find the first non-empty slot.
+ * This covers decreases in the group's start time, but what about
+ * increases of the start time ?
+ * Here too we should make sure that i is less than 32
+ */
+static void qfq_slot_rotate(struct qfq_group *grp, u64 roundedS)
+{
+	unsigned int i = (grp->S - roundedS) >> grp->slot_shift;
+
+	grp->full_slots <<= i;
+	grp->front = (grp->front - i) % QFQ_MAX_SLOTS;
+}
+
+static void qfq_update_eligible(struct qfq_sched *q, u64 old_V)
+{
+	struct qfq_group *grp;
+	unsigned long ineligible;
+
+	ineligible = q->bitmaps[IR] | q->bitmaps[IB];
+	if (ineligible) {
+		if (!q->bitmaps[ER]) {
+			grp = qfq_ffs(q, ineligible);
+			if (qfq_gt(grp->S, q->V))
+				q->V = grp->S;
+		}
+		qfq_make_eligible(q, old_V);
+	}
+}
+
+/* What is length of next packet in queue (0 if queue is empty) */
+static unsigned int qdisc_peek_len(struct Qdisc *sch)
+{
+	struct sk_buff *skb;
+
+	skb = sch->ops->peek(sch);
+	return skb ? qdisc_pkt_len(skb) : 0;
+}
+
+/*
+ * Updates the class, returns true if also the group needs to be updated.
+ */
+static bool qfq_update_class(struct qfq_group *grp, struct qfq_class *cl)
+{
+	unsigned int len = qdisc_peek_len(cl->qdisc);
+
+	cl->S = cl->F;
+	if (!len)
+		qfq_front_slot_remove(grp);	/* queue is empty */
+	else {
+		u64 roundedS;
+
+		cl->F = cl->S + (u64)len * cl->inv_w;
+		roundedS = qfq_round_down(cl->S, grp->slot_shift);
+		if (roundedS == grp->S)
+			return false;
+
+		qfq_front_slot_remove(grp);
+		qfq_slot_insert(grp, cl, roundedS);
+	}
+
+	return true;
+}
+
+static struct sk_buff *qfq_dequeue(struct Qdisc *sch)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_group *grp;
+	struct qfq_class *cl;
+	struct sk_buff *skb;
+	unsigned int len;
+	u64 old_V;
+
+	if (!q->bitmaps[ER])
+		return NULL;
+
+	grp = qfq_ffs(q, q->bitmaps[ER]);
+
+	cl = grp->slots[grp->front];
+	skb = qdisc_dequeue_peeked(cl->qdisc);
+	if (!skb) {
+		WARN_ONCE(1, "qfq_dequeue: non-workconserving leaf\n");
+		return NULL;
+	}
+
+	sch->q.qlen--;
+
+	old_V = q->V;
+	len = qdisc_pkt_len(skb);
+	q->V += (u64)len * IWSUM;
+	pr_debug("qfq enqueue: len %u F %lld now %lld\n",
+		 len, (unsigned long long) cl->F, (unsigned long long) q->V);
+
+	if (qfq_update_class(grp, cl)) {
+		u64 old_F = grp->F;
+
+		cl = qfq_slot_scan(grp);
+		if (!cl)
+			__clear_bit(grp->index, &q->bitmaps[ER]);
+		else {
+			u64 roundedS = qfq_round_down(cl->S, grp->slot_shift);
+			unsigned int s;
+
+			if (grp->S == roundedS)
+				goto skip_unblock;
+			grp->S = roundedS;
+			grp->F = roundedS + (2ULL << grp->slot_shift);
+			__clear_bit(grp->index, &q->bitmaps[ER]);
+			s = qfq_calc_state(q, grp);
+			__set_bit(grp->index, &q->bitmaps[s]);
+		}
+
+		qfq_unblock_groups(q, grp->index, old_F);
+	}
+
+skip_unblock:
+	qfq_update_eligible(q, old_V);
+
+	return skb;
+}
+
+/*
+ * Assign a reasonable start time for a new flow k in group i.
+ * Admissible values for \hat(F) are multiples of \sigma_i
+ * no greater than V+\sigma_i . Larger values mean that
+ * we had a wraparound so we consider the timestamp to be stale.
+ *
+ * If F is not stale and F >= V then we set S = F.
+ * Otherwise we should assign S = V, but this may violate
+ * the ordering in ER. So, if we have groups in ER, set S to
+ * the F_j of the first group j which would be blocking us.
+ * We are guaranteed not to move S backward because
+ * otherwise our group i would still be blocked.
+ */
+static void qfq_update_start(struct qfq_sched *q, struct qfq_class *cl)
+{
+	unsigned long mask;
+	uint32_t limit, roundedF;
+	int slot_shift = cl->grp->slot_shift;
+
+	roundedF = qfq_round_down(cl->F, slot_shift);
+	limit = qfq_round_down(q->V, slot_shift) + (1UL << slot_shift);
+
+	if (!qfq_gt(cl->F, q->V) || qfq_gt(roundedF, limit)) {
+		/* timestamp was stale */
+		mask = mask_from(q->bitmaps[ER], cl->grp->index);
+		if (mask) {
+			struct qfq_group *next = qfq_ffs(q, mask);
+			if (qfq_gt(roundedF, next->F)) {
+				cl->S = next->F;
+				return;
+			}
+		}
+		cl->S = q->V;
+	} else { /* timestamp is not stale */
+		cl->S = cl->F;
+	}
+}
+
+static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_group *grp;
+	struct qfq_class *cl;
+	unsigned int len;
+	int err;
+	u64 roundedS;
+	int s;
+
+	cl = qfq_classify(skb, sch, &err);
+	if (cl == NULL || cl->qdisc->q.qlen > 80) {
+		if (err & __NET_XMIT_BYPASS)
+			sch->qstats.drops++;
+		kfree_skb(skb);
+		return err;
+	}
+
+	len = qdisc_pkt_len(skb);
+	err = qdisc_enqueue(skb, cl->qdisc);
+	if (unlikely(err != NET_XMIT_SUCCESS)) {
+		if (net_xmit_drop_count(err)) {
+			cl->qstats.drops++;
+			sch->qstats.drops++;
+		}
+		return err;
+	}
+
+	cl->bstats.packets += skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1;
+	cl->bstats.bytes += qdisc_pkt_len(skb);
+
+	sch->q.qlen++;
+	sch->bstats.packets += skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1;
+	sch->bstats.bytes += qdisc_pkt_len(skb);
+
+	if (qdisc_peek_head(sch) != skb)
+		return err;
+
+	/* If reach this point, queue q was idle */
+	grp = cl->grp;
+	qfq_update_start(q, cl);
+
+	/* compute new finish time and rounded start. */
+	cl->F = cl->S + (u64)qdisc_pkt_len(skb) * cl->inv_w;
+	roundedS = qfq_round_down(cl->S, grp->slot_shift);
+
+	/*
+	 * insert cl in the correct bucket.
+	 * If cl->S >= grp->S we don't need to adjust the
+	 * bucket list and simply go to the insertion phase.
+	 * Otherwise grp->S is decreasing, we must make room
+	 * in the bucket list, and also recompute the group state.
+	 * Finally, if there were no flows in this group and nobody
+	 * was in ER make sure to adjust V.
+	 */
+	if (grp->full_slots) {
+		if (!qfq_gt(grp->S, cl->S))
+			goto skip_update;
+
+		/* create a slot for this cl->S */
+		qfq_slot_rotate(grp, roundedS);
+		/* group was surely ineligible, remove */
+		__clear_bit(grp->index, &q->bitmaps[IR]);
+		__clear_bit(grp->index, &q->bitmaps[IB]);
+	} else if (!q->bitmaps[ER] && qfq_gt(roundedS, q->V))
+		q->V = roundedS;
+
+	grp->S = roundedS;
+	grp->F = roundedS + (2ULL << grp->slot_shift);
+	s = qfq_calc_state(q, grp);
+	__set_bit(grp->index, &q->bitmaps[s]);
+
+	pr_debug("qfq enqueue: new state %d %#lx S %lld F %lld V %lld\n",
+		 s, q->bitmaps[s],
+		 (unsigned long long) cl->S,
+		 (unsigned long long) cl->F,
+		 (unsigned long long) q->V);
+
+skip_update:
+	qfq_slot_insert(grp, cl, roundedS);
+
+	return err;
+}
+
+
+static void qfq_slot_remove(struct qfq_sched *q, struct qfq_group *grp,
+			    struct qfq_class *cl, struct qfq_class **pprev)
+{
+	unsigned int i, offset;
+	u64 roundedS;
+
+	roundedS = qfq_round_down(cl->S, grp->slot_shift);
+	offset = (roundedS - grp->S) >> grp->slot_shift;
+	i = (grp->front + offset) % QFQ_MAX_SLOTS;
+
+	if (!pprev) {
+		pprev = &grp->slots[i];
+		while (*pprev && *pprev != cl)
+			pprev = &(*pprev)->next;
+	}
+
+	*pprev = cl->next;
+	if (!grp->slots[i])
+		__clear_bit(offset, &grp->full_slots);
+}
+
+/*
+ * called to forcibly destroy a queue.
+ * If the queue is not in the front bucket, or if it has
+ * other queues in the front bucket, we can simply remove
+ * the queue with no other side effects.
+ * Otherwise we must propagate the event up.
+ */
+static void qfq_deactivate_class(struct qfq_sched *q, struct qfq_class *cl,
+				 struct qfq_class **pprev)
+{
+	struct qfq_group *grp = cl->grp;
+	unsigned long mask;
+	u64 roundedS;
+	int s;
+
+	cl->F = cl->S;
+	qfq_slot_remove(q, grp, cl, pprev);
+
+	if (!grp->full_slots) {
+		__clear_bit(grp->index, &q->bitmaps[IR]);
+		__clear_bit(grp->index, &q->bitmaps[EB]);
+		__clear_bit(grp->index, &q->bitmaps[IB]);
+
+		if (test_bit(grp->index, &q->bitmaps[ER]) &&
+		    !(q->bitmaps[ER] & ~((1UL << grp->index) - 1))) {
+			mask = q->bitmaps[ER] & ((1UL << grp->index) - 1);
+			if (mask)
+				mask = ~((1UL << __fls(mask)) - 1);
+			else
+				mask = ~0UL;
+			qfq_move_groups(q, mask, EB, ER);
+			qfq_move_groups(q, mask, IB, IR);
+		}
+		__clear_bit(grp->index, &q->bitmaps[ER]);
+	} else if (!grp->slots[grp->front]) {
+		cl = qfq_slot_scan(grp);
+		roundedS = qfq_round_down(cl->S, grp->slot_shift);
+		if (grp->S != roundedS) {
+			__clear_bit(grp->index, &q->bitmaps[ER]);
+			__clear_bit(grp->index, &q->bitmaps[IR]);
+			__clear_bit(grp->index, &q->bitmaps[EB]);
+			__clear_bit(grp->index, &q->bitmaps[IB]);
+			grp->S = roundedS;
+			grp->F = roundedS + (2ULL << grp->slot_shift);
+			s = qfq_calc_state(q, grp);
+			__set_bit(grp->index, &q->bitmaps[s]);
+		}
+	}
+
+	qfq_update_eligible(q, q->V);
+}
+
+static void qfq_qlen_notify(struct Qdisc *sch, unsigned long arg)
+{
+	struct qfq_sched *q = (struct qfq_sched *)sch;
+	struct qfq_class *cl = (struct qfq_class *)arg;
+
+	if (cl->qdisc->q.qlen == 0)
+		qfq_deactivate_class(q, cl, NULL);
+}
+
+static unsigned int qfq_drop(struct Qdisc *sch)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_group *grp;
+	struct qfq_class *cl, **pp;
+	unsigned int i, j, len;
+
+	for (i = 0; i <= QFQ_MAX_INDEX; i++) {
+		grp = &q->groups[i];
+		for (j = 0; j < QFQ_MAX_SLOTS; j++) {
+			for (pp = &grp->slots[j]; *pp; pp = &(*pp)->next) {
+				cl = *pp;
+				if (!cl->qdisc->ops->drop)
+					continue;
+
+				len = cl->qdisc->ops->drop(cl->qdisc);
+				if (len > 0) {
+					sch->q.qlen--;
+					if (!cl->qdisc->q.qlen)
+						qfq_deactivate_class(q, cl, pp);
+
+					return len;
+				}
+			}
+		}
+	}
+
+	return 0;
+}
+
+static int qfq_init_qdisc(struct Qdisc *sch, struct nlattr *opt)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_group *grp;
+	int i, err;
+
+	err = qdisc_class_hash_init(&q->clhash);
+	if (err < 0)
+		return err;
+
+	for (i = 0; i <= QFQ_MAX_INDEX; i++) {
+		grp = &q->groups[i];
+		grp->index = i;
+	}
+
+	return 0;
+}
+
+static void qfq_reset_qdisc(struct Qdisc *sch)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_group *grp;
+	struct qfq_class *cl, **pp;
+	struct hlist_node *n;
+	unsigned int i, j;
+
+	for (i = 0; i <= QFQ_MAX_INDEX; i++) {
+		grp = &q->groups[i];
+		for (j = 0; j < QFQ_MAX_SLOTS; j++) {
+			for (pp = &grp->slots[j]; *pp; pp = &(*pp)->next) {
+				cl = *pp;
+				if (cl->qdisc->q.qlen)
+					qfq_deactivate_class(q, cl, pp);
+			}
+		}
+	}
+
+	for (i = 0; i < q->clhash.hashsize; i++) {
+		hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode)
+			qdisc_reset(cl->qdisc);
+	}
+	sch->q.qlen = 0;
+}
+
+static void qfq_destroy_qdisc(struct Qdisc *sch)
+{
+	struct qfq_sched *q = qdisc_priv(sch);
+	struct qfq_class *cl;
+	struct hlist_node *n, *next;
+	unsigned int i;
+
+	tcf_destroy_chain(&q->filter_list);
+
+	for (i = 0; i < q->clhash.hashsize; i++) {
+		hlist_for_each_entry_safe(cl, n, next, &q->clhash.hash[i],
+					  common.hnode)
+			qfq_destroy_class(sch, cl);
+	}
+	qdisc_class_hash_destroy(&q->clhash);
+}
+
+static const struct Qdisc_class_ops qfq_class_ops = {
+	.change		= qfq_change_class,
+	.delete		= qfq_delete_class,
+	.get		= qfq_get_class,
+	.put		= qfq_put_class,
+	.tcf_chain	= qfq_tcf_chain,
+	.bind_tcf	= qfq_bind_tcf,
+	.unbind_tcf	= qfq_unbind_tcf,
+	.graft		= qfq_graft_class,
+	.leaf		= qfq_class_leaf,
+	.qlen_notify	= qfq_qlen_notify,
+	.dump		= qfq_dump_class,
+	.dump_stats	= qfq_dump_class_stats,
+	.walk		= qfq_walk,
+};
+
+static struct Qdisc_ops qfq_qdisc_ops __read_mostly = {
+	.cl_ops		= &qfq_class_ops,
+	.id		= "qfq",
+	.priv_size	= sizeof(struct qfq_sched),
+	.enqueue	= qfq_enqueue,
+	.dequeue	= qfq_dequeue,
+	.peek		= qdisc_peek_dequeued,
+	.drop		= qfq_drop,
+	.init		= qfq_init_qdisc,
+	.reset		= qfq_reset_qdisc,
+	.destroy	= qfq_destroy_qdisc,
+	.owner		= THIS_MODULE,
+};
+
+static int __init qfq_init(void)
+{
+	return register_qdisc(&qfq_qdisc_ops);
+}
+
+static void __exit qfq_exit(void)
+{
+	unregister_qdisc(&qfq_qdisc_ops);
+}
+
+module_init(qfq_init);
+module_exit(qfq_exit);
+MODULE_LICENSE("GPL");