aboutsummaryrefslogtreecommitdiffstats
path: root/net
diff options
context:
space:
mode:
authorEric Dumazet <edumazet@google.com>2012-05-11 05:30:50 -0400
committerDavid S. Miller <davem@davemloft.net>2012-05-12 15:53:42 -0400
commit4b549a2ef4bef9965d97cbd992ba67930cd3e0fe (patch)
tree2b30a5af713817396ca0c259cddbb2e0972a03e7 /net
parent536edd67109df5e0cdb2c4ee759e9bade7976367 (diff)
fq_codel: Fair Queue Codel AQM
Fair Queue Codel packet scheduler Principles : - Packets are classified (internal classifier or external) on flows. - This is a Stochastic model (as we use a hash, several flows might be hashed on same slot) - Each flow has a CoDel 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 (CoDel uses a FIFO) - head drops only. - ECN capability is on by default. - Very low memory footprint (64 bytes per flow) tc qdisc ... fq_codel [ limit PACKETS ] [ flows number ] [ target TIME ] [ interval TIME ] [ noecn ] [ quantum BYTES ] defaults : 1024 flows, 10240 packets limit, quantum : device MTU target : 5ms (CoDel default) interval : 100ms (CoDel default) Impressive results on load : class htb 1:1 root leaf 10: prio 0 quantum 1514 rate 200000Kbit ceil 200000Kbit burst 1475b/8 mpu 0b overhead 0b cburst 1475b/8 mpu 0b overhead 0b level 0 Sent 43304920109 bytes 33063109 pkt (dropped 0, overlimits 0 requeues 0) rate 201691Kbit 28595pps backlog 0b 312p requeues 0 lended: 33063109 borrowed: 0 giants: 0 tokens: -912 ctokens: -912 class fq_codel 10:1735 parent 10: (dropped 1292, overlimits 0 requeues 0) backlog 15140b 10p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 7.1ms class fq_codel 10:4524 parent 10: (dropped 1291, overlimits 0 requeues 0) backlog 16654b 11p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 7.1ms class fq_codel 10:4e74 parent 10: (dropped 1290, overlimits 0 requeues 0) backlog 6056b 4p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 6.4ms dropping drop_next 92.0ms class fq_codel 10:628a parent 10: (dropped 1289, overlimits 0 requeues 0) backlog 7570b 5p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 5.4ms dropping drop_next 90.9ms class fq_codel 10:a4b3 parent 10: (dropped 302, overlimits 0 requeues 0) backlog 16654b 11p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 7.1ms class fq_codel 10:c3c2 parent 10: (dropped 1284, overlimits 0 requeues 0) backlog 13626b 9p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 5.9ms class fq_codel 10:d331 parent 10: (dropped 299, overlimits 0 requeues 0) backlog 15140b 10p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 7.0ms class fq_codel 10:d526 parent 10: (dropped 12160, overlimits 0 requeues 0) backlog 35870b 211p requeues 0 deficit 1508 count 12160 lastcount 1 ldelay 15.3ms dropping drop_next 247us class fq_codel 10:e2c6 parent 10: (dropped 1288, overlimits 0 requeues 0) backlog 15140b 10p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 7.1ms class fq_codel 10:eab5 parent 10: (dropped 1285, overlimits 0 requeues 0) backlog 16654b 11p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 5.9ms class fq_codel 10:f220 parent 10: (dropped 1289, overlimits 0 requeues 0) backlog 15140b 10p requeues 0 deficit 1514 count 1 lastcount 1 ldelay 7.1ms qdisc htb 1: root refcnt 6 r2q 10 default 1 direct_packets_stat 0 ver 3.17 Sent 43331086547 bytes 33092812 pkt (dropped 0, overlimits 66063544 requeues 71) rate 201697Kbit 28602pps backlog 0b 260p requeues 71 qdisc fq_codel 10: parent 1:1 limit 10240p flows 65536 target 5.0ms interval 100.0ms ecn Sent 43331086547 bytes 33092812 pkt (dropped 949359, overlimits 0 requeues 0) rate 201697Kbit 28602pps backlog 189352b 260p requeues 0 maxpacket 1514 drop_overlimit 0 new_flow_count 5582 ecn_mark 125593 new_flows_len 0 old_flows_len 11 PING 172.30.42.18 (172.30.42.18) 56(84) bytes of data. 64 bytes from 172.30.42.18: icmp_req=1 ttl=64 time=0.227 ms 64 bytes from 172.30.42.18: icmp_req=2 ttl=64 time=0.165 ms 64 bytes from 172.30.42.18: icmp_req=3 ttl=64 time=0.166 ms 64 bytes from 172.30.42.18: icmp_req=4 ttl=64 time=0.151 ms 64 bytes from 172.30.42.18: icmp_req=5 ttl=64 time=0.164 ms 64 bytes from 172.30.42.18: icmp_req=6 ttl=64 time=0.172 ms 64 bytes from 172.30.42.18: icmp_req=7 ttl=64 time=0.175 ms 64 bytes from 172.30.42.18: icmp_req=8 ttl=64 time=0.183 ms 64 bytes from 172.30.42.18: icmp_req=9 ttl=64 time=0.158 ms 64 bytes from 172.30.42.18: icmp_req=10 ttl=64 time=0.200 ms 10 packets transmitted, 10 received, 0% packet loss, time 8999ms rtt min/avg/max/mdev = 0.151/0.176/0.227/0.022 ms Much better than SFQ because of priority given to new flows, and fast path dirtying less cache lines. Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net')
-rw-r--r--net/sched/Kconfig11
-rw-r--r--net/sched/Makefile1
-rw-r--r--net/sched/sch_fq_codel.c624
3 files changed, 636 insertions, 0 deletions
diff --git a/net/sched/Kconfig b/net/sched/Kconfig
index fadd2522053d..e7a8976bf25c 100644
--- a/net/sched/Kconfig
+++ b/net/sched/Kconfig
@@ -261,6 +261,17 @@ config NET_SCH_CODEL
261 261
262 If unsure, say N. 262 If unsure, say N.
263 263
264config NET_SCH_FQ_CODEL
265 tristate "Fair Queue Controlled Delay AQM (FQ_CODEL)"
266 help
267 Say Y here if you want to use the FQ Controlled Delay (FQ_CODEL)
268 packet scheduling algorithm.
269
270 To compile this driver as a module, choose M here: the module
271 will be called sch_fq_codel.
272
273 If unsure, say N.
274
264config NET_SCH_INGRESS 275config NET_SCH_INGRESS
265 tristate "Ingress Qdisc" 276 tristate "Ingress Qdisc"
266 depends on NET_CLS_ACT 277 depends on NET_CLS_ACT
diff --git a/net/sched/Makefile b/net/sched/Makefile
index 30fab03b8516..5940a1992f0d 100644
--- a/net/sched/Makefile
+++ b/net/sched/Makefile
@@ -38,6 +38,7 @@ obj-$(CONFIG_NET_SCH_MQPRIO) += sch_mqprio.o
38obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o 38obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o
39obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o 39obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o
40obj-$(CONFIG_NET_SCH_CODEL) += sch_codel.o 40obj-$(CONFIG_NET_SCH_CODEL) += sch_codel.o
41obj-$(CONFIG_NET_SCH_FQ_CODEL) += sch_fq_codel.o
41 42
42obj-$(CONFIG_NET_CLS_U32) += cls_u32.o 43obj-$(CONFIG_NET_CLS_U32) += cls_u32.o
43obj-$(CONFIG_NET_CLS_ROUTE4) += cls_route.o 44obj-$(CONFIG_NET_CLS_ROUTE4) += cls_route.o
diff --git a/net/sched/sch_fq_codel.c b/net/sched/sch_fq_codel.c
new file mode 100644
index 000000000000..a7b3754d21c8
--- /dev/null
+++ b/net/sched/sch_fq_codel.c
@@ -0,0 +1,624 @@
1/*
2 * Fair Queue CoDel discipline
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Copyright (C) 2012 Eric Dumazet <edumazet@google.com>
10 */
11
12#include <linux/module.h>
13#include <linux/types.h>
14#include <linux/kernel.h>
15#include <linux/jiffies.h>
16#include <linux/string.h>
17#include <linux/in.h>
18#include <linux/errno.h>
19#include <linux/init.h>
20#include <linux/skbuff.h>
21#include <linux/jhash.h>
22#include <linux/slab.h>
23#include <linux/vmalloc.h>
24#include <net/netlink.h>
25#include <net/pkt_sched.h>
26#include <net/flow_keys.h>
27#include <net/codel.h>
28
29/* Fair Queue CoDel.
30 *
31 * Principles :
32 * Packets are classified (internal classifier or external) on flows.
33 * This is a Stochastic model (as we use a hash, several flows
34 * might be hashed on same slot)
35 * Each flow has a CoDel managed queue.
36 * Flows are linked onto two (Round Robin) lists,
37 * so that new flows have priority on old ones.
38 *
39 * For a given flow, packets are not reordered (CoDel uses a FIFO)
40 * head drops only.
41 * ECN capability is on by default.
42 * Low memory footprint (64 bytes per flow)
43 */
44
45struct fq_codel_flow {
46 struct sk_buff *head;
47 struct sk_buff *tail;
48 struct list_head flowchain;
49 int deficit;
50 u32 dropped; /* number of drops (or ECN marks) on this flow */
51 struct codel_vars cvars;
52}; /* please try to keep this structure <= 64 bytes */
53
54struct fq_codel_sched_data {
55 struct tcf_proto *filter_list; /* optional external classifier */
56 struct fq_codel_flow *flows; /* Flows table [flows_cnt] */
57 u32 *backlogs; /* backlog table [flows_cnt] */
58 u32 flows_cnt; /* number of flows */
59 u32 perturbation; /* hash perturbation */
60 u32 quantum; /* psched_mtu(qdisc_dev(sch)); */
61 struct codel_params cparams;
62 struct codel_stats cstats;
63 u32 drop_overlimit;
64 u32 new_flow_count;
65
66 struct list_head new_flows; /* list of new flows */
67 struct list_head old_flows; /* list of old flows */
68};
69
70static unsigned int fq_codel_hash(const struct fq_codel_sched_data *q,
71 const struct sk_buff *skb)
72{
73 struct flow_keys keys;
74 unsigned int hash;
75
76 skb_flow_dissect(skb, &keys);
77 hash = jhash_3words((__force u32)keys.dst,
78 (__force u32)keys.src ^ keys.ip_proto,
79 (__force u32)keys.ports, q->perturbation);
80 return ((u64)hash * q->flows_cnt) >> 32;
81}
82
83static unsigned int fq_codel_classify(struct sk_buff *skb, struct Qdisc *sch,
84 int *qerr)
85{
86 struct fq_codel_sched_data *q = qdisc_priv(sch);
87 struct tcf_result res;
88 int result;
89
90 if (TC_H_MAJ(skb->priority) == sch->handle &&
91 TC_H_MIN(skb->priority) > 0 &&
92 TC_H_MIN(skb->priority) <= q->flows_cnt)
93 return TC_H_MIN(skb->priority);
94
95 if (!q->filter_list)
96 return fq_codel_hash(q, skb) + 1;
97
98 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
99 result = tc_classify(skb, q->filter_list, &res);
100 if (result >= 0) {
101#ifdef CONFIG_NET_CLS_ACT
102 switch (result) {
103 case TC_ACT_STOLEN:
104 case TC_ACT_QUEUED:
105 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
106 case TC_ACT_SHOT:
107 return 0;
108 }
109#endif
110 if (TC_H_MIN(res.classid) <= q->flows_cnt)
111 return TC_H_MIN(res.classid);
112 }
113 return 0;
114}
115
116/* helper functions : might be changed when/if skb use a standard list_head */
117
118/* remove one skb from head of slot queue */
119static inline struct sk_buff *dequeue_head(struct fq_codel_flow *flow)
120{
121 struct sk_buff *skb = flow->head;
122
123 flow->head = skb->next;
124 skb->next = NULL;
125 return skb;
126}
127
128/* add skb to flow queue (tail add) */
129static inline void flow_queue_add(struct fq_codel_flow *flow,
130 struct sk_buff *skb)
131{
132 if (flow->head == NULL)
133 flow->head = skb;
134 else
135 flow->tail->next = skb;
136 flow->tail = skb;
137 skb->next = NULL;
138}
139
140static unsigned int fq_codel_drop(struct Qdisc *sch)
141{
142 struct fq_codel_sched_data *q = qdisc_priv(sch);
143 struct sk_buff *skb;
144 unsigned int maxbacklog = 0, idx = 0, i, len;
145 struct fq_codel_flow *flow;
146
147 /* Queue is full! Find the fat flow and drop packet from it.
148 * This might sound expensive, but with 1024 flows, we scan
149 * 4KB of memory, and we dont need to handle a complex tree
150 * in fast path (packet queue/enqueue) with many cache misses.
151 */
152 for (i = 0; i < q->flows_cnt; i++) {
153 if (q->backlogs[i] > maxbacklog) {
154 maxbacklog = q->backlogs[i];
155 idx = i;
156 }
157 }
158 flow = &q->flows[idx];
159 skb = dequeue_head(flow);
160 len = qdisc_pkt_len(skb);
161 q->backlogs[idx] -= len;
162 kfree_skb(skb);
163 sch->q.qlen--;
164 sch->qstats.drops++;
165 sch->qstats.backlog -= len;
166 flow->dropped++;
167 return idx;
168}
169
170static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch)
171{
172 struct fq_codel_sched_data *q = qdisc_priv(sch);
173 unsigned int idx;
174 struct fq_codel_flow *flow;
175 int uninitialized_var(ret);
176
177 idx = fq_codel_classify(skb, sch, &ret);
178 if (idx == 0) {
179 if (ret & __NET_XMIT_BYPASS)
180 sch->qstats.drops++;
181 kfree_skb(skb);
182 return ret;
183 }
184 idx--;
185
186 codel_set_enqueue_time(skb);
187 flow = &q->flows[idx];
188 flow_queue_add(flow, skb);
189 q->backlogs[idx] += qdisc_pkt_len(skb);
190 sch->qstats.backlog += qdisc_pkt_len(skb);
191
192 if (list_empty(&flow->flowchain)) {
193 list_add_tail(&flow->flowchain, &q->new_flows);
194 codel_vars_init(&flow->cvars);
195 q->new_flow_count++;
196 flow->deficit = q->quantum;
197 flow->dropped = 0;
198 }
199 if (++sch->q.qlen < sch->limit)
200 return NET_XMIT_SUCCESS;
201
202 q->drop_overlimit++;
203 /* Return Congestion Notification only if we dropped a packet
204 * from this flow.
205 */
206 if (fq_codel_drop(sch) == idx)
207 return NET_XMIT_CN;
208
209 /* As we dropped a packet, better let upper stack know this */
210 qdisc_tree_decrease_qlen(sch, 1);
211 return NET_XMIT_SUCCESS;
212}
213
214/* This is the specific function called from codel_dequeue()
215 * to dequeue a packet from queue. Note: backlog is handled in
216 * codel, we dont need to reduce it here.
217 */
218static struct sk_buff *dequeue(struct codel_vars *vars, struct Qdisc *sch)
219{
220 struct fq_codel_flow *flow;
221 struct sk_buff *skb = NULL;
222
223 flow = container_of(vars, struct fq_codel_flow, cvars);
224 if (flow->head) {
225 skb = dequeue_head(flow);
226 sch->qstats.backlog -= qdisc_pkt_len(skb);
227 sch->q.qlen--;
228 }
229 return skb;
230}
231
232static struct sk_buff *fq_codel_dequeue(struct Qdisc *sch)
233{
234 struct fq_codel_sched_data *q = qdisc_priv(sch);
235 struct sk_buff *skb;
236 struct fq_codel_flow *flow;
237 struct list_head *head;
238 u32 prev_drop_count, prev_ecn_mark;
239
240begin:
241 head = &q->new_flows;
242 if (list_empty(head)) {
243 head = &q->old_flows;
244 if (list_empty(head))
245 return NULL;
246 }
247 flow = list_first_entry(head, struct fq_codel_flow, flowchain);
248
249 if (flow->deficit <= 0) {
250 flow->deficit += q->quantum;
251 list_move_tail(&flow->flowchain, &q->old_flows);
252 goto begin;
253 }
254
255 prev_drop_count = q->cstats.drop_count;
256 prev_ecn_mark = q->cstats.ecn_mark;
257
258 skb = codel_dequeue(sch, &q->cparams, &flow->cvars, &q->cstats,
259 dequeue, &q->backlogs[flow - q->flows]);
260
261 flow->dropped += q->cstats.drop_count - prev_drop_count;
262 flow->dropped += q->cstats.ecn_mark - prev_ecn_mark;
263
264 if (!skb) {
265 /* force a pass through old_flows to prevent starvation */
266 if ((head == &q->new_flows) && !list_empty(&q->old_flows))
267 list_move_tail(&flow->flowchain, &q->old_flows);
268 else
269 list_del_init(&flow->flowchain);
270 goto begin;
271 }
272 qdisc_bstats_update(sch, skb);
273 flow->deficit -= qdisc_pkt_len(skb);
274 /* We cant call qdisc_tree_decrease_qlen() if our qlen is 0,
275 * or HTB crashes. Defer it for next round.
276 */
277 if (q->cstats.drop_count && sch->q.qlen) {
278 qdisc_tree_decrease_qlen(sch, q->cstats.drop_count);
279 q->cstats.drop_count = 0;
280 }
281 return skb;
282}
283
284static void fq_codel_reset(struct Qdisc *sch)
285{
286 struct sk_buff *skb;
287
288 while ((skb = fq_codel_dequeue(sch)) != NULL)
289 kfree_skb(skb);
290}
291
292static const struct nla_policy fq_codel_policy[TCA_FQ_CODEL_MAX + 1] = {
293 [TCA_FQ_CODEL_TARGET] = { .type = NLA_U32 },
294 [TCA_FQ_CODEL_LIMIT] = { .type = NLA_U32 },
295 [TCA_FQ_CODEL_INTERVAL] = { .type = NLA_U32 },
296 [TCA_FQ_CODEL_ECN] = { .type = NLA_U32 },
297 [TCA_FQ_CODEL_FLOWS] = { .type = NLA_U32 },
298 [TCA_FQ_CODEL_QUANTUM] = { .type = NLA_U32 },
299};
300
301static int fq_codel_change(struct Qdisc *sch, struct nlattr *opt)
302{
303 struct fq_codel_sched_data *q = qdisc_priv(sch);
304 struct nlattr *tb[TCA_FQ_CODEL_MAX + 1];
305 int err;
306
307 if (!opt)
308 return -EINVAL;
309
310 err = nla_parse_nested(tb, TCA_FQ_CODEL_MAX, opt, fq_codel_policy);
311 if (err < 0)
312 return err;
313 if (tb[TCA_FQ_CODEL_FLOWS]) {
314 if (q->flows)
315 return -EINVAL;
316 q->flows_cnt = nla_get_u32(tb[TCA_FQ_CODEL_FLOWS]);
317 if (!q->flows_cnt ||
318 q->flows_cnt > 65536)
319 return -EINVAL;
320 }
321 sch_tree_lock(sch);
322
323 if (tb[TCA_FQ_CODEL_TARGET]) {
324 u64 target = nla_get_u32(tb[TCA_FQ_CODEL_TARGET]);
325
326 q->cparams.target = (target * NSEC_PER_USEC) >> CODEL_SHIFT;
327 }
328
329 if (tb[TCA_FQ_CODEL_INTERVAL]) {
330 u64 interval = nla_get_u32(tb[TCA_FQ_CODEL_INTERVAL]);
331
332 q->cparams.interval = (interval * NSEC_PER_USEC) >> CODEL_SHIFT;
333 }
334
335 if (tb[TCA_FQ_CODEL_LIMIT])
336 sch->limit = nla_get_u32(tb[TCA_FQ_CODEL_LIMIT]);
337
338 if (tb[TCA_FQ_CODEL_ECN])
339 q->cparams.ecn = !!nla_get_u32(tb[TCA_FQ_CODEL_ECN]);
340
341 if (tb[TCA_FQ_CODEL_QUANTUM])
342 q->quantum = max(256U, nla_get_u32(tb[TCA_FQ_CODEL_QUANTUM]));
343
344 while (sch->q.qlen > sch->limit) {
345 struct sk_buff *skb = fq_codel_dequeue(sch);
346
347 kfree_skb(skb);
348 q->cstats.drop_count++;
349 }
350 qdisc_tree_decrease_qlen(sch, q->cstats.drop_count);
351 q->cstats.drop_count = 0;
352
353 sch_tree_unlock(sch);
354 return 0;
355}
356
357static void *fq_codel_zalloc(size_t sz)
358{
359 void *ptr = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN);
360
361 if (!ptr)
362 ptr = vzalloc(sz);
363 return ptr;
364}
365
366static void fq_codel_free(void *addr)
367{
368 if (addr) {
369 if (is_vmalloc_addr(addr))
370 vfree(addr);
371 else
372 kfree(addr);
373 }
374}
375
376static void fq_codel_destroy(struct Qdisc *sch)
377{
378 struct fq_codel_sched_data *q = qdisc_priv(sch);
379
380 tcf_destroy_chain(&q->filter_list);
381 fq_codel_free(q->backlogs);
382 fq_codel_free(q->flows);
383}
384
385static int fq_codel_init(struct Qdisc *sch, struct nlattr *opt)
386{
387 struct fq_codel_sched_data *q = qdisc_priv(sch);
388 int i;
389
390 sch->limit = 10*1024;
391 q->flows_cnt = 1024;
392 q->quantum = psched_mtu(qdisc_dev(sch));
393 q->perturbation = net_random();
394 INIT_LIST_HEAD(&q->new_flows);
395 INIT_LIST_HEAD(&q->old_flows);
396 codel_params_init(&q->cparams);
397 codel_stats_init(&q->cstats);
398 q->cparams.ecn = true;
399
400 if (opt) {
401 int err = fq_codel_change(sch, opt);
402 if (err)
403 return err;
404 }
405
406 if (!q->flows) {
407 q->flows = fq_codel_zalloc(q->flows_cnt *
408 sizeof(struct fq_codel_flow));
409 if (!q->flows)
410 return -ENOMEM;
411 q->backlogs = fq_codel_zalloc(q->flows_cnt * sizeof(u32));
412 if (!q->backlogs) {
413 fq_codel_free(q->flows);
414 return -ENOMEM;
415 }
416 for (i = 0; i < q->flows_cnt; i++) {
417 struct fq_codel_flow *flow = q->flows + i;
418
419 INIT_LIST_HEAD(&flow->flowchain);
420 }
421 }
422 if (sch->limit >= 1)
423 sch->flags |= TCQ_F_CAN_BYPASS;
424 else
425 sch->flags &= ~TCQ_F_CAN_BYPASS;
426 return 0;
427}
428
429static int fq_codel_dump(struct Qdisc *sch, struct sk_buff *skb)
430{
431 struct fq_codel_sched_data *q = qdisc_priv(sch);
432 struct nlattr *opts;
433
434 opts = nla_nest_start(skb, TCA_OPTIONS);
435 if (opts == NULL)
436 goto nla_put_failure;
437
438 if (nla_put_u32(skb, TCA_FQ_CODEL_TARGET,
439 codel_time_to_us(q->cparams.target)) ||
440 nla_put_u32(skb, TCA_FQ_CODEL_LIMIT,
441 sch->limit) ||
442 nla_put_u32(skb, TCA_FQ_CODEL_INTERVAL,
443 codel_time_to_us(q->cparams.interval)) ||
444 nla_put_u32(skb, TCA_FQ_CODEL_ECN,
445 q->cparams.ecn) ||
446 nla_put_u32(skb, TCA_FQ_CODEL_QUANTUM,
447 q->quantum) ||
448 nla_put_u32(skb, TCA_FQ_CODEL_FLOWS,
449 q->flows_cnt))
450 goto nla_put_failure;
451
452 nla_nest_end(skb, opts);
453 return skb->len;
454
455nla_put_failure:
456 return -1;
457}
458
459static int fq_codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
460{
461 struct fq_codel_sched_data *q = qdisc_priv(sch);
462 struct tc_fq_codel_xstats st = {
463 .type = TCA_FQ_CODEL_XSTATS_QDISC,
464 .qdisc_stats.maxpacket = q->cstats.maxpacket,
465 .qdisc_stats.drop_overlimit = q->drop_overlimit,
466 .qdisc_stats.ecn_mark = q->cstats.ecn_mark,
467 .qdisc_stats.new_flow_count = q->new_flow_count,
468 };
469 struct list_head *pos;
470
471 list_for_each(pos, &q->new_flows)
472 st.qdisc_stats.new_flows_len++;
473
474 list_for_each(pos, &q->old_flows)
475 st.qdisc_stats.old_flows_len++;
476
477 return gnet_stats_copy_app(d, &st, sizeof(st));
478}
479
480static struct Qdisc *fq_codel_leaf(struct Qdisc *sch, unsigned long arg)
481{
482 return NULL;
483}
484
485static unsigned long fq_codel_get(struct Qdisc *sch, u32 classid)
486{
487 return 0;
488}
489
490static unsigned long fq_codel_bind(struct Qdisc *sch, unsigned long parent,
491 u32 classid)
492{
493 /* we cannot bypass queue discipline anymore */
494 sch->flags &= ~TCQ_F_CAN_BYPASS;
495 return 0;
496}
497
498static void fq_codel_put(struct Qdisc *q, unsigned long cl)
499{
500}
501
502static struct tcf_proto **fq_codel_find_tcf(struct Qdisc *sch, unsigned long cl)
503{
504 struct fq_codel_sched_data *q = qdisc_priv(sch);
505
506 if (cl)
507 return NULL;
508 return &q->filter_list;
509}
510
511static int fq_codel_dump_class(struct Qdisc *sch, unsigned long cl,
512 struct sk_buff *skb, struct tcmsg *tcm)
513{
514 tcm->tcm_handle |= TC_H_MIN(cl);
515 return 0;
516}
517
518static int fq_codel_dump_class_stats(struct Qdisc *sch, unsigned long cl,
519 struct gnet_dump *d)
520{
521 struct fq_codel_sched_data *q = qdisc_priv(sch);
522 u32 idx = cl - 1;
523 struct gnet_stats_queue qs = { 0 };
524 struct tc_fq_codel_xstats xstats;
525
526 if (idx < q->flows_cnt) {
527 const struct fq_codel_flow *flow = &q->flows[idx];
528 const struct sk_buff *skb = flow->head;
529
530 memset(&xstats, 0, sizeof(xstats));
531 xstats.type = TCA_FQ_CODEL_XSTATS_CLASS;
532 xstats.class_stats.deficit = flow->deficit;
533 xstats.class_stats.ldelay =
534 codel_time_to_us(flow->cvars.ldelay);
535 xstats.class_stats.count = flow->cvars.count;
536 xstats.class_stats.lastcount = flow->cvars.lastcount;
537 xstats.class_stats.dropping = flow->cvars.dropping;
538 if (flow->cvars.dropping) {
539 codel_tdiff_t delta = flow->cvars.drop_next -
540 codel_get_time();
541
542 xstats.class_stats.drop_next = (delta >= 0) ?
543 codel_time_to_us(delta) :
544 -codel_time_to_us(-delta);
545 }
546 while (skb) {
547 qs.qlen++;
548 skb = skb->next;
549 }
550 qs.backlog = q->backlogs[idx];
551 qs.drops = flow->dropped;
552 }
553 if (gnet_stats_copy_queue(d, &qs) < 0)
554 return -1;
555 if (idx < q->flows_cnt)
556 return gnet_stats_copy_app(d, &xstats, sizeof(xstats));
557 return 0;
558}
559
560static void fq_codel_walk(struct Qdisc *sch, struct qdisc_walker *arg)
561{
562 struct fq_codel_sched_data *q = qdisc_priv(sch);
563 unsigned int i;
564
565 if (arg->stop)
566 return;
567
568 for (i = 0; i < q->flows_cnt; i++) {
569 if (list_empty(&q->flows[i].flowchain) ||
570 arg->count < arg->skip) {
571 arg->count++;
572 continue;
573 }
574 if (arg->fn(sch, i + 1, arg) < 0) {
575 arg->stop = 1;
576 break;
577 }
578 arg->count++;
579 }
580}
581
582static const struct Qdisc_class_ops fq_codel_class_ops = {
583 .leaf = fq_codel_leaf,
584 .get = fq_codel_get,
585 .put = fq_codel_put,
586 .tcf_chain = fq_codel_find_tcf,
587 .bind_tcf = fq_codel_bind,
588 .unbind_tcf = fq_codel_put,
589 .dump = fq_codel_dump_class,
590 .dump_stats = fq_codel_dump_class_stats,
591 .walk = fq_codel_walk,
592};
593
594static struct Qdisc_ops fq_codel_qdisc_ops __read_mostly = {
595 .cl_ops = &fq_codel_class_ops,
596 .id = "fq_codel",
597 .priv_size = sizeof(struct fq_codel_sched_data),
598 .enqueue = fq_codel_enqueue,
599 .dequeue = fq_codel_dequeue,
600 .peek = qdisc_peek_dequeued,
601 .drop = fq_codel_drop,
602 .init = fq_codel_init,
603 .reset = fq_codel_reset,
604 .destroy = fq_codel_destroy,
605 .change = fq_codel_change,
606 .dump = fq_codel_dump,
607 .dump_stats = fq_codel_dump_stats,
608 .owner = THIS_MODULE,
609};
610
611static int __init fq_codel_module_init(void)
612{
613 return register_qdisc(&fq_codel_qdisc_ops);
614}
615
616static void __exit fq_codel_module_exit(void)
617{
618 unregister_qdisc(&fq_codel_qdisc_ops);
619}
620
621module_init(fq_codel_module_init)
622module_exit(fq_codel_module_exit)
623MODULE_AUTHOR("Eric Dumazet");
624MODULE_LICENSE("GPL");
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-26 14:41:42 -0500