diff options
author | Alexey Kuznetsov <kaber@ms2.inr.ac.ru> | 2007-09-30 20:51:33 -0400 |
---|---|---|
committer | David S. Miller <davem@sunset.davemloft.net> | 2007-10-02 00:01:23 -0400 |
commit | 32740ddc1095e5e320bf961dda146bf97bc28adb (patch) | |
tree | 613fccb815ed1c0a8e7712848a0ddea5d0d52b5e /net | |
parent | 3146b39c185f8a436d430132457e84fa1d8f8208 (diff) |
[SFQ]: Remove artificial limitation for queue limit.
This is followup to Patrick's patch. A little optimization to enqueue
routine allows to remove artificial limitation on queue length.
Plus, testing showed that hash function used by SFQ is too bad or even worse.
It does not even sweep the whole range of hash values.
Switched to Jenkins' hash.
Signed-off-by: Alexey Kuznetsov <kaber@ms2.inr.ac.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net')
-rw-r--r-- | net/sched/sch_sfq.c | 47 |
1 files changed, 31 insertions, 16 deletions
diff --git a/net/sched/sch_sfq.c b/net/sched/sch_sfq.c index 3a23e30bc79e..b542c875e154 100644 --- a/net/sched/sch_sfq.c +++ b/net/sched/sch_sfq.c | |||
@@ -19,6 +19,7 @@ | |||
19 | #include <linux/init.h> | 19 | #include <linux/init.h> |
20 | #include <linux/ipv6.h> | 20 | #include <linux/ipv6.h> |
21 | #include <linux/skbuff.h> | 21 | #include <linux/skbuff.h> |
22 | #include <linux/jhash.h> | ||
22 | #include <net/ip.h> | 23 | #include <net/ip.h> |
23 | #include <net/netlink.h> | 24 | #include <net/netlink.h> |
24 | #include <net/pkt_sched.h> | 25 | #include <net/pkt_sched.h> |
@@ -95,7 +96,7 @@ struct sfq_sched_data | |||
95 | 96 | ||
96 | /* Variables */ | 97 | /* Variables */ |
97 | struct timer_list perturb_timer; | 98 | struct timer_list perturb_timer; |
98 | int perturbation; | 99 | u32 perturbation; |
99 | sfq_index tail; /* Index of current slot in round */ | 100 | sfq_index tail; /* Index of current slot in round */ |
100 | sfq_index max_depth; /* Maximal depth */ | 101 | sfq_index max_depth; /* Maximal depth */ |
101 | 102 | ||
@@ -109,12 +110,7 @@ struct sfq_sched_data | |||
109 | 110 | ||
110 | static __inline__ unsigned sfq_fold_hash(struct sfq_sched_data *q, u32 h, u32 h1) | 111 | static __inline__ unsigned sfq_fold_hash(struct sfq_sched_data *q, u32 h, u32 h1) |
111 | { | 112 | { |
112 | int pert = q->perturbation; | 113 | return jhash_2words(h, h1, q->perturbation) & (SFQ_HASH_DIVISOR - 1); |
113 | |||
114 | /* Have we any rotation primitives? If not, WHY? */ | ||
115 | h ^= (h1<<pert) ^ (h1>>(0x1F - pert)); | ||
116 | h ^= h>>10; | ||
117 | return h & 0x3FF; | ||
118 | } | 114 | } |
119 | 115 | ||
120 | static unsigned sfq_hash(struct sfq_sched_data *q, struct sk_buff *skb) | 116 | static unsigned sfq_hash(struct sfq_sched_data *q, struct sk_buff *skb) |
@@ -256,6 +252,13 @@ sfq_enqueue(struct sk_buff *skb, struct Qdisc* sch) | |||
256 | q->ht[hash] = x = q->dep[SFQ_DEPTH].next; | 252 | q->ht[hash] = x = q->dep[SFQ_DEPTH].next; |
257 | q->hash[x] = hash; | 253 | q->hash[x] = hash; |
258 | } | 254 | } |
255 | /* If selected queue has length q->limit, this means that | ||
256 | * all another queues are empty and that we do simple tail drop, | ||
257 | * i.e. drop _this_ packet. | ||
258 | */ | ||
259 | if (q->qs[x].qlen >= q->limit) | ||
260 | return qdisc_drop(skb, sch); | ||
261 | |||
259 | sch->qstats.backlog += skb->len; | 262 | sch->qstats.backlog += skb->len; |
260 | __skb_queue_tail(&q->qs[x], skb); | 263 | __skb_queue_tail(&q->qs[x], skb); |
261 | sfq_inc(q, x); | 264 | sfq_inc(q, x); |
@@ -294,6 +297,19 @@ sfq_requeue(struct sk_buff *skb, struct Qdisc* sch) | |||
294 | } | 297 | } |
295 | sch->qstats.backlog += skb->len; | 298 | sch->qstats.backlog += skb->len; |
296 | __skb_queue_head(&q->qs[x], skb); | 299 | __skb_queue_head(&q->qs[x], skb); |
300 | /* If selected queue has length q->limit+1, this means that | ||
301 | * all another queues are empty and we do simple tail drop. | ||
302 | * This packet is still requeued at head of queue, tail packet | ||
303 | * is dropped. | ||
304 | */ | ||
305 | if (q->qs[x].qlen > q->limit) { | ||
306 | skb = q->qs[x].prev; | ||
307 | __skb_unlink(skb, &q->qs[x]); | ||
308 | sch->qstats.drops++; | ||
309 | sch->qstats.backlog -= skb->len; | ||
310 | kfree_skb(skb); | ||
311 | return NET_XMIT_CN; | ||
312 | } | ||
297 | sfq_inc(q, x); | 313 | sfq_inc(q, x); |
298 | if (q->qs[x].qlen == 1) { /* The flow is new */ | 314 | if (q->qs[x].qlen == 1) { /* The flow is new */ |
299 | if (q->tail == SFQ_DEPTH) { /* It is the first flow */ | 315 | if (q->tail == SFQ_DEPTH) { /* It is the first flow */ |
@@ -370,12 +386,10 @@ static void sfq_perturbation(unsigned long arg) | |||
370 | struct Qdisc *sch = (struct Qdisc*)arg; | 386 | struct Qdisc *sch = (struct Qdisc*)arg; |
371 | struct sfq_sched_data *q = qdisc_priv(sch); | 387 | struct sfq_sched_data *q = qdisc_priv(sch); |
372 | 388 | ||
373 | q->perturbation = net_random()&0x1F; | 389 | get_random_bytes(&q->perturbation, 4); |
374 | 390 | ||
375 | if (q->perturb_period) { | 391 | if (q->perturb_period) |
376 | q->perturb_timer.expires = jiffies + q->perturb_period; | 392 | mod_timer(&q->perturb_timer, jiffies + q->perturb_period); |
377 | add_timer(&q->perturb_timer); | ||
378 | } | ||
379 | } | 393 | } |
380 | 394 | ||
381 | static int sfq_change(struct Qdisc *sch, struct rtattr *opt) | 395 | static int sfq_change(struct Qdisc *sch, struct rtattr *opt) |
@@ -391,7 +405,7 @@ static int sfq_change(struct Qdisc *sch, struct rtattr *opt) | |||
391 | q->quantum = ctl->quantum ? : psched_mtu(sch->dev); | 405 | q->quantum = ctl->quantum ? : psched_mtu(sch->dev); |
392 | q->perturb_period = ctl->perturb_period*HZ; | 406 | q->perturb_period = ctl->perturb_period*HZ; |
393 | if (ctl->limit) | 407 | if (ctl->limit) |
394 | q->limit = min_t(u32, ctl->limit, SFQ_DEPTH - 2); | 408 | q->limit = min_t(u32, ctl->limit, SFQ_DEPTH - 1); |
395 | 409 | ||
396 | qlen = sch->q.qlen; | 410 | qlen = sch->q.qlen; |
397 | while (sch->q.qlen > q->limit) | 411 | while (sch->q.qlen > q->limit) |
@@ -400,8 +414,8 @@ static int sfq_change(struct Qdisc *sch, struct rtattr *opt) | |||
400 | 414 | ||
401 | del_timer(&q->perturb_timer); | 415 | del_timer(&q->perturb_timer); |
402 | if (q->perturb_period) { | 416 | if (q->perturb_period) { |
403 | q->perturb_timer.expires = jiffies + q->perturb_period; | 417 | mod_timer(&q->perturb_timer, jiffies + q->perturb_period); |
404 | add_timer(&q->perturb_timer); | 418 | get_random_bytes(&q->perturbation, 4); |
405 | } | 419 | } |
406 | sch_tree_unlock(sch); | 420 | sch_tree_unlock(sch); |
407 | return 0; | 421 | return 0; |
@@ -423,12 +437,13 @@ static int sfq_init(struct Qdisc *sch, struct rtattr *opt) | |||
423 | q->dep[i+SFQ_DEPTH].next = i+SFQ_DEPTH; | 437 | q->dep[i+SFQ_DEPTH].next = i+SFQ_DEPTH; |
424 | q->dep[i+SFQ_DEPTH].prev = i+SFQ_DEPTH; | 438 | q->dep[i+SFQ_DEPTH].prev = i+SFQ_DEPTH; |
425 | } | 439 | } |
426 | q->limit = SFQ_DEPTH - 2; | 440 | q->limit = SFQ_DEPTH - 1; |
427 | q->max_depth = 0; | 441 | q->max_depth = 0; |
428 | q->tail = SFQ_DEPTH; | 442 | q->tail = SFQ_DEPTH; |
429 | if (opt == NULL) { | 443 | if (opt == NULL) { |
430 | q->quantum = psched_mtu(sch->dev); | 444 | q->quantum = psched_mtu(sch->dev); |
431 | q->perturb_period = 0; | 445 | q->perturb_period = 0; |
446 | get_random_bytes(&q->perturbation, 4); | ||
432 | } else { | 447 | } else { |
433 | int err = sfq_change(sch, opt); | 448 | int err = sfq_change(sch, opt); |
434 | if (err) | 449 | if (err) |