Linux-2.6.12-rc2
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sched / sch_dsmark.c
blob8a3db9d95bab238b99a7018a73b5e9ddbd189835
1 /* net/sched/sch_dsmark.c - Differentiated Services field marker */
3 /* Written 1998-2000 by Werner Almesberger, EPFL ICA */
6 #include <linux/config.h>
7 #include <linux/module.h>
8 #include <linux/init.h>
9 #include <linux/types.h>
10 #include <linux/string.h>
11 #include <linux/errno.h>
12 #include <linux/skbuff.h>
13 #include <linux/netdevice.h> /* for pkt_sched */
14 #include <linux/rtnetlink.h>
15 #include <net/pkt_sched.h>
16 #include <net/dsfield.h>
17 #include <net/inet_ecn.h>
18 #include <asm/byteorder.h>
21 #if 1 /* control */
22 #define DPRINTK(format,args...) printk(KERN_DEBUG format,##args)
23 #else
24 #define DPRINTK(format,args...)
25 #endif
27 #if 0 /* data */
28 #define D2PRINTK(format,args...) printk(KERN_DEBUG format,##args)
29 #else
30 #define D2PRINTK(format,args...)
31 #endif
34 #define PRIV(sch) qdisc_priv(sch)
38 * classid class marking
39 * ------- ----- -------
40 * n/a 0 n/a
41 * x:0 1 use entry [0]
42 * ... ... ...
43 * x:y y>0 y+1 use entry [y]
44 * ... ... ...
45 * x:indices-1 indices use entry [indices-1]
46 * ... ... ...
47 * x:y y+1 use entry [y & (indices-1)]
48 * ... ... ...
49 * 0xffff 0x10000 use entry [indices-1]
53 #define NO_DEFAULT_INDEX (1 << 16)
55 struct dsmark_qdisc_data {
56 struct Qdisc *q;
57 struct tcf_proto *filter_list;
58 __u8 *mask; /* "owns" the array */
59 __u8 *value;
60 __u16 indices;
61 __u32 default_index; /* index range is 0...0xffff */
62 int set_tc_index;
66 /* ------------------------- Class/flow operations ------------------------- */
69 static int dsmark_graft(struct Qdisc *sch,unsigned long arg,
70 struct Qdisc *new,struct Qdisc **old)
72 struct dsmark_qdisc_data *p = PRIV(sch);
74 DPRINTK("dsmark_graft(sch %p,[qdisc %p],new %p,old %p)\n",sch,p,new,
75 old);
76 if (!new)
77 new = &noop_qdisc;
78 sch_tree_lock(sch);
79 *old = xchg(&p->q,new);
80 if (*old)
81 qdisc_reset(*old);
82 sch->q.qlen = 0;
83 sch_tree_unlock(sch); /* @@@ move up ? */
84 return 0;
88 static struct Qdisc *dsmark_leaf(struct Qdisc *sch, unsigned long arg)
90 struct dsmark_qdisc_data *p = PRIV(sch);
92 return p->q;
96 static unsigned long dsmark_get(struct Qdisc *sch,u32 classid)
98 struct dsmark_qdisc_data *p __attribute__((unused)) = PRIV(sch);
100 DPRINTK("dsmark_get(sch %p,[qdisc %p],classid %x)\n",sch,p,classid);
101 return TC_H_MIN(classid)+1;
105 static unsigned long dsmark_bind_filter(struct Qdisc *sch,
106 unsigned long parent, u32 classid)
108 return dsmark_get(sch,classid);
112 static void dsmark_put(struct Qdisc *sch, unsigned long cl)
117 static int dsmark_change(struct Qdisc *sch, u32 classid, u32 parent,
118 struct rtattr **tca, unsigned long *arg)
120 struct dsmark_qdisc_data *p = PRIV(sch);
121 struct rtattr *opt = tca[TCA_OPTIONS-1];
122 struct rtattr *tb[TCA_DSMARK_MAX];
124 DPRINTK("dsmark_change(sch %p,[qdisc %p],classid %x,parent %x),"
125 "arg 0x%lx\n",sch,p,classid,parent,*arg);
126 if (*arg > p->indices)
127 return -ENOENT;
128 if (!opt || rtattr_parse_nested(tb, TCA_DSMARK_MAX, opt))
129 return -EINVAL;
130 if (tb[TCA_DSMARK_MASK-1]) {
131 if (!RTA_PAYLOAD(tb[TCA_DSMARK_MASK-1]))
132 return -EINVAL;
133 p->mask[*arg-1] = *(__u8 *) RTA_DATA(tb[TCA_DSMARK_MASK-1]);
135 if (tb[TCA_DSMARK_VALUE-1]) {
136 if (!RTA_PAYLOAD(tb[TCA_DSMARK_VALUE-1]))
137 return -EINVAL;
138 p->value[*arg-1] = *(__u8 *) RTA_DATA(tb[TCA_DSMARK_VALUE-1]);
140 return 0;
144 static int dsmark_delete(struct Qdisc *sch,unsigned long arg)
146 struct dsmark_qdisc_data *p = PRIV(sch);
148 if (!arg || arg > p->indices)
149 return -EINVAL;
150 p->mask[arg-1] = 0xff;
151 p->value[arg-1] = 0;
152 return 0;
156 static void dsmark_walk(struct Qdisc *sch,struct qdisc_walker *walker)
158 struct dsmark_qdisc_data *p = PRIV(sch);
159 int i;
161 DPRINTK("dsmark_walk(sch %p,[qdisc %p],walker %p)\n",sch,p,walker);
162 if (walker->stop)
163 return;
164 for (i = 0; i < p->indices; i++) {
165 if (p->mask[i] == 0xff && !p->value[i])
166 continue;
167 if (walker->count >= walker->skip) {
168 if (walker->fn(sch, i+1, walker) < 0) {
169 walker->stop = 1;
170 break;
173 walker->count++;
178 static struct tcf_proto **dsmark_find_tcf(struct Qdisc *sch,unsigned long cl)
180 struct dsmark_qdisc_data *p = PRIV(sch);
182 return &p->filter_list;
186 /* --------------------------- Qdisc operations ---------------------------- */
189 static int dsmark_enqueue(struct sk_buff *skb,struct Qdisc *sch)
191 struct dsmark_qdisc_data *p = PRIV(sch);
192 struct tcf_result res;
193 int result;
194 int ret = NET_XMIT_POLICED;
196 D2PRINTK("dsmark_enqueue(skb %p,sch %p,[qdisc %p])\n",skb,sch,p);
197 if (p->set_tc_index) {
198 /* FIXME: Safe with non-linear skbs? --RR */
199 switch (skb->protocol) {
200 case __constant_htons(ETH_P_IP):
201 skb->tc_index = ipv4_get_dsfield(skb->nh.iph)
202 & ~INET_ECN_MASK;
203 break;
204 case __constant_htons(ETH_P_IPV6):
205 skb->tc_index = ipv6_get_dsfield(skb->nh.ipv6h)
206 & ~INET_ECN_MASK;
207 break;
208 default:
209 skb->tc_index = 0;
210 break;
213 result = TC_POLICE_OK; /* be nice to gcc */
214 if (TC_H_MAJ(skb->priority) == sch->handle) {
215 skb->tc_index = TC_H_MIN(skb->priority);
216 } else {
217 result = tc_classify(skb,p->filter_list,&res);
218 D2PRINTK("result %d class 0x%04x\n",result,res.classid);
219 switch (result) {
220 #ifdef CONFIG_NET_CLS_POLICE
221 case TC_POLICE_SHOT:
222 kfree_skb(skb);
223 break;
224 #if 0
225 case TC_POLICE_RECLASSIFY:
226 /* FIXME: what to do here ??? */
227 #endif
228 #endif
229 case TC_POLICE_OK:
230 skb->tc_index = TC_H_MIN(res.classid);
231 break;
232 case TC_POLICE_UNSPEC:
233 /* fall through */
234 default:
235 if (p->default_index != NO_DEFAULT_INDEX)
236 skb->tc_index = p->default_index;
237 break;
240 if (
241 #ifdef CONFIG_NET_CLS_POLICE
242 result == TC_POLICE_SHOT ||
243 #endif
245 ((ret = p->q->enqueue(skb,p->q)) != 0)) {
246 sch->qstats.drops++;
247 return ret;
249 sch->bstats.bytes += skb->len;
250 sch->bstats.packets++;
251 sch->q.qlen++;
252 return ret;
256 static struct sk_buff *dsmark_dequeue(struct Qdisc *sch)
258 struct dsmark_qdisc_data *p = PRIV(sch);
259 struct sk_buff *skb;
260 int index;
262 D2PRINTK("dsmark_dequeue(sch %p,[qdisc %p])\n",sch,p);
263 skb = p->q->ops->dequeue(p->q);
264 if (!skb)
265 return NULL;
266 sch->q.qlen--;
267 index = skb->tc_index & (p->indices-1);
268 D2PRINTK("index %d->%d\n",skb->tc_index,index);
269 switch (skb->protocol) {
270 case __constant_htons(ETH_P_IP):
271 ipv4_change_dsfield(skb->nh.iph,
272 p->mask[index],p->value[index]);
273 break;
274 case __constant_htons(ETH_P_IPV6):
275 ipv6_change_dsfield(skb->nh.ipv6h,
276 p->mask[index],p->value[index]);
277 break;
278 default:
280 * Only complain if a change was actually attempted.
281 * This way, we can send non-IP traffic through dsmark
282 * and don't need yet another qdisc as a bypass.
284 if (p->mask[index] != 0xff || p->value[index])
285 printk(KERN_WARNING "dsmark_dequeue: "
286 "unsupported protocol %d\n",
287 htons(skb->protocol));
288 break;
290 return skb;
294 static int dsmark_requeue(struct sk_buff *skb,struct Qdisc *sch)
296 int ret;
297 struct dsmark_qdisc_data *p = PRIV(sch);
299 D2PRINTK("dsmark_requeue(skb %p,sch %p,[qdisc %p])\n",skb,sch,p);
300 if ((ret = p->q->ops->requeue(skb, p->q)) == 0) {
301 sch->q.qlen++;
302 sch->qstats.requeues++;
303 return 0;
305 sch->qstats.drops++;
306 return ret;
310 static unsigned int dsmark_drop(struct Qdisc *sch)
312 struct dsmark_qdisc_data *p = PRIV(sch);
313 unsigned int len;
315 DPRINTK("dsmark_reset(sch %p,[qdisc %p])\n",sch,p);
316 if (!p->q->ops->drop)
317 return 0;
318 if (!(len = p->q->ops->drop(p->q)))
319 return 0;
320 sch->q.qlen--;
321 return len;
325 static int dsmark_init(struct Qdisc *sch,struct rtattr *opt)
327 struct dsmark_qdisc_data *p = PRIV(sch);
328 struct rtattr *tb[TCA_DSMARK_MAX];
329 __u16 tmp;
331 DPRINTK("dsmark_init(sch %p,[qdisc %p],opt %p)\n",sch,p,opt);
332 if (!opt ||
333 rtattr_parse(tb,TCA_DSMARK_MAX,RTA_DATA(opt),RTA_PAYLOAD(opt)) < 0 ||
334 !tb[TCA_DSMARK_INDICES-1] ||
335 RTA_PAYLOAD(tb[TCA_DSMARK_INDICES-1]) < sizeof(__u16))
336 return -EINVAL;
337 p->indices = *(__u16 *) RTA_DATA(tb[TCA_DSMARK_INDICES-1]);
338 if (!p->indices)
339 return -EINVAL;
340 for (tmp = p->indices; tmp != 1; tmp >>= 1) {
341 if (tmp & 1)
342 return -EINVAL;
344 p->default_index = NO_DEFAULT_INDEX;
345 if (tb[TCA_DSMARK_DEFAULT_INDEX-1]) {
346 if (RTA_PAYLOAD(tb[TCA_DSMARK_DEFAULT_INDEX-1]) < sizeof(__u16))
347 return -EINVAL;
348 p->default_index =
349 *(__u16 *) RTA_DATA(tb[TCA_DSMARK_DEFAULT_INDEX-1]);
351 p->set_tc_index = !!tb[TCA_DSMARK_SET_TC_INDEX-1];
352 p->mask = kmalloc(p->indices*2,GFP_KERNEL);
353 if (!p->mask)
354 return -ENOMEM;
355 p->value = p->mask+p->indices;
356 memset(p->mask,0xff,p->indices);
357 memset(p->value,0,p->indices);
358 if (!(p->q = qdisc_create_dflt(sch->dev, &pfifo_qdisc_ops)))
359 p->q = &noop_qdisc;
360 DPRINTK("dsmark_init: qdisc %p\n",&p->q);
361 return 0;
365 static void dsmark_reset(struct Qdisc *sch)
367 struct dsmark_qdisc_data *p = PRIV(sch);
369 DPRINTK("dsmark_reset(sch %p,[qdisc %p])\n",sch,p);
370 qdisc_reset(p->q);
371 sch->q.qlen = 0;
375 static void dsmark_destroy(struct Qdisc *sch)
377 struct dsmark_qdisc_data *p = PRIV(sch);
378 struct tcf_proto *tp;
380 DPRINTK("dsmark_destroy(sch %p,[qdisc %p])\n",sch,p);
381 while (p->filter_list) {
382 tp = p->filter_list;
383 p->filter_list = tp->next;
384 tcf_destroy(tp);
386 qdisc_destroy(p->q);
387 kfree(p->mask);
391 static int dsmark_dump_class(struct Qdisc *sch, unsigned long cl,
392 struct sk_buff *skb, struct tcmsg *tcm)
394 struct dsmark_qdisc_data *p = PRIV(sch);
395 unsigned char *b = skb->tail;
396 struct rtattr *rta;
398 DPRINTK("dsmark_dump_class(sch %p,[qdisc %p],class %ld\n",sch,p,cl);
399 if (!cl || cl > p->indices)
400 return -EINVAL;
401 tcm->tcm_handle = TC_H_MAKE(TC_H_MAJ(sch->handle),cl-1);
402 rta = (struct rtattr *) b;
403 RTA_PUT(skb,TCA_OPTIONS,0,NULL);
404 RTA_PUT(skb,TCA_DSMARK_MASK,1,&p->mask[cl-1]);
405 RTA_PUT(skb,TCA_DSMARK_VALUE,1,&p->value[cl-1]);
406 rta->rta_len = skb->tail-b;
407 return skb->len;
409 rtattr_failure:
410 skb_trim(skb,b-skb->data);
411 return -1;
414 static int dsmark_dump(struct Qdisc *sch, struct sk_buff *skb)
416 struct dsmark_qdisc_data *p = PRIV(sch);
417 unsigned char *b = skb->tail;
418 struct rtattr *rta;
420 rta = (struct rtattr *) b;
421 RTA_PUT(skb,TCA_OPTIONS,0,NULL);
422 RTA_PUT(skb,TCA_DSMARK_INDICES,sizeof(__u16),&p->indices);
423 if (p->default_index != NO_DEFAULT_INDEX) {
424 __u16 tmp = p->default_index;
426 RTA_PUT(skb,TCA_DSMARK_DEFAULT_INDEX, sizeof(__u16), &tmp);
428 if (p->set_tc_index)
429 RTA_PUT(skb, TCA_DSMARK_SET_TC_INDEX, 0, NULL);
430 rta->rta_len = skb->tail-b;
431 return skb->len;
433 rtattr_failure:
434 skb_trim(skb,b-skb->data);
435 return -1;
438 static struct Qdisc_class_ops dsmark_class_ops = {
439 .graft = dsmark_graft,
440 .leaf = dsmark_leaf,
441 .get = dsmark_get,
442 .put = dsmark_put,
443 .change = dsmark_change,
444 .delete = dsmark_delete,
445 .walk = dsmark_walk,
446 .tcf_chain = dsmark_find_tcf,
447 .bind_tcf = dsmark_bind_filter,
448 .unbind_tcf = dsmark_put,
449 .dump = dsmark_dump_class,
452 static struct Qdisc_ops dsmark_qdisc_ops = {
453 .next = NULL,
454 .cl_ops = &dsmark_class_ops,
455 .id = "dsmark",
456 .priv_size = sizeof(struct dsmark_qdisc_data),
457 .enqueue = dsmark_enqueue,
458 .dequeue = dsmark_dequeue,
459 .requeue = dsmark_requeue,
460 .drop = dsmark_drop,
461 .init = dsmark_init,
462 .reset = dsmark_reset,
463 .destroy = dsmark_destroy,
464 .change = NULL,
465 .dump = dsmark_dump,
466 .owner = THIS_MODULE,
469 static int __init dsmark_module_init(void)
471 return register_qdisc(&dsmark_qdisc_ops);
473 static void __exit dsmark_module_exit(void)
475 unregister_qdisc(&dsmark_qdisc_ops);
477 module_init(dsmark_module_init)
478 module_exit(dsmark_module_exit)
479 MODULE_LICENSE("GPL");