1 /* 2 * Copyright (c) 2008, Intel Corporation. 3 * 4 * This program is free software; you can redistribute it and/or modify it 5 * under the terms and conditions of the GNU General Public License, 6 * version 2, as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope it will be useful, but WITHOUT 9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for 11 * more details. 12 * 13 * You should have received a copy of the GNU General Public License along with 14 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple 15 * Place - Suite 330, Boston, MA 02111-1307 USA. 16 * 17 * Author: Alexander Duyck <alexander.h.duyck@intel.com> 18 */ 19 20 #include <linux/module.h> 21 #include <linux/slab.h> 22 #include <linux/types.h> 23 #include <linux/kernel.h> 24 #include <linux/string.h> 25 #include <linux/errno.h> 26 #include <linux/skbuff.h> 27 #include <net/netlink.h> 28 #include <net/pkt_sched.h> 29 30 31 struct multiq_sched_data { 32 u16 bands; 33 u16 max_bands; 34 u16 curband; 35 struct tcf_proto *filter_list; 36 struct Qdisc **queues; 37 }; 38 39 40 static struct Qdisc * 41 multiq_classify(struct sk_buff *skb, struct Qdisc *sch, int *qerr) 42 { 43 struct multiq_sched_data *q = qdisc_priv(sch); 44 u32 band; 45 struct tcf_result res; 46 int err; 47 48 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; 49 err = tc_classify(skb, q->filter_list, &res); 50 #ifdef CONFIG_NET_CLS_ACT 51 switch (err) { 52 case TC_ACT_STOLEN: 53 case TC_ACT_QUEUED: 54 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN; 55 case TC_ACT_SHOT: 56 return NULL; 57 } 58 #endif 59 band = skb_get_queue_mapping(skb); 60 61 if (band >= q->bands) 62 return q->queues[0]; 63 64 return q->queues[band]; 65 } 66 67 static int 68 multiq_enqueue(struct sk_buff *skb, struct Qdisc *sch) 69 { 70 struct Qdisc *qdisc; 71 int ret; 72 73 qdisc = multiq_classify(skb, sch, &ret); 74 #ifdef CONFIG_NET_CLS_ACT 75 if (qdisc == NULL) { 76 77 if (ret & __NET_XMIT_BYPASS) 78 sch->qstats.drops++; 79 kfree_skb(skb); 80 return ret; 81 } 82 #endif 83 84 ret = qdisc_enqueue(skb, qdisc); 85 if (ret == NET_XMIT_SUCCESS) { 86 sch->bstats.bytes += qdisc_pkt_len(skb); 87 sch->bstats.packets++; 88 sch->q.qlen++; 89 return NET_XMIT_SUCCESS; 90 } 91 if (net_xmit_drop_count(ret)) 92 sch->qstats.drops++; 93 return ret; 94 } 95 96 static struct sk_buff *multiq_dequeue(struct Qdisc *sch) 97 { 98 struct multiq_sched_data *q = qdisc_priv(sch); 99 struct Qdisc *qdisc; 100 struct sk_buff *skb; 101 int band; 102 103 for (band = 0; band < q->bands; band++) { 104 /* cycle through bands to ensure fairness */ 105 q->curband++; 106 if (q->curband >= q->bands) 107 q->curband = 0; 108 109 /* Check that target subqueue is available before 110 * pulling an skb to avoid head-of-line blocking. 111 */ 112 if (!__netif_subqueue_stopped(qdisc_dev(sch), q->curband)) { 113 qdisc = q->queues[q->curband]; 114 skb = qdisc->dequeue(qdisc); 115 if (skb) { 116 sch->q.qlen--; 117 return skb; 118 } 119 } 120 } 121 return NULL; 122 123 } 124 125 static struct sk_buff *multiq_peek(struct Qdisc *sch) 126 { 127 struct multiq_sched_data *q = qdisc_priv(sch); 128 unsigned int curband = q->curband; 129 struct Qdisc *qdisc; 130 struct sk_buff *skb; 131 int band; 132 133 for (band = 0; band < q->bands; band++) { 134 /* cycle through bands to ensure fairness */ 135 curband++; 136 if (curband >= q->bands) 137 curband = 0; 138 139 /* Check that target subqueue is available before 140 * pulling an skb to avoid head-of-line blocking. 141 */ 142 if (!__netif_subqueue_stopped(qdisc_dev(sch), curband)) { 143 qdisc = q->queues[curband]; 144 skb = qdisc->ops->peek(qdisc); 145 if (skb) 146 return skb; 147 } 148 } 149 return NULL; 150 151 } 152 153 static unsigned int multiq_drop(struct Qdisc *sch) 154 { 155 struct multiq_sched_data *q = qdisc_priv(sch); 156 int band; 157 unsigned int len; 158 struct Qdisc *qdisc; 159 160 for (band = q->bands-1; band >= 0; band--) { 161 qdisc = q->queues[band]; 162 if (qdisc->ops->drop) { 163 len = qdisc->ops->drop(qdisc); 164 if (len != 0) { 165 sch->q.qlen--; 166 return len; 167 } 168 } 169 } 170 return 0; 171 } 172 173 174 static void 175 multiq_reset(struct Qdisc *sch) 176 { 177 u16 band; 178 struct multiq_sched_data *q = qdisc_priv(sch); 179 180 for (band = 0; band < q->bands; band++) 181 qdisc_reset(q->queues[band]); 182 sch->q.qlen = 0; 183 q->curband = 0; 184 } 185 186 static void 187 multiq_destroy(struct Qdisc *sch) 188 { 189 int band; 190 struct multiq_sched_data *q = qdisc_priv(sch); 191 192 tcf_destroy_chain(&q->filter_list); 193 for (band = 0; band < q->bands; band++) 194 qdisc_destroy(q->queues[band]); 195 196 kfree(q->queues); 197 } 198 199 static int multiq_tune(struct Qdisc *sch, struct nlattr *opt) 200 { 201 struct multiq_sched_data *q = qdisc_priv(sch); 202 struct tc_multiq_qopt *qopt; 203 int i; 204 205 if (!netif_is_multiqueue(qdisc_dev(sch))) 206 return -EOPNOTSUPP; 207 if (nla_len(opt) < sizeof(*qopt)) 208 return -EINVAL; 209 210 qopt = nla_data(opt); 211 212 qopt->bands = qdisc_dev(sch)->real_num_tx_queues; 213 214 sch_tree_lock(sch); 215 q->bands = qopt->bands; 216 for (i = q->bands; i < q->max_bands; i++) { 217 if (q->queues[i] != &noop_qdisc) { 218 struct Qdisc *child = q->queues[i]; 219 q->queues[i] = &noop_qdisc; 220 qdisc_tree_decrease_qlen(child, child->q.qlen); 221 qdisc_destroy(child); 222 } 223 } 224 225 sch_tree_unlock(sch); 226 227 for (i = 0; i < q->bands; i++) { 228 if (q->queues[i] == &noop_qdisc) { 229 struct Qdisc *child, *old; 230 child = qdisc_create_dflt(qdisc_dev(sch), 231 sch->dev_queue, 232 &pfifo_qdisc_ops, 233 TC_H_MAKE(sch->handle, 234 i + 1)); 235 if (child) { 236 sch_tree_lock(sch); 237 old = q->queues[i]; 238 q->queues[i] = child; 239 240 if (old != &noop_qdisc) { 241 qdisc_tree_decrease_qlen(old, 242 old->q.qlen); 243 qdisc_destroy(old); 244 } 245 sch_tree_unlock(sch); 246 } 247 } 248 } 249 return 0; 250 } 251 252 static int multiq_init(struct Qdisc *sch, struct nlattr *opt) 253 { 254 struct multiq_sched_data *q = qdisc_priv(sch); 255 int i, err; 256 257 q->queues = NULL; 258 259 if (opt == NULL) 260 return -EINVAL; 261 262 q->max_bands = qdisc_dev(sch)->num_tx_queues; 263 264 q->queues = kcalloc(q->max_bands, sizeof(struct Qdisc *), GFP_KERNEL); 265 if (!q->queues) 266 return -ENOBUFS; 267 for (i = 0; i < q->max_bands; i++) 268 q->queues[i] = &noop_qdisc; 269 270 err = multiq_tune(sch,opt); 271 272 if (err) 273 kfree(q->queues); 274 275 return err; 276 } 277 278 static int multiq_dump(struct Qdisc *sch, struct sk_buff *skb) 279 { 280 struct multiq_sched_data *q = qdisc_priv(sch); 281 unsigned char *b = skb_tail_pointer(skb); 282 struct tc_multiq_qopt opt; 283 284 opt.bands = q->bands; 285 opt.max_bands = q->max_bands; 286 287 NLA_PUT(skb, TCA_OPTIONS, sizeof(opt), &opt); 288 289 return skb->len; 290 291 nla_put_failure: 292 nlmsg_trim(skb, b); 293 return -1; 294 } 295 296 static int multiq_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, 297 struct Qdisc **old) 298 { 299 struct multiq_sched_data *q = qdisc_priv(sch); 300 unsigned long band = arg - 1; 301 302 if (new == NULL) 303 new = &noop_qdisc; 304 305 sch_tree_lock(sch); 306 *old = q->queues[band]; 307 q->queues[band] = new; 308 qdisc_tree_decrease_qlen(*old, (*old)->q.qlen); 309 qdisc_reset(*old); 310 sch_tree_unlock(sch); 311 312 return 0; 313 } 314 315 static struct Qdisc * 316 multiq_leaf(struct Qdisc *sch, unsigned long arg) 317 { 318 struct multiq_sched_data *q = qdisc_priv(sch); 319 unsigned long band = arg - 1; 320 321 return q->queues[band]; 322 } 323 324 static unsigned long multiq_get(struct Qdisc *sch, u32 classid) 325 { 326 struct multiq_sched_data *q = qdisc_priv(sch); 327 unsigned long band = TC_H_MIN(classid); 328 329 if (band - 1 >= q->bands) 330 return 0; 331 return band; 332 } 333 334 static unsigned long multiq_bind(struct Qdisc *sch, unsigned long parent, 335 u32 classid) 336 { 337 return multiq_get(sch, classid); 338 } 339 340 341 static void multiq_put(struct Qdisc *q, unsigned long cl) 342 { 343 } 344 345 static int multiq_dump_class(struct Qdisc *sch, unsigned long cl, 346 struct sk_buff *skb, struct tcmsg *tcm) 347 { 348 struct multiq_sched_data *q = qdisc_priv(sch); 349 350 tcm->tcm_handle |= TC_H_MIN(cl); 351 tcm->tcm_info = q->queues[cl-1]->handle; 352 return 0; 353 } 354 355 static int multiq_dump_class_stats(struct Qdisc *sch, unsigned long cl, 356 struct gnet_dump *d) 357 { 358 struct multiq_sched_data *q = qdisc_priv(sch); 359 struct Qdisc *cl_q; 360 361 cl_q = q->queues[cl - 1]; 362 cl_q->qstats.qlen = cl_q->q.qlen; 363 if (gnet_stats_copy_basic(d, &cl_q->bstats) < 0 || 364 gnet_stats_copy_queue(d, &cl_q->qstats) < 0) 365 return -1; 366 367 return 0; 368 } 369 370 static void multiq_walk(struct Qdisc *sch, struct qdisc_walker *arg) 371 { 372 struct multiq_sched_data *q = qdisc_priv(sch); 373 int band; 374 375 if (arg->stop) 376 return; 377 378 for (band = 0; band < q->bands; band++) { 379 if (arg->count < arg->skip) { 380 arg->count++; 381 continue; 382 } 383 if (arg->fn(sch, band+1, arg) < 0) { 384 arg->stop = 1; 385 break; 386 } 387 arg->count++; 388 } 389 } 390 391 static struct tcf_proto **multiq_find_tcf(struct Qdisc *sch, unsigned long cl) 392 { 393 struct multiq_sched_data *q = qdisc_priv(sch); 394 395 if (cl) 396 return NULL; 397 return &q->filter_list; 398 } 399 400 static const struct Qdisc_class_ops multiq_class_ops = { 401 .graft = multiq_graft, 402 .leaf = multiq_leaf, 403 .get = multiq_get, 404 .put = multiq_put, 405 .walk = multiq_walk, 406 .tcf_chain = multiq_find_tcf, 407 .bind_tcf = multiq_bind, 408 .unbind_tcf = multiq_put, 409 .dump = multiq_dump_class, 410 .dump_stats = multiq_dump_class_stats, 411 }; 412 413 static struct Qdisc_ops multiq_qdisc_ops __read_mostly = { 414 .next = NULL, 415 .cl_ops = &multiq_class_ops, 416 .id = "multiq", 417 .priv_size = sizeof(struct multiq_sched_data), 418 .enqueue = multiq_enqueue, 419 .dequeue = multiq_dequeue, 420 .peek = multiq_peek, 421 .drop = multiq_drop, 422 .init = multiq_init, 423 .reset = multiq_reset, 424 .destroy = multiq_destroy, 425 .change = multiq_tune, 426 .dump = multiq_dump, 427 .owner = THIS_MODULE, 428 }; 429 430 static int __init multiq_module_init(void) 431 { 432 return register_qdisc(&multiq_qdisc_ops); 433 } 434 435 static void __exit multiq_module_exit(void) 436 { 437 unregister_qdisc(&multiq_qdisc_ops); 438 } 439 440 module_init(multiq_module_init) 441 module_exit(multiq_module_exit) 442 443 MODULE_LICENSE("GPL"); 444