1/* 2 * Copyright (c) 2006 Mellanox Technologies. All rights reserved 3 * 4 * This software is available to you under a choice of one of two 5 * licenses. You may choose to be licensed under the terms of the GNU 6 * General Public License (GPL) Version 2, available from the file 7 * COPYING in the main directory of this source tree, or the 8 * OpenIB.org BSD license below: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * 18 * - Redistributions in binary form must reproduce the above 19 * copyright notice, this list of conditions and the following 20 * disclaimer in the documentation and/or other materials 21 * provided with the distribution. 22 * 23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 30 * SOFTWARE. 31 */ 32 33#include <rdma/ib_cm.h> 34#include <net/dst.h> 35#include <net/icmp.h> 36#include <linux/icmpv6.h> 37#include <linux/delay.h> 38#include <linux/slab.h> 39#include <linux/vmalloc.h> 40#include <linux/moduleparam.h> 41#include <linux/sched/signal.h> 42#include <linux/sched/mm.h> 43 44#include "ipoib.h" 45 46int ipoib_max_conn_qp = 128; 47 48module_param_named(max_nonsrq_conn_qp, ipoib_max_conn_qp, int, 0444); 49MODULE_PARM_DESC(max_nonsrq_conn_qp, 50 "Max number of connected-mode QPs per interface " 51 "(applied only if shared receive queue is not available)"); 52 53#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG_DATA 54static int data_debug_level; 55 56module_param_named(cm_data_debug_level, data_debug_level, int, 0644); 57MODULE_PARM_DESC(cm_data_debug_level, 58 "Enable data path debug tracing for connected mode if > 0"); 59#endif 60 61#define IPOIB_CM_IETF_ID 0x1000000000000000ULL 62 63#define IPOIB_CM_RX_UPDATE_TIME (256 * HZ) 64#define IPOIB_CM_RX_TIMEOUT (2 * 256 * HZ) 65#define IPOIB_CM_RX_DELAY (3 * 256 * HZ) 66#define IPOIB_CM_RX_UPDATE_MASK (0x3) 67 68#define IPOIB_CM_RX_RESERVE (ALIGN(IPOIB_HARD_LEN, 16) - IPOIB_ENCAP_LEN) 69 70static struct ib_qp_attr ipoib_cm_err_attr = { 71 .qp_state = IB_QPS_ERR 72}; 73 74#define IPOIB_CM_RX_DRAIN_WRID 0xffffffff 75 76static struct ib_send_wr ipoib_cm_rx_drain_wr = { 77 .opcode = IB_WR_SEND, 78}; 79 80static int ipoib_cm_tx_handler(struct ib_cm_id *cm_id, 81 const struct ib_cm_event *event); 82 83static void ipoib_cm_dma_unmap_rx(struct ipoib_dev_priv *priv, int frags, 84 u64 mapping[IPOIB_CM_RX_SG]) 85{ 86 int i; 87 88 ib_dma_unmap_single(priv->ca, mapping[0], IPOIB_CM_HEAD_SIZE, DMA_FROM_DEVICE); 89 90 for (i = 0; i < frags; ++i) 91 ib_dma_unmap_page(priv->ca, mapping[i + 1], PAGE_SIZE, DMA_FROM_DEVICE); 92} 93 94static int ipoib_cm_post_receive_srq(struct net_device *dev, int id) 95{ 96 struct ipoib_dev_priv *priv = ipoib_priv(dev); 97 int i, ret; 98 99 priv->cm.rx_wr.wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV; 100 101 for (i = 0; i < priv->cm.num_frags; ++i) 102 priv->cm.rx_sge[i].addr = priv->cm.srq_ring[id].mapping[i]; 103 104 ret = ib_post_srq_recv(priv->cm.srq, &priv->cm.rx_wr, NULL); 105 if (unlikely(ret)) { 106 ipoib_warn(priv, "post srq failed for buf %d (%d)\n", id, ret); 107 ipoib_cm_dma_unmap_rx(priv, priv->cm.num_frags - 1, 108 priv->cm.srq_ring[id].mapping); 109 dev_kfree_skb_any(priv->cm.srq_ring[id].skb); 110 priv->cm.srq_ring[id].skb = NULL; 111 } 112 113 return ret; 114} 115 116static int ipoib_cm_post_receive_nonsrq(struct net_device *dev, 117 struct ipoib_cm_rx *rx, 118 struct ib_recv_wr *wr, 119 struct ib_sge *sge, int id) 120{ 121 struct ipoib_dev_priv *priv = ipoib_priv(dev); 122 int i, ret; 123 124 wr->wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV; 125 126 for (i = 0; i < IPOIB_CM_RX_SG; ++i) 127 sge[i].addr = rx->rx_ring[id].mapping[i]; 128 129 ret = ib_post_recv(rx->qp, wr, NULL); 130 if (unlikely(ret)) { 131 ipoib_warn(priv, "post recv failed for buf %d (%d)\n", id, ret); 132 ipoib_cm_dma_unmap_rx(priv, IPOIB_CM_RX_SG - 1, 133 rx->rx_ring[id].mapping); 134 dev_kfree_skb_any(rx->rx_ring[id].skb); 135 rx->rx_ring[id].skb = NULL; 136 } 137 138 return ret; 139} 140 141static struct sk_buff *ipoib_cm_alloc_rx_skb(struct net_device *dev, 142 struct ipoib_cm_rx_buf *rx_ring, 143 int id, int frags, 144 u64 mapping[IPOIB_CM_RX_SG], 145 gfp_t gfp) 146{ 147 struct ipoib_dev_priv *priv = ipoib_priv(dev); 148 struct sk_buff *skb; 149 int i; 150 151 skb = dev_alloc_skb(ALIGN(IPOIB_CM_HEAD_SIZE + IPOIB_PSEUDO_LEN, 16)); 152 if (unlikely(!skb)) 153 return NULL; 154 155 /* 156 * IPoIB adds a IPOIB_ENCAP_LEN byte header, this will align the 157 * IP header to a multiple of 16. 158 */ 159 skb_reserve(skb, IPOIB_CM_RX_RESERVE); 160 161 mapping[0] = ib_dma_map_single(priv->ca, skb->data, IPOIB_CM_HEAD_SIZE, 162 DMA_FROM_DEVICE); 163 if (unlikely(ib_dma_mapping_error(priv->ca, mapping[0]))) { 164 dev_kfree_skb_any(skb); 165 return NULL; 166 } 167 168 for (i = 0; i < frags; i++) { 169 struct page *page = alloc_page(gfp); 170 171 if (!page) 172 goto partial_error; 173 skb_fill_page_desc(skb, i, page, 0, PAGE_SIZE); 174 175 mapping[i + 1] = ib_dma_map_page(priv->ca, page, 176 0, PAGE_SIZE, DMA_FROM_DEVICE); 177 if (unlikely(ib_dma_mapping_error(priv->ca, mapping[i + 1]))) 178 goto partial_error; 179 } 180 181 rx_ring[id].skb = skb; 182 return skb; 183 184partial_error: 185 186 ib_dma_unmap_single(priv->ca, mapping[0], IPOIB_CM_HEAD_SIZE, DMA_FROM_DEVICE); 187 188 for (; i > 0; --i) 189 ib_dma_unmap_page(priv->ca, mapping[i], PAGE_SIZE, DMA_FROM_DEVICE); 190 191 dev_kfree_skb_any(skb); 192 return NULL; 193} 194 195static void ipoib_cm_free_rx_ring(struct net_device *dev, 196 struct ipoib_cm_rx_buf *rx_ring) 197{ 198 struct ipoib_dev_priv *priv = ipoib_priv(dev); 199 int i; 200 201 for (i = 0; i < ipoib_recvq_size; ++i) 202 if (rx_ring[i].skb) { 203 ipoib_cm_dma_unmap_rx(priv, IPOIB_CM_RX_SG - 1, 204 rx_ring[i].mapping); 205 dev_kfree_skb_any(rx_ring[i].skb); 206 } 207 208 vfree(rx_ring); 209} 210 211static void ipoib_cm_start_rx_drain(struct ipoib_dev_priv *priv) 212{ 213 struct ipoib_cm_rx *p; 214 215 /* We only reserved 1 extra slot in CQ for drain WRs, so 216 * make sure we have at most 1 outstanding WR. */ 217 if (list_empty(&priv->cm.rx_flush_list) || 218 !list_empty(&priv->cm.rx_drain_list)) 219 return; 220 221 /* 222 * QPs on flush list are error state. This way, a "flush 223 * error" WC will be immediately generated for each WR we post. 224 */ 225 p = list_entry(priv->cm.rx_flush_list.next, typeof(*p), list); 226 ipoib_cm_rx_drain_wr.wr_id = IPOIB_CM_RX_DRAIN_WRID; 227 if (ib_post_send(p->qp, &ipoib_cm_rx_drain_wr, NULL)) 228 ipoib_warn(priv, "failed to post drain wr\n"); 229 230 list_splice_init(&priv->cm.rx_flush_list, &priv->cm.rx_drain_list); 231} 232 233static void ipoib_cm_rx_event_handler(struct ib_event *event, void *ctx) 234{ 235 struct ipoib_cm_rx *p = ctx; 236 struct ipoib_dev_priv *priv = ipoib_priv(p->dev); 237 unsigned long flags; 238 239 if (event->event != IB_EVENT_QP_LAST_WQE_REACHED) 240 return; 241 242 spin_lock_irqsave(&priv->lock, flags); 243 list_move(&p->list, &priv->cm.rx_flush_list); 244 p->state = IPOIB_CM_RX_FLUSH; 245 ipoib_cm_start_rx_drain(priv); 246 spin_unlock_irqrestore(&priv->lock, flags); 247} 248 249static struct ib_qp *ipoib_cm_create_rx_qp(struct net_device *dev, 250 struct ipoib_cm_rx *p) 251{ 252 struct ipoib_dev_priv *priv = ipoib_priv(dev); 253 struct ib_qp_init_attr attr = { 254 .event_handler = ipoib_cm_rx_event_handler, 255 .send_cq = priv->recv_cq, /* For drain WR */ 256 .recv_cq = priv->recv_cq, 257 .srq = priv->cm.srq, 258 .cap.max_send_wr = 1, /* For drain WR */ 259 .cap.max_send_sge = 1, /* FIXME: 0 Seems not to work */ 260 .sq_sig_type = IB_SIGNAL_ALL_WR, 261 .qp_type = IB_QPT_RC, 262 .qp_context = p, 263 }; 264 265 if (!ipoib_cm_has_srq(dev)) { 266 attr.cap.max_recv_wr = ipoib_recvq_size; 267 attr.cap.max_recv_sge = IPOIB_CM_RX_SG; 268 } 269 270 return ib_create_qp(priv->pd, &attr); 271} 272 273static int ipoib_cm_modify_rx_qp(struct net_device *dev, 274 struct ib_cm_id *cm_id, struct ib_qp *qp, 275 unsigned int psn) 276{ 277 struct ipoib_dev_priv *priv = ipoib_priv(dev); 278 struct ib_qp_attr qp_attr; 279 int qp_attr_mask, ret; 280 281 qp_attr.qp_state = IB_QPS_INIT; 282 ret = ib_cm_init_qp_attr(cm_id, &qp_attr, &qp_attr_mask); 283 if (ret) { 284 ipoib_warn(priv, "failed to init QP attr for INIT: %d\n", ret); 285 return ret; 286 } 287 ret = ib_modify_qp(qp, &qp_attr, qp_attr_mask); 288 if (ret) { 289 ipoib_warn(priv, "failed to modify QP to INIT: %d\n", ret); 290 return ret; 291 } 292 qp_attr.qp_state = IB_QPS_RTR; 293 ret = ib_cm_init_qp_attr(cm_id, &qp_attr, &qp_attr_mask); 294 if (ret) { 295 ipoib_warn(priv, "failed to init QP attr for RTR: %d\n", ret); 296 return ret; 297 } 298 qp_attr.rq_psn = psn; 299 ret = ib_modify_qp(qp, &qp_attr, qp_attr_mask); 300 if (ret) { 301 ipoib_warn(priv, "failed to modify QP to RTR: %d\n", ret); 302 return ret; 303 } 304 305 /* 306 * Current Mellanox HCA firmware won't generate completions 307 * with error for drain WRs unless the QP has been moved to 308 * RTS first. This work-around leaves a window where a QP has 309 * moved to error asynchronously, but this will eventually get 310 * fixed in firmware, so let's not error out if modify QP 311 * fails. 312 */ 313 qp_attr.qp_state = IB_QPS_RTS; 314 ret = ib_cm_init_qp_attr(cm_id, &qp_attr, &qp_attr_mask); 315 if (ret) { 316 ipoib_warn(priv, "failed to init QP attr for RTS: %d\n", ret); 317 return 0; 318 } 319 ret = ib_modify_qp(qp, &qp_attr, qp_attr_mask); 320 if (ret) { 321 ipoib_warn(priv, "failed to modify QP to RTS: %d\n", ret); 322 return 0; 323 } 324 325 return 0; 326} 327 328static void ipoib_cm_init_rx_wr(struct net_device *dev, 329 struct ib_recv_wr *wr, 330 struct ib_sge *sge) 331{ 332 struct ipoib_dev_priv *priv = ipoib_priv(dev); 333 int i; 334 335 for (i = 0; i < priv->cm.num_frags; ++i) 336 sge[i].lkey = priv->pd->local_dma_lkey; 337 338 sge[0].length = IPOIB_CM_HEAD_SIZE; 339 for (i = 1; i < priv->cm.num_frags; ++i) 340 sge[i].length = PAGE_SIZE; 341 342 wr->next = NULL; 343 wr->sg_list = sge; 344 wr->num_sge = priv->cm.num_frags; 345} 346 347static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_id, 348 struct ipoib_cm_rx *rx) 349{ 350 struct ipoib_dev_priv *priv = ipoib_priv(dev); 351 struct { 352 struct ib_recv_wr wr; 353 struct ib_sge sge[IPOIB_CM_RX_SG]; 354 } *t; 355 int ret; 356 int i; 357 358 rx->rx_ring = vzalloc(array_size(ipoib_recvq_size, 359 sizeof(*rx->rx_ring))); 360 if (!rx->rx_ring) 361 return -ENOMEM; 362 363 t = kmalloc(sizeof(*t), GFP_KERNEL); 364 if (!t) { 365 ret = -ENOMEM; 366 goto err_free_1; 367 } 368 369 ipoib_cm_init_rx_wr(dev, &t->wr, t->sge); 370 371 spin_lock_irq(&priv->lock); 372 373 if (priv->cm.nonsrq_conn_qp >= ipoib_max_conn_qp) { 374 spin_unlock_irq(&priv->lock); 375 ib_send_cm_rej(cm_id, IB_CM_REJ_NO_QP, NULL, 0, NULL, 0); 376 ret = -EINVAL; 377 goto err_free; 378 } else 379 ++priv->cm.nonsrq_conn_qp; 380 381 spin_unlock_irq(&priv->lock); 382 383 for (i = 0; i < ipoib_recvq_size; ++i) { 384 if (!ipoib_cm_alloc_rx_skb(dev, rx->rx_ring, i, IPOIB_CM_RX_SG - 1, 385 rx->rx_ring[i].mapping, 386 GFP_KERNEL)) { 387 ipoib_warn(priv, "failed to allocate receive buffer %d\n", i); 388 ret = -ENOMEM; 389 goto err_count; 390 } 391 ret = ipoib_cm_post_receive_nonsrq(dev, rx, &t->wr, t->sge, i); 392 if (ret) { 393 ipoib_warn(priv, "ipoib_cm_post_receive_nonsrq " 394 "failed for buf %d\n", i); 395 ret = -EIO; 396 goto err_count; 397 } 398 } 399 400 rx->recv_count = ipoib_recvq_size; 401 402 kfree(t); 403 404 return 0; 405 406err_count: 407 spin_lock_irq(&priv->lock); 408 --priv->cm.nonsrq_conn_qp; 409 spin_unlock_irq(&priv->lock); 410 411err_free: 412 kfree(t); 413 414err_free_1: 415 ipoib_cm_free_rx_ring(dev, rx->rx_ring); 416 417 return ret; 418} 419 420static int ipoib_cm_send_rep(struct net_device *dev, struct ib_cm_id *cm_id, 421 struct ib_qp *qp, 422 const struct ib_cm_req_event_param *req, 423 unsigned int psn) 424{ 425 struct ipoib_dev_priv *priv = ipoib_priv(dev); 426 struct ipoib_cm_data data = {}; 427 struct ib_cm_rep_param rep = {}; 428 429 data.qpn = cpu_to_be32(priv->qp->qp_num); 430 data.mtu = cpu_to_be32(IPOIB_CM_BUF_SIZE); 431 432 rep.private_data = &data; 433 rep.private_data_len = sizeof(data); 434 rep.flow_control = 0; 435 rep.rnr_retry_count = req->rnr_retry_count; 436 rep.srq = ipoib_cm_has_srq(dev); 437 rep.qp_num = qp->qp_num; 438 rep.starting_psn = psn; 439 return ib_send_cm_rep(cm_id, &rep); 440} 441 442static int ipoib_cm_req_handler(struct ib_cm_id *cm_id, 443 const struct ib_cm_event *event) 444{ 445 struct net_device *dev = cm_id->context; 446 struct ipoib_dev_priv *priv = ipoib_priv(dev); 447 struct ipoib_cm_rx *p; 448 unsigned int psn; 449 int ret; 450 451 ipoib_dbg(priv, "REQ arrived\n"); 452 p = kzalloc(sizeof(*p), GFP_KERNEL); 453 if (!p) 454 return -ENOMEM; 455 p->dev = dev; 456 p->id = cm_id; 457 cm_id->context = p; 458 p->state = IPOIB_CM_RX_LIVE; 459 p->jiffies = jiffies; 460 INIT_LIST_HEAD(&p->list); 461 462 p->qp = ipoib_cm_create_rx_qp(dev, p); 463 if (IS_ERR(p->qp)) { 464 ret = PTR_ERR(p->qp); 465 goto err_qp; 466 } 467 468 psn = prandom_u32() & 0xffffff; 469 ret = ipoib_cm_modify_rx_qp(dev, cm_id, p->qp, psn); 470 if (ret) 471 goto err_modify; 472 473 if (!ipoib_cm_has_srq(dev)) { 474 ret = ipoib_cm_nonsrq_init_rx(dev, cm_id, p); 475 if (ret) 476 goto err_modify; 477 } 478 479 spin_lock_irq(&priv->lock); 480 queue_delayed_work(priv->wq, 481 &priv->cm.stale_task, IPOIB_CM_RX_DELAY); 482 /* Add this entry to passive ids list head, but do not re-add it 483 * if IB_EVENT_QP_LAST_WQE_REACHED has moved it to flush list. */ 484 p->jiffies = jiffies; 485 if (p->state == IPOIB_CM_RX_LIVE) 486 list_move(&p->list, &priv->cm.passive_ids); 487 spin_unlock_irq(&priv->lock); 488 489 ret = ipoib_cm_send_rep(dev, cm_id, p->qp, &event->param.req_rcvd, psn); 490 if (ret) { 491 ipoib_warn(priv, "failed to send REP: %d\n", ret); 492 if (ib_modify_qp(p->qp, &ipoib_cm_err_attr, IB_QP_STATE)) 493 ipoib_warn(priv, "unable to move qp to error state\n"); 494 } 495 return 0; 496 497err_modify: 498 ib_destroy_qp(p->qp); 499err_qp: 500 kfree(p); 501 return ret; 502} 503 504static int ipoib_cm_rx_handler(struct ib_cm_id *cm_id, 505 const struct ib_cm_event *event) 506{ 507 struct ipoib_cm_rx *p; 508 struct ipoib_dev_priv *priv; 509 510 switch (event->event) { 511 case IB_CM_REQ_RECEIVED: 512 return ipoib_cm_req_handler(cm_id, event); 513 case IB_CM_DREQ_RECEIVED: 514 ib_send_cm_drep(cm_id, NULL, 0); 515 fallthrough; 516 case IB_CM_REJ_RECEIVED: 517 p = cm_id->context; 518 priv = ipoib_priv(p->dev); 519 if (ib_modify_qp(p->qp, &ipoib_cm_err_attr, IB_QP_STATE)) 520 ipoib_warn(priv, "unable to move qp to error state\n"); 521 fallthrough; 522 default: 523 return 0; 524 } 525} 526/* Adjust length of skb with fragments to match received data */ 527static void skb_put_frags(struct sk_buff *skb, unsigned int hdr_space, 528 unsigned int length, struct sk_buff *toskb) 529{ 530 int i, num_frags; 531 unsigned int size; 532 533 /* put header into skb */ 534 size = min(length, hdr_space); 535 skb->tail += size; 536 skb->len += size; 537 length -= size; 538 539 num_frags = skb_shinfo(skb)->nr_frags; 540 for (i = 0; i < num_frags; i++) { 541 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 542 543 if (length == 0) { 544 /* don't need this page */ 545 skb_fill_page_desc(toskb, i, skb_frag_page(frag), 546 0, PAGE_SIZE); 547 --skb_shinfo(skb)->nr_frags; 548 } else { 549 size = min_t(unsigned int, length, PAGE_SIZE); 550 551 skb_frag_size_set(frag, size); 552 skb->data_len += size; 553 skb->truesize += size; 554 skb->len += size; 555 length -= size; 556 } 557 } 558} 559 560void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc) 561{ 562 struct ipoib_dev_priv *priv = ipoib_priv(dev); 563 struct ipoib_cm_rx_buf *rx_ring; 564 unsigned int wr_id = wc->wr_id & ~(IPOIB_OP_CM | IPOIB_OP_RECV); 565 struct sk_buff *skb, *newskb; 566 struct ipoib_cm_rx *p; 567 unsigned long flags; 568 u64 mapping[IPOIB_CM_RX_SG]; 569 int frags; 570 int has_srq; 571 struct sk_buff *small_skb; 572 573 ipoib_dbg_data(priv, "cm recv completion: id %d, status: %d\n", 574 wr_id, wc->status); 575 576 if (unlikely(wr_id >= ipoib_recvq_size)) { 577 if (wr_id == (IPOIB_CM_RX_DRAIN_WRID & ~(IPOIB_OP_CM | IPOIB_OP_RECV))) { 578 spin_lock_irqsave(&priv->lock, flags); 579 list_splice_init(&priv->cm.rx_drain_list, &priv->cm.rx_reap_list); 580 ipoib_cm_start_rx_drain(priv); 581 queue_work(priv->wq, &priv->cm.rx_reap_task); 582 spin_unlock_irqrestore(&priv->lock, flags); 583 } else 584 ipoib_warn(priv, "cm recv completion event with wrid %d (> %d)\n", 585 wr_id, ipoib_recvq_size); 586 return; 587 } 588 589 p = wc->qp->qp_context; 590 591 has_srq = ipoib_cm_has_srq(dev); 592 rx_ring = has_srq ? priv->cm.srq_ring : p->rx_ring; 593 594 skb = rx_ring[wr_id].skb; 595 596 if (unlikely(wc->status != IB_WC_SUCCESS)) { 597 ipoib_dbg(priv, 598 "cm recv error (status=%d, wrid=%d vend_err %#x)\n", 599 wc->status, wr_id, wc->vendor_err); 600 ++dev->stats.rx_dropped; 601 if (has_srq) 602 goto repost; 603 else { 604 if (!--p->recv_count) { 605 spin_lock_irqsave(&priv->lock, flags); 606 list_move(&p->list, &priv->cm.rx_reap_list); 607 spin_unlock_irqrestore(&priv->lock, flags); 608 queue_work(priv->wq, &priv->cm.rx_reap_task); 609 } 610 return; 611 } 612 } 613 614 if (unlikely(!(wr_id & IPOIB_CM_RX_UPDATE_MASK))) { 615 if (p && time_after_eq(jiffies, p->jiffies + IPOIB_CM_RX_UPDATE_TIME)) { 616 spin_lock_irqsave(&priv->lock, flags); 617 p->jiffies = jiffies; 618 /* Move this entry to list head, but do not re-add it 619 * if it has been moved out of list. */ 620 if (p->state == IPOIB_CM_RX_LIVE) 621 list_move(&p->list, &priv->cm.passive_ids); 622 spin_unlock_irqrestore(&priv->lock, flags); 623 } 624 } 625 626 if (wc->byte_len < IPOIB_CM_COPYBREAK) { 627 int dlen = wc->byte_len; 628 629 small_skb = dev_alloc_skb(dlen + IPOIB_CM_RX_RESERVE); 630 if (small_skb) { 631 skb_reserve(small_skb, IPOIB_CM_RX_RESERVE); 632 ib_dma_sync_single_for_cpu(priv->ca, rx_ring[wr_id].mapping[0], 633 dlen, DMA_FROM_DEVICE); 634 skb_copy_from_linear_data(skb, small_skb->data, dlen); 635 ib_dma_sync_single_for_device(priv->ca, rx_ring[wr_id].mapping[0], 636 dlen, DMA_FROM_DEVICE); 637 skb_put(small_skb, dlen); 638 skb = small_skb; 639 goto copied; 640 } 641 } 642 643 frags = PAGE_ALIGN(wc->byte_len - 644 min_t(u32, wc->byte_len, IPOIB_CM_HEAD_SIZE)) / 645 PAGE_SIZE; 646 647 newskb = ipoib_cm_alloc_rx_skb(dev, rx_ring, wr_id, frags, 648 mapping, GFP_ATOMIC); 649 if (unlikely(!newskb)) { 650 /* 651 * If we can't allocate a new RX buffer, dump 652 * this packet and reuse the old buffer. 653 */ 654 ipoib_dbg(priv, "failed to allocate receive buffer %d\n", wr_id); 655 ++dev->stats.rx_dropped; 656 goto repost; 657 } 658 659 ipoib_cm_dma_unmap_rx(priv, frags, rx_ring[wr_id].mapping); 660 memcpy(rx_ring[wr_id].mapping, mapping, (frags + 1) * sizeof(*mapping)); 661 662 ipoib_dbg_data(priv, "received %d bytes, SLID 0x%04x\n", 663 wc->byte_len, wc->slid); 664 665 skb_put_frags(skb, IPOIB_CM_HEAD_SIZE, wc->byte_len, newskb); 666 667copied: 668 skb->protocol = ((struct ipoib_header *) skb->data)->proto; 669 skb_add_pseudo_hdr(skb); 670 671 ++dev->stats.rx_packets; 672 dev->stats.rx_bytes += skb->len; 673 674 skb->dev = dev; 675 /* XXX get correct PACKET_ type here */ 676 skb->pkt_type = PACKET_HOST; 677 netif_receive_skb(skb); 678 679repost: 680 if (has_srq) { 681 if (unlikely(ipoib_cm_post_receive_srq(dev, wr_id))) 682 ipoib_warn(priv, "ipoib_cm_post_receive_srq failed " 683 "for buf %d\n", wr_id); 684 } else { 685 if (unlikely(ipoib_cm_post_receive_nonsrq(dev, p, 686 &priv->cm.rx_wr, 687 priv->cm.rx_sge, 688 wr_id))) { 689 --p->recv_count; 690 ipoib_warn(priv, "ipoib_cm_post_receive_nonsrq failed " 691 "for buf %d\n", wr_id); 692 } 693 } 694} 695 696static inline int post_send(struct ipoib_dev_priv *priv, 697 struct ipoib_cm_tx *tx, 698 unsigned int wr_id, 699 struct ipoib_tx_buf *tx_req) 700{ 701 ipoib_build_sge(priv, tx_req); 702 703 priv->tx_wr.wr.wr_id = wr_id | IPOIB_OP_CM; 704 705 return ib_post_send(tx->qp, &priv->tx_wr.wr, NULL); 706} 707 708void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_tx *tx) 709{ 710 struct ipoib_dev_priv *priv = ipoib_priv(dev); 711 struct ipoib_tx_buf *tx_req; 712 int rc; 713 unsigned int usable_sge = tx->max_send_sge - !!skb_headlen(skb); 714 715 if (unlikely(skb->len > tx->mtu)) { 716 ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n", 717 skb->len, tx->mtu); 718 ++dev->stats.tx_dropped; 719 ++dev->stats.tx_errors; 720 ipoib_cm_skb_too_long(dev, skb, tx->mtu - IPOIB_ENCAP_LEN); 721 return; 722 } 723 if (skb_shinfo(skb)->nr_frags > usable_sge) { 724 if (skb_linearize(skb) < 0) { 725 ipoib_warn(priv, "skb could not be linearized\n"); 726 ++dev->stats.tx_dropped; 727 ++dev->stats.tx_errors; 728 dev_kfree_skb_any(skb); 729 return; 730 } 731 /* Does skb_linearize return ok without reducing nr_frags? */ 732 if (skb_shinfo(skb)->nr_frags > usable_sge) { 733 ipoib_warn(priv, "too many frags after skb linearize\n"); 734 ++dev->stats.tx_dropped; 735 ++dev->stats.tx_errors; 736 dev_kfree_skb_any(skb); 737 return; 738 } 739 } 740 ipoib_dbg_data(priv, "sending packet: head 0x%x length %d connection 0x%x\n", 741 tx->tx_head, skb->len, tx->qp->qp_num); 742 743 /* 744 * We put the skb into the tx_ring _before_ we call post_send() 745 * because it's entirely possible that the completion handler will 746 * run before we execute anything after the post_send(). That 747 * means we have to make sure everything is properly recorded and 748 * our state is consistent before we call post_send(). 749 */ 750 tx_req = &tx->tx_ring[tx->tx_head & (ipoib_sendq_size - 1)]; 751 tx_req->skb = skb; 752 753 if (unlikely(ipoib_dma_map_tx(priv->ca, tx_req))) { 754 ++dev->stats.tx_errors; 755 dev_kfree_skb_any(skb); 756 return; 757 } 758 759 if ((priv->global_tx_head - priv->global_tx_tail) == 760 ipoib_sendq_size - 1) { 761 ipoib_dbg(priv, "TX ring 0x%x full, stopping kernel net queue\n", 762 tx->qp->qp_num); 763 netif_stop_queue(dev); 764 } 765 766 skb_orphan(skb); 767 skb_dst_drop(skb); 768 769 if (netif_queue_stopped(dev)) { 770 rc = ib_req_notify_cq(priv->send_cq, IB_CQ_NEXT_COMP | 771 IB_CQ_REPORT_MISSED_EVENTS); 772 if (unlikely(rc < 0)) 773 ipoib_warn(priv, "IPoIB/CM:request notify on send CQ failed\n"); 774 else if (rc) 775 napi_schedule(&priv->send_napi); 776 } 777 778 rc = post_send(priv, tx, tx->tx_head & (ipoib_sendq_size - 1), tx_req); 779 if (unlikely(rc)) { 780 ipoib_warn(priv, "IPoIB/CM:post_send failed, error %d\n", rc); 781 ++dev->stats.tx_errors; 782 ipoib_dma_unmap_tx(priv, tx_req); 783 dev_kfree_skb_any(skb); 784 785 if (netif_queue_stopped(dev)) 786 netif_wake_queue(dev); 787 } else { 788 netif_trans_update(dev); 789 ++tx->tx_head; 790 ++priv->global_tx_head; 791 } 792} 793 794void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc) 795{ 796 struct ipoib_dev_priv *priv = ipoib_priv(dev); 797 struct ipoib_cm_tx *tx = wc->qp->qp_context; 798 unsigned int wr_id = wc->wr_id & ~IPOIB_OP_CM; 799 struct ipoib_tx_buf *tx_req; 800 unsigned long flags; 801 802 ipoib_dbg_data(priv, "cm send completion: id %d, status: %d\n", 803 wr_id, wc->status); 804 805 if (unlikely(wr_id >= ipoib_sendq_size)) { 806 ipoib_warn(priv, "cm send completion event with wrid %d (> %d)\n", 807 wr_id, ipoib_sendq_size); 808 return; 809 } 810 811 tx_req = &tx->tx_ring[wr_id]; 812 813 ipoib_dma_unmap_tx(priv, tx_req); 814 815 /* FIXME: is this right? Shouldn't we only increment on success? */ 816 ++dev->stats.tx_packets; 817 dev->stats.tx_bytes += tx_req->skb->len; 818 819 dev_kfree_skb_any(tx_req->skb); 820 821 netif_tx_lock(dev); 822 823 ++tx->tx_tail; 824 ++priv->global_tx_tail; 825 826 if (unlikely(netif_queue_stopped(dev) && 827 ((priv->global_tx_head - priv->global_tx_tail) <= 828 ipoib_sendq_size >> 1) && 829 test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags))) 830 netif_wake_queue(dev); 831 832 if (wc->status != IB_WC_SUCCESS && 833 wc->status != IB_WC_WR_FLUSH_ERR) { 834 struct ipoib_neigh *neigh; 835 836 /* IB_WC[_RNR]_RETRY_EXC_ERR error is part of the life cycle, 837 * so don't make waves. 838 */ 839 if (wc->status == IB_WC_RNR_RETRY_EXC_ERR || 840 wc->status == IB_WC_RETRY_EXC_ERR) 841 ipoib_dbg(priv, 842 "%s: failed cm send event (status=%d, wrid=%d vend_err %#x)\n", 843 __func__, wc->status, wr_id, wc->vendor_err); 844 else 845 ipoib_warn(priv, 846 "%s: failed cm send event (status=%d, wrid=%d vend_err %#x)\n", 847 __func__, wc->status, wr_id, wc->vendor_err); 848 849 spin_lock_irqsave(&priv->lock, flags); 850 neigh = tx->neigh; 851 852 if (neigh) { 853 neigh->cm = NULL; 854 ipoib_neigh_free(neigh); 855 856 tx->neigh = NULL; 857 } 858 859 if (test_and_clear_bit(IPOIB_FLAG_INITIALIZED, &tx->flags)) { 860 list_move(&tx->list, &priv->cm.reap_list); 861 queue_work(priv->wq, &priv->cm.reap_task); 862 } 863 864 clear_bit(IPOIB_FLAG_OPER_UP, &tx->flags); 865 866 spin_unlock_irqrestore(&priv->lock, flags); 867 } 868 869 netif_tx_unlock(dev); 870} 871 872int ipoib_cm_dev_open(struct net_device *dev) 873{ 874 struct ipoib_dev_priv *priv = ipoib_priv(dev); 875 int ret; 876 877 if (!IPOIB_CM_SUPPORTED(dev->dev_addr)) 878 return 0; 879 880 priv->cm.id = ib_create_cm_id(priv->ca, ipoib_cm_rx_handler, dev); 881 if (IS_ERR(priv->cm.id)) { 882 pr_warn("%s: failed to create CM ID\n", priv->ca->name); 883 ret = PTR_ERR(priv->cm.id); 884 goto err_cm; 885 } 886 887 ret = ib_cm_listen(priv->cm.id, cpu_to_be64(IPOIB_CM_IETF_ID | priv->qp->qp_num), 888 0); 889 if (ret) { 890 pr_warn("%s: failed to listen on ID 0x%llx\n", priv->ca->name, 891 IPOIB_CM_IETF_ID | priv->qp->qp_num); 892 goto err_listen; 893 } 894 895 return 0; 896 897err_listen: 898 ib_destroy_cm_id(priv->cm.id); 899err_cm: 900 priv->cm.id = NULL; 901 return ret; 902} 903 904static void ipoib_cm_free_rx_reap_list(struct net_device *dev) 905{ 906 struct ipoib_dev_priv *priv = ipoib_priv(dev); 907 struct ipoib_cm_rx *rx, *n; 908 LIST_HEAD(list); 909 910 spin_lock_irq(&priv->lock); 911 list_splice_init(&priv->cm.rx_reap_list, &list); 912 spin_unlock_irq(&priv->lock); 913 914 list_for_each_entry_safe(rx, n, &list, list) { 915 ib_destroy_cm_id(rx->id); 916 ib_destroy_qp(rx->qp); 917 if (!ipoib_cm_has_srq(dev)) { 918 ipoib_cm_free_rx_ring(priv->dev, rx->rx_ring); 919 spin_lock_irq(&priv->lock); 920 --priv->cm.nonsrq_conn_qp; 921 spin_unlock_irq(&priv->lock); 922 } 923 kfree(rx); 924 } 925} 926 927void ipoib_cm_dev_stop(struct net_device *dev) 928{ 929 struct ipoib_dev_priv *priv = ipoib_priv(dev); 930 struct ipoib_cm_rx *p; 931 unsigned long begin; 932 int ret; 933 934 if (!IPOIB_CM_SUPPORTED(dev->dev_addr) || !priv->cm.id) 935 return; 936 937 ib_destroy_cm_id(priv->cm.id); 938 priv->cm.id = NULL; 939 940 spin_lock_irq(&priv->lock); 941 while (!list_empty(&priv->cm.passive_ids)) { 942 p = list_entry(priv->cm.passive_ids.next, typeof(*p), list); 943 list_move(&p->list, &priv->cm.rx_error_list); 944 p->state = IPOIB_CM_RX_ERROR; 945 spin_unlock_irq(&priv->lock); 946 ret = ib_modify_qp(p->qp, &ipoib_cm_err_attr, IB_QP_STATE); 947 if (ret) 948 ipoib_warn(priv, "unable to move qp to error state: %d\n", ret); 949 spin_lock_irq(&priv->lock); 950 } 951 952 /* Wait for all RX to be drained */ 953 begin = jiffies; 954 955 while (!list_empty(&priv->cm.rx_error_list) || 956 !list_empty(&priv->cm.rx_flush_list) || 957 !list_empty(&priv->cm.rx_drain_list)) { 958 if (time_after(jiffies, begin + 5 * HZ)) { 959 ipoib_warn(priv, "RX drain timing out\n"); 960 961 /* 962 * assume the HW is wedged and just free up everything. 963 */ 964 list_splice_init(&priv->cm.rx_flush_list, 965 &priv->cm.rx_reap_list); 966 list_splice_init(&priv->cm.rx_error_list, 967 &priv->cm.rx_reap_list); 968 list_splice_init(&priv->cm.rx_drain_list, 969 &priv->cm.rx_reap_list); 970 break; 971 } 972 spin_unlock_irq(&priv->lock); 973 usleep_range(1000, 2000); 974 ipoib_drain_cq(dev); 975 spin_lock_irq(&priv->lock); 976 } 977 978 spin_unlock_irq(&priv->lock); 979 980 ipoib_cm_free_rx_reap_list(dev); 981 982 cancel_delayed_work(&priv->cm.stale_task); 983} 984 985static int ipoib_cm_rep_handler(struct ib_cm_id *cm_id, 986 const struct ib_cm_event *event) 987{ 988 struct ipoib_cm_tx *p = cm_id->context; 989 struct ipoib_dev_priv *priv = ipoib_priv(p->dev); 990 struct ipoib_cm_data *data = event->private_data; 991 struct sk_buff_head skqueue; 992 struct ib_qp_attr qp_attr; 993 int qp_attr_mask, ret; 994 struct sk_buff *skb; 995 996 p->mtu = be32_to_cpu(data->mtu); 997 998 if (p->mtu <= IPOIB_ENCAP_LEN) { 999 ipoib_warn(priv, "Rejecting connection: mtu %d <= %d\n", 1000 p->mtu, IPOIB_ENCAP_LEN); 1001 return -EINVAL; 1002 } 1003 1004 qp_attr.qp_state = IB_QPS_RTR; 1005 ret = ib_cm_init_qp_attr(cm_id, &qp_attr, &qp_attr_mask); 1006 if (ret) { 1007 ipoib_warn(priv, "failed to init QP attr for RTR: %d\n", ret); 1008 return ret; 1009 } 1010 1011 qp_attr.rq_psn = 0 /* FIXME */; 1012 ret = ib_modify_qp(p->qp, &qp_attr, qp_attr_mask); 1013 if (ret) { 1014 ipoib_warn(priv, "failed to modify QP to RTR: %d\n", ret); 1015 return ret; 1016 } 1017 1018 qp_attr.qp_state = IB_QPS_RTS; 1019 ret = ib_cm_init_qp_attr(cm_id, &qp_attr, &qp_attr_mask); 1020 if (ret) { 1021 ipoib_warn(priv, "failed to init QP attr for RTS: %d\n", ret); 1022 return ret; 1023 } 1024 ret = ib_modify_qp(p->qp, &qp_attr, qp_attr_mask); 1025 if (ret) { 1026 ipoib_warn(priv, "failed to modify QP to RTS: %d\n", ret); 1027 return ret; 1028 } 1029 1030 skb_queue_head_init(&skqueue); 1031 1032 netif_tx_lock_bh(p->dev); 1033 spin_lock_irq(&priv->lock); 1034 set_bit(IPOIB_FLAG_OPER_UP, &p->flags); 1035 if (p->neigh) 1036 while ((skb = __skb_dequeue(&p->neigh->queue))) 1037 __skb_queue_tail(&skqueue, skb); 1038 spin_unlock_irq(&priv->lock); 1039 netif_tx_unlock_bh(p->dev); 1040 1041 while ((skb = __skb_dequeue(&skqueue))) { 1042 skb->dev = p->dev; 1043 ret = dev_queue_xmit(skb); 1044 if (ret) 1045 ipoib_warn(priv, "%s:dev_queue_xmit failed to re-queue packet, ret:%d\n", 1046 __func__, ret); 1047 } 1048 1049 ret = ib_send_cm_rtu(cm_id, NULL, 0); 1050 if (ret) { 1051 ipoib_warn(priv, "failed to send RTU: %d\n", ret); 1052 return ret; 1053 } 1054 return 0; 1055} 1056 1057static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_cm_tx *tx) 1058{ 1059 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1060 struct ib_qp_init_attr attr = { 1061 .send_cq = priv->send_cq, 1062 .recv_cq = priv->recv_cq, 1063 .srq = priv->cm.srq, 1064 .cap.max_send_wr = ipoib_sendq_size, 1065 .cap.max_send_sge = 1, 1066 .sq_sig_type = IB_SIGNAL_ALL_WR, 1067 .qp_type = IB_QPT_RC, 1068 .qp_context = tx, 1069 .create_flags = 0 1070 }; 1071 struct ib_qp *tx_qp; 1072 1073 if (dev->features & NETIF_F_SG) 1074 attr.cap.max_send_sge = min_t(u32, priv->ca->attrs.max_send_sge, 1075 MAX_SKB_FRAGS + 1); 1076 1077 tx_qp = ib_create_qp(priv->pd, &attr); 1078 tx->max_send_sge = attr.cap.max_send_sge; 1079 return tx_qp; 1080} 1081 1082static int ipoib_cm_send_req(struct net_device *dev, 1083 struct ib_cm_id *id, struct ib_qp *qp, 1084 u32 qpn, 1085 struct sa_path_rec *pathrec) 1086{ 1087 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1088 struct ipoib_cm_data data = {}; 1089 struct ib_cm_req_param req = {}; 1090 1091 data.qpn = cpu_to_be32(priv->qp->qp_num); 1092 data.mtu = cpu_to_be32(IPOIB_CM_BUF_SIZE); 1093 1094 req.primary_path = pathrec; 1095 req.alternate_path = NULL; 1096 req.service_id = cpu_to_be64(IPOIB_CM_IETF_ID | qpn); 1097 req.qp_num = qp->qp_num; 1098 req.qp_type = qp->qp_type; 1099 req.private_data = &data; 1100 req.private_data_len = sizeof(data); 1101 req.flow_control = 0; 1102 1103 req.starting_psn = 0; /* FIXME */ 1104 1105 /* 1106 * Pick some arbitrary defaults here; we could make these 1107 * module parameters if anyone cared about setting them. 1108 */ 1109 req.responder_resources = 4; 1110 req.remote_cm_response_timeout = 20; 1111 req.local_cm_response_timeout = 20; 1112 req.retry_count = 0; /* RFC draft warns against retries */ 1113 req.rnr_retry_count = 0; /* RFC draft warns against retries */ 1114 req.max_cm_retries = 15; 1115 req.srq = ipoib_cm_has_srq(dev); 1116 return ib_send_cm_req(id, &req); 1117} 1118 1119static int ipoib_cm_modify_tx_init(struct net_device *dev, 1120 struct ib_cm_id *cm_id, struct ib_qp *qp) 1121{ 1122 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1123 struct ib_qp_attr qp_attr; 1124 int qp_attr_mask, ret; 1125 ret = ib_find_pkey(priv->ca, priv->port, priv->pkey, &qp_attr.pkey_index); 1126 if (ret) { 1127 ipoib_warn(priv, "pkey 0x%x not found: %d\n", priv->pkey, ret); 1128 return ret; 1129 } 1130 1131 qp_attr.qp_state = IB_QPS_INIT; 1132 qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE; 1133 qp_attr.port_num = priv->port; 1134 qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS | IB_QP_PKEY_INDEX | IB_QP_PORT; 1135 1136 ret = ib_modify_qp(qp, &qp_attr, qp_attr_mask); 1137 if (ret) { 1138 ipoib_warn(priv, "failed to modify tx QP to INIT: %d\n", ret); 1139 return ret; 1140 } 1141 return 0; 1142} 1143 1144static int ipoib_cm_tx_init(struct ipoib_cm_tx *p, u32 qpn, 1145 struct sa_path_rec *pathrec) 1146{ 1147 struct ipoib_dev_priv *priv = ipoib_priv(p->dev); 1148 unsigned int noio_flag; 1149 int ret; 1150 1151 noio_flag = memalloc_noio_save(); 1152 p->tx_ring = vzalloc(array_size(ipoib_sendq_size, sizeof(*p->tx_ring))); 1153 if (!p->tx_ring) { 1154 memalloc_noio_restore(noio_flag); 1155 ret = -ENOMEM; 1156 goto err_tx; 1157 } 1158 1159 p->qp = ipoib_cm_create_tx_qp(p->dev, p); 1160 memalloc_noio_restore(noio_flag); 1161 if (IS_ERR(p->qp)) { 1162 ret = PTR_ERR(p->qp); 1163 ipoib_warn(priv, "failed to create tx qp: %d\n", ret); 1164 goto err_qp; 1165 } 1166 1167 p->id = ib_create_cm_id(priv->ca, ipoib_cm_tx_handler, p); 1168 if (IS_ERR(p->id)) { 1169 ret = PTR_ERR(p->id); 1170 ipoib_warn(priv, "failed to create tx cm id: %d\n", ret); 1171 goto err_id; 1172 } 1173 1174 ret = ipoib_cm_modify_tx_init(p->dev, p->id, p->qp); 1175 if (ret) { 1176 ipoib_warn(priv, "failed to modify tx qp to rtr: %d\n", ret); 1177 goto err_modify_send; 1178 } 1179 1180 ret = ipoib_cm_send_req(p->dev, p->id, p->qp, qpn, pathrec); 1181 if (ret) { 1182 ipoib_warn(priv, "failed to send cm req: %d\n", ret); 1183 goto err_modify_send; 1184 } 1185 1186 ipoib_dbg(priv, "Request connection 0x%x for gid %pI6 qpn 0x%x\n", 1187 p->qp->qp_num, pathrec->dgid.raw, qpn); 1188 1189 return 0; 1190 1191err_modify_send: 1192 ib_destroy_cm_id(p->id); 1193err_id: 1194 p->id = NULL; 1195 ib_destroy_qp(p->qp); 1196err_qp: 1197 p->qp = NULL; 1198 vfree(p->tx_ring); 1199err_tx: 1200 return ret; 1201} 1202 1203static void ipoib_cm_tx_destroy(struct ipoib_cm_tx *p) 1204{ 1205 struct ipoib_dev_priv *priv = ipoib_priv(p->dev); 1206 struct ipoib_tx_buf *tx_req; 1207 unsigned long begin; 1208 1209 ipoib_dbg(priv, "Destroy active connection 0x%x head 0x%x tail 0x%x\n", 1210 p->qp ? p->qp->qp_num : 0, p->tx_head, p->tx_tail); 1211 1212 if (p->id) 1213 ib_destroy_cm_id(p->id); 1214 1215 if (p->tx_ring) { 1216 /* Wait for all sends to complete */ 1217 begin = jiffies; 1218 while ((int) p->tx_tail - (int) p->tx_head < 0) { 1219 if (time_after(jiffies, begin + 5 * HZ)) { 1220 ipoib_warn(priv, "timing out; %d sends not completed\n", 1221 p->tx_head - p->tx_tail); 1222 goto timeout; 1223 } 1224 1225 usleep_range(1000, 2000); 1226 } 1227 } 1228 1229timeout: 1230 1231 while ((int) p->tx_tail - (int) p->tx_head < 0) { 1232 tx_req = &p->tx_ring[p->tx_tail & (ipoib_sendq_size - 1)]; 1233 ipoib_dma_unmap_tx(priv, tx_req); 1234 dev_kfree_skb_any(tx_req->skb); 1235 netif_tx_lock_bh(p->dev); 1236 ++p->tx_tail; 1237 ++priv->global_tx_tail; 1238 if (unlikely((priv->global_tx_head - priv->global_tx_tail) <= 1239 ipoib_sendq_size >> 1) && 1240 netif_queue_stopped(p->dev) && 1241 test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags)) 1242 netif_wake_queue(p->dev); 1243 netif_tx_unlock_bh(p->dev); 1244 } 1245 1246 if (p->qp) 1247 ib_destroy_qp(p->qp); 1248 1249 vfree(p->tx_ring); 1250 kfree(p); 1251} 1252 1253static int ipoib_cm_tx_handler(struct ib_cm_id *cm_id, 1254 const struct ib_cm_event *event) 1255{ 1256 struct ipoib_cm_tx *tx = cm_id->context; 1257 struct ipoib_dev_priv *priv = ipoib_priv(tx->dev); 1258 struct net_device *dev = priv->dev; 1259 struct ipoib_neigh *neigh; 1260 unsigned long flags; 1261 int ret; 1262 1263 switch (event->event) { 1264 case IB_CM_DREQ_RECEIVED: 1265 ipoib_dbg(priv, "DREQ received.\n"); 1266 ib_send_cm_drep(cm_id, NULL, 0); 1267 break; 1268 case IB_CM_REP_RECEIVED: 1269 ipoib_dbg(priv, "REP received.\n"); 1270 ret = ipoib_cm_rep_handler(cm_id, event); 1271 if (ret) 1272 ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED, 1273 NULL, 0, NULL, 0); 1274 break; 1275 case IB_CM_REQ_ERROR: 1276 case IB_CM_REJ_RECEIVED: 1277 case IB_CM_TIMEWAIT_EXIT: 1278 ipoib_dbg(priv, "CM error %d.\n", event->event); 1279 netif_tx_lock_bh(dev); 1280 spin_lock_irqsave(&priv->lock, flags); 1281 neigh = tx->neigh; 1282 1283 if (neigh) { 1284 neigh->cm = NULL; 1285 ipoib_neigh_free(neigh); 1286 1287 tx->neigh = NULL; 1288 } 1289 1290 if (test_and_clear_bit(IPOIB_FLAG_INITIALIZED, &tx->flags)) { 1291 list_move(&tx->list, &priv->cm.reap_list); 1292 queue_work(priv->wq, &priv->cm.reap_task); 1293 } 1294 1295 spin_unlock_irqrestore(&priv->lock, flags); 1296 netif_tx_unlock_bh(dev); 1297 break; 1298 default: 1299 break; 1300 } 1301 1302 return 0; 1303} 1304 1305struct ipoib_cm_tx *ipoib_cm_create_tx(struct net_device *dev, struct ipoib_path *path, 1306 struct ipoib_neigh *neigh) 1307{ 1308 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1309 struct ipoib_cm_tx *tx; 1310 1311 tx = kzalloc(sizeof(*tx), GFP_ATOMIC); 1312 if (!tx) 1313 return NULL; 1314 1315 neigh->cm = tx; 1316 tx->neigh = neigh; 1317 tx->dev = dev; 1318 list_add(&tx->list, &priv->cm.start_list); 1319 set_bit(IPOIB_FLAG_INITIALIZED, &tx->flags); 1320 queue_work(priv->wq, &priv->cm.start_task); 1321 return tx; 1322} 1323 1324void ipoib_cm_destroy_tx(struct ipoib_cm_tx *tx) 1325{ 1326 struct ipoib_dev_priv *priv = ipoib_priv(tx->dev); 1327 unsigned long flags; 1328 if (test_and_clear_bit(IPOIB_FLAG_INITIALIZED, &tx->flags)) { 1329 spin_lock_irqsave(&priv->lock, flags); 1330 list_move(&tx->list, &priv->cm.reap_list); 1331 queue_work(priv->wq, &priv->cm.reap_task); 1332 ipoib_dbg(priv, "Reap connection for gid %pI6\n", 1333 tx->neigh->daddr + 4); 1334 tx->neigh = NULL; 1335 spin_unlock_irqrestore(&priv->lock, flags); 1336 } 1337} 1338 1339#define QPN_AND_OPTIONS_OFFSET 4 1340 1341static void ipoib_cm_tx_start(struct work_struct *work) 1342{ 1343 struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv, 1344 cm.start_task); 1345 struct net_device *dev = priv->dev; 1346 struct ipoib_neigh *neigh; 1347 struct ipoib_cm_tx *p; 1348 unsigned long flags; 1349 struct ipoib_path *path; 1350 int ret; 1351 1352 struct sa_path_rec pathrec; 1353 u32 qpn; 1354 1355 netif_tx_lock_bh(dev); 1356 spin_lock_irqsave(&priv->lock, flags); 1357 1358 while (!list_empty(&priv->cm.start_list)) { 1359 p = list_entry(priv->cm.start_list.next, typeof(*p), list); 1360 list_del_init(&p->list); 1361 neigh = p->neigh; 1362 1363 qpn = IPOIB_QPN(neigh->daddr); 1364 /* 1365 * As long as the search is with these 2 locks, 1366 * path existence indicates its validity. 1367 */ 1368 path = __path_find(dev, neigh->daddr + QPN_AND_OPTIONS_OFFSET); 1369 if (!path) { 1370 pr_info("%s ignore not valid path %pI6\n", 1371 __func__, 1372 neigh->daddr + QPN_AND_OPTIONS_OFFSET); 1373 goto free_neigh; 1374 } 1375 memcpy(&pathrec, &path->pathrec, sizeof(pathrec)); 1376 1377 spin_unlock_irqrestore(&priv->lock, flags); 1378 netif_tx_unlock_bh(dev); 1379 1380 ret = ipoib_cm_tx_init(p, qpn, &pathrec); 1381 1382 netif_tx_lock_bh(dev); 1383 spin_lock_irqsave(&priv->lock, flags); 1384 1385 if (ret) { 1386free_neigh: 1387 neigh = p->neigh; 1388 if (neigh) { 1389 neigh->cm = NULL; 1390 ipoib_neigh_free(neigh); 1391 } 1392 list_del(&p->list); 1393 kfree(p); 1394 } 1395 } 1396 1397 spin_unlock_irqrestore(&priv->lock, flags); 1398 netif_tx_unlock_bh(dev); 1399} 1400 1401static void ipoib_cm_tx_reap(struct work_struct *work) 1402{ 1403 struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv, 1404 cm.reap_task); 1405 struct net_device *dev = priv->dev; 1406 struct ipoib_cm_tx *p; 1407 unsigned long flags; 1408 1409 netif_tx_lock_bh(dev); 1410 spin_lock_irqsave(&priv->lock, flags); 1411 1412 while (!list_empty(&priv->cm.reap_list)) { 1413 p = list_entry(priv->cm.reap_list.next, typeof(*p), list); 1414 list_del_init(&p->list); 1415 spin_unlock_irqrestore(&priv->lock, flags); 1416 netif_tx_unlock_bh(dev); 1417 ipoib_cm_tx_destroy(p); 1418 netif_tx_lock_bh(dev); 1419 spin_lock_irqsave(&priv->lock, flags); 1420 } 1421 1422 spin_unlock_irqrestore(&priv->lock, flags); 1423 netif_tx_unlock_bh(dev); 1424} 1425 1426static void ipoib_cm_skb_reap(struct work_struct *work) 1427{ 1428 struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv, 1429 cm.skb_task); 1430 struct net_device *dev = priv->dev; 1431 struct sk_buff *skb; 1432 unsigned long flags; 1433 unsigned int mtu = priv->mcast_mtu; 1434 1435 netif_tx_lock_bh(dev); 1436 spin_lock_irqsave(&priv->lock, flags); 1437 1438 while ((skb = skb_dequeue(&priv->cm.skb_queue))) { 1439 spin_unlock_irqrestore(&priv->lock, flags); 1440 netif_tx_unlock_bh(dev); 1441 1442 if (skb->protocol == htons(ETH_P_IP)) { 1443 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 1444 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu)); 1445 } 1446#if IS_ENABLED(CONFIG_IPV6) 1447 else if (skb->protocol == htons(ETH_P_IPV6)) { 1448 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 1449 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1450 } 1451#endif 1452 dev_kfree_skb_any(skb); 1453 1454 netif_tx_lock_bh(dev); 1455 spin_lock_irqsave(&priv->lock, flags); 1456 } 1457 1458 spin_unlock_irqrestore(&priv->lock, flags); 1459 netif_tx_unlock_bh(dev); 1460} 1461 1462void ipoib_cm_skb_too_long(struct net_device *dev, struct sk_buff *skb, 1463 unsigned int mtu) 1464{ 1465 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1466 int e = skb_queue_empty(&priv->cm.skb_queue); 1467 1468 skb_dst_update_pmtu(skb, mtu); 1469 1470 skb_queue_tail(&priv->cm.skb_queue, skb); 1471 if (e) 1472 queue_work(priv->wq, &priv->cm.skb_task); 1473} 1474 1475static void ipoib_cm_rx_reap(struct work_struct *work) 1476{ 1477 ipoib_cm_free_rx_reap_list(container_of(work, struct ipoib_dev_priv, 1478 cm.rx_reap_task)->dev); 1479} 1480 1481static void ipoib_cm_stale_task(struct work_struct *work) 1482{ 1483 struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv, 1484 cm.stale_task.work); 1485 struct ipoib_cm_rx *p; 1486 int ret; 1487 1488 spin_lock_irq(&priv->lock); 1489 while (!list_empty(&priv->cm.passive_ids)) { 1490 /* List is sorted by LRU, start from tail, 1491 * stop when we see a recently used entry */ 1492 p = list_entry(priv->cm.passive_ids.prev, typeof(*p), list); 1493 if (time_before_eq(jiffies, p->jiffies + IPOIB_CM_RX_TIMEOUT)) 1494 break; 1495 list_move(&p->list, &priv->cm.rx_error_list); 1496 p->state = IPOIB_CM_RX_ERROR; 1497 spin_unlock_irq(&priv->lock); 1498 ret = ib_modify_qp(p->qp, &ipoib_cm_err_attr, IB_QP_STATE); 1499 if (ret) 1500 ipoib_warn(priv, "unable to move qp to error state: %d\n", ret); 1501 spin_lock_irq(&priv->lock); 1502 } 1503 1504 if (!list_empty(&priv->cm.passive_ids)) 1505 queue_delayed_work(priv->wq, 1506 &priv->cm.stale_task, IPOIB_CM_RX_DELAY); 1507 spin_unlock_irq(&priv->lock); 1508} 1509 1510static ssize_t show_mode(struct device *d, struct device_attribute *attr, 1511 char *buf) 1512{ 1513 struct net_device *dev = to_net_dev(d); 1514 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1515 1516 if (test_bit(IPOIB_FLAG_ADMIN_CM, &priv->flags)) 1517 return sprintf(buf, "connected\n"); 1518 else 1519 return sprintf(buf, "datagram\n"); 1520} 1521 1522static ssize_t set_mode(struct device *d, struct device_attribute *attr, 1523 const char *buf, size_t count) 1524{ 1525 struct net_device *dev = to_net_dev(d); 1526 int ret; 1527 1528 if (!rtnl_trylock()) { 1529 return restart_syscall(); 1530 } 1531 1532 if (dev->reg_state != NETREG_REGISTERED) { 1533 rtnl_unlock(); 1534 return -EPERM; 1535 } 1536 1537 ret = ipoib_set_mode(dev, buf); 1538 1539 /* The assumption is that the function ipoib_set_mode returned 1540 * with the rtnl held by it, if not the value -EBUSY returned, 1541 * then no need to rtnl_unlock 1542 */ 1543 if (ret != -EBUSY) 1544 rtnl_unlock(); 1545 1546 return (!ret || ret == -EBUSY) ? count : ret; 1547} 1548 1549static DEVICE_ATTR(mode, S_IWUSR | S_IRUGO, show_mode, set_mode); 1550 1551int ipoib_cm_add_mode_attr(struct net_device *dev) 1552{ 1553 return device_create_file(&dev->dev, &dev_attr_mode); 1554} 1555 1556static void ipoib_cm_create_srq(struct net_device *dev, int max_sge) 1557{ 1558 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1559 struct ib_srq_init_attr srq_init_attr = { 1560 .srq_type = IB_SRQT_BASIC, 1561 .attr = { 1562 .max_wr = ipoib_recvq_size, 1563 .max_sge = max_sge 1564 } 1565 }; 1566 1567 priv->cm.srq = ib_create_srq(priv->pd, &srq_init_attr); 1568 if (IS_ERR(priv->cm.srq)) { 1569 if (PTR_ERR(priv->cm.srq) != -EOPNOTSUPP) 1570 pr_warn("%s: failed to allocate SRQ, error %ld\n", 1571 priv->ca->name, PTR_ERR(priv->cm.srq)); 1572 priv->cm.srq = NULL; 1573 return; 1574 } 1575 1576 priv->cm.srq_ring = vzalloc(array_size(ipoib_recvq_size, 1577 sizeof(*priv->cm.srq_ring))); 1578 if (!priv->cm.srq_ring) { 1579 ib_destroy_srq(priv->cm.srq); 1580 priv->cm.srq = NULL; 1581 return; 1582 } 1583 1584} 1585 1586int ipoib_cm_dev_init(struct net_device *dev) 1587{ 1588 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1589 int max_srq_sge, i; 1590 1591 INIT_LIST_HEAD(&priv->cm.passive_ids); 1592 INIT_LIST_HEAD(&priv->cm.reap_list); 1593 INIT_LIST_HEAD(&priv->cm.start_list); 1594 INIT_LIST_HEAD(&priv->cm.rx_error_list); 1595 INIT_LIST_HEAD(&priv->cm.rx_flush_list); 1596 INIT_LIST_HEAD(&priv->cm.rx_drain_list); 1597 INIT_LIST_HEAD(&priv->cm.rx_reap_list); 1598 INIT_WORK(&priv->cm.start_task, ipoib_cm_tx_start); 1599 INIT_WORK(&priv->cm.reap_task, ipoib_cm_tx_reap); 1600 INIT_WORK(&priv->cm.skb_task, ipoib_cm_skb_reap); 1601 INIT_WORK(&priv->cm.rx_reap_task, ipoib_cm_rx_reap); 1602 INIT_DELAYED_WORK(&priv->cm.stale_task, ipoib_cm_stale_task); 1603 1604 skb_queue_head_init(&priv->cm.skb_queue); 1605 1606 ipoib_dbg(priv, "max_srq_sge=%d\n", priv->ca->attrs.max_srq_sge); 1607 1608 max_srq_sge = min_t(int, IPOIB_CM_RX_SG, priv->ca->attrs.max_srq_sge); 1609 ipoib_cm_create_srq(dev, max_srq_sge); 1610 if (ipoib_cm_has_srq(dev)) { 1611 priv->cm.max_cm_mtu = max_srq_sge * PAGE_SIZE - 0x10; 1612 priv->cm.num_frags = max_srq_sge; 1613 ipoib_dbg(priv, "max_cm_mtu = 0x%x, num_frags=%d\n", 1614 priv->cm.max_cm_mtu, priv->cm.num_frags); 1615 } else { 1616 priv->cm.max_cm_mtu = IPOIB_CM_MTU; 1617 priv->cm.num_frags = IPOIB_CM_RX_SG; 1618 } 1619 1620 ipoib_cm_init_rx_wr(dev, &priv->cm.rx_wr, priv->cm.rx_sge); 1621 1622 if (ipoib_cm_has_srq(dev)) { 1623 for (i = 0; i < ipoib_recvq_size; ++i) { 1624 if (!ipoib_cm_alloc_rx_skb(dev, priv->cm.srq_ring, i, 1625 priv->cm.num_frags - 1, 1626 priv->cm.srq_ring[i].mapping, 1627 GFP_KERNEL)) { 1628 ipoib_warn(priv, "failed to allocate " 1629 "receive buffer %d\n", i); 1630 ipoib_cm_dev_cleanup(dev); 1631 return -ENOMEM; 1632 } 1633 1634 if (ipoib_cm_post_receive_srq(dev, i)) { 1635 ipoib_warn(priv, "ipoib_cm_post_receive_srq " 1636 "failed for buf %d\n", i); 1637 ipoib_cm_dev_cleanup(dev); 1638 return -EIO; 1639 } 1640 } 1641 } 1642 1643 priv->dev->dev_addr[0] = IPOIB_FLAGS_RC; 1644 return 0; 1645} 1646 1647void ipoib_cm_dev_cleanup(struct net_device *dev) 1648{ 1649 struct ipoib_dev_priv *priv = ipoib_priv(dev); 1650 1651 if (!priv->cm.srq) 1652 return; 1653 1654 ipoib_dbg(priv, "Cleanup ipoib connected mode.\n"); 1655 1656 ib_destroy_srq(priv->cm.srq); 1657 priv->cm.srq = NULL; 1658 if (!priv->cm.srq_ring) 1659 return; 1660 1661 ipoib_cm_free_rx_ring(dev, priv->cm.srq_ring); 1662 priv->cm.srq_ring = NULL; 1663} 1664