blob: 3d479067d54dbd02810f9c6b1bf2d373d9388cbf [file] [log] [blame]
Andy Groverfcd8b7c2009-02-24 15:30:36 +00001/*
2 * Copyright (c) 2006 Oracle. 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 <linux/kernel.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090034#include <linux/slab.h>
Andy Groverfcd8b7c2009-02-24 15:30:36 +000035#include <linux/pci.h>
36#include <linux/dma-mapping.h>
37#include <rdma/rdma_cm.h>
38
39#include "rds.h"
40#include "iw.h"
41
42static struct kmem_cache *rds_iw_incoming_slab;
43static struct kmem_cache *rds_iw_frag_slab;
44static atomic_t rds_iw_allocation = ATOMIC_INIT(0);
45
46static void rds_iw_frag_drop_page(struct rds_page_frag *frag)
47{
48 rdsdebug("frag %p page %p\n", frag, frag->f_page);
49 __free_page(frag->f_page);
50 frag->f_page = NULL;
51}
52
53static void rds_iw_frag_free(struct rds_page_frag *frag)
54{
55 rdsdebug("frag %p page %p\n", frag, frag->f_page);
56 BUG_ON(frag->f_page != NULL);
57 kmem_cache_free(rds_iw_frag_slab, frag);
58}
59
60/*
61 * We map a page at a time. Its fragments are posted in order. This
62 * is called in fragment order as the fragments get send completion events.
63 * Only the last frag in the page performs the unmapping.
64 *
65 * It's OK for ring cleanup to call this in whatever order it likes because
66 * DMA is not in flight and so we can unmap while other ring entries still
67 * hold page references in their frags.
68 */
69static void rds_iw_recv_unmap_page(struct rds_iw_connection *ic,
70 struct rds_iw_recv_work *recv)
71{
72 struct rds_page_frag *frag = recv->r_frag;
73
74 rdsdebug("recv %p frag %p page %p\n", recv, frag, frag->f_page);
75 if (frag->f_mapped)
76 ib_dma_unmap_page(ic->i_cm_id->device,
77 frag->f_mapped,
78 RDS_FRAG_SIZE, DMA_FROM_DEVICE);
79 frag->f_mapped = 0;
80}
81
82void rds_iw_recv_init_ring(struct rds_iw_connection *ic)
83{
84 struct rds_iw_recv_work *recv;
85 u32 i;
86
87 for (i = 0, recv = ic->i_recvs; i < ic->i_recv_ring.w_nr; i++, recv++) {
88 struct ib_sge *sge;
89
90 recv->r_iwinc = NULL;
91 recv->r_frag = NULL;
92
93 recv->r_wr.next = NULL;
94 recv->r_wr.wr_id = i;
95 recv->r_wr.sg_list = recv->r_sge;
96 recv->r_wr.num_sge = RDS_IW_RECV_SGE;
97
98 sge = rds_iw_data_sge(ic, recv->r_sge);
99 sge->addr = 0;
100 sge->length = RDS_FRAG_SIZE;
101 sge->lkey = 0;
102
103 sge = rds_iw_header_sge(ic, recv->r_sge);
104 sge->addr = ic->i_recv_hdrs_dma + (i * sizeof(struct rds_header));
105 sge->length = sizeof(struct rds_header);
106 sge->lkey = 0;
107 }
108}
109
110static void rds_iw_recv_clear_one(struct rds_iw_connection *ic,
111 struct rds_iw_recv_work *recv)
112{
113 if (recv->r_iwinc) {
114 rds_inc_put(&recv->r_iwinc->ii_inc);
115 recv->r_iwinc = NULL;
116 }
117 if (recv->r_frag) {
118 rds_iw_recv_unmap_page(ic, recv);
119 if (recv->r_frag->f_page)
120 rds_iw_frag_drop_page(recv->r_frag);
121 rds_iw_frag_free(recv->r_frag);
122 recv->r_frag = NULL;
123 }
124}
125
126void rds_iw_recv_clear_ring(struct rds_iw_connection *ic)
127{
128 u32 i;
129
130 for (i = 0; i < ic->i_recv_ring.w_nr; i++)
131 rds_iw_recv_clear_one(ic, &ic->i_recvs[i]);
132
133 if (ic->i_frag.f_page)
134 rds_iw_frag_drop_page(&ic->i_frag);
135}
136
137static int rds_iw_recv_refill_one(struct rds_connection *conn,
138 struct rds_iw_recv_work *recv,
139 gfp_t kptr_gfp, gfp_t page_gfp)
140{
141 struct rds_iw_connection *ic = conn->c_transport_data;
142 dma_addr_t dma_addr;
143 struct ib_sge *sge;
144 int ret = -ENOMEM;
145
146 if (recv->r_iwinc == NULL) {
Andy Grover86357b12009-10-30 08:51:54 +0000147 if (!atomic_add_unless(&rds_iw_allocation, 1, rds_iw_sysctl_max_recv_allocation)) {
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000148 rds_iw_stats_inc(s_iw_rx_alloc_limit);
149 goto out;
150 }
151 recv->r_iwinc = kmem_cache_alloc(rds_iw_incoming_slab,
152 kptr_gfp);
Andy Grover86357b12009-10-30 08:51:54 +0000153 if (recv->r_iwinc == NULL) {
154 atomic_dec(&rds_iw_allocation);
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000155 goto out;
Andy Grover86357b12009-10-30 08:51:54 +0000156 }
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000157 INIT_LIST_HEAD(&recv->r_iwinc->ii_frags);
158 rds_inc_init(&recv->r_iwinc->ii_inc, conn, conn->c_faddr);
159 }
160
161 if (recv->r_frag == NULL) {
162 recv->r_frag = kmem_cache_alloc(rds_iw_frag_slab, kptr_gfp);
163 if (recv->r_frag == NULL)
164 goto out;
165 INIT_LIST_HEAD(&recv->r_frag->f_item);
166 recv->r_frag->f_page = NULL;
167 }
168
169 if (ic->i_frag.f_page == NULL) {
170 ic->i_frag.f_page = alloc_page(page_gfp);
171 if (ic->i_frag.f_page == NULL)
172 goto out;
173 ic->i_frag.f_offset = 0;
174 }
175
176 dma_addr = ib_dma_map_page(ic->i_cm_id->device,
177 ic->i_frag.f_page,
178 ic->i_frag.f_offset,
179 RDS_FRAG_SIZE,
180 DMA_FROM_DEVICE);
181 if (ib_dma_mapping_error(ic->i_cm_id->device, dma_addr))
182 goto out;
183
184 /*
185 * Once we get the RDS_PAGE_LAST_OFF frag then rds_iw_frag_unmap()
186 * must be called on this recv. This happens as completions hit
187 * in order or on connection shutdown.
188 */
189 recv->r_frag->f_page = ic->i_frag.f_page;
190 recv->r_frag->f_offset = ic->i_frag.f_offset;
191 recv->r_frag->f_mapped = dma_addr;
192
193 sge = rds_iw_data_sge(ic, recv->r_sge);
194 sge->addr = dma_addr;
195 sge->length = RDS_FRAG_SIZE;
196
197 sge = rds_iw_header_sge(ic, recv->r_sge);
198 sge->addr = ic->i_recv_hdrs_dma + (recv - ic->i_recvs) * sizeof(struct rds_header);
199 sge->length = sizeof(struct rds_header);
200
201 get_page(recv->r_frag->f_page);
202
203 if (ic->i_frag.f_offset < RDS_PAGE_LAST_OFF) {
204 ic->i_frag.f_offset += RDS_FRAG_SIZE;
205 } else {
206 put_page(ic->i_frag.f_page);
207 ic->i_frag.f_page = NULL;
208 ic->i_frag.f_offset = 0;
209 }
210
211 ret = 0;
212out:
213 return ret;
214}
215
216/*
217 * This tries to allocate and post unused work requests after making sure that
218 * they have all the allocations they need to queue received fragments into
219 * sockets. The i_recv_mutex is held here so that ring_alloc and _unalloc
220 * pairs don't go unmatched.
221 *
222 * -1 is returned if posting fails due to temporary resource exhaustion.
223 */
224int rds_iw_recv_refill(struct rds_connection *conn, gfp_t kptr_gfp,
225 gfp_t page_gfp, int prefill)
226{
227 struct rds_iw_connection *ic = conn->c_transport_data;
228 struct rds_iw_recv_work *recv;
229 struct ib_recv_wr *failed_wr;
230 unsigned int posted = 0;
231 int ret = 0;
232 u32 pos;
233
Joe Perchesf64f9e72009-11-29 16:55:45 -0800234 while ((prefill || rds_conn_up(conn)) &&
235 rds_iw_ring_alloc(&ic->i_recv_ring, 1, &pos)) {
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000236 if (pos >= ic->i_recv_ring.w_nr) {
237 printk(KERN_NOTICE "Argh - ring alloc returned pos=%u\n",
238 pos);
239 ret = -EINVAL;
240 break;
241 }
242
243 recv = &ic->i_recvs[pos];
244 ret = rds_iw_recv_refill_one(conn, recv, kptr_gfp, page_gfp);
245 if (ret) {
246 ret = -1;
247 break;
248 }
249
250 /* XXX when can this fail? */
251 ret = ib_post_recv(ic->i_cm_id->qp, &recv->r_wr, &failed_wr);
252 rdsdebug("recv %p iwinc %p page %p addr %lu ret %d\n", recv,
253 recv->r_iwinc, recv->r_frag->f_page,
254 (long) recv->r_frag->f_mapped, ret);
255 if (ret) {
256 rds_iw_conn_error(conn, "recv post on "
257 "%pI4 returned %d, disconnecting and "
258 "reconnecting\n", &conn->c_faddr,
259 ret);
260 ret = -1;
261 break;
262 }
263
264 posted++;
265 }
266
267 /* We're doing flow control - update the window. */
268 if (ic->i_flowctl && posted)
269 rds_iw_advertise_credits(conn, posted);
270
271 if (ret)
272 rds_iw_ring_unalloc(&ic->i_recv_ring, 1);
273 return ret;
274}
275
276void rds_iw_inc_purge(struct rds_incoming *inc)
277{
278 struct rds_iw_incoming *iwinc;
279 struct rds_page_frag *frag;
280 struct rds_page_frag *pos;
281
282 iwinc = container_of(inc, struct rds_iw_incoming, ii_inc);
283 rdsdebug("purging iwinc %p inc %p\n", iwinc, inc);
284
285 list_for_each_entry_safe(frag, pos, &iwinc->ii_frags, f_item) {
286 list_del_init(&frag->f_item);
287 rds_iw_frag_drop_page(frag);
288 rds_iw_frag_free(frag);
289 }
290}
291
292void rds_iw_inc_free(struct rds_incoming *inc)
293{
294 struct rds_iw_incoming *iwinc;
295
296 iwinc = container_of(inc, struct rds_iw_incoming, ii_inc);
297
298 rds_iw_inc_purge(inc);
299 rdsdebug("freeing iwinc %p inc %p\n", iwinc, inc);
300 BUG_ON(!list_empty(&iwinc->ii_frags));
301 kmem_cache_free(rds_iw_incoming_slab, iwinc);
302 atomic_dec(&rds_iw_allocation);
303 BUG_ON(atomic_read(&rds_iw_allocation) < 0);
304}
305
306int rds_iw_inc_copy_to_user(struct rds_incoming *inc, struct iovec *first_iov,
307 size_t size)
308{
309 struct rds_iw_incoming *iwinc;
310 struct rds_page_frag *frag;
311 struct iovec *iov = first_iov;
312 unsigned long to_copy;
313 unsigned long frag_off = 0;
314 unsigned long iov_off = 0;
315 int copied = 0;
316 int ret;
317 u32 len;
318
319 iwinc = container_of(inc, struct rds_iw_incoming, ii_inc);
320 frag = list_entry(iwinc->ii_frags.next, struct rds_page_frag, f_item);
321 len = be32_to_cpu(inc->i_hdr.h_len);
322
323 while (copied < size && copied < len) {
324 if (frag_off == RDS_FRAG_SIZE) {
325 frag = list_entry(frag->f_item.next,
326 struct rds_page_frag, f_item);
327 frag_off = 0;
328 }
329 while (iov_off == iov->iov_len) {
330 iov_off = 0;
331 iov++;
332 }
333
334 to_copy = min(iov->iov_len - iov_off, RDS_FRAG_SIZE - frag_off);
335 to_copy = min_t(size_t, to_copy, size - copied);
336 to_copy = min_t(unsigned long, to_copy, len - copied);
337
338 rdsdebug("%lu bytes to user [%p, %zu] + %lu from frag "
339 "[%p, %lu] + %lu\n",
340 to_copy, iov->iov_base, iov->iov_len, iov_off,
341 frag->f_page, frag->f_offset, frag_off);
342
343 /* XXX needs + offset for multiple recvs per page */
344 ret = rds_page_copy_to_user(frag->f_page,
345 frag->f_offset + frag_off,
346 iov->iov_base + iov_off,
347 to_copy);
348 if (ret) {
349 copied = ret;
350 break;
351 }
352
353 iov_off += to_copy;
354 frag_off += to_copy;
355 copied += to_copy;
356 }
357
358 return copied;
359}
360
361/* ic starts out kzalloc()ed */
362void rds_iw_recv_init_ack(struct rds_iw_connection *ic)
363{
364 struct ib_send_wr *wr = &ic->i_ack_wr;
365 struct ib_sge *sge = &ic->i_ack_sge;
366
367 sge->addr = ic->i_ack_dma;
368 sge->length = sizeof(struct rds_header);
369 sge->lkey = rds_iw_local_dma_lkey(ic);
370
371 wr->sg_list = sge;
372 wr->num_sge = 1;
373 wr->opcode = IB_WR_SEND;
374 wr->wr_id = RDS_IW_ACK_WR_ID;
375 wr->send_flags = IB_SEND_SIGNALED | IB_SEND_SOLICITED;
376}
377
378/*
379 * You'd think that with reliable IB connections you wouldn't need to ack
380 * messages that have been received. The problem is that IB hardware generates
381 * an ack message before it has DMAed the message into memory. This creates a
382 * potential message loss if the HCA is disabled for any reason between when it
383 * sends the ack and before the message is DMAed and processed. This is only a
384 * potential issue if another HCA is available for fail-over.
385 *
386 * When the remote host receives our ack they'll free the sent message from
387 * their send queue. To decrease the latency of this we always send an ack
388 * immediately after we've received messages.
389 *
390 * For simplicity, we only have one ack in flight at a time. This puts
391 * pressure on senders to have deep enough send queues to absorb the latency of
392 * a single ack frame being in flight. This might not be good enough.
393 *
394 * This is implemented by have a long-lived send_wr and sge which point to a
395 * statically allocated ack frame. This ack wr does not fall under the ring
396 * accounting that the tx and rx wrs do. The QP attribute specifically makes
397 * room for it beyond the ring size. Send completion notices its special
398 * wr_id and avoids working with the ring in that case.
399 */
Andy Grover8cbd9602009-04-01 08:20:20 +0000400#ifndef KERNEL_HAS_ATOMIC64
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000401static void rds_iw_set_ack(struct rds_iw_connection *ic, u64 seq,
402 int ack_required)
403{
Andy Grover8cbd9602009-04-01 08:20:20 +0000404 unsigned long flags;
405
406 spin_lock_irqsave(&ic->i_ack_lock, flags);
407 ic->i_ack_next = seq;
408 if (ack_required)
409 set_bit(IB_ACK_REQUESTED, &ic->i_ack_flags);
410 spin_unlock_irqrestore(&ic->i_ack_lock, flags);
411}
412
413static u64 rds_iw_get_ack(struct rds_iw_connection *ic)
414{
415 unsigned long flags;
416 u64 seq;
417
418 clear_bit(IB_ACK_REQUESTED, &ic->i_ack_flags);
419
420 spin_lock_irqsave(&ic->i_ack_lock, flags);
421 seq = ic->i_ack_next;
422 spin_unlock_irqrestore(&ic->i_ack_lock, flags);
423
424 return seq;
425}
426#else
427static void rds_iw_set_ack(struct rds_iw_connection *ic, u64 seq,
428 int ack_required)
429{
430 atomic64_set(&ic->i_ack_next, seq);
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000431 if (ack_required) {
432 smp_mb__before_clear_bit();
433 set_bit(IB_ACK_REQUESTED, &ic->i_ack_flags);
434 }
435}
436
437static u64 rds_iw_get_ack(struct rds_iw_connection *ic)
438{
439 clear_bit(IB_ACK_REQUESTED, &ic->i_ack_flags);
440 smp_mb__after_clear_bit();
441
Andy Grover8cbd9602009-04-01 08:20:20 +0000442 return atomic64_read(&ic->i_ack_next);
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000443}
Andy Grover8cbd9602009-04-01 08:20:20 +0000444#endif
445
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000446
447static void rds_iw_send_ack(struct rds_iw_connection *ic, unsigned int adv_credits)
448{
449 struct rds_header *hdr = ic->i_ack;
450 struct ib_send_wr *failed_wr;
451 u64 seq;
452 int ret;
453
454 seq = rds_iw_get_ack(ic);
455
456 rdsdebug("send_ack: ic %p ack %llu\n", ic, (unsigned long long) seq);
457 rds_message_populate_header(hdr, 0, 0, 0);
458 hdr->h_ack = cpu_to_be64(seq);
459 hdr->h_credit = adv_credits;
460 rds_message_make_checksum(hdr);
461 ic->i_ack_queued = jiffies;
462
463 ret = ib_post_send(ic->i_cm_id->qp, &ic->i_ack_wr, &failed_wr);
464 if (unlikely(ret)) {
465 /* Failed to send. Release the WR, and
466 * force another ACK.
467 */
468 clear_bit(IB_ACK_IN_FLIGHT, &ic->i_ack_flags);
469 set_bit(IB_ACK_REQUESTED, &ic->i_ack_flags);
470
471 rds_iw_stats_inc(s_iw_ack_send_failure);
Andy Grover735f61e2010-03-11 13:49:55 +0000472
473 rds_iw_conn_error(ic->conn, "sending ack failed\n");
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000474 } else
475 rds_iw_stats_inc(s_iw_ack_sent);
476}
477
478/*
479 * There are 3 ways of getting acknowledgements to the peer:
480 * 1. We call rds_iw_attempt_ack from the recv completion handler
481 * to send an ACK-only frame.
482 * However, there can be only one such frame in the send queue
483 * at any time, so we may have to postpone it.
484 * 2. When another (data) packet is transmitted while there's
485 * an ACK in the queue, we piggyback the ACK sequence number
486 * on the data packet.
487 * 3. If the ACK WR is done sending, we get called from the
488 * send queue completion handler, and check whether there's
489 * another ACK pending (postponed because the WR was on the
490 * queue). If so, we transmit it.
491 *
492 * We maintain 2 variables:
493 * - i_ack_flags, which keeps track of whether the ACK WR
494 * is currently in the send queue or not (IB_ACK_IN_FLIGHT)
495 * - i_ack_next, which is the last sequence number we received
496 *
497 * Potentially, send queue and receive queue handlers can run concurrently.
Andy Grover8cbd9602009-04-01 08:20:20 +0000498 * It would be nice to not have to use a spinlock to synchronize things,
499 * but the one problem that rules this out is that 64bit updates are
500 * not atomic on all platforms. Things would be a lot simpler if
501 * we had atomic64 or maybe cmpxchg64 everywhere.
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000502 *
503 * Reconnecting complicates this picture just slightly. When we
504 * reconnect, we may be seeing duplicate packets. The peer
505 * is retransmitting them, because it hasn't seen an ACK for
506 * them. It is important that we ACK these.
507 *
508 * ACK mitigation adds a header flag "ACK_REQUIRED"; any packet with
509 * this flag set *MUST* be acknowledged immediately.
510 */
511
512/*
513 * When we get here, we're called from the recv queue handler.
514 * Check whether we ought to transmit an ACK.
515 */
516void rds_iw_attempt_ack(struct rds_iw_connection *ic)
517{
518 unsigned int adv_credits;
519
520 if (!test_bit(IB_ACK_REQUESTED, &ic->i_ack_flags))
521 return;
522
523 if (test_and_set_bit(IB_ACK_IN_FLIGHT, &ic->i_ack_flags)) {
524 rds_iw_stats_inc(s_iw_ack_send_delayed);
525 return;
526 }
527
528 /* Can we get a send credit? */
Steve Wise7b70d032009-04-09 14:09:39 +0000529 if (!rds_iw_send_grab_credits(ic, 1, &adv_credits, 0, RDS_MAX_ADV_CREDIT)) {
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000530 rds_iw_stats_inc(s_iw_tx_throttle);
531 clear_bit(IB_ACK_IN_FLIGHT, &ic->i_ack_flags);
532 return;
533 }
534
535 clear_bit(IB_ACK_REQUESTED, &ic->i_ack_flags);
536 rds_iw_send_ack(ic, adv_credits);
537}
538
539/*
540 * We get here from the send completion handler, when the
541 * adapter tells us the ACK frame was sent.
542 */
543void rds_iw_ack_send_complete(struct rds_iw_connection *ic)
544{
545 clear_bit(IB_ACK_IN_FLIGHT, &ic->i_ack_flags);
546 rds_iw_attempt_ack(ic);
547}
548
549/*
550 * This is called by the regular xmit code when it wants to piggyback
551 * an ACK on an outgoing frame.
552 */
553u64 rds_iw_piggyb_ack(struct rds_iw_connection *ic)
554{
555 if (test_and_clear_bit(IB_ACK_REQUESTED, &ic->i_ack_flags))
556 rds_iw_stats_inc(s_iw_ack_send_piggybacked);
557 return rds_iw_get_ack(ic);
558}
559
560/*
561 * It's kind of lame that we're copying from the posted receive pages into
562 * long-lived bitmaps. We could have posted the bitmaps and rdma written into
563 * them. But receiving new congestion bitmaps should be a *rare* event, so
564 * hopefully we won't need to invest that complexity in making it more
565 * efficient. By copying we can share a simpler core with TCP which has to
566 * copy.
567 */
568static void rds_iw_cong_recv(struct rds_connection *conn,
569 struct rds_iw_incoming *iwinc)
570{
571 struct rds_cong_map *map;
572 unsigned int map_off;
573 unsigned int map_page;
574 struct rds_page_frag *frag;
575 unsigned long frag_off;
576 unsigned long to_copy;
577 unsigned long copied;
578 uint64_t uncongested = 0;
579 void *addr;
580
581 /* catch completely corrupt packets */
582 if (be32_to_cpu(iwinc->ii_inc.i_hdr.h_len) != RDS_CONG_MAP_BYTES)
583 return;
584
585 map = conn->c_fcong;
586 map_page = 0;
587 map_off = 0;
588
589 frag = list_entry(iwinc->ii_frags.next, struct rds_page_frag, f_item);
590 frag_off = 0;
591
592 copied = 0;
593
594 while (copied < RDS_CONG_MAP_BYTES) {
595 uint64_t *src, *dst;
596 unsigned int k;
597
598 to_copy = min(RDS_FRAG_SIZE - frag_off, PAGE_SIZE - map_off);
599 BUG_ON(to_copy & 7); /* Must be 64bit aligned. */
600
601 addr = kmap_atomic(frag->f_page, KM_SOFTIRQ0);
602
603 src = addr + frag_off;
604 dst = (void *)map->m_page_addrs[map_page] + map_off;
605 for (k = 0; k < to_copy; k += 8) {
606 /* Record ports that became uncongested, ie
607 * bits that changed from 0 to 1. */
608 uncongested |= ~(*src) & *dst;
609 *dst++ = *src++;
610 }
611 kunmap_atomic(addr, KM_SOFTIRQ0);
612
613 copied += to_copy;
614
615 map_off += to_copy;
616 if (map_off == PAGE_SIZE) {
617 map_off = 0;
618 map_page++;
619 }
620
621 frag_off += to_copy;
622 if (frag_off == RDS_FRAG_SIZE) {
623 frag = list_entry(frag->f_item.next,
624 struct rds_page_frag, f_item);
625 frag_off = 0;
626 }
627 }
628
629 /* the congestion map is in little endian order */
630 uncongested = le64_to_cpu(uncongested);
631
632 rds_cong_map_updated(map, uncongested);
633}
634
635/*
636 * Rings are posted with all the allocations they'll need to queue the
637 * incoming message to the receiving socket so this can't fail.
638 * All fragments start with a header, so we can make sure we're not receiving
639 * garbage, and we can tell a small 8 byte fragment from an ACK frame.
640 */
641struct rds_iw_ack_state {
642 u64 ack_next;
643 u64 ack_recv;
644 unsigned int ack_required:1;
645 unsigned int ack_next_valid:1;
646 unsigned int ack_recv_valid:1;
647};
648
649static void rds_iw_process_recv(struct rds_connection *conn,
650 struct rds_iw_recv_work *recv, u32 byte_len,
651 struct rds_iw_ack_state *state)
652{
653 struct rds_iw_connection *ic = conn->c_transport_data;
654 struct rds_iw_incoming *iwinc = ic->i_iwinc;
655 struct rds_header *ihdr, *hdr;
656
657 /* XXX shut down the connection if port 0,0 are seen? */
658
659 rdsdebug("ic %p iwinc %p recv %p byte len %u\n", ic, iwinc, recv,
660 byte_len);
661
662 if (byte_len < sizeof(struct rds_header)) {
663 rds_iw_conn_error(conn, "incoming message "
664 "from %pI4 didn't inclue a "
665 "header, disconnecting and "
666 "reconnecting\n",
667 &conn->c_faddr);
668 return;
669 }
670 byte_len -= sizeof(struct rds_header);
671
672 ihdr = &ic->i_recv_hdrs[recv - ic->i_recvs];
673
674 /* Validate the checksum. */
675 if (!rds_message_verify_checksum(ihdr)) {
676 rds_iw_conn_error(conn, "incoming message "
677 "from %pI4 has corrupted header - "
678 "forcing a reconnect\n",
679 &conn->c_faddr);
680 rds_stats_inc(s_recv_drop_bad_checksum);
681 return;
682 }
683
684 /* Process the ACK sequence which comes with every packet */
685 state->ack_recv = be64_to_cpu(ihdr->h_ack);
686 state->ack_recv_valid = 1;
687
688 /* Process the credits update if there was one */
689 if (ihdr->h_credit)
690 rds_iw_send_add_credits(conn, ihdr->h_credit);
691
692 if (ihdr->h_sport == 0 && ihdr->h_dport == 0 && byte_len == 0) {
693 /* This is an ACK-only packet. The fact that it gets
694 * special treatment here is that historically, ACKs
695 * were rather special beasts.
696 */
697 rds_iw_stats_inc(s_iw_ack_received);
698
699 /*
700 * Usually the frags make their way on to incs and are then freed as
701 * the inc is freed. We don't go that route, so we have to drop the
702 * page ref ourselves. We can't just leave the page on the recv
703 * because that confuses the dma mapping of pages and each recv's use
704 * of a partial page. We can leave the frag, though, it will be
705 * reused.
706 *
707 * FIXME: Fold this into the code path below.
708 */
709 rds_iw_frag_drop_page(recv->r_frag);
710 return;
711 }
712
713 /*
714 * If we don't already have an inc on the connection then this
715 * fragment has a header and starts a message.. copy its header
716 * into the inc and save the inc so we can hang upcoming fragments
717 * off its list.
718 */
719 if (iwinc == NULL) {
720 iwinc = recv->r_iwinc;
721 recv->r_iwinc = NULL;
722 ic->i_iwinc = iwinc;
723
724 hdr = &iwinc->ii_inc.i_hdr;
725 memcpy(hdr, ihdr, sizeof(*hdr));
726 ic->i_recv_data_rem = be32_to_cpu(hdr->h_len);
727
728 rdsdebug("ic %p iwinc %p rem %u flag 0x%x\n", ic, iwinc,
729 ic->i_recv_data_rem, hdr->h_flags);
730 } else {
731 hdr = &iwinc->ii_inc.i_hdr;
732 /* We can't just use memcmp here; fragments of a
733 * single message may carry different ACKs */
Joe Perchesf64f9e72009-11-29 16:55:45 -0800734 if (hdr->h_sequence != ihdr->h_sequence ||
735 hdr->h_len != ihdr->h_len ||
736 hdr->h_sport != ihdr->h_sport ||
737 hdr->h_dport != ihdr->h_dport) {
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000738 rds_iw_conn_error(conn,
739 "fragment header mismatch; forcing reconnect\n");
740 return;
741 }
742 }
743
744 list_add_tail(&recv->r_frag->f_item, &iwinc->ii_frags);
745 recv->r_frag = NULL;
746
747 if (ic->i_recv_data_rem > RDS_FRAG_SIZE)
748 ic->i_recv_data_rem -= RDS_FRAG_SIZE;
749 else {
750 ic->i_recv_data_rem = 0;
751 ic->i_iwinc = NULL;
752
753 if (iwinc->ii_inc.i_hdr.h_flags == RDS_FLAG_CONG_BITMAP)
754 rds_iw_cong_recv(conn, iwinc);
755 else {
756 rds_recv_incoming(conn, conn->c_faddr, conn->c_laddr,
757 &iwinc->ii_inc, GFP_ATOMIC,
758 KM_SOFTIRQ0);
759 state->ack_next = be64_to_cpu(hdr->h_sequence);
760 state->ack_next_valid = 1;
761 }
762
763 /* Evaluate the ACK_REQUIRED flag *after* we received
764 * the complete frame, and after bumping the next_rx
765 * sequence. */
766 if (hdr->h_flags & RDS_FLAG_ACK_REQUIRED) {
767 rds_stats_inc(s_recv_ack_required);
768 state->ack_required = 1;
769 }
770
771 rds_inc_put(&iwinc->ii_inc);
772 }
773}
774
775/*
776 * Plucking the oldest entry from the ring can be done concurrently with
777 * the thread refilling the ring. Each ring operation is protected by
778 * spinlocks and the transient state of refilling doesn't change the
779 * recording of which entry is oldest.
780 *
781 * This relies on IB only calling one cq comp_handler for each cq so that
782 * there will only be one caller of rds_recv_incoming() per RDS connection.
783 */
784void rds_iw_recv_cq_comp_handler(struct ib_cq *cq, void *context)
785{
786 struct rds_connection *conn = context;
787 struct rds_iw_connection *ic = conn->c_transport_data;
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000788
789 rdsdebug("conn %p cq %p\n", conn, cq);
790
791 rds_iw_stats_inc(s_iw_rx_cq_call);
792
Andy Groverd521b632009-10-30 08:51:57 +0000793 tasklet_schedule(&ic->i_recv_tasklet);
794}
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000795
Andy Groverd521b632009-10-30 08:51:57 +0000796static inline void rds_poll_cq(struct rds_iw_connection *ic,
797 struct rds_iw_ack_state *state)
798{
799 struct rds_connection *conn = ic->conn;
800 struct ib_wc wc;
801 struct rds_iw_recv_work *recv;
802
803 while (ib_poll_cq(ic->i_recv_cq, 1, &wc) > 0) {
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000804 rdsdebug("wc wr_id 0x%llx status %u byte_len %u imm_data %u\n",
805 (unsigned long long)wc.wr_id, wc.status, wc.byte_len,
806 be32_to_cpu(wc.ex.imm_data));
807 rds_iw_stats_inc(s_iw_rx_cq_event);
808
809 recv = &ic->i_recvs[rds_iw_ring_oldest(&ic->i_recv_ring)];
810
811 rds_iw_recv_unmap_page(ic, recv);
812
813 /*
814 * Also process recvs in connecting state because it is possible
815 * to get a recv completion _before_ the rdmacm ESTABLISHED
816 * event is processed.
817 */
818 if (rds_conn_up(conn) || rds_conn_connecting(conn)) {
819 /* We expect errors as the qp is drained during shutdown */
820 if (wc.status == IB_WC_SUCCESS) {
Andy Groverd521b632009-10-30 08:51:57 +0000821 rds_iw_process_recv(conn, recv, wc.byte_len, state);
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000822 } else {
823 rds_iw_conn_error(conn, "recv completion on "
824 "%pI4 had status %u, disconnecting and "
825 "reconnecting\n", &conn->c_faddr,
826 wc.status);
827 }
828 }
829
830 rds_iw_ring_free(&ic->i_recv_ring, 1);
831 }
Andy Groverd521b632009-10-30 08:51:57 +0000832}
833
834void rds_iw_recv_tasklet_fn(unsigned long data)
835{
836 struct rds_iw_connection *ic = (struct rds_iw_connection *) data;
837 struct rds_connection *conn = ic->conn;
838 struct rds_iw_ack_state state = { 0, };
839
840 rds_poll_cq(ic, &state);
841 ib_req_notify_cq(ic->i_recv_cq, IB_CQ_SOLICITED);
842 rds_poll_cq(ic, &state);
Andy Groverfcd8b7c2009-02-24 15:30:36 +0000843
844 if (state.ack_next_valid)
845 rds_iw_set_ack(ic, state.ack_next, state.ack_required);
846 if (state.ack_recv_valid && state.ack_recv > ic->i_ack_recv) {
847 rds_send_drop_acked(conn, state.ack_recv, NULL);
848 ic->i_ack_recv = state.ack_recv;
849 }
850 if (rds_conn_up(conn))
851 rds_iw_attempt_ack(ic);
852
853 /* If we ever end up with a really empty receive ring, we're
854 * in deep trouble, as the sender will definitely see RNR
855 * timeouts. */
856 if (rds_iw_ring_empty(&ic->i_recv_ring))
857 rds_iw_stats_inc(s_iw_rx_ring_empty);
858
859 /*
860 * If the ring is running low, then schedule the thread to refill.
861 */
862 if (rds_iw_ring_low(&ic->i_recv_ring))
863 queue_delayed_work(rds_wq, &conn->c_recv_w, 0);
864}
865
866int rds_iw_recv(struct rds_connection *conn)
867{
868 struct rds_iw_connection *ic = conn->c_transport_data;
869 int ret = 0;
870
871 rdsdebug("conn %p\n", conn);
872
873 /*
874 * If we get a temporary posting failure in this context then
875 * we're really low and we want the caller to back off for a bit.
876 */
877 mutex_lock(&ic->i_recv_mutex);
878 if (rds_iw_recv_refill(conn, GFP_KERNEL, GFP_HIGHUSER, 0))
879 ret = -ENOMEM;
880 else
881 rds_iw_stats_inc(s_iw_rx_refill_from_thread);
882 mutex_unlock(&ic->i_recv_mutex);
883
884 if (rds_conn_up(conn))
885 rds_iw_attempt_ack(ic);
886
887 return ret;
888}
889
890int __init rds_iw_recv_init(void)
891{
892 struct sysinfo si;
893 int ret = -ENOMEM;
894
895 /* Default to 30% of all available RAM for recv memory */
896 si_meminfo(&si);
897 rds_iw_sysctl_max_recv_allocation = si.totalram / 3 * PAGE_SIZE / RDS_FRAG_SIZE;
898
899 rds_iw_incoming_slab = kmem_cache_create("rds_iw_incoming",
900 sizeof(struct rds_iw_incoming),
901 0, 0, NULL);
902 if (rds_iw_incoming_slab == NULL)
903 goto out;
904
905 rds_iw_frag_slab = kmem_cache_create("rds_iw_frag",
906 sizeof(struct rds_page_frag),
907 0, 0, NULL);
908 if (rds_iw_frag_slab == NULL)
909 kmem_cache_destroy(rds_iw_incoming_slab);
910 else
911 ret = 0;
912out:
913 return ret;
914}
915
916void rds_iw_recv_exit(void)
917{
918 kmem_cache_destroy(rds_iw_incoming_slab);
919 kmem_cache_destroy(rds_iw_frag_slab);
920}