]>
Commit | Line | Data |
---|---|---|
5c115590 AG |
1 | /* |
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> | |
d9b93842 | 34 | #include <linux/moduleparam.h> |
5a0e3ad6 | 35 | #include <linux/gfp.h> |
5c115590 AG |
36 | #include <net/sock.h> |
37 | #include <linux/in.h> | |
38 | #include <linux/list.h> | |
cb0a6056 | 39 | #include <linux/ratelimit.h> |
bc3b2d7f | 40 | #include <linux/export.h> |
4bebdd7a | 41 | #include <linux/sizes.h> |
5c115590 AG |
42 | |
43 | #include "rds.h" | |
5c115590 AG |
44 | |
45 | /* When transmitting messages in rds_send_xmit, we need to emerge from | |
46 | * time to time and briefly release the CPU. Otherwise the softlock watchdog | |
47 | * will kick our shin. | |
48 | * Also, it seems fairer to not let one busy connection stall all the | |
49 | * others. | |
50 | * | |
51 | * send_batch_count is the number of times we'll loop in send_xmit. Setting | |
52 | * it to 0 will restore the old behavior (where we looped until we had | |
53 | * drained the queue). | |
54 | */ | |
4bebdd7a | 55 | static int send_batch_count = SZ_1K; |
5c115590 AG |
56 | module_param(send_batch_count, int, 0444); |
57 | MODULE_PARM_DESC(send_batch_count, " batch factor when working the send queue"); | |
58 | ||
ff51bf84 | 59 | static void rds_send_remove_from_sock(struct list_head *messages, int status); |
60 | ||
5c115590 | 61 | /* |
0f4b1c7e ZB |
62 | * Reset the send state. Callers must ensure that this doesn't race with |
63 | * rds_send_xmit(). | |
5c115590 | 64 | */ |
d769ef81 | 65 | void rds_send_path_reset(struct rds_conn_path *cp) |
5c115590 AG |
66 | { |
67 | struct rds_message *rm, *tmp; | |
68 | unsigned long flags; | |
69 | ||
4e9b551c SV |
70 | if (cp->cp_xmit_rm) { |
71 | rm = cp->cp_xmit_rm; | |
72 | cp->cp_xmit_rm = NULL; | |
5c115590 AG |
73 | /* Tell the user the RDMA op is no longer mapped by the |
74 | * transport. This isn't entirely true (it's flushed out | |
75 | * independently) but as the connection is down, there's | |
76 | * no ongoing RDMA to/from that memory */ | |
7e3f2952 | 77 | rds_message_unmapped(rm); |
7e3f2952 | 78 | rds_message_put(rm); |
5c115590 | 79 | } |
7e3f2952 | 80 | |
4e9b551c SV |
81 | cp->cp_xmit_sg = 0; |
82 | cp->cp_xmit_hdr_off = 0; | |
83 | cp->cp_xmit_data_off = 0; | |
84 | cp->cp_xmit_atomic_sent = 0; | |
85 | cp->cp_xmit_rdma_sent = 0; | |
86 | cp->cp_xmit_data_sent = 0; | |
5c115590 | 87 | |
4e9b551c | 88 | cp->cp_conn->c_map_queued = 0; |
5c115590 | 89 | |
4e9b551c SV |
90 | cp->cp_unacked_packets = rds_sysctl_max_unacked_packets; |
91 | cp->cp_unacked_bytes = rds_sysctl_max_unacked_bytes; | |
5c115590 AG |
92 | |
93 | /* Mark messages as retransmissions, and move them to the send q */ | |
4e9b551c SV |
94 | spin_lock_irqsave(&cp->cp_lock, flags); |
95 | list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) { | |
5c115590 AG |
96 | set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags); |
97 | set_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags); | |
98 | } | |
4e9b551c SV |
99 | list_splice_init(&cp->cp_retrans, &cp->cp_send_queue); |
100 | spin_unlock_irqrestore(&cp->cp_lock, flags); | |
101 | } | |
d769ef81 | 102 | EXPORT_SYMBOL_GPL(rds_send_path_reset); |
5c115590 | 103 | |
1f9ecd7e | 104 | static int acquire_in_xmit(struct rds_conn_path *cp) |
0f4b1c7e | 105 | { |
1f9ecd7e | 106 | return test_and_set_bit(RDS_IN_XMIT, &cp->cp_flags) == 0; |
0f4b1c7e ZB |
107 | } |
108 | ||
1f9ecd7e | 109 | static void release_in_xmit(struct rds_conn_path *cp) |
0f4b1c7e | 110 | { |
1f9ecd7e | 111 | clear_bit(RDS_IN_XMIT, &cp->cp_flags); |
4e857c58 | 112 | smp_mb__after_atomic(); |
0f4b1c7e ZB |
113 | /* |
114 | * We don't use wait_on_bit()/wake_up_bit() because our waking is in a | |
115 | * hot path and finding waiters is very rare. We don't want to walk | |
116 | * the system-wide hashed waitqueue buckets in the fast path only to | |
117 | * almost never find waiters. | |
118 | */ | |
1f9ecd7e SV |
119 | if (waitqueue_active(&cp->cp_waitq)) |
120 | wake_up_all(&cp->cp_waitq); | |
0f4b1c7e ZB |
121 | } |
122 | ||
5c115590 | 123 | /* |
25985edc | 124 | * We're making the conscious trade-off here to only send one message |
5c115590 AG |
125 | * down the connection at a time. |
126 | * Pro: | |
127 | * - tx queueing is a simple fifo list | |
128 | * - reassembly is optional and easily done by transports per conn | |
129 | * - no per flow rx lookup at all, straight to the socket | |
130 | * - less per-frag memory and wire overhead | |
131 | * Con: | |
132 | * - queued acks can be delayed behind large messages | |
133 | * Depends: | |
134 | * - small message latency is higher behind queued large messages | |
135 | * - large message latency isn't starved by intervening small sends | |
136 | */ | |
1f9ecd7e | 137 | int rds_send_xmit(struct rds_conn_path *cp) |
5c115590 | 138 | { |
1f9ecd7e | 139 | struct rds_connection *conn = cp->cp_conn; |
5c115590 AG |
140 | struct rds_message *rm; |
141 | unsigned long flags; | |
142 | unsigned int tmp; | |
5c115590 AG |
143 | struct scatterlist *sg; |
144 | int ret = 0; | |
5c115590 | 145 | LIST_HEAD(to_be_dropped); |
443be0e5 SV |
146 | int batch_count; |
147 | unsigned long send_gen = 0; | |
5c115590 | 148 | |
fcc5450c | 149 | restart: |
443be0e5 | 150 | batch_count = 0; |
049ee3f5 | 151 | |
5c115590 AG |
152 | /* |
153 | * sendmsg calls here after having queued its message on the send | |
154 | * queue. We only have one task feeding the connection at a time. If | |
155 | * another thread is already feeding the queue then we back off. This | |
156 | * avoids blocking the caller and trading per-connection data between | |
157 | * caches per message. | |
5c115590 | 158 | */ |
1f9ecd7e | 159 | if (!acquire_in_xmit(cp)) { |
049ee3f5 | 160 | rds_stats_inc(s_send_lock_contention); |
5c115590 AG |
161 | ret = -ENOMEM; |
162 | goto out; | |
163 | } | |
0f4b1c7e | 164 | |
443be0e5 SV |
165 | /* |
166 | * we record the send generation after doing the xmit acquire. | |
167 | * if someone else manages to jump in and do some work, we'll use | |
168 | * this to avoid a goto restart farther down. | |
169 | * | |
170 | * The acquire_in_xmit() check above ensures that only one | |
171 | * caller can increment c_send_gen at any time. | |
172 | */ | |
e623a48e HB |
173 | send_gen = READ_ONCE(cp->cp_send_gen) + 1; |
174 | WRITE_ONCE(cp->cp_send_gen, send_gen); | |
443be0e5 | 175 | |
0f4b1c7e ZB |
176 | /* |
177 | * rds_conn_shutdown() sets the conn state and then tests RDS_IN_XMIT, | |
178 | * we do the opposite to avoid races. | |
179 | */ | |
1f9ecd7e SV |
180 | if (!rds_conn_path_up(cp)) { |
181 | release_in_xmit(cp); | |
0f4b1c7e ZB |
182 | ret = 0; |
183 | goto out; | |
184 | } | |
5c115590 | 185 | |
226f7a7d SV |
186 | if (conn->c_trans->xmit_path_prepare) |
187 | conn->c_trans->xmit_path_prepare(cp); | |
5c115590 AG |
188 | |
189 | /* | |
190 | * spin trying to push headers and data down the connection until | |
5b2366bd | 191 | * the connection doesn't make forward progress. |
5c115590 | 192 | */ |
fcc5450c | 193 | while (1) { |
5c115590 | 194 | |
1f9ecd7e | 195 | rm = cp->cp_xmit_rm; |
5c115590 | 196 | |
5b2366bd AG |
197 | /* |
198 | * If between sending messages, we can send a pending congestion | |
199 | * map update. | |
5c115590 | 200 | */ |
8690bfa1 | 201 | if (!rm && test_and_clear_bit(0, &conn->c_map_queued)) { |
77dd550e AG |
202 | rm = rds_cong_update_alloc(conn); |
203 | if (IS_ERR(rm)) { | |
204 | ret = PTR_ERR(rm); | |
205 | break; | |
5b2366bd | 206 | } |
77dd550e | 207 | rm->data.op_active = 1; |
1f9ecd7e SV |
208 | rm->m_inc.i_conn_path = cp; |
209 | rm->m_inc.i_conn = cp->cp_conn; | |
77dd550e | 210 | |
1f9ecd7e | 211 | cp->cp_xmit_rm = rm; |
5c115590 AG |
212 | } |
213 | ||
214 | /* | |
5b2366bd | 215 | * If not already working on one, grab the next message. |
5c115590 | 216 | * |
1f9ecd7e | 217 | * cp_xmit_rm holds a ref while we're sending this message down |
5c115590 AG |
218 | * the connction. We can use this ref while holding the |
219 | * send_sem.. rds_send_reset() is serialized with it. | |
220 | */ | |
8690bfa1 | 221 | if (!rm) { |
5c115590 AG |
222 | unsigned int len; |
223 | ||
443be0e5 SV |
224 | batch_count++; |
225 | ||
226 | /* we want to process as big a batch as we can, but | |
227 | * we also want to avoid softlockups. If we've been | |
228 | * through a lot of messages, lets back off and see | |
229 | * if anyone else jumps in | |
230 | */ | |
4bebdd7a | 231 | if (batch_count >= send_batch_count) |
443be0e5 SV |
232 | goto over_batch; |
233 | ||
1f9ecd7e | 234 | spin_lock_irqsave(&cp->cp_lock, flags); |
5c115590 | 235 | |
1f9ecd7e SV |
236 | if (!list_empty(&cp->cp_send_queue)) { |
237 | rm = list_entry(cp->cp_send_queue.next, | |
5c115590 AG |
238 | struct rds_message, |
239 | m_conn_item); | |
240 | rds_message_addref(rm); | |
241 | ||
242 | /* | |
243 | * Move the message from the send queue to the retransmit | |
244 | * list right away. | |
245 | */ | |
1f9ecd7e SV |
246 | list_move_tail(&rm->m_conn_item, |
247 | &cp->cp_retrans); | |
5c115590 AG |
248 | } |
249 | ||
1f9ecd7e | 250 | spin_unlock_irqrestore(&cp->cp_lock, flags); |
5c115590 | 251 | |
fcc5450c | 252 | if (!rm) |
5c115590 | 253 | break; |
5c115590 AG |
254 | |
255 | /* Unfortunately, the way Infiniband deals with | |
256 | * RDMA to a bad MR key is by moving the entire | |
257 | * queue pair to error state. We cold possibly | |
258 | * recover from that, but right now we drop the | |
259 | * connection. | |
260 | * Therefore, we never retransmit messages with RDMA ops. | |
261 | */ | |
905dd418 SV |
262 | if (test_bit(RDS_MSG_FLUSH, &rm->m_flags) || |
263 | (rm->rdma.op_active && | |
264 | test_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags))) { | |
1f9ecd7e | 265 | spin_lock_irqsave(&cp->cp_lock, flags); |
5c115590 AG |
266 | if (test_and_clear_bit(RDS_MSG_ON_CONN, &rm->m_flags)) |
267 | list_move(&rm->m_conn_item, &to_be_dropped); | |
1f9ecd7e | 268 | spin_unlock_irqrestore(&cp->cp_lock, flags); |
5c115590 AG |
269 | continue; |
270 | } | |
271 | ||
272 | /* Require an ACK every once in a while */ | |
273 | len = ntohl(rm->m_inc.i_hdr.h_len); | |
1f9ecd7e SV |
274 | if (cp->cp_unacked_packets == 0 || |
275 | cp->cp_unacked_bytes < len) { | |
5c115590 AG |
276 | __set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags); |
277 | ||
1f9ecd7e SV |
278 | cp->cp_unacked_packets = |
279 | rds_sysctl_max_unacked_packets; | |
280 | cp->cp_unacked_bytes = | |
281 | rds_sysctl_max_unacked_bytes; | |
5c115590 AG |
282 | rds_stats_inc(s_send_ack_required); |
283 | } else { | |
1f9ecd7e SV |
284 | cp->cp_unacked_bytes -= len; |
285 | cp->cp_unacked_packets--; | |
5c115590 AG |
286 | } |
287 | ||
1f9ecd7e | 288 | cp->cp_xmit_rm = rm; |
5c115590 AG |
289 | } |
290 | ||
2c3a5f9a | 291 | /* The transport either sends the whole rdma or none of it */ |
1f9ecd7e | 292 | if (rm->rdma.op_active && !cp->cp_xmit_rdma_sent) { |
ff3d7d36 | 293 | rm->m_final_op = &rm->rdma; |
4f73113c | 294 | /* The transport owns the mapped memory for now. |
295 | * You can't unmap it while it's on the send queue | |
296 | */ | |
297 | set_bit(RDS_MSG_MAPPED, &rm->m_flags); | |
2c3a5f9a | 298 | ret = conn->c_trans->xmit_rdma(conn, &rm->rdma); |
4f73113c | 299 | if (ret) { |
300 | clear_bit(RDS_MSG_MAPPED, &rm->m_flags); | |
301 | wake_up_interruptible(&rm->m_flush_wait); | |
15133f6e | 302 | break; |
4f73113c | 303 | } |
1f9ecd7e | 304 | cp->cp_xmit_rdma_sent = 1; |
2c3a5f9a | 305 | |
15133f6e AG |
306 | } |
307 | ||
1f9ecd7e | 308 | if (rm->atomic.op_active && !cp->cp_xmit_atomic_sent) { |
ff3d7d36 | 309 | rm->m_final_op = &rm->atomic; |
4f73113c | 310 | /* The transport owns the mapped memory for now. |
311 | * You can't unmap it while it's on the send queue | |
312 | */ | |
313 | set_bit(RDS_MSG_MAPPED, &rm->m_flags); | |
ff3d7d36 | 314 | ret = conn->c_trans->xmit_atomic(conn, &rm->atomic); |
4f73113c | 315 | if (ret) { |
316 | clear_bit(RDS_MSG_MAPPED, &rm->m_flags); | |
317 | wake_up_interruptible(&rm->m_flush_wait); | |
5c115590 | 318 | break; |
4f73113c | 319 | } |
1f9ecd7e | 320 | cp->cp_xmit_atomic_sent = 1; |
ff3d7d36 | 321 | |
5c115590 AG |
322 | } |
323 | ||
2c3a5f9a AG |
324 | /* |
325 | * A number of cases require an RDS header to be sent | |
326 | * even if there is no data. | |
327 | * We permit 0-byte sends; rds-ping depends on this. | |
328 | * However, if there are exclusively attached silent ops, | |
329 | * we skip the hdr/data send, to enable silent operation. | |
330 | */ | |
331 | if (rm->data.op_nents == 0) { | |
332 | int ops_present; | |
333 | int all_ops_are_silent = 1; | |
334 | ||
335 | ops_present = (rm->atomic.op_active || rm->rdma.op_active); | |
336 | if (rm->atomic.op_active && !rm->atomic.op_silent) | |
337 | all_ops_are_silent = 0; | |
338 | if (rm->rdma.op_active && !rm->rdma.op_silent) | |
339 | all_ops_are_silent = 0; | |
340 | ||
341 | if (ops_present && all_ops_are_silent | |
342 | && !rm->m_rdma_cookie) | |
343 | rm->data.op_active = 0; | |
344 | } | |
345 | ||
1f9ecd7e | 346 | if (rm->data.op_active && !cp->cp_xmit_data_sent) { |
ff3d7d36 | 347 | rm->m_final_op = &rm->data; |
1f9ecd7e | 348 | |
5c115590 | 349 | ret = conn->c_trans->xmit(conn, rm, |
1f9ecd7e SV |
350 | cp->cp_xmit_hdr_off, |
351 | cp->cp_xmit_sg, | |
352 | cp->cp_xmit_data_off); | |
5c115590 AG |
353 | if (ret <= 0) |
354 | break; | |
355 | ||
1f9ecd7e | 356 | if (cp->cp_xmit_hdr_off < sizeof(struct rds_header)) { |
5c115590 AG |
357 | tmp = min_t(int, ret, |
358 | sizeof(struct rds_header) - | |
1f9ecd7e SV |
359 | cp->cp_xmit_hdr_off); |
360 | cp->cp_xmit_hdr_off += tmp; | |
5c115590 AG |
361 | ret -= tmp; |
362 | } | |
363 | ||
1f9ecd7e | 364 | sg = &rm->data.op_sg[cp->cp_xmit_sg]; |
5c115590 AG |
365 | while (ret) { |
366 | tmp = min_t(int, ret, sg->length - | |
1f9ecd7e SV |
367 | cp->cp_xmit_data_off); |
368 | cp->cp_xmit_data_off += tmp; | |
5c115590 | 369 | ret -= tmp; |
1f9ecd7e SV |
370 | if (cp->cp_xmit_data_off == sg->length) { |
371 | cp->cp_xmit_data_off = 0; | |
5c115590 | 372 | sg++; |
1f9ecd7e SV |
373 | cp->cp_xmit_sg++; |
374 | BUG_ON(ret != 0 && cp->cp_xmit_sg == | |
375 | rm->data.op_nents); | |
5c115590 AG |
376 | } |
377 | } | |
5b2366bd | 378 | |
1f9ecd7e SV |
379 | if (cp->cp_xmit_hdr_off == sizeof(struct rds_header) && |
380 | (cp->cp_xmit_sg == rm->data.op_nents)) | |
381 | cp->cp_xmit_data_sent = 1; | |
5b2366bd AG |
382 | } |
383 | ||
384 | /* | |
385 | * A rm will only take multiple times through this loop | |
386 | * if there is a data op. Thus, if the data is sent (or there was | |
387 | * none), then we're done with the rm. | |
388 | */ | |
1f9ecd7e SV |
389 | if (!rm->data.op_active || cp->cp_xmit_data_sent) { |
390 | cp->cp_xmit_rm = NULL; | |
391 | cp->cp_xmit_sg = 0; | |
392 | cp->cp_xmit_hdr_off = 0; | |
393 | cp->cp_xmit_data_off = 0; | |
394 | cp->cp_xmit_rdma_sent = 0; | |
395 | cp->cp_xmit_atomic_sent = 0; | |
396 | cp->cp_xmit_data_sent = 0; | |
5b2366bd AG |
397 | |
398 | rds_message_put(rm); | |
5c115590 AG |
399 | } |
400 | } | |
401 | ||
443be0e5 | 402 | over_batch: |
226f7a7d SV |
403 | if (conn->c_trans->xmit_path_complete) |
404 | conn->c_trans->xmit_path_complete(cp); | |
1f9ecd7e | 405 | release_in_xmit(cp); |
5c115590 | 406 | |
2ad8099b AG |
407 | /* Nuke any messages we decided not to retransmit. */ |
408 | if (!list_empty(&to_be_dropped)) { | |
409 | /* irqs on here, so we can put(), unlike above */ | |
410 | list_for_each_entry(rm, &to_be_dropped, m_conn_item) | |
411 | rds_message_put(rm); | |
412 | rds_send_remove_from_sock(&to_be_dropped, RDS_RDMA_DROPPED); | |
413 | } | |
414 | ||
fcc5450c | 415 | /* |
0f4b1c7e ZB |
416 | * Other senders can queue a message after we last test the send queue |
417 | * but before we clear RDS_IN_XMIT. In that case they'd back off and | |
418 | * not try and send their newly queued message. We need to check the | |
419 | * send queue after having cleared RDS_IN_XMIT so that their message | |
420 | * doesn't get stuck on the send queue. | |
fcc5450c AG |
421 | * |
422 | * If the transport cannot continue (i.e ret != 0), then it must | |
423 | * call us when more room is available, such as from the tx | |
424 | * completion handler. | |
443be0e5 SV |
425 | * |
426 | * We have an extra generation check here so that if someone manages | |
427 | * to jump in after our release_in_xmit, we'll see that they have done | |
428 | * some work and we will skip our goto | |
fcc5450c AG |
429 | */ |
430 | if (ret == 0) { | |
9e29db0e | 431 | smp_mb(); |
0c484240 | 432 | if ((test_bit(0, &conn->c_map_queued) || |
1f9ecd7e | 433 | !list_empty(&cp->cp_send_queue)) && |
e623a48e | 434 | send_gen == READ_ONCE(cp->cp_send_gen)) { |
049ee3f5 | 435 | rds_stats_inc(s_send_lock_queue_raced); |
4bebdd7a SS |
436 | if (batch_count < send_batch_count) |
437 | goto restart; | |
1f9ecd7e | 438 | queue_delayed_work(rds_wq, &cp->cp_send_w, 1); |
5c115590 | 439 | } |
5c115590 AG |
440 | } |
441 | out: | |
442 | return ret; | |
443 | } | |
0c28c045 | 444 | EXPORT_SYMBOL_GPL(rds_send_xmit); |
5c115590 AG |
445 | |
446 | static void rds_send_sndbuf_remove(struct rds_sock *rs, struct rds_message *rm) | |
447 | { | |
448 | u32 len = be32_to_cpu(rm->m_inc.i_hdr.h_len); | |
449 | ||
450 | assert_spin_locked(&rs->rs_lock); | |
451 | ||
452 | BUG_ON(rs->rs_snd_bytes < len); | |
453 | rs->rs_snd_bytes -= len; | |
454 | ||
455 | if (rs->rs_snd_bytes == 0) | |
456 | rds_stats_inc(s_send_queue_empty); | |
457 | } | |
458 | ||
459 | static inline int rds_send_is_acked(struct rds_message *rm, u64 ack, | |
460 | is_acked_func is_acked) | |
461 | { | |
462 | if (is_acked) | |
463 | return is_acked(rm, ack); | |
464 | return be64_to_cpu(rm->m_inc.i_hdr.h_sequence) <= ack; | |
465 | } | |
466 | ||
5c115590 AG |
467 | /* |
468 | * This is pretty similar to what happens below in the ACK | |
469 | * handling code - except that we call here as soon as we get | |
470 | * the IB send completion on the RDMA op and the accompanying | |
471 | * message. | |
472 | */ | |
473 | void rds_rdma_send_complete(struct rds_message *rm, int status) | |
474 | { | |
475 | struct rds_sock *rs = NULL; | |
f8b3aaf2 | 476 | struct rm_rdma_op *ro; |
5c115590 | 477 | struct rds_notifier *notifier; |
9de0864c | 478 | unsigned long flags; |
941f8d55 | 479 | unsigned int notify = 0; |
5c115590 | 480 | |
9de0864c | 481 | spin_lock_irqsave(&rm->m_rs_lock, flags); |
5c115590 | 482 | |
941f8d55 | 483 | notify = rm->rdma.op_notify | rm->data.op_notify; |
f8b3aaf2 | 484 | ro = &rm->rdma; |
f64f9e71 | 485 | if (test_bit(RDS_MSG_ON_SOCK, &rm->m_flags) && |
941f8d55 | 486 | ro->op_active && notify && ro->op_notifier) { |
f8b3aaf2 | 487 | notifier = ro->op_notifier; |
5c115590 AG |
488 | rs = rm->m_rs; |
489 | sock_hold(rds_rs_to_sk(rs)); | |
490 | ||
491 | notifier->n_status = status; | |
492 | spin_lock(&rs->rs_lock); | |
493 | list_add_tail(¬ifier->n_list, &rs->rs_notify_queue); | |
494 | spin_unlock(&rs->rs_lock); | |
495 | ||
f8b3aaf2 | 496 | ro->op_notifier = NULL; |
5c115590 AG |
497 | } |
498 | ||
9de0864c | 499 | spin_unlock_irqrestore(&rm->m_rs_lock, flags); |
5c115590 AG |
500 | |
501 | if (rs) { | |
502 | rds_wake_sk_sleep(rs); | |
503 | sock_put(rds_rs_to_sk(rs)); | |
504 | } | |
505 | } | |
616b757a | 506 | EXPORT_SYMBOL_GPL(rds_rdma_send_complete); |
5c115590 | 507 | |
15133f6e AG |
508 | /* |
509 | * Just like above, except looks at atomic op | |
510 | */ | |
511 | void rds_atomic_send_complete(struct rds_message *rm, int status) | |
512 | { | |
513 | struct rds_sock *rs = NULL; | |
514 | struct rm_atomic_op *ao; | |
515 | struct rds_notifier *notifier; | |
cf4b7389 | 516 | unsigned long flags; |
15133f6e | 517 | |
cf4b7389 | 518 | spin_lock_irqsave(&rm->m_rs_lock, flags); |
15133f6e AG |
519 | |
520 | ao = &rm->atomic; | |
521 | if (test_bit(RDS_MSG_ON_SOCK, &rm->m_flags) | |
522 | && ao->op_active && ao->op_notify && ao->op_notifier) { | |
523 | notifier = ao->op_notifier; | |
524 | rs = rm->m_rs; | |
525 | sock_hold(rds_rs_to_sk(rs)); | |
526 | ||
527 | notifier->n_status = status; | |
528 | spin_lock(&rs->rs_lock); | |
529 | list_add_tail(¬ifier->n_list, &rs->rs_notify_queue); | |
530 | spin_unlock(&rs->rs_lock); | |
531 | ||
532 | ao->op_notifier = NULL; | |
533 | } | |
534 | ||
cf4b7389 | 535 | spin_unlock_irqrestore(&rm->m_rs_lock, flags); |
15133f6e AG |
536 | |
537 | if (rs) { | |
538 | rds_wake_sk_sleep(rs); | |
539 | sock_put(rds_rs_to_sk(rs)); | |
540 | } | |
541 | } | |
542 | EXPORT_SYMBOL_GPL(rds_atomic_send_complete); | |
543 | ||
5c115590 AG |
544 | /* |
545 | * This is the same as rds_rdma_send_complete except we | |
546 | * don't do any locking - we have all the ingredients (message, | |
547 | * socket, socket lock) and can just move the notifier. | |
548 | */ | |
549 | static inline void | |
940786eb | 550 | __rds_send_complete(struct rds_sock *rs, struct rds_message *rm, int status) |
5c115590 | 551 | { |
f8b3aaf2 | 552 | struct rm_rdma_op *ro; |
940786eb | 553 | struct rm_atomic_op *ao; |
5c115590 | 554 | |
f8b3aaf2 AG |
555 | ro = &rm->rdma; |
556 | if (ro->op_active && ro->op_notify && ro->op_notifier) { | |
557 | ro->op_notifier->n_status = status; | |
558 | list_add_tail(&ro->op_notifier->n_list, &rs->rs_notify_queue); | |
559 | ro->op_notifier = NULL; | |
5c115590 AG |
560 | } |
561 | ||
940786eb AG |
562 | ao = &rm->atomic; |
563 | if (ao->op_active && ao->op_notify && ao->op_notifier) { | |
564 | ao->op_notifier->n_status = status; | |
565 | list_add_tail(&ao->op_notifier->n_list, &rs->rs_notify_queue); | |
566 | ao->op_notifier = NULL; | |
567 | } | |
568 | ||
5c115590 AG |
569 | /* No need to wake the app - caller does this */ |
570 | } | |
571 | ||
5c115590 AG |
572 | /* |
573 | * This removes messages from the socket's list if they're on it. The list | |
574 | * argument must be private to the caller, we must be able to modify it | |
575 | * without locks. The messages must have a reference held for their | |
576 | * position on the list. This function will drop that reference after | |
577 | * removing the messages from the 'messages' list regardless of if it found | |
578 | * the messages on the socket list or not. | |
579 | */ | |
ff51bf84 | 580 | static void rds_send_remove_from_sock(struct list_head *messages, int status) |
5c115590 | 581 | { |
561c7df6 | 582 | unsigned long flags; |
5c115590 AG |
583 | struct rds_sock *rs = NULL; |
584 | struct rds_message *rm; | |
585 | ||
5c115590 | 586 | while (!list_empty(messages)) { |
561c7df6 AG |
587 | int was_on_sock = 0; |
588 | ||
5c115590 AG |
589 | rm = list_entry(messages->next, struct rds_message, |
590 | m_conn_item); | |
591 | list_del_init(&rm->m_conn_item); | |
592 | ||
593 | /* | |
594 | * If we see this flag cleared then we're *sure* that someone | |
595 | * else beat us to removing it from the sock. If we race | |
596 | * with their flag update we'll get the lock and then really | |
597 | * see that the flag has been cleared. | |
598 | * | |
599 | * The message spinlock makes sure nobody clears rm->m_rs | |
600 | * while we're messing with it. It does not prevent the | |
601 | * message from being removed from the socket, though. | |
602 | */ | |
561c7df6 | 603 | spin_lock_irqsave(&rm->m_rs_lock, flags); |
5c115590 AG |
604 | if (!test_bit(RDS_MSG_ON_SOCK, &rm->m_flags)) |
605 | goto unlock_and_drop; | |
606 | ||
607 | if (rs != rm->m_rs) { | |
608 | if (rs) { | |
5c115590 AG |
609 | rds_wake_sk_sleep(rs); |
610 | sock_put(rds_rs_to_sk(rs)); | |
611 | } | |
612 | rs = rm->m_rs; | |
593cbb3e HK |
613 | if (rs) |
614 | sock_hold(rds_rs_to_sk(rs)); | |
5c115590 | 615 | } |
593cbb3e HK |
616 | if (!rs) |
617 | goto unlock_and_drop; | |
048c15e6 | 618 | spin_lock(&rs->rs_lock); |
5c115590 AG |
619 | |
620 | if (test_and_clear_bit(RDS_MSG_ON_SOCK, &rm->m_flags)) { | |
f8b3aaf2 | 621 | struct rm_rdma_op *ro = &rm->rdma; |
5c115590 AG |
622 | struct rds_notifier *notifier; |
623 | ||
624 | list_del_init(&rm->m_sock_item); | |
625 | rds_send_sndbuf_remove(rs, rm); | |
626 | ||
f8b3aaf2 AG |
627 | if (ro->op_active && ro->op_notifier && |
628 | (ro->op_notify || (ro->op_recverr && status))) { | |
629 | notifier = ro->op_notifier; | |
5c115590 AG |
630 | list_add_tail(¬ifier->n_list, |
631 | &rs->rs_notify_queue); | |
632 | if (!notifier->n_status) | |
633 | notifier->n_status = status; | |
f8b3aaf2 | 634 | rm->rdma.op_notifier = NULL; |
5c115590 | 635 | } |
561c7df6 | 636 | was_on_sock = 1; |
5c115590 AG |
637 | rm->m_rs = NULL; |
638 | } | |
048c15e6 | 639 | spin_unlock(&rs->rs_lock); |
5c115590 AG |
640 | |
641 | unlock_and_drop: | |
561c7df6 | 642 | spin_unlock_irqrestore(&rm->m_rs_lock, flags); |
5c115590 | 643 | rds_message_put(rm); |
561c7df6 AG |
644 | if (was_on_sock) |
645 | rds_message_put(rm); | |
5c115590 AG |
646 | } |
647 | ||
648 | if (rs) { | |
5c115590 AG |
649 | rds_wake_sk_sleep(rs); |
650 | sock_put(rds_rs_to_sk(rs)); | |
651 | } | |
5c115590 AG |
652 | } |
653 | ||
654 | /* | |
655 | * Transports call here when they've determined that the receiver queued | |
656 | * messages up to, and including, the given sequence number. Messages are | |
657 | * moved to the retrans queue when rds_send_xmit picks them off the send | |
658 | * queue. This means that in the TCP case, the message may not have been | |
659 | * assigned the m_ack_seq yet - but that's fine as long as tcp_is_acked | |
660 | * checks the RDS_MSG_HAS_ACK_SEQ bit. | |
5c115590 | 661 | */ |
5c3d274c SV |
662 | void rds_send_path_drop_acked(struct rds_conn_path *cp, u64 ack, |
663 | is_acked_func is_acked) | |
5c115590 AG |
664 | { |
665 | struct rds_message *rm, *tmp; | |
666 | unsigned long flags; | |
667 | LIST_HEAD(list); | |
668 | ||
5c3d274c | 669 | spin_lock_irqsave(&cp->cp_lock, flags); |
5c115590 | 670 | |
5c3d274c | 671 | list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) { |
5c115590 AG |
672 | if (!rds_send_is_acked(rm, ack, is_acked)) |
673 | break; | |
674 | ||
675 | list_move(&rm->m_conn_item, &list); | |
676 | clear_bit(RDS_MSG_ON_CONN, &rm->m_flags); | |
677 | } | |
678 | ||
679 | /* order flag updates with spin locks */ | |
680 | if (!list_empty(&list)) | |
4e857c58 | 681 | smp_mb__after_atomic(); |
5c115590 | 682 | |
5c3d274c | 683 | spin_unlock_irqrestore(&cp->cp_lock, flags); |
5c115590 AG |
684 | |
685 | /* now remove the messages from the sock list as needed */ | |
686 | rds_send_remove_from_sock(&list, RDS_RDMA_SUCCESS); | |
687 | } | |
5c3d274c SV |
688 | EXPORT_SYMBOL_GPL(rds_send_path_drop_acked); |
689 | ||
690 | void rds_send_drop_acked(struct rds_connection *conn, u64 ack, | |
691 | is_acked_func is_acked) | |
692 | { | |
693 | WARN_ON(conn->c_trans->t_mp_capable); | |
694 | rds_send_path_drop_acked(&conn->c_path[0], ack, is_acked); | |
695 | } | |
616b757a | 696 | EXPORT_SYMBOL_GPL(rds_send_drop_acked); |
5c115590 AG |
697 | |
698 | void rds_send_drop_to(struct rds_sock *rs, struct sockaddr_in *dest) | |
699 | { | |
700 | struct rds_message *rm, *tmp; | |
701 | struct rds_connection *conn; | |
01ff34ed | 702 | struct rds_conn_path *cp; |
7c82eaf0 | 703 | unsigned long flags; |
5c115590 | 704 | LIST_HEAD(list); |
5c115590 AG |
705 | |
706 | /* get all the messages we're dropping under the rs lock */ | |
707 | spin_lock_irqsave(&rs->rs_lock, flags); | |
708 | ||
709 | list_for_each_entry_safe(rm, tmp, &rs->rs_send_queue, m_sock_item) { | |
710 | if (dest && (dest->sin_addr.s_addr != rm->m_daddr || | |
711 | dest->sin_port != rm->m_inc.i_hdr.h_dport)) | |
712 | continue; | |
713 | ||
5c115590 AG |
714 | list_move(&rm->m_sock_item, &list); |
715 | rds_send_sndbuf_remove(rs, rm); | |
716 | clear_bit(RDS_MSG_ON_SOCK, &rm->m_flags); | |
5c115590 AG |
717 | } |
718 | ||
719 | /* order flag updates with the rs lock */ | |
4e857c58 | 720 | smp_mb__after_atomic(); |
5c115590 AG |
721 | |
722 | spin_unlock_irqrestore(&rs->rs_lock, flags); | |
723 | ||
7c82eaf0 AG |
724 | if (list_empty(&list)) |
725 | return; | |
5c115590 | 726 | |
7c82eaf0 | 727 | /* Remove the messages from the conn */ |
5c115590 | 728 | list_for_each_entry(rm, &list, m_sock_item) { |
7c82eaf0 AG |
729 | |
730 | conn = rm->m_inc.i_conn; | |
01ff34ed SV |
731 | if (conn->c_trans->t_mp_capable) |
732 | cp = rm->m_inc.i_conn_path; | |
733 | else | |
734 | cp = &conn->c_path[0]; | |
5c115590 | 735 | |
01ff34ed | 736 | spin_lock_irqsave(&cp->cp_lock, flags); |
5c115590 | 737 | /* |
7c82eaf0 AG |
738 | * Maybe someone else beat us to removing rm from the conn. |
739 | * If we race with their flag update we'll get the lock and | |
740 | * then really see that the flag has been cleared. | |
5c115590 | 741 | */ |
7c82eaf0 | 742 | if (!test_and_clear_bit(RDS_MSG_ON_CONN, &rm->m_flags)) { |
01ff34ed | 743 | spin_unlock_irqrestore(&cp->cp_lock, flags); |
593cbb3e HK |
744 | spin_lock_irqsave(&rm->m_rs_lock, flags); |
745 | rm->m_rs = NULL; | |
746 | spin_unlock_irqrestore(&rm->m_rs_lock, flags); | |
5c115590 | 747 | continue; |
5c115590 | 748 | } |
9de0864c | 749 | list_del_init(&rm->m_conn_item); |
01ff34ed | 750 | spin_unlock_irqrestore(&cp->cp_lock, flags); |
5c115590 | 751 | |
7c82eaf0 AG |
752 | /* |
753 | * Couldn't grab m_rs_lock in top loop (lock ordering), | |
754 | * but we can now. | |
755 | */ | |
9de0864c | 756 | spin_lock_irqsave(&rm->m_rs_lock, flags); |
5c115590 | 757 | |
7c82eaf0 | 758 | spin_lock(&rs->rs_lock); |
940786eb | 759 | __rds_send_complete(rs, rm, RDS_RDMA_CANCELED); |
7c82eaf0 AG |
760 | spin_unlock(&rs->rs_lock); |
761 | ||
762 | rm->m_rs = NULL; | |
9de0864c | 763 | spin_unlock_irqrestore(&rm->m_rs_lock, flags); |
7c82eaf0 | 764 | |
7c82eaf0 | 765 | rds_message_put(rm); |
7c82eaf0 | 766 | } |
5c115590 | 767 | |
7c82eaf0 | 768 | rds_wake_sk_sleep(rs); |
550a8002 | 769 | |
5c115590 AG |
770 | while (!list_empty(&list)) { |
771 | rm = list_entry(list.next, struct rds_message, m_sock_item); | |
772 | list_del_init(&rm->m_sock_item); | |
5c115590 | 773 | rds_message_wait(rm); |
dfcec251 | 774 | |
775 | /* just in case the code above skipped this message | |
776 | * because RDS_MSG_ON_CONN wasn't set, run it again here | |
777 | * taking m_rs_lock is the only thing that keeps us | |
778 | * from racing with ack processing. | |
779 | */ | |
780 | spin_lock_irqsave(&rm->m_rs_lock, flags); | |
781 | ||
782 | spin_lock(&rs->rs_lock); | |
783 | __rds_send_complete(rs, rm, RDS_RDMA_CANCELED); | |
784 | spin_unlock(&rs->rs_lock); | |
785 | ||
786 | rm->m_rs = NULL; | |
787 | spin_unlock_irqrestore(&rm->m_rs_lock, flags); | |
788 | ||
5c115590 AG |
789 | rds_message_put(rm); |
790 | } | |
791 | } | |
792 | ||
793 | /* | |
794 | * we only want this to fire once so we use the callers 'queued'. It's | |
795 | * possible that another thread can race with us and remove the | |
796 | * message from the flow with RDS_CANCEL_SENT_TO. | |
797 | */ | |
798 | static int rds_send_queue_rm(struct rds_sock *rs, struct rds_connection *conn, | |
780a6d9e | 799 | struct rds_conn_path *cp, |
5c115590 AG |
800 | struct rds_message *rm, __be16 sport, |
801 | __be16 dport, int *queued) | |
802 | { | |
803 | unsigned long flags; | |
804 | u32 len; | |
805 | ||
806 | if (*queued) | |
807 | goto out; | |
808 | ||
809 | len = be32_to_cpu(rm->m_inc.i_hdr.h_len); | |
810 | ||
811 | /* this is the only place which holds both the socket's rs_lock | |
812 | * and the connection's c_lock */ | |
813 | spin_lock_irqsave(&rs->rs_lock, flags); | |
814 | ||
815 | /* | |
816 | * If there is a little space in sndbuf, we don't queue anything, | |
817 | * and userspace gets -EAGAIN. But poll() indicates there's send | |
818 | * room. This can lead to bad behavior (spinning) if snd_bytes isn't | |
819 | * freed up by incoming acks. So we check the *old* value of | |
820 | * rs_snd_bytes here to allow the last msg to exceed the buffer, | |
821 | * and poll() now knows no more data can be sent. | |
822 | */ | |
823 | if (rs->rs_snd_bytes < rds_sk_sndbuf(rs)) { | |
824 | rs->rs_snd_bytes += len; | |
825 | ||
826 | /* let recv side know we are close to send space exhaustion. | |
827 | * This is probably not the optimal way to do it, as this | |
828 | * means we set the flag on *all* messages as soon as our | |
829 | * throughput hits a certain threshold. | |
830 | */ | |
831 | if (rs->rs_snd_bytes >= rds_sk_sndbuf(rs) / 2) | |
832 | __set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags); | |
833 | ||
834 | list_add_tail(&rm->m_sock_item, &rs->rs_send_queue); | |
835 | set_bit(RDS_MSG_ON_SOCK, &rm->m_flags); | |
836 | rds_message_addref(rm); | |
837 | rm->m_rs = rs; | |
838 | ||
839 | /* The code ordering is a little weird, but we're | |
840 | trying to minimize the time we hold c_lock */ | |
841 | rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport, 0); | |
842 | rm->m_inc.i_conn = conn; | |
780a6d9e | 843 | rm->m_inc.i_conn_path = cp; |
5c115590 AG |
844 | rds_message_addref(rm); |
845 | ||
780a6d9e SV |
846 | spin_lock(&cp->cp_lock); |
847 | rm->m_inc.i_hdr.h_sequence = cpu_to_be64(cp->cp_next_tx_seq++); | |
848 | list_add_tail(&rm->m_conn_item, &cp->cp_send_queue); | |
5c115590 | 849 | set_bit(RDS_MSG_ON_CONN, &rm->m_flags); |
780a6d9e | 850 | spin_unlock(&cp->cp_lock); |
5c115590 AG |
851 | |
852 | rdsdebug("queued msg %p len %d, rs %p bytes %d seq %llu\n", | |
853 | rm, len, rs, rs->rs_snd_bytes, | |
854 | (unsigned long long)be64_to_cpu(rm->m_inc.i_hdr.h_sequence)); | |
855 | ||
856 | *queued = 1; | |
857 | } | |
858 | ||
859 | spin_unlock_irqrestore(&rs->rs_lock, flags); | |
860 | out: | |
861 | return *queued; | |
862 | } | |
863 | ||
fc445084 AG |
864 | /* |
865 | * rds_message is getting to be quite complicated, and we'd like to allocate | |
866 | * it all in one go. This figures out how big it needs to be up front. | |
867 | */ | |
868 | static int rds_rm_size(struct msghdr *msg, int data_len) | |
869 | { | |
ff87e97a | 870 | struct cmsghdr *cmsg; |
fc445084 | 871 | int size = 0; |
aa0a4ef4 | 872 | int cmsg_groups = 0; |
ff87e97a AG |
873 | int retval; |
874 | ||
f95b414e | 875 | for_each_cmsghdr(cmsg, msg) { |
ff87e97a AG |
876 | if (!CMSG_OK(msg, cmsg)) |
877 | return -EINVAL; | |
878 | ||
879 | if (cmsg->cmsg_level != SOL_RDS) | |
880 | continue; | |
881 | ||
882 | switch (cmsg->cmsg_type) { | |
883 | case RDS_CMSG_RDMA_ARGS: | |
aa0a4ef4 | 884 | cmsg_groups |= 1; |
ff87e97a AG |
885 | retval = rds_rdma_extra_size(CMSG_DATA(cmsg)); |
886 | if (retval < 0) | |
887 | return retval; | |
888 | size += retval; | |
aa0a4ef4 | 889 | |
ff87e97a AG |
890 | break; |
891 | ||
892 | case RDS_CMSG_RDMA_DEST: | |
893 | case RDS_CMSG_RDMA_MAP: | |
aa0a4ef4 | 894 | cmsg_groups |= 2; |
ff87e97a AG |
895 | /* these are valid but do no add any size */ |
896 | break; | |
897 | ||
15133f6e AG |
898 | case RDS_CMSG_ATOMIC_CSWP: |
899 | case RDS_CMSG_ATOMIC_FADD: | |
20c72bd5 AG |
900 | case RDS_CMSG_MASKED_ATOMIC_CSWP: |
901 | case RDS_CMSG_MASKED_ATOMIC_FADD: | |
aa0a4ef4 | 902 | cmsg_groups |= 1; |
15133f6e AG |
903 | size += sizeof(struct scatterlist); |
904 | break; | |
905 | ||
ff87e97a AG |
906 | default: |
907 | return -EINVAL; | |
908 | } | |
909 | ||
910 | } | |
fc445084 | 911 | |
ff87e97a | 912 | size += ceil(data_len, PAGE_SIZE) * sizeof(struct scatterlist); |
fc445084 | 913 | |
aa0a4ef4 AG |
914 | /* Ensure (DEST, MAP) are never used with (ARGS, ATOMIC) */ |
915 | if (cmsg_groups == 3) | |
916 | return -EINVAL; | |
917 | ||
fc445084 AG |
918 | return size; |
919 | } | |
920 | ||
5c115590 AG |
921 | static int rds_cmsg_send(struct rds_sock *rs, struct rds_message *rm, |
922 | struct msghdr *msg, int *allocated_mr) | |
923 | { | |
924 | struct cmsghdr *cmsg; | |
925 | int ret = 0; | |
926 | ||
f95b414e | 927 | for_each_cmsghdr(cmsg, msg) { |
5c115590 AG |
928 | if (!CMSG_OK(msg, cmsg)) |
929 | return -EINVAL; | |
930 | ||
931 | if (cmsg->cmsg_level != SOL_RDS) | |
932 | continue; | |
933 | ||
934 | /* As a side effect, RDMA_DEST and RDMA_MAP will set | |
15133f6e | 935 | * rm->rdma.m_rdma_cookie and rm->rdma.m_rdma_mr. |
5c115590 AG |
936 | */ |
937 | switch (cmsg->cmsg_type) { | |
938 | case RDS_CMSG_RDMA_ARGS: | |
939 | ret = rds_cmsg_rdma_args(rs, rm, cmsg); | |
940 | break; | |
941 | ||
942 | case RDS_CMSG_RDMA_DEST: | |
943 | ret = rds_cmsg_rdma_dest(rs, rm, cmsg); | |
944 | break; | |
945 | ||
946 | case RDS_CMSG_RDMA_MAP: | |
947 | ret = rds_cmsg_rdma_map(rs, rm, cmsg); | |
948 | if (!ret) | |
949 | *allocated_mr = 1; | |
584a8279 SS |
950 | else if (ret == -ENODEV) |
951 | /* Accommodate the get_mr() case which can fail | |
952 | * if connection isn't established yet. | |
953 | */ | |
954 | ret = -EAGAIN; | |
5c115590 | 955 | break; |
15133f6e AG |
956 | case RDS_CMSG_ATOMIC_CSWP: |
957 | case RDS_CMSG_ATOMIC_FADD: | |
20c72bd5 AG |
958 | case RDS_CMSG_MASKED_ATOMIC_CSWP: |
959 | case RDS_CMSG_MASKED_ATOMIC_FADD: | |
15133f6e AG |
960 | ret = rds_cmsg_atomic(rs, rm, cmsg); |
961 | break; | |
5c115590 AG |
962 | |
963 | default: | |
964 | return -EINVAL; | |
965 | } | |
966 | ||
967 | if (ret) | |
968 | break; | |
969 | } | |
970 | ||
971 | return ret; | |
972 | } | |
973 | ||
5916e2c1 SV |
974 | static int rds_send_mprds_hash(struct rds_sock *rs, struct rds_connection *conn) |
975 | { | |
976 | int hash; | |
977 | ||
978 | if (conn->c_npaths == 0) | |
979 | hash = RDS_MPATH_HASH(rs, RDS_MPATH_WORKERS); | |
980 | else | |
981 | hash = RDS_MPATH_HASH(rs, conn->c_npaths); | |
982 | if (conn->c_npaths == 0 && hash != 0) { | |
69b92b5b | 983 | rds_send_ping(conn, 0); |
5916e2c1 SV |
984 | |
985 | if (conn->c_npaths == 0) { | |
986 | wait_event_interruptible(conn->c_hs_waitq, | |
987 | (conn->c_npaths != 0)); | |
988 | } | |
989 | if (conn->c_npaths == 1) | |
990 | hash = 0; | |
991 | } | |
992 | return hash; | |
993 | } | |
994 | ||
f9fb69ad AR |
995 | static int rds_rdma_bytes(struct msghdr *msg, size_t *rdma_bytes) |
996 | { | |
997 | struct rds_rdma_args *args; | |
998 | struct cmsghdr *cmsg; | |
999 | ||
1000 | for_each_cmsghdr(cmsg, msg) { | |
1001 | if (!CMSG_OK(msg, cmsg)) | |
1002 | return -EINVAL; | |
1003 | ||
1004 | if (cmsg->cmsg_level != SOL_RDS) | |
1005 | continue; | |
1006 | ||
1007 | if (cmsg->cmsg_type == RDS_CMSG_RDMA_ARGS) { | |
1008 | args = CMSG_DATA(cmsg); | |
1009 | *rdma_bytes += args->remote_vec.bytes; | |
1010 | } | |
1011 | } | |
1012 | return 0; | |
1013 | } | |
1014 | ||
1b784140 | 1015 | int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len) |
5c115590 AG |
1016 | { |
1017 | struct sock *sk = sock->sk; | |
1018 | struct rds_sock *rs = rds_sk_to_rs(sk); | |
342dfc30 | 1019 | DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); |
5c115590 AG |
1020 | __be32 daddr; |
1021 | __be16 dport; | |
1022 | struct rds_message *rm = NULL; | |
1023 | struct rds_connection *conn; | |
1024 | int ret = 0; | |
1025 | int queued = 0, allocated_mr = 0; | |
1026 | int nonblock = msg->msg_flags & MSG_DONTWAIT; | |
1123fd73 | 1027 | long timeo = sock_sndtimeo(sk, nonblock); |
780a6d9e | 1028 | struct rds_conn_path *cpath; |
f9fb69ad | 1029 | size_t total_payload_len = payload_len, rdma_payload_len = 0; |
5c115590 AG |
1030 | |
1031 | /* Mirror Linux UDP mirror of BSD error message compatibility */ | |
1032 | /* XXX: Perhaps MSG_MORE someday */ | |
1033 | if (msg->msg_flags & ~(MSG_DONTWAIT | MSG_CMSG_COMPAT)) { | |
5c115590 AG |
1034 | ret = -EOPNOTSUPP; |
1035 | goto out; | |
1036 | } | |
1037 | ||
1038 | if (msg->msg_namelen) { | |
1039 | /* XXX fail non-unicast destination IPs? */ | |
1040 | if (msg->msg_namelen < sizeof(*usin) || usin->sin_family != AF_INET) { | |
1041 | ret = -EINVAL; | |
1042 | goto out; | |
1043 | } | |
1044 | daddr = usin->sin_addr.s_addr; | |
1045 | dport = usin->sin_port; | |
1046 | } else { | |
1047 | /* We only care about consistency with ->connect() */ | |
1048 | lock_sock(sk); | |
1049 | daddr = rs->rs_conn_addr; | |
1050 | dport = rs->rs_conn_port; | |
1051 | release_sock(sk); | |
1052 | } | |
1053 | ||
8c7188b2 | 1054 | lock_sock(sk); |
5c115590 | 1055 | if (daddr == 0 || rs->rs_bound_addr == 0) { |
8c7188b2 | 1056 | release_sock(sk); |
5c115590 AG |
1057 | ret = -ENOTCONN; /* XXX not a great errno */ |
1058 | goto out; | |
1059 | } | |
8c7188b2 | 1060 | release_sock(sk); |
5c115590 | 1061 | |
f9fb69ad AR |
1062 | ret = rds_rdma_bytes(msg, &rdma_payload_len); |
1063 | if (ret) | |
1064 | goto out; | |
1065 | ||
1066 | total_payload_len += rdma_payload_len; | |
1067 | if (max_t(size_t, payload_len, rdma_payload_len) > RDS_MAX_MSG_SIZE) { | |
1068 | ret = -EMSGSIZE; | |
1069 | goto out; | |
1070 | } | |
1071 | ||
06e8941e MK |
1072 | if (payload_len > rds_sk_sndbuf(rs)) { |
1073 | ret = -EMSGSIZE; | |
1074 | goto out; | |
1075 | } | |
1076 | ||
fc445084 AG |
1077 | /* size of rm including all sgs */ |
1078 | ret = rds_rm_size(msg, payload_len); | |
1079 | if (ret < 0) | |
1080 | goto out; | |
1081 | ||
1082 | rm = rds_message_alloc(ret, GFP_KERNEL); | |
1083 | if (!rm) { | |
1084 | ret = -ENOMEM; | |
5c115590 AG |
1085 | goto out; |
1086 | } | |
1087 | ||
372cd7de AG |
1088 | /* Attach data to the rm */ |
1089 | if (payload_len) { | |
1090 | rm->data.op_sg = rds_message_alloc_sgs(rm, ceil(payload_len, PAGE_SIZE)); | |
d139ff09 AG |
1091 | if (!rm->data.op_sg) { |
1092 | ret = -ENOMEM; | |
1093 | goto out; | |
1094 | } | |
c0371da6 | 1095 | ret = rds_message_copy_from_user(rm, &msg->msg_iter); |
372cd7de AG |
1096 | if (ret) |
1097 | goto out; | |
1098 | } | |
1099 | rm->data.op_active = 1; | |
fc445084 | 1100 | |
5c115590 AG |
1101 | rm->m_daddr = daddr; |
1102 | ||
5c115590 AG |
1103 | /* rds_conn_create has a spinlock that runs with IRQ off. |
1104 | * Caching the conn in the socket helps a lot. */ | |
1105 | if (rs->rs_conn && rs->rs_conn->c_faddr == daddr) | |
1106 | conn = rs->rs_conn; | |
1107 | else { | |
d5a8ac28 SV |
1108 | conn = rds_conn_create_outgoing(sock_net(sock->sk), |
1109 | rs->rs_bound_addr, daddr, | |
5c115590 AG |
1110 | rs->rs_transport, |
1111 | sock->sk->sk_allocation); | |
1112 | if (IS_ERR(conn)) { | |
1113 | ret = PTR_ERR(conn); | |
1114 | goto out; | |
1115 | } | |
1116 | rs->rs_conn = conn; | |
1117 | } | |
1118 | ||
49f69691 AG |
1119 | /* Parse any control messages the user may have included. */ |
1120 | ret = rds_cmsg_send(rs, rm, msg, &allocated_mr); | |
584a8279 SS |
1121 | if (ret) { |
1122 | /* Trigger connection so that its ready for the next retry */ | |
1123 | if (ret == -EAGAIN) | |
1124 | rds_conn_connect_if_down(conn); | |
49f69691 | 1125 | goto out; |
584a8279 | 1126 | } |
49f69691 | 1127 | |
2c3a5f9a | 1128 | if (rm->rdma.op_active && !conn->c_trans->xmit_rdma) { |
cb0a6056 | 1129 | printk_ratelimited(KERN_NOTICE "rdma_op %p conn xmit_rdma %p\n", |
f8b3aaf2 | 1130 | &rm->rdma, conn->c_trans->xmit_rdma); |
15133f6e AG |
1131 | ret = -EOPNOTSUPP; |
1132 | goto out; | |
1133 | } | |
1134 | ||
1135 | if (rm->atomic.op_active && !conn->c_trans->xmit_atomic) { | |
cb0a6056 | 1136 | printk_ratelimited(KERN_NOTICE "atomic_op %p conn xmit_atomic %p\n", |
15133f6e | 1137 | &rm->atomic, conn->c_trans->xmit_atomic); |
5c115590 AG |
1138 | ret = -EOPNOTSUPP; |
1139 | goto out; | |
1140 | } | |
1141 | ||
5916e2c1 SV |
1142 | if (conn->c_trans->t_mp_capable) |
1143 | cpath = &conn->c_path[rds_send_mprds_hash(rs, conn)]; | |
1144 | else | |
1145 | cpath = &conn->c_path[0]; | |
3c0a5900 SV |
1146 | |
1147 | rds_conn_path_connect_if_down(cpath); | |
5c115590 AG |
1148 | |
1149 | ret = rds_cong_wait(conn->c_fcong, dport, nonblock, rs); | |
b98ba52f AG |
1150 | if (ret) { |
1151 | rs->rs_seen_congestion = 1; | |
5c115590 | 1152 | goto out; |
b98ba52f | 1153 | } |
780a6d9e | 1154 | while (!rds_send_queue_rm(rs, conn, cpath, rm, rs->rs_bound_port, |
5c115590 AG |
1155 | dport, &queued)) { |
1156 | rds_stats_inc(s_send_queue_full); | |
06e8941e | 1157 | |
5c115590 AG |
1158 | if (nonblock) { |
1159 | ret = -EAGAIN; | |
1160 | goto out; | |
1161 | } | |
1162 | ||
aa395145 | 1163 | timeo = wait_event_interruptible_timeout(*sk_sleep(sk), |
780a6d9e | 1164 | rds_send_queue_rm(rs, conn, cpath, rm, |
5c115590 AG |
1165 | rs->rs_bound_port, |
1166 | dport, | |
1167 | &queued), | |
1168 | timeo); | |
1169 | rdsdebug("sendmsg woke queued %d timeo %ld\n", queued, timeo); | |
1170 | if (timeo > 0 || timeo == MAX_SCHEDULE_TIMEOUT) | |
1171 | continue; | |
1172 | ||
1173 | ret = timeo; | |
1174 | if (ret == 0) | |
1175 | ret = -ETIMEDOUT; | |
1176 | goto out; | |
1177 | } | |
1178 | ||
1179 | /* | |
1180 | * By now we've committed to the send. We reuse rds_send_worker() | |
1181 | * to retry sends in the rds thread if the transport asks us to. | |
1182 | */ | |
1183 | rds_stats_inc(s_send_queued); | |
1184 | ||
1f9ecd7e | 1185 | ret = rds_send_xmit(cpath); |
db6526dc | 1186 | if (ret == -ENOMEM || ret == -EAGAIN) |
1f9ecd7e | 1187 | queue_delayed_work(rds_wq, &cpath->cp_send_w, 1); |
5c115590 AG |
1188 | |
1189 | rds_message_put(rm); | |
1190 | return payload_len; | |
1191 | ||
1192 | out: | |
1193 | /* If the user included a RDMA_MAP cmsg, we allocated a MR on the fly. | |
1194 | * If the sendmsg goes through, we keep the MR. If it fails with EAGAIN | |
1195 | * or in any other way, we need to destroy the MR again */ | |
1196 | if (allocated_mr) | |
1197 | rds_rdma_unuse(rs, rds_rdma_cookie_key(rm->m_rdma_cookie), 1); | |
1198 | ||
1199 | if (rm) | |
1200 | rds_message_put(rm); | |
1201 | return ret; | |
1202 | } | |
1203 | ||
1204 | /* | |
5916e2c1 SV |
1205 | * send out a probe. Can be shared by rds_send_ping, |
1206 | * rds_send_pong, rds_send_hb. | |
1207 | * rds_send_hb should use h_flags | |
1208 | * RDS_FLAG_HB_PING|RDS_FLAG_ACK_REQUIRED | |
1209 | * or | |
1210 | * RDS_FLAG_HB_PONG|RDS_FLAG_ACK_REQUIRED | |
5c115590 | 1211 | */ |
bb789763 | 1212 | static int |
5916e2c1 SV |
1213 | rds_send_probe(struct rds_conn_path *cp, __be16 sport, |
1214 | __be16 dport, u8 h_flags) | |
5c115590 AG |
1215 | { |
1216 | struct rds_message *rm; | |
1217 | unsigned long flags; | |
1218 | int ret = 0; | |
1219 | ||
1220 | rm = rds_message_alloc(0, GFP_ATOMIC); | |
8690bfa1 | 1221 | if (!rm) { |
5c115590 AG |
1222 | ret = -ENOMEM; |
1223 | goto out; | |
1224 | } | |
1225 | ||
45997e9e | 1226 | rm->m_daddr = cp->cp_conn->c_faddr; |
acfcd4d4 | 1227 | rm->data.op_active = 1; |
5c115590 | 1228 | |
3c0a5900 | 1229 | rds_conn_path_connect_if_down(cp); |
5c115590 | 1230 | |
45997e9e | 1231 | ret = rds_cong_wait(cp->cp_conn->c_fcong, dport, 1, NULL); |
5c115590 AG |
1232 | if (ret) |
1233 | goto out; | |
1234 | ||
45997e9e SV |
1235 | spin_lock_irqsave(&cp->cp_lock, flags); |
1236 | list_add_tail(&rm->m_conn_item, &cp->cp_send_queue); | |
5c115590 AG |
1237 | set_bit(RDS_MSG_ON_CONN, &rm->m_flags); |
1238 | rds_message_addref(rm); | |
45997e9e SV |
1239 | rm->m_inc.i_conn = cp->cp_conn; |
1240 | rm->m_inc.i_conn_path = cp; | |
5c115590 | 1241 | |
5916e2c1 | 1242 | rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport, |
45997e9e | 1243 | cp->cp_next_tx_seq); |
5916e2c1 | 1244 | rm->m_inc.i_hdr.h_flags |= h_flags; |
45997e9e | 1245 | cp->cp_next_tx_seq++; |
5916e2c1 | 1246 | |
00354de5 SV |
1247 | if (RDS_HS_PROBE(be16_to_cpu(sport), be16_to_cpu(dport)) && |
1248 | cp->cp_conn->c_trans->t_mp_capable) { | |
1249 | u16 npaths = cpu_to_be16(RDS_MPATH_WORKERS); | |
1250 | u32 my_gen_num = cpu_to_be32(cp->cp_conn->c_my_gen_num); | |
5916e2c1 SV |
1251 | |
1252 | rds_message_add_extension(&rm->m_inc.i_hdr, | |
1253 | RDS_EXTHDR_NPATHS, &npaths, | |
1254 | sizeof(npaths)); | |
905dd418 SV |
1255 | rds_message_add_extension(&rm->m_inc.i_hdr, |
1256 | RDS_EXTHDR_GEN_NUM, | |
00354de5 | 1257 | &my_gen_num, |
905dd418 | 1258 | sizeof(u32)); |
5916e2c1 | 1259 | } |
45997e9e | 1260 | spin_unlock_irqrestore(&cp->cp_lock, flags); |
5c115590 AG |
1261 | |
1262 | rds_stats_inc(s_send_queued); | |
1263 | rds_stats_inc(s_send_pong); | |
1264 | ||
7b4b0009 | 1265 | /* schedule the send work on rds_wq */ |
45997e9e | 1266 | queue_delayed_work(rds_wq, &cp->cp_send_w, 1); |
acfcd4d4 | 1267 | |
5c115590 AG |
1268 | rds_message_put(rm); |
1269 | return 0; | |
1270 | ||
1271 | out: | |
1272 | if (rm) | |
1273 | rds_message_put(rm); | |
1274 | return ret; | |
1275 | } | |
5916e2c1 SV |
1276 | |
1277 | int | |
1278 | rds_send_pong(struct rds_conn_path *cp, __be16 dport) | |
1279 | { | |
1280 | return rds_send_probe(cp, 0, dport, 0); | |
1281 | } | |
1282 | ||
69b92b5b SV |
1283 | void |
1284 | rds_send_ping(struct rds_connection *conn, int cp_index) | |
5916e2c1 SV |
1285 | { |
1286 | unsigned long flags; | |
69b92b5b | 1287 | struct rds_conn_path *cp = &conn->c_path[cp_index]; |
5916e2c1 SV |
1288 | |
1289 | spin_lock_irqsave(&cp->cp_lock, flags); | |
1290 | if (conn->c_ping_triggered) { | |
1291 | spin_unlock_irqrestore(&cp->cp_lock, flags); | |
1292 | return; | |
1293 | } | |
1294 | conn->c_ping_triggered = 1; | |
1295 | spin_unlock_irqrestore(&cp->cp_lock, flags); | |
69b92b5b | 1296 | rds_send_probe(cp, cpu_to_be16(RDS_FLAG_PROBE_PORT), 0, 0); |
5916e2c1 | 1297 | } |
69b92b5b | 1298 | EXPORT_SYMBOL_GPL(rds_send_ping); |