1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
|
/*
* VMware vSockets Driver
*
* Copyright (C) 2009-2013 VMware, Inc. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the Free
* Software Foundation version 2 and no later version.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*/
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/stddef.h>
#include <net/sock.h>
#include "vmci_transport_notify.h"
#define PKT_FIELD(vsk, field_name) \
(vmci_trans(vsk)->notify.pkt_q_state.field_name)
static bool vmci_transport_notify_waiting_write(struct vsock_sock *vsk)
{
bool retval;
u64 notify_limit;
if (!PKT_FIELD(vsk, peer_waiting_write))
return false;
/* When the sender blocks, we take that as a sign that the sender is
* faster than the receiver. To reduce the transmit rate of the sender,
* we delay the sending of the read notification by decreasing the
* write_notify_window. The notification is delayed until the number of
* bytes used in the queue drops below the write_notify_window.
*/
if (!PKT_FIELD(vsk, peer_waiting_write_detected)) {
PKT_FIELD(vsk, peer_waiting_write_detected) = true;
if (PKT_FIELD(vsk, write_notify_window) < PAGE_SIZE) {
PKT_FIELD(vsk, write_notify_window) =
PKT_FIELD(vsk, write_notify_min_window);
} else {
PKT_FIELD(vsk, write_notify_window) -= PAGE_SIZE;
if (PKT_FIELD(vsk, write_notify_window) <
PKT_FIELD(vsk, write_notify_min_window))
PKT_FIELD(vsk, write_notify_window) =
PKT_FIELD(vsk, write_notify_min_window);
}
}
notify_limit = vmci_trans(vsk)->consume_size -
PKT_FIELD(vsk, write_notify_window);
/* The notify_limit is used to delay notifications in the case where
* flow control is enabled. Below the test is expressed in terms of
* free space in the queue: if free_space > ConsumeSize -
* write_notify_window then notify An alternate way of expressing this
* is to rewrite the expression to use the data ready in the receive
* queue: if write_notify_window > bufferReady then notify as
* free_space == ConsumeSize - bufferReady.
*/
retval = vmci_qpair_consume_free_space(vmci_trans(vsk)->qpair) >
notify_limit;
if (retval) {
/* Once we notify the peer, we reset the detected flag so the
* next wait will again cause a decrease in the window size.
*/
PKT_FIELD(vsk, peer_waiting_write_detected) = false;
}
return retval;
}
static void
vmci_transport_handle_read(struct sock *sk,
struct vmci_transport_packet *pkt,
bool bottom_half,
struct sockaddr_vm *dst, struct sockaddr_vm *src)
{
sk->sk_write_space(sk);
}
static void
vmci_transport_handle_wrote(struct sock *sk,
struct vmci_transport_packet *pkt,
bool bottom_half,
struct sockaddr_vm *dst, struct sockaddr_vm *src)
{
sk->sk_data_ready(sk);
}
static void vsock_block_update_write_window(struct sock *sk)
{
struct vsock_sock *vsk = vsock_sk(sk);
if (PKT_FIELD(vsk, write_notify_window) < vmci_trans(vsk)->consume_size)
PKT_FIELD(vsk, write_notify_window) =
min(PKT_FIELD(vsk, write_notify_window) + PAGE_SIZE,
vmci_trans(vsk)->consume_size);
}
static int vmci_transport_send_read_notification(struct sock *sk)
{
struct vsock_sock *vsk;
bool sent_read;
unsigned int retries;
int err;
vsk = vsock_sk(sk);
sent_read = false;
retries = 0;
err = 0;
if (vmci_transport_notify_waiting_write(vsk)) {
/* Notify the peer that we have read, retrying the send on
* failure up to our maximum value. XXX For now we just log
* the failure, but later we should schedule a work item to
* handle the resend until it succeeds. That would require
* keeping track of work items in the vsk and cleaning them up
* upon socket close.
*/
while (!(vsk->peer_shutdown & RCV_SHUTDOWN) &&
!sent_read &&
retries < VMCI_TRANSPORT_MAX_DGRAM_RESENDS) {
err = vmci_transport_send_read(sk);
if (err >= 0)
sent_read = true;
retries++;
}
if (retries >= VMCI_TRANSPORT_MAX_DGRAM_RESENDS && !sent_read)
pr_err("%p unable to send read notification to peer\n",
sk);
else
PKT_FIELD(vsk, peer_waiting_write) = false;
}
return err;
}
static void vmci_transport_notify_pkt_socket_init(struct sock *sk)
{
struct vsock_sock *vsk = vsock_sk(sk);
PKT_FIELD(vsk, write_notify_window) = PAGE_SIZE;
PKT_FIELD(vsk, write_notify_min_window) = PAGE_SIZE;
PKT_FIELD(vsk, peer_waiting_write) = false;
PKT_FIELD(vsk, peer_waiting_write_detected) = false;
}
static void vmci_transport_notify_pkt_socket_destruct(struct vsock_sock *vsk)
{
PKT_FIELD(vsk, write_notify_window) = PAGE_SIZE;
PKT_FIELD(vsk, write_notify_min_window) = PAGE_SIZE;
PKT_FIELD(vsk, peer_waiting_write) = false;
PKT_FIELD(vsk, peer_waiting_write_detected) = false;
}
static int
vmci_transport_notify_pkt_poll_in(struct sock *sk,
size_t target, bool *data_ready_now)
{
struct vsock_sock *vsk = vsock_sk(sk);
if (vsock_stream_has_data(vsk)) {
*data_ready_now = true;
} else {
/* We can't read right now because there is nothing in the
* queue. Ask for notifications when there is something to
* read.
*/
if (sk->sk_state == TCP_ESTABLISHED)
vsock_block_update_write_window(sk);
*data_ready_now = false;
}
return 0;
}
static int
vmci_transport_notify_pkt_poll_out(struct sock *sk,
size_t target, bool *space_avail_now)
{
s64 produce_q_free_space;
struct vsock_sock *vsk = vsock_sk(sk);
produce_q_free_space = vsock_stream_has_space(vsk);
if (produce_q_free_space > 0) {
*space_avail_now = true;
return 0;
} else if (produce_q_free_space == 0) {
/* This is a connected socket but we can't currently send data.
* Nothing else to do.
*/
*space_avail_now = false;
}
return 0;
}
static int
vmci_transport_notify_pkt_recv_init(
struct sock *sk,
size_t target,
struct vmci_transport_recv_notify_data *data)
{
struct vsock_sock *vsk = vsock_sk(sk);
data->consume_head = 0;
data->produce_tail = 0;
data->notify_on_block = false;
if (PKT_FIELD(vsk, write_notify_min_window) < target + 1) {
PKT_FIELD(vsk, write_notify_min_window) = target + 1;
if (PKT_FIELD(vsk, write_notify_window) <
PKT_FIELD(vsk, write_notify_min_window)) {
/* If the current window is smaller than the new
* minimal window size, we need to reevaluate whether
* we need to notify the sender. If the number of ready
* bytes are smaller than the new window, we need to
* send a notification to the sender before we block.
*/
PKT_FIELD(vsk, write_notify_window) =
PKT_FIELD(vsk, write_notify_min_window);
data->notify_on_block = true;
}
}
return 0;
}
static int
vmci_transport_notify_pkt_recv_pre_block(
struct sock *sk,
size_t target,
struct vmci_transport_recv_notify_data *data)
{
int err = 0;
vsock_block_update_write_window(sk);
if (data->notify_on_block) {
err = vmci_transport_send_read_notification(sk);
if (err < 0)
return err;
data->notify_on_block = false;
}
return err;
}
static int
vmci_transport_notify_pkt_recv_post_dequeue(
struct sock *sk,
size_t target,
ssize_t copied,
bool data_read,
struct vmci_transport_recv_notify_data *data)
{
struct vsock_sock *vsk;
int err;
bool was_full = false;
u64 free_space;
vsk = vsock_sk(sk);
err = 0;
if (data_read) {
smp_mb();
free_space =
vmci_qpair_consume_free_space(vmci_trans(vsk)->qpair);
was_full = free_space == copied;
if (was_full)
PKT_FIELD(vsk, peer_waiting_write) = true;
err = vmci_transport_send_read_notification(sk);
if (err < 0)
return err;
|