summaryrefslogtreecommitdiff
path: root/source3/modules/vfs_aio_linux.c
blob: b685cdcad2f7e30ab38ea362341f63e2444dbc41 (plain)
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
/*
 * Simulate Posix AIO using Linux kernel AIO.
 *
 * Copyright (C) Jeremy Allison 2012
 * Copyright (C) Volker Lendecke 2012
 *
 * 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; either version 3 of the License, or
 * (at your option) any 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.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#include "includes.h"
#include "system/filesys.h"
#include "smbd/smbd.h"
#include "smbd/globals.h"
#include "lib/util/tevent_unix.h"
#include <sys/eventfd.h>
#include <libaio.h>

static int event_fd = -1;
static io_context_t io_ctx;
static struct fd_event *aio_read_event;
static bool used;
static unsigned num_busy;

static void aio_linux_done(struct event_context *event_ctx,
			   struct fd_event *event,
			   uint16 flags, void *private_data);

/************************************************************************
 Housekeeping. Cleanup if no activity for 30 seconds.
***********************************************************************/

static void aio_linux_housekeeping(struct tevent_context *event_ctx,
                                        struct tevent_timer *te,
                                        struct timeval now,
                                        void *private_data)
{
	/* Remove this timed event handler. */
	TALLOC_FREE(te);

	if ((num_busy != 0) || used) {
		used = false;

		/* Still busy. Look again in 30 seconds. */
		(void)tevent_add_timer(event_ctx,
					NULL,
					timeval_current_ofs(30, 0),
					aio_linux_housekeeping,
					NULL);
		return;
	}

	/* No activity for 30 seconds. Close out kernel resources. */
	io_queue_release(io_ctx);
	memset(&io_ctx, '\0', sizeof(io_ctx));

	if (event_fd != -1) {
		close(event_fd);
		event_fd = -1;
	}

	TALLOC_FREE(aio_read_event);
}

/************************************************************************
 Ensure event fd and aio context are initialized.
***********************************************************************/

static bool init_aio_linux(struct vfs_handle_struct *handle)
{
	struct tevent_timer *te = NULL;

	if (event_fd != -1) {
		/* Already initialized. */
		return true;
	}

	/* Schedule a shutdown event for 30 seconds from now. */
	te = tevent_add_timer(handle->conn->sconn->ev_ctx,
				NULL,
				timeval_current_ofs(30, 0),
				aio_linux_housekeeping,
				NULL);

	if (te == NULL) {
		goto fail;
	}

	event_fd = eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC);
	if (event_fd == -1) {
		goto fail;
	}

	aio_read_event = tevent_add_fd(server_event_context(),
				NULL,
				event_fd,
				TEVENT_FD_READ,
				aio_linux_done,
				NULL);
	if (aio_read_event == NULL) {
		goto fail;
	}

	if (io_queue_init(aio_pending_size, &io_ctx)) {
		goto fail;
	}

	DEBUG(10,("init_aio_linux: initialized with up to %d events\n",
		  aio_pending_size));

	return true;

  fail:

	DEBUG(10,("init_aio_linux: initialization failed\n"));

	TALLOC_FREE(te);
	TALLOC_FREE(aio_read_event);
	if (event_fd != -1) {
		close(event_fd);
		event_fd = -1;
	}
	memset(&io_ctx, '\0', sizeof(io_ctx));
	return false;
}

struct aio_linux_state {
	struct iocb event_iocb;
	ssize_t ret;
	int err;
};

static struct tevent_req *aio_linux_pread_send(
	struct vfs_handle_struct *handle, TALLOC_CTX *mem_ctx,
	struct tevent_context *ev, struct files_struct *fsp,
	void *data, size_t n, off_t offset)
{
	struct tevent_req *req;
	struct aio_linux_state *state;
	struct iocb *piocb;
	int ret;

	req = tevent_req_create(mem_ctx, &state, struct aio_linux_state);
	if (req == NULL) {
		return NULL;
	}
	if (!init_aio_linux(handle)) {
		tevent_req_error(req, EIO);
		return tevent_req_post(req, ev);
	}

	io_prep_pread(&state->event_iocb, fsp->fh->fd, data, n, offset);
	io_set_eventfd(&state->event_iocb, event_fd);
	state->event_iocb.data = req;

	piocb = &state->event_iocb;

	ret = io_submit(io_ctx, 1, &piocb);
	if (ret < 0) {
		tevent_req_error(req, -ret);
		return tevent_req_post(req, ev);
	}
	num_busy += 1;
	used = true;
	return req;
}

static struct tevent_req *aio_linux_pwrite_send(
	struct vfs_handle_struct *handle, TALLOC_CTX *mem_ctx,
	struct tevent_context *ev, struct files_struct *fsp,
	const void *data, size_t n, off_t offset)
{
	struct tevent_req *req;
	struct aio_linux_state *state;
	struct iocb *piocb;
	int ret;

	req = tevent_req_create(mem_ctx, &state, struct aio_linux_state);
	if (req == NULL) {
		return NULL;
	}
	if (!init_aio_linux(handle)) {
		tevent_req_error(req, EIO);
		return tevent_req_post(req, ev);
	}

	io_prep_pwrite(&state->event_iocb, fsp->fh->fd, discard_const(data),
		       n, offset);
	io_set_eventfd(&state->event_iocb, event_fd);
	state->event_iocb.data = req;

	piocb = &state->event_iocb;

	ret = io_submit(io_ctx, 1, &piocb);
	if (ret < 0) {
		tevent_req_error(req, -ret);
		return tevent_req_post(req, ev);
	}
	num_busy += 1;
	used = true;
	return req;
}

static void aio_linux_done(struct event_context *event_ctx,
			   struct fd_event *event,
			   uint16 flags, void *private_data)
{
	uint64_t num_events = 0;

	DEBUG(10, ("aio_linux_done called with flags=%d\n",
		   (int)flags));

	/* Read the number of events available. */
	if (sys_read(event_fd, &num_events, sizeof(num_events)) !=
			sizeof(num_events)) {
		smb_panic("aio_linux_handle_completion: invalid read");
	}

	while (num_events > 0) {
		struct timespec ts = { 0, };
		struct io_event finished;
		struct tevent_req *req;
		struct aio_linux_state *state;
		int ret;

		ret = io_getevents(io_ctx, 1, 1, &finished, &ts);
		if (ret < 0) {
			DEBUG(1, ("aio_linux_done: io_getevents returned %s\n",
				  strerror(-ret)));
			return;
		}
		if (ret == 0) {
			DEBUG(10, ("aio_linux_done: io_getvents returned "
				   "0\n"));
			continue;
		}

		num_busy -= 1;

		req = talloc_get_type_abort(finished.data,
					    struct tevent_req);
		state = tevent_req_data(req, struct aio_linux_state);

		if (finished.res < 0) {
			state->ret = -1;
			state->err = -finished.res;
		} else {
			state->ret = finished.res;
			state->err = 0;
		}
		tevent_req_done(req);
		num_events -= 1;
	}
}

static ssize_t aio_linux_recv(struct tevent_req *req, int *err)
{
	struct aio_linux_state *state = tevent_req_data(
		req, struct aio_linux_state);

	if (tevent_req_is_unix_error(req, err)) {
		return -1;
	}
	if (state->ret == -1) {
		*err = state->err;
	}
	return state->ret;
}

static int aio_linux_connect(vfs_handle_struct *handle, const char *service,
			       const char *user)
{
	/*********************************************************************
	 * How many io_events to initialize ?
	 * 128 per process seems insane as a default until you realize that
	 * (a) Throttling is done in SMB2 via the crediting algorithm.
	 * (b) SMB1 clients are limited to max_mux (50) outstanding
	 *     requests and Windows clients don't use this anyway.
	 * Essentially we want this to be unlimited unless smb.conf
	 * says different.
	 *********************************************************************/
	aio_pending_size = lp_parm_int(
		SNUM(handle->conn), "aio_linux", "aio num events", 128);
	return SMB_VFS_NEXT_CONNECT(handle, service, user);
}

static struct vfs_fn_pointers vfs_aio_linux_fns = {
	.connect_fn = aio_linux_connect,
	.pread_send_fn = aio_linux_pread_send,
	.pread_recv_fn = aio_linux_recv,
	.pwrite_send_fn = aio_linux_pwrite_send,
	.pwrite_recv_fn = aio_linux_recv,
};

NTSTATUS vfs_aio_linux_init(void)
{
	return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
				"aio_linux", &vfs_aio_linux_fns);
}