summaryrefslogtreecommitdiff
path: root/src/io_uring.c
blob: 21ce4f9c712aea30d8a676bcfc5286507147f95f (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
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include <unistd.h>
#include <errno.h>
#include <string.h>

#include "io_uring.h"
#include "liburing.h"
#include "barrier.h"

static int __io_uring_get_completion(int fd, struct io_uring_cq *cq,
				     struct io_uring_event **ev_ptr, int wait)
{
	const unsigned mask = *cq->kring_mask;
	unsigned head;
	int ret;

	*ev_ptr = NULL;
	head = *cq->khead;
	do {
		read_barrier();
		if (head != *cq->ktail) {
			*ev_ptr = &cq->events[head & mask];
			break;
		}
		if (!wait)
			break;
		ret = io_uring_enter(fd, 0, 1, IORING_ENTER_GETEVENTS);
		if (ret < 0)
			return -errno;
	} while (1);

	if (*ev_ptr) {
		*cq->khead = head + 1;
		write_barrier();
	}

	return 0;
}

/*
 * Return an IO completion, if one is readily available
 */
int io_uring_get_completion(int fd, struct io_uring_cq *cq,
			    struct io_uring_event **ev_ptr)
{
	return __io_uring_get_completion(fd, cq, ev_ptr, 0);
}

/*
 * Return an IO completion, waiting for it if necessary
 */
int io_uring_wait_completion(int fd, struct io_uring_cq *cq,
			     struct io_uring_event **ev_ptr)
{
	return __io_uring_get_completion(fd, cq, ev_ptr, 1);
}

/*
 * Submit iocbs acquired from io_uring_get_iocb() to the kernel.
 *
 * Returns number of iocbs submitted
 */
int io_uring_submit(int fd, struct io_uring_sq *sq)
{
	const unsigned mask = *sq->kring_mask;
	unsigned ktail, ktail_next, submitted;

	/*
	 * If we have pending IO in the kring, submit it first
	 */
	read_barrier();
	if (*sq->khead != *sq->ktail) {
		submitted = *sq->kring_entries;
		goto submit;
	}

	if (sq->iocb_head == sq->iocb_tail)
		return 0;

	/*
	 * Fill in iocbs that we have queued up, adding them to the kernel ring
	 */
	submitted = 0;
	ktail = ktail_next = *sq->ktail;
	while (sq->iocb_head < sq->iocb_tail) {
		ktail_next++;
		read_barrier();
		if (ktail_next == *sq->khead)
			break;

		sq->array[ktail & mask] = sq->iocb_head & mask;
		ktail = ktail_next;

		sq->iocb_head++;
		submitted++;
	}

	if (!submitted)
		return 0;

	if (*sq->ktail != ktail) {
		write_barrier();
		*sq->ktail = ktail;
		write_barrier();
	}

submit:
	return io_uring_enter(fd, submitted, 0, IORING_ENTER_GETEVENTS);
}

/*
 * Return an iocb to fill. Application must later call io_uring_submit()
 * when it's ready to tell the kernel about it. The caller may call this
 * function multiple times before calling io_uring_submit().
 *
 * Returns a vacant iocb, or NULL if we're full.
 */
struct io_uring_iocb *io_uring_get_iocb(struct io_uring_sq *sq)
{
	unsigned next = sq->iocb_tail + 1;
	struct io_uring_iocb *iocb;

	/*
	 * All iocbs are used
	 */
	if (next - sq->iocb_head > *sq->kring_entries)
		return NULL;

	iocb = &sq->iocbs[sq->iocb_tail & *sq->kring_mask];
	sq->iocb_tail = next;
	return iocb;
}

static int io_uring_mmap(int fd, struct io_uring_params *p,
			 struct io_uring_sq *sq, struct io_uring_cq *cq)
{
	size_t size;
	void *ptr;
	int ret;

	sq->ring_sz = p->sq_off.array + p->sq_entries * sizeof(unsigned);
	ptr = mmap(0, sq->ring_sz, PROT_READ | PROT_WRITE,
			MAP_SHARED | MAP_POPULATE, fd, IORING_OFF_SQ_RING);
	if (ptr == MAP_FAILED)
		return -errno;
	sq->khead = ptr + p->sq_off.head;
	sq->ktail = ptr + p->sq_off.tail;
	sq->kring_mask = ptr + p->sq_off.ring_mask;
	sq->kring_entries = ptr + p->sq_off.ring_entries;
	sq->kflags = ptr + p->sq_off.flags;
	sq->kdropped = ptr + p->sq_off.dropped;
	sq->array = ptr + p->sq_off.array;

	size = p->sq_entries * sizeof(struct io_uring_iocb);
	sq->iocbs = mmap(0, size, PROT_READ | PROT_WRITE,
				MAP_SHARED | MAP_POPULATE, fd,
				IORING_OFF_IOCB);
	if (sq->iocbs == MAP_FAILED) {
		ret = -errno;
err:
		munmap(sq->khead, sq->ring_sz);
		return ret;
	}

	cq->ring_sz = p->cq_off.events + p->cq_entries * sizeof(struct io_uring_event);
	ptr = mmap(0, cq->ring_sz, PROT_READ | PROT_WRITE,
			MAP_SHARED | MAP_POPULATE, fd, IORING_OFF_CQ_RING);
	if (ptr == MAP_FAILED) {
		ret = -errno;
		munmap(sq->iocbs, p->sq_entries * sizeof(struct io_uring_iocb));
		goto err;
	}
	cq->khead = ptr + p->cq_off.head;
	cq->ktail = ptr + p->cq_off.tail;
	cq->kring_mask = ptr + p->cq_off.ring_mask;
	cq->kring_entries = ptr + p->cq_off.ring_entries;
	cq->koverflow = ptr + p->cq_off.overflow;
	cq->events = ptr + p->cq_off.events;
	return fd;
}

/*
 * Returns -1 on error, or an 'fd' on success. On success, 'sq' and 'cq'
 * contain the necessary information to read/write to the rings.
 */
int io_uring_queue_init(unsigned entries, struct io_uring_params *p,
			struct iovec *iovecs, struct io_uring_sq *sq,
			struct io_uring_cq *cq)
{
	int fd;

	fd = io_uring_setup(entries, iovecs, p);
	if (fd < 0)
		return fd;

	memset(sq, 0, sizeof(*sq));
	memset(cq, 0, sizeof(*cq));
	return io_uring_mmap(fd, p, sq, cq);
}

void io_uring_queue_exit(int fd, struct io_uring_sq *sq, struct io_uring_cq *cq)
{
	munmap(sq->iocbs, *sq->kring_entries * sizeof(struct io_uring_iocb));
	munmap(sq->khead, sq->ring_sz);
	munmap(cq->khead, cq->ring_sz);
	close(fd);
}