1/*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (C) 2012-2013 Intel Corporation
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29#include <sys/param.h>
30#include <sys/bio.h>
31#include <sys/conf.h>
32#include <sys/fcntl.h>
33#include <sys/kthread.h>
34#include <sys/malloc.h>
35#include <sys/module.h>
36#include <sys/proc.h>
37#include <sys/syscallsubr.h>
38#include <sys/sysctl.h>
39#include <sys/sysproto.h>
40#include <sys/systm.h>
41#include <sys/unistd.h>
42
43#include <geom/geom.h>
44
45#include "nvme_private.h"
46
47struct nvme_io_test_thread {
48	uint32_t		idx;
49	struct nvme_namespace	*ns;
50	enum nvme_nvm_opcode	opc;
51	struct timeval		start;
52	void			*buf;
53	uint32_t		size;
54	uint32_t		time;
55	uint64_t		io_completed;
56};
57
58struct nvme_io_test_internal {
59	struct nvme_namespace	*ns;
60	enum nvme_nvm_opcode	opc;
61	struct timeval		start;
62	uint32_t		time;
63	uint32_t		size;
64	uint32_t		td_active;
65	uint32_t		td_idx;
66	uint32_t		flags;
67	uint64_t		io_completed[NVME_TEST_MAX_THREADS];
68};
69
70static void
71nvme_ns_bio_test_cb(struct bio *bio)
72{
73	struct mtx *mtx;
74
75	mtx = mtx_pool_find(mtxpool_sleep, bio);
76	mtx_lock(mtx);
77	wakeup(bio);
78	mtx_unlock(mtx);
79}
80
81static void
82nvme_ns_bio_test(void *arg)
83{
84	struct nvme_io_test_internal	*io_test = arg;
85	struct cdevsw			*csw;
86	struct mtx			*mtx;
87	struct bio			*bio;
88	struct cdev			*dev;
89	void				*buf;
90	struct timeval			t;
91	uint64_t			io_completed = 0, offset;
92	uint32_t			idx;
93	int				ref;
94
95	buf = malloc(io_test->size, M_NVME, M_WAITOK);
96	idx = atomic_fetchadd_int(&io_test->td_idx, 1);
97	dev = io_test->ns->cdev;
98
99	offset = idx * 2048ULL * nvme_ns_get_sector_size(io_test->ns);
100
101	while (1) {
102		bio = g_alloc_bio();
103
104		memset(bio, 0, sizeof(*bio));
105		bio->bio_cmd = (io_test->opc == NVME_OPC_READ) ?
106		    BIO_READ : BIO_WRITE;
107		bio->bio_done = nvme_ns_bio_test_cb;
108		bio->bio_dev = dev;
109		bio->bio_offset = offset;
110		bio->bio_data = buf;
111		bio->bio_bcount = io_test->size;
112
113		if (io_test->flags & NVME_TEST_FLAG_REFTHREAD) {
114			csw = dev_refthread(dev, &ref);
115		} else
116			csw = dev->si_devsw;
117
118		if (csw == NULL)
119			panic("Unable to retrieve device switch");
120		mtx = mtx_pool_find(mtxpool_sleep, bio);
121		mtx_lock(mtx);
122		(*csw->d_strategy)(bio);
123		msleep(bio, mtx, PRIBIO, "biotestwait", 0);
124		mtx_unlock(mtx);
125
126		if (io_test->flags & NVME_TEST_FLAG_REFTHREAD) {
127			dev_relthread(dev, ref);
128		}
129
130		if ((bio->bio_flags & BIO_ERROR) || (bio->bio_resid > 0))
131			break;
132
133		g_destroy_bio(bio);
134
135		io_completed++;
136
137		getmicrouptime(&t);
138		timevalsub(&t, &io_test->start);
139
140		if (t.tv_sec >= io_test->time)
141			break;
142
143		offset += io_test->size;
144		if ((offset + io_test->size) > nvme_ns_get_size(io_test->ns))
145			offset = 0;
146	}
147
148	io_test->io_completed[idx] = io_completed;
149	wakeup_one(io_test);
150
151	free(buf, M_NVME);
152
153	atomic_subtract_int(&io_test->td_active, 1);
154	mb();
155
156	kthread_exit();
157}
158
159static void
160nvme_ns_io_test_cb(void *arg, const struct nvme_completion *cpl)
161{
162	struct nvme_io_test_thread	*tth = arg;
163	struct timeval			t;
164
165	tth->io_completed++;
166
167	if (nvme_completion_is_error(cpl)) {
168		printf("%s: error occurred\n", __func__);
169		wakeup_one(tth);
170		return;
171	}
172
173	getmicrouptime(&t);
174	timevalsub(&t, &tth->start);
175
176	if (t.tv_sec >= tth->time) {
177		wakeup_one(tth);
178		return;
179	}
180
181	switch (tth->opc) {
182	case NVME_OPC_WRITE:
183		nvme_ns_cmd_write(tth->ns, tth->buf, tth->idx * 2048,
184		    tth->size/nvme_ns_get_sector_size(tth->ns),
185		    nvme_ns_io_test_cb, tth);
186		break;
187	case NVME_OPC_READ:
188		nvme_ns_cmd_read(tth->ns, tth->buf, tth->idx * 2048,
189		    tth->size/nvme_ns_get_sector_size(tth->ns),
190		    nvme_ns_io_test_cb, tth);
191		break;
192	default:
193		break;
194	}
195}
196
197static void
198nvme_ns_io_test(void *arg)
199{
200	struct nvme_io_test_internal	*io_test = arg;
201	struct nvme_io_test_thread	*tth;
202	struct nvme_completion		cpl;
203	int				error;
204
205	tth = malloc(sizeof(*tth), M_NVME, M_WAITOK | M_ZERO);
206	tth->ns = io_test->ns;
207	tth->opc = io_test->opc;
208	memcpy(&tth->start, &io_test->start, sizeof(tth->start));
209	tth->buf = malloc(io_test->size, M_NVME, M_WAITOK);
210	tth->size = io_test->size;
211	tth->time = io_test->time;
212	tth->idx = atomic_fetchadd_int(&io_test->td_idx, 1);
213
214	memset(&cpl, 0, sizeof(cpl));
215
216	nvme_ns_io_test_cb(tth, &cpl);
217
218	error = tsleep(tth, 0, "test_wait", tth->time*hz*2);
219
220	if (error)
221		printf("%s: error = %d\n", __func__, error);
222
223	io_test->io_completed[tth->idx] = tth->io_completed;
224	wakeup_one(io_test);
225
226	free(tth->buf, M_NVME);
227	free(tth, M_NVME);
228
229	atomic_subtract_int(&io_test->td_active, 1);
230	mb();
231
232	kthread_exit();
233}
234
235void
236nvme_ns_test(struct nvme_namespace *ns, u_long cmd, caddr_t arg)
237{
238	struct nvme_io_test		*io_test;
239	struct nvme_io_test_internal	*io_test_internal;
240	void				(*fn)(void *);
241	int				i;
242
243	io_test = (struct nvme_io_test *)arg;
244
245	if ((io_test->opc != NVME_OPC_READ) &&
246	    (io_test->opc != NVME_OPC_WRITE))
247		return;
248
249	if (io_test->size % nvme_ns_get_sector_size(ns))
250		return;
251
252	io_test_internal = malloc(sizeof(*io_test_internal), M_NVME,
253	    M_WAITOK | M_ZERO);
254	io_test_internal->opc = io_test->opc;
255	io_test_internal->ns = ns;
256	io_test_internal->td_active = io_test->num_threads;
257	io_test_internal->time = io_test->time;
258	io_test_internal->size = io_test->size;
259	io_test_internal->flags = io_test->flags;
260
261	if (cmd == NVME_IO_TEST)
262		fn = nvme_ns_io_test;
263	else
264		fn = nvme_ns_bio_test;
265
266	getmicrouptime(&io_test_internal->start);
267
268	for (i = 0; i < io_test->num_threads; i++)
269		kthread_add(fn, io_test_internal,
270		    NULL, NULL, 0, 0, "nvme_io_test[%d]", i);
271
272	tsleep(io_test_internal, 0, "nvme_test", io_test->time * 2 * hz);
273
274	while (io_test_internal->td_active > 0)
275		DELAY(10);
276
277	memcpy(io_test->io_completed, io_test_internal->io_completed,
278	    sizeof(io_test->io_completed));
279
280	free(io_test_internal, M_NVME);
281}
282