nvd.c revision 240616 
1/*-
2 * Copyright (C) 2012 Intel Corporation
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD: head/sys/dev/nvd/nvd.c 240616 2012-09-17 19:23:01Z jimharris $");
29
30#include <sys/param.h>
31#include <sys/bio.h>
32#include <sys/kernel.h>
33#include <sys/malloc.h>
34#include <sys/module.h>
35#include <sys/systm.h>
36#include <sys/taskqueue.h>
37
38#include <geom/geom.h>
39#include <geom/geom_disk.h>
40
41#include <dev/nvme/nvme.h>
42
43struct nvd_disk;
44
45static disk_ioctl_t nvd_ioctl;
46static disk_strategy_t nvd_strategy;
47
48static void create_geom_disk(void *, struct nvme_namespace *ns);
49static void destroy_geom_disk(struct nvd_disk *ndisk);
50
51static int nvd_load(void);
52static void nvd_unload(void);
53
54MALLOC_DEFINE(M_NVD, "nvd", "nvd(4) allocations");
55
56struct nvme_consumer *consumer_handle;
57
58struct nvd_disk {
59
60	struct bio_queue_head	bioq;
61	struct task		bioqtask;
62	struct mtx		bioqlock;
63
64	struct disk		*disk;
65	struct taskqueue	*tq;
66	struct nvme_namespace	*ns;
67
68	uint32_t		cur_depth;
69
70	TAILQ_ENTRY(nvd_disk)	tailq;
71};
72
73TAILQ_HEAD(, nvd_disk)	nvd_head;
74
75static int nvd_modevent(module_t mod, int type, void *arg)
76{
77	int error = 0;
78
79	switch (type) {
80	case MOD_LOAD:
81		error = nvd_load();
82		break;
83	case MOD_UNLOAD:
84		nvd_unload();
85		break;
86	default:
87		break;
88	}
89
90	return (error);
91}
92
93moduledata_t nvd_mod = {
94	"nvd",
95	(modeventhand_t)nvd_modevent,
96	0
97};
98
99DECLARE_MODULE(nvd, nvd_mod, SI_SUB_DRIVERS, SI_ORDER_ANY);
100MODULE_VERSION(nvd, 1);
101MODULE_DEPEND(nvd, nvme, 1, 1, 1);
102
103static int
104nvd_load()
105{
106
107	TAILQ_INIT(&nvd_head);
108	consumer_handle = nvme_register_consumer(create_geom_disk, NULL);
109
110	return (consumer_handle != NULL ? 0 : -1);
111}
112
113static void
114nvd_unload()
115{
116	struct nvd_disk *nvd;
117
118	while (!TAILQ_EMPTY(&nvd_head)) {
119		nvd = TAILQ_FIRST(&nvd_head);
120		TAILQ_REMOVE(&nvd_head, nvd, tailq);
121		destroy_geom_disk(nvd);
122		free(nvd, M_NVD);
123	}
124
125	nvme_unregister_consumer(consumer_handle);
126}
127
128static void
129nvd_strategy(struct bio *bp)
130{
131	struct nvd_disk *ndisk;
132
133	ndisk = (struct nvd_disk *)bp->bio_disk->d_drv1;
134
135	mtx_lock(&ndisk->bioqlock);
136	bioq_insert_tail(&ndisk->bioq, bp);
137	mtx_unlock(&ndisk->bioqlock);
138	taskqueue_enqueue(ndisk->tq, &ndisk->bioqtask);
139}
140
141static int
142nvd_ioctl(struct disk *ndisk, u_long cmd, void *data, int fflag,
143    struct thread *td)
144{
145	int ret = 0;
146
147	switch (cmd) {
148	default:
149		ret = EIO;
150	}
151
152	return (ret);
153}
154
155static void
156nvd_done(void *arg, const struct nvme_completion *status)
157{
158	struct bio *bp;
159	struct nvd_disk *ndisk;
160
161	bp = (struct bio *)arg;
162
163	ndisk = bp->bio_disk->d_drv1;
164
165	if (atomic_fetchadd_int(&ndisk->cur_depth, -1) == NVME_QD)
166		taskqueue_enqueue(ndisk->tq, &ndisk->bioqtask);
167
168	/*
169	 * TODO: add more extensive translation of NVMe status codes
170	 *  to different bio error codes (i.e. EIO, EINVAL, etc.)
171	 */
172	if (status->sf_sc || status->sf_sct) {
173		bp->bio_error = EIO;
174		bp->bio_flags |= BIO_ERROR;
175		bp->bio_resid = bp->bio_bcount;
176	} else
177		bp->bio_resid = 0;
178
179	biodone(bp);
180}
181
182static void
183nvd_bioq_process(void *arg, int pending)
184{
185	struct nvd_disk *ndisk = arg;
186	struct bio *bp;
187	int err;
188
189	for (;;) {
190		if (atomic_load_acq_int(&ndisk->cur_depth) >= NVME_QD)
191			break;
192
193		mtx_lock(&ndisk->bioqlock);
194		bp = bioq_takefirst(&ndisk->bioq);
195		mtx_unlock(&ndisk->bioqlock);
196		if (bp == NULL)
197			break;
198
199#ifdef BIO_ORDERED
200		/*
201		 * BIO_ORDERED flag dictates that all outstanding bios
202		 *  must be completed before processing the bio with
203		 *  BIO_ORDERED flag set.
204		 */
205		if (bp->bio_flags & BIO_ORDERED) {
206			while (ndisk->cur_depth > 0) {
207				pause("nvd flush", 1);
208			}
209		}
210#endif
211
212		bp->bio_driver1 = NULL;
213		atomic_add_acq_int(&ndisk->cur_depth, 1);
214
215		err = nvme_ns_bio_process(ndisk->ns, bp, nvd_done);
216
217		if (err) {
218			atomic_add_acq_int(&ndisk->cur_depth, -1);
219			bp->bio_error = EIO;
220			bp->bio_flags |= BIO_ERROR;
221			bp->bio_resid = bp->bio_bcount;
222			biodone(bp);
223		}
224
225#ifdef BIO_ORDERED
226		/*
227		 * BIO_ORDERED flag dictates that the bio with BIO_ORDERED
228		 *  flag set must be completed before proceeding with
229		 *  additional bios.
230		 */
231		if (bp->bio_flags & BIO_ORDERED) {
232			while (ndisk->cur_depth > 0) {
233				pause("nvd flush", 1);
234			}
235		}
236#endif
237	}
238}
239
240static void
241create_geom_disk(void *arg, struct nvme_namespace *ns)
242{
243	struct nvd_disk *ndisk;
244	struct disk *disk;
245
246	ndisk = malloc(sizeof(struct nvd_disk), M_NVD, M_ZERO | M_NOWAIT);
247
248	disk = disk_alloc();
249	disk->d_strategy = nvd_strategy;
250	disk->d_ioctl = nvd_ioctl;
251	disk->d_name = "nvd";
252	disk->d_drv1 = ndisk;
253
254	disk->d_maxsize = nvme_ns_get_max_io_xfer_size(ns);
255	disk->d_sectorsize = nvme_ns_get_sector_size(ns);
256	disk->d_mediasize = (off_t)nvme_ns_get_size(ns);
257
258	if (TAILQ_EMPTY(&nvd_head))
259		disk->d_unit = 0;
260	else
261		disk->d_unit = TAILQ_FIRST(&nvd_head)->disk->d_unit + 1;
262
263	disk->d_flags = 0;
264
265	if (nvme_ns_get_flags(ns) & NVME_NS_DEALLOCATE_SUPPORTED)
266		disk->d_flags |= DISKFLAG_CANDELETE;
267
268	if (nvme_ns_get_flags(ns) & NVME_NS_FLUSH_SUPPORTED)
269		disk->d_flags |= DISKFLAG_CANFLUSHCACHE;
270
271	strlcpy(disk->d_ident, nvme_ns_get_serial_number(ns),
272	    sizeof(disk->d_ident));
273
274#if __FreeBSD_version >= 900034
275	strlcpy(disk->d_descr, nvme_ns_get_model_number(ns),
276	    sizeof(disk->d_descr));
277#endif
278
279	disk_create(disk, DISK_VERSION);
280
281	ndisk->ns = ns;
282	ndisk->disk = disk;
283	ndisk->cur_depth = 0;
284
285	mtx_init(&ndisk->bioqlock, "NVD bioq lock", NULL, MTX_DEF);
286	bioq_init(&ndisk->bioq);
287
288	TASK_INIT(&ndisk->bioqtask, 0, nvd_bioq_process, ndisk);
289	ndisk->tq = taskqueue_create("nvd_taskq", M_WAITOK,
290	    taskqueue_thread_enqueue, &ndisk->tq);
291	taskqueue_start_threads(&ndisk->tq, 1, PI_DISK, "nvd taskq");
292
293	TAILQ_INSERT_HEAD(&nvd_head, ndisk, tailq);
294}
295
296static void
297destroy_geom_disk(struct nvd_disk *ndisk)
298{
299	struct bio *bp;
300
301	taskqueue_free(ndisk->tq);
302	disk_destroy(ndisk->disk);
303
304	mtx_lock(&ndisk->bioqlock);
305	for (;;) {
306		bp = bioq_takefirst(&ndisk->bioq);
307		if (bp == NULL)
308			break;
309		bp->bio_error = EIO;
310		bp->bio_flags |= BIO_ERROR;
311		bp->bio_resid = bp->bio_bcount;
312
313		biodone(bp);
314	}
315	mtx_unlock(&ndisk->bioqlock);
316
317	mtx_destroy(&ndisk->bioqlock);
318}
319