xref: /freebsd/sys/dev/nvd/nvd.c (revision ec84ecbb) 
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("$FreeBSD$");
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 
43 struct nvd_disk;
44 
45 static disk_ioctl_t nvd_ioctl;
46 static disk_strategy_t nvd_strategy;
47 
48 static void *nvd_new_disk(struct nvme_namespace *ns, void *ctrlr);
49 static void destroy_geom_disk(struct nvd_disk *ndisk);
50 
51 static void *nvd_new_controller(struct nvme_controller *ctrlr);
52 
53 static int nvd_load(void);
54 static void nvd_unload(void);
55 
56 MALLOC_DEFINE(M_NVD, "nvd", "nvd(4) allocations");
57 
58 struct nvme_consumer *consumer_handle;
59 
60 struct nvd_disk {
61 
62 	struct bio_queue_head	bioq;
63 	struct task		bioqtask;
64 	struct mtx		bioqlock;
65 
66 	struct disk		*disk;
67 	struct taskqueue	*tq;
68 	struct nvme_namespace	*ns;
69 
70 	uint32_t		cur_depth;
71 
72 	TAILQ_ENTRY(nvd_disk)	global_tailq;
73 	TAILQ_ENTRY(nvd_disk)	ctrlr_tailq;
74 };
75 
76 struct nvd_controller {
77 
78 	TAILQ_ENTRY(nvd_controller)	tailq;
79 	TAILQ_HEAD(, nvd_disk)		disk_head;
80 };
81 
82 static TAILQ_HEAD(, nvd_controller)	ctrlr_head;
83 static TAILQ_HEAD(disk_list, nvd_disk)	disk_head;
84 
85 static int nvd_modevent(module_t mod, int type, void *arg)
86 {
87 	int error = 0;
88 
89 	switch (type) {
90 	case MOD_LOAD:
91 		error = nvd_load();
92 		break;
93 	case MOD_UNLOAD:
94 		nvd_unload();
95 		break;
96 	default:
97 		break;
98 	}
99 
100 	return (error);
101 }
102 
103 moduledata_t nvd_mod = {
104 	"nvd",
105 	(modeventhand_t)nvd_modevent,
106 	0
107 };
108 
109 DECLARE_MODULE(nvd, nvd_mod, SI_SUB_DRIVERS, SI_ORDER_ANY);
110 MODULE_VERSION(nvd, 1);
111 MODULE_DEPEND(nvd, nvme, 1, 1, 1);
112 
113 static int
114 nvd_load()
115 {
116 
117 	TAILQ_INIT(&ctrlr_head);
118 	TAILQ_INIT(&disk_head);
119 
120 	consumer_handle = nvme_register_consumer(nvd_new_disk,
121 	    nvd_new_controller, NULL);
122 
123 	return (consumer_handle != NULL ? 0 : -1);
124 }
125 
126 static void
127 nvd_unload()
128 {
129 	struct nvd_controller	*ctrlr;
130 	struct nvd_disk		*disk;
131 
132 	while (!TAILQ_EMPTY(&ctrlr_head)) {
133 		ctrlr = TAILQ_FIRST(&ctrlr_head);
134 		TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq);
135 		free(ctrlr, M_NVD);
136 	}
137 
138 	while (!TAILQ_EMPTY(&disk_head)) {
139 		disk = TAILQ_FIRST(&disk_head);
140 		TAILQ_REMOVE(&disk_head, disk, global_tailq);
141 		destroy_geom_disk(disk);
142 		free(disk, M_NVD);
143 	}
144 
145 	nvme_unregister_consumer(consumer_handle);
146 }
147 
148 static void
149 nvd_strategy(struct bio *bp)
150 {
151 	struct nvd_disk *ndisk;
152 
153 	ndisk = (struct nvd_disk *)bp->bio_disk->d_drv1;
154 
155 	mtx_lock(&ndisk->bioqlock);
156 	bioq_insert_tail(&ndisk->bioq, bp);
157 	mtx_unlock(&ndisk->bioqlock);
158 	taskqueue_enqueue(ndisk->tq, &ndisk->bioqtask);
159 }
160 
161 static int
162 nvd_ioctl(struct disk *ndisk, u_long cmd, void *data, int fflag,
163     struct thread *td)
164 {
165 	int ret = 0;
166 
167 	switch (cmd) {
168 	default:
169 		ret = EIO;
170 	}
171 
172 	return (ret);
173 }
174 
175 static void
176 nvd_done(void *arg, const struct nvme_completion *cpl)
177 {
178 	struct bio *bp;
179 	struct nvd_disk *ndisk;
180 
181 	bp = (struct bio *)arg;
182 
183 	ndisk = bp->bio_disk->d_drv1;
184 
185 	atomic_add_int(&ndisk->cur_depth, -1);
186 
187 	/*
188 	 * TODO: add more extensive translation of NVMe status codes
189 	 *  to different bio error codes (i.e. EIO, EINVAL, etc.)
190 	 */
191 	if (nvme_completion_is_error(cpl)) {
192 		bp->bio_error = EIO;
193 		bp->bio_flags |= BIO_ERROR;
194 		bp->bio_resid = bp->bio_bcount;
195 	} else
196 		bp->bio_resid = 0;
197 
198 	biodone(bp);
199 }
200 
201 static void
202 nvd_bioq_process(void *arg, int pending)
203 {
204 	struct nvd_disk *ndisk = arg;
205 	struct bio *bp;
206 	int err;
207 
208 	for (;;) {
209 		mtx_lock(&ndisk->bioqlock);
210 		bp = bioq_takefirst(&ndisk->bioq);
211 		mtx_unlock(&ndisk->bioqlock);
212 		if (bp == NULL)
213 			break;
214 
215 #ifdef BIO_ORDERED
216 		/*
217 		 * BIO_ORDERED flag dictates that all outstanding bios
218 		 *  must be completed before processing the bio with
219 		 *  BIO_ORDERED flag set.
220 		 */
221 		if (bp->bio_flags & BIO_ORDERED) {
222 			while (ndisk->cur_depth > 0) {
223 				pause("nvd flush", 1);
224 			}
225 		}
226 #endif
227 
228 		bp->bio_driver1 = NULL;
229 		atomic_add_int(&ndisk->cur_depth, 1);
230 
231 		err = nvme_ns_bio_process(ndisk->ns, bp, nvd_done);
232 
233 		if (err) {
234 			atomic_add_int(&ndisk->cur_depth, -1);
235 			bp->bio_error = err;
236 			bp->bio_flags |= BIO_ERROR;
237 			bp->bio_resid = bp->bio_bcount;
238 			biodone(bp);
239 		}
240 
241 #ifdef BIO_ORDERED
242 		/*
243 		 * BIO_ORDERED flag dictates that the bio with BIO_ORDERED
244 		 *  flag set must be completed before proceeding with
245 		 *  additional bios.
246 		 */
247 		if (bp->bio_flags & BIO_ORDERED) {
248 			while (ndisk->cur_depth > 0) {
249 				pause("nvd flush", 1);
250 			}
251 		}
252 #endif
253 	}
254 }
255 
256 static void *
257 nvd_new_controller(struct nvme_controller *ctrlr)
258 {
259 	struct nvd_controller	*nvd_ctrlr;
260 
261 	nvd_ctrlr = malloc(sizeof(struct nvd_controller), M_NVD,
262 	    M_ZERO | M_NOWAIT);
263 
264 	TAILQ_INIT(&nvd_ctrlr->disk_head);
265 	TAILQ_INSERT_TAIL(&ctrlr_head, nvd_ctrlr, tailq);
266 
267 	return (nvd_ctrlr);
268 }
269 
270 static void *
271 nvd_new_disk(struct nvme_namespace *ns, void *ctrlr_arg)
272 {
273 	struct nvd_disk		*ndisk;
274 	struct disk		*disk;
275 	struct nvd_controller	*ctrlr = ctrlr_arg;
276 
277 	ndisk = malloc(sizeof(struct nvd_disk), M_NVD, M_ZERO | M_NOWAIT);
278 
279 	disk = disk_alloc();
280 	disk->d_strategy = nvd_strategy;
281 	disk->d_ioctl = nvd_ioctl;
282 	disk->d_name = "nvd";
283 	disk->d_drv1 = ndisk;
284 
285 	disk->d_maxsize = nvme_ns_get_max_io_xfer_size(ns);
286 	disk->d_sectorsize = nvme_ns_get_sector_size(ns);
287 	disk->d_mediasize = (off_t)nvme_ns_get_size(ns);
288 
289 	if (TAILQ_EMPTY(&disk_head))
290 		disk->d_unit = 0;
291 	else
292 		disk->d_unit =
293 		    TAILQ_LAST(&disk_head, disk_list)->disk->d_unit + 1;
294 
295 	disk->d_flags = 0;
296 
297 	if (nvme_ns_get_flags(ns) & NVME_NS_DEALLOCATE_SUPPORTED)
298 		disk->d_flags |= DISKFLAG_CANDELETE;
299 
300 	if (nvme_ns_get_flags(ns) & NVME_NS_FLUSH_SUPPORTED)
301 		disk->d_flags |= DISKFLAG_CANFLUSHCACHE;
302 
303 	strlcpy(disk->d_ident, nvme_ns_get_serial_number(ns),
304 	    sizeof(disk->d_ident));
305 
306 #if __FreeBSD_version >= 900034
307 	strlcpy(disk->d_descr, nvme_ns_get_model_number(ns),
308 	    sizeof(disk->d_descr));
309 #endif
310 
311 	disk_create(disk, DISK_VERSION);
312 
313 	ndisk->ns = ns;
314 	ndisk->disk = disk;
315 	ndisk->cur_depth = 0;
316 
317 	mtx_init(&ndisk->bioqlock, "NVD bioq lock", NULL, MTX_DEF);
318 	bioq_init(&ndisk->bioq);
319 
320 	TASK_INIT(&ndisk->bioqtask, 0, nvd_bioq_process, ndisk);
321 	ndisk->tq = taskqueue_create("nvd_taskq", M_WAITOK,
322 	    taskqueue_thread_enqueue, &ndisk->tq);
323 	taskqueue_start_threads(&ndisk->tq, 1, PI_DISK, "nvd taskq");
324 
325 	TAILQ_INSERT_TAIL(&disk_head, ndisk, global_tailq);
326 	TAILQ_INSERT_TAIL(&ctrlr->disk_head, ndisk, ctrlr_tailq);
327 
328 	return (NULL);
329 }
330 
331 static void
332 destroy_geom_disk(struct nvd_disk *ndisk)
333 {
334 	struct bio *bp;
335 
336 	taskqueue_free(ndisk->tq);
337 	disk_destroy(ndisk->disk);
338 
339 	mtx_lock(&ndisk->bioqlock);
340 	for (;;) {
341 		bp = bioq_takefirst(&ndisk->bioq);
342 		if (bp == NULL)
343 			break;
344 		bp->bio_error = EIO;
345 		bp->bio_flags |= BIO_ERROR;
346 		bp->bio_resid = bp->bio_bcount;
347 
348 		biodone(bp);
349 	}
350 	mtx_unlock(&ndisk->bioqlock);
351 
352 	mtx_destroy(&ndisk->bioqlock);
353 }
354