xref: /dragonfly/sys/dev/virtual/vkernel/disk/vdisk.c (revision 9d626b29)
1 /*
2  * Copyright (c) 2006,2016 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Matthew Dillon <dillon@backplane.com>
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  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in
15  *    the documentation and/or other materials provided with the
16  *    distribution.
17  * 3. Neither the name of The DragonFly Project nor the names of its
18  *    contributors may be used to endorse or promote products derived
19  *    from this software without specific, prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 /*
37  * Virtual disk driver
38  */
39 #include <sys/types.h>
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/malloc.h>
44 #include <sys/conf.h>
45 #include <sys/bus.h>
46 #include <sys/buf.h>
47 #include <sys/devicestat.h>
48 #include <sys/disk.h>
49 #include <machine/cothread.h>
50 #include <machine/md_var.h>
51 
52 #include <sys/buf2.h>
53 
54 #include <sys/stat.h>
55 #include <unistd.h>
56 
57 struct vkd_softc {
58 	struct bio_queue_head bio_queue;
59 	struct devstat stats;
60 	struct disk disk;
61 	cothread_t	cotd;
62 	TAILQ_HEAD(, bio) cotd_queue;
63 	TAILQ_HEAD(, bio) cotd_done;
64 	cdev_t dev;
65 	int unit;
66 	int fd;
67 	int flags;
68 	off_t	size;		/* in bytes */
69 	char	*map_buf;	/* COW mode only */
70 };
71 
72 static void vkd_io_thread(cothread_t cotd);
73 static void vkd_io_intr(cothread_t cotd);
74 static void vkd_doio(struct vkd_softc *sc, struct bio *bio);
75 
76 static d_strategy_t	vkdstrategy;
77 static d_open_t		vkdopen;
78 
79 static struct dev_ops vkd_ops = {
80 	{ "vkd", 0, D_DISK },
81         .d_open =	vkdopen,
82         .d_close =	nullclose,
83         .d_read =	physread,
84         .d_write =	physwrite,
85         .d_strategy =	vkdstrategy,
86 };
87 
88 static void
89 vkdinit(void *dummy __unused)
90 {
91 	struct vkdisk_info *dsk;
92 	struct vkd_softc *sc;
93 	struct disk_info info;
94 	struct stat st;
95 	int i;
96 
97 	for (i = 0; i < DiskNum; i++) {
98 		/* check that the 'bus device' has been initialized */
99 		dsk = &DiskInfo[i];
100 		if (dsk == NULL || dsk->type != VKD_DISK)
101 			continue;
102 		if (dsk->fd < 0 || fstat(dsk->fd, &st) < 0)
103 			continue;
104 
105 		/*
106 		 * Devices may return a st_size of 0, try to use
107 		 * lseek.
108 		 */
109 		if (st.st_size == 0) {
110 			st.st_size = lseek(dsk->fd, 0L, SEEK_END);
111 			if (st.st_size == -1)
112 				st.st_size = 0;
113 		}
114 
115 		/* and create a new device */
116 		sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
117 		sc->unit = dsk->unit;
118 		sc->fd = dsk->fd;
119 		sc->size = st.st_size;
120 		sc->flags = dsk->flags;
121 		bioq_init(&sc->bio_queue);
122 		devstat_add_entry(&sc->stats, "vkd", sc->unit, DEV_BSIZE,
123 				  DEVSTAT_NO_ORDERED_TAGS,
124 				  DEVSTAT_TYPE_DIRECT | DEVSTAT_TYPE_IF_OTHER,
125 				  DEVSTAT_PRIORITY_DISK);
126 		sc->dev = disk_create(sc->unit, &sc->disk, &vkd_ops);
127 		sc->dev->si_drv1 = sc;
128 		sc->dev->si_iosize_max = min(MAXPHYS,256*1024);
129 
130 		/*
131 		 * Use a private mmap if COW mode is requested.
132 		 */
133 		if (sc->flags & 1) {
134 			sc->map_buf = mmap(NULL, sc->size,
135 					   PROT_READ|PROT_WRITE,
136 					   MAP_PRIVATE,
137 					   sc->fd, 0);
138 			if ((void *)sc->map_buf == MAP_FAILED) {
139 				panic("vkd: cannot mmap %jd MBytes\n",
140 				      (intmax_t)sc->size / (1024 * 1024));
141 			}
142 			kprintf("vkd%d: COW disk\n", sc->unit);
143 		}
144 
145 		TAILQ_INIT(&sc->cotd_queue);
146 		TAILQ_INIT(&sc->cotd_done);
147 		sc->cotd = cothread_create(vkd_io_thread, vkd_io_intr,
148 					   sc, "vkd");
149 
150 		bzero(&info, sizeof(info));
151 		info.d_media_blksize = DEV_BSIZE;
152 		info.d_media_blocks = st.st_size / info.d_media_blksize;
153 
154 		info.d_nheads = 1;
155 		info.d_ncylinders = 1;
156 		info.d_secpertrack = info.d_media_blocks;
157 		info.d_secpercyl = info.d_secpertrack * info.d_nheads;
158 
159 		if (dsk->serno) {
160 			info.d_serialno =
161 				kmalloc(SERNOLEN, M_TEMP, M_WAITOK | M_ZERO);
162 			strlcpy(info.d_serialno, dsk->serno, SERNOLEN);
163 		}
164 		disk_setdiskinfo(&sc->disk, &info);
165 
166 		/* Announce disk details */
167 		kprintf("vkd%d: <Virtual disk> ", i);
168 		if (info.d_serialno)
169 			kprintf("Serial Number %s", info.d_serialno);
170 		kprintf("\nvkd%d: %dMB (%ju %d byte sectors)\n",
171 			i, (int)(st.st_size / 1024 / 1024),
172 			info.d_media_blocks, info.d_media_blksize);
173 	}
174 }
175 
176 SYSINIT(vkdisk, SI_SUB_DRIVERS, SI_ORDER_FIRST, vkdinit, NULL);
177 
178 static int
179 vkdopen(struct dev_open_args *ap)
180 {
181 	struct vkd_softc *sc;
182 	/* struct disk_info info; */
183 	struct stat st;
184 	cdev_t dev;
185 
186 	dev = ap->a_head.a_dev;
187 	sc = dev->si_drv1;
188 	if (fstat(sc->fd, &st) < 0)
189 		return(ENXIO);
190 
191 	/*
192 	 * Devices may return a st_size of 0, try to use
193 	 * lseek.
194 	 */
195 	if (st.st_size == 0) {
196 		st.st_size = lseek(sc->fd, 0L, SEEK_END);
197 		if (st.st_size == -1)
198 			st.st_size = 0;
199 	}
200 	if (st.st_size == 0)
201 		return(ENXIO);
202 
203 /*
204 	bzero(&info, sizeof(info));
205 	info.d_media_blksize = DEV_BSIZE;
206 	info.d_media_blocks = st.st_size / info.d_media_blksize;
207 
208 	info.d_nheads = 1;
209 	info.d_ncylinders = 1;
210 	info.d_secpertrack = info.d_media_blocks;
211 	info.d_secpercyl = info.d_secpertrack * info.d_nheads;
212 
213 	disk_setdiskinfo(&sc->disk, &info); */
214 	return(0);
215 }
216 
217 static int
218 vkdstrategy(struct dev_strategy_args *ap)
219 {
220 	struct bio *bio = ap->a_bio;
221 	struct vkd_softc *sc;
222 	cdev_t dev;
223 
224 	dev = ap->a_head.a_dev;
225 	sc = dev->si_drv1;
226 
227 	devstat_start_transaction(&sc->stats);
228 	cothread_lock(sc->cotd, 0);
229 	TAILQ_INSERT_TAIL(&sc->cotd_queue, bio, bio_act);
230 	cothread_signal(sc->cotd);
231 	cothread_unlock(sc->cotd, 0);
232 
233 	return(0);
234 }
235 
236 static
237 void
238 vkd_io_intr(cothread_t cotd)
239 {
240 	struct vkd_softc *sc;
241 	struct bio *bio;
242 	TAILQ_HEAD(, bio) tmpq;
243 
244 	sc = cotd->arg;
245 
246 	/*
247 	 * We can't call back into the kernel while holding cothread!
248 	 */
249 	TAILQ_INIT(&tmpq);
250 	cothread_lock(cotd, 0);
251 	while ((bio = TAILQ_FIRST(&sc->cotd_done)) != NULL) {
252 		TAILQ_REMOVE(&sc->cotd_done, bio, bio_act);
253 		TAILQ_INSERT_TAIL(&tmpq, bio, bio_act);
254 	}
255 	cothread_unlock(cotd, 0);
256 
257 	while ((bio = TAILQ_FIRST(&tmpq)) != NULL) {
258 		TAILQ_REMOVE(&tmpq, bio, bio_act);
259 		devstat_end_transaction_buf(&sc->stats, bio->bio_buf);
260 		biodone(bio);
261 	}
262 }
263 
264 /*
265  * WARNING!  This runs as a cothread and has no access to mycpu nor can it
266  * make vkernel-specific calls other then cothread_*() calls.
267  *
268  * WARNING!  A signal can occur and be discarded prior to our initial
269  * call to cothread_lock().  Process pending I/O before waiting.
270  */
271 static
272 void
273 vkd_io_thread(cothread_t cotd)
274 {
275 	struct bio *bio;
276 	struct vkd_softc *sc = cotd->arg;
277 	int count;
278 
279 	cothread_lock(cotd, 1);
280 	for (;;) {
281 		count = 0;
282 		while ((bio = TAILQ_FIRST(&sc->cotd_queue)) != NULL) {
283 			TAILQ_REMOVE(&sc->cotd_queue, bio, bio_act);
284 			cothread_unlock(cotd, 1);
285 			vkd_doio(sc, bio);
286 			cothread_lock(cotd, 1);
287 			TAILQ_INSERT_TAIL(&sc->cotd_done, bio, bio_act);
288 			if (++count == 8) {
289 				cothread_intr(cotd);
290 				count = 0;
291 			}
292 		}
293 		if (count)
294 			cothread_intr(cotd);
295 		cothread_wait(cotd);	/* interlocks cothread lock */
296 	}
297 	/* NOT REACHED */
298 	cothread_unlock(cotd, 1);
299 }
300 
301 static
302 void
303 vkd_doio(struct vkd_softc *sc, struct bio *bio)
304 {
305 	struct buf *bp = bio->bio_buf;
306 	int n;
307 
308 	switch(bp->b_cmd) {
309 	case BUF_CMD_READ:
310 		if (sc->map_buf) {
311 			bcopy(sc->map_buf + bio->bio_offset,
312 			      bp->b_data,
313 			      bp->b_bcount);
314 			n = bp->b_bcount;
315 		} else {
316 			n = pread(sc->fd, bp->b_data, bp->b_bcount,
317 				  bio->bio_offset);
318 		}
319 		break;
320 	case BUF_CMD_WRITE:
321 		/* XXX HANDLE SHORT WRITE XXX */
322 		if (sc->map_buf) {
323 			bcopy(bp->b_data,
324 			      sc->map_buf + bio->bio_offset,
325 			      bp->b_bcount);
326 			n = bp->b_bcount;
327 		} else {
328 			n = pwrite(sc->fd, bp->b_data, bp->b_bcount,
329 				   bio->bio_offset);
330 		}
331 		break;
332 	case BUF_CMD_FLUSH:
333 		if (sc->map_buf == NULL && fsync(sc->fd) < 0)
334 			n = -1;
335 		else
336 			n = bp->b_bcount;
337 		break;
338 	default:
339 		panic("vkd: bad b_cmd %d", bp->b_cmd);
340 		break; /* not reached */
341 	}
342 	if (bp->b_bcount != n) {
343 		bp->b_error = EIO;
344 		bp->b_flags |= B_ERROR;
345 	}
346 	bp->b_resid = bp->b_bcount - n;
347 }
348