xref: /dragonfly/sys/dev/virtual/vkernel/disk/vdisk.c (revision 52f9f0d9)
1 /*
2  * Copyright (c) 2006 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 };
68 
69 static void vkd_io_thread(cothread_t cotd);
70 static void vkd_io_intr(cothread_t cotd);
71 static void vkd_doio(struct vkd_softc *sc, struct bio *bio);
72 
73 static d_strategy_t	vkdstrategy;
74 static d_open_t		vkdopen;
75 
76 static struct dev_ops vkd_ops = {
77 	{ "vkd", 0, D_DISK },
78         .d_open =	vkdopen,
79         .d_close =	nullclose,
80         .d_read =	physread,
81         .d_write =	physwrite,
82         .d_strategy =	vkdstrategy,
83 };
84 
85 static void
86 vkdinit(void *dummy __unused)
87 {
88 	struct vkdisk_info *dsk;
89 	struct vkd_softc *sc;
90 	struct disk_info info;
91 	struct stat st;
92 	int i;
93 
94 	for (i = 0; i < DiskNum; i++) {
95 		/* check that the 'bus device' has been initialized */
96 		dsk = &DiskInfo[i];
97 		if (dsk == NULL || dsk->type != VKD_DISK)
98 			continue;
99 		if (dsk->fd < 0 || fstat(dsk->fd, &st) < 0)
100 			continue;
101 
102 		/*
103 		 * Devices may return a st_size of 0, try to use
104 		 * lseek.
105 		 */
106 		if (st.st_size == 0) {
107 			st.st_size = lseek(dsk->fd, 0L, SEEK_END);
108 			if (st.st_size == -1)
109 				st.st_size = 0;
110 		}
111 
112 		/* and create a new device */
113 		sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
114 		sc->unit = dsk->unit;
115 		sc->fd = dsk->fd;
116 		bioq_init(&sc->bio_queue);
117 		devstat_add_entry(&sc->stats, "vkd", sc->unit, DEV_BSIZE,
118 				  DEVSTAT_NO_ORDERED_TAGS,
119 				  DEVSTAT_TYPE_DIRECT | DEVSTAT_TYPE_IF_OTHER,
120 				  DEVSTAT_PRIORITY_DISK);
121 		sc->dev = disk_create(sc->unit, &sc->disk, &vkd_ops);
122 		sc->dev->si_drv1 = sc;
123 		sc->dev->si_iosize_max = min(MAXPHYS,256*1024);
124 
125 		TAILQ_INIT(&sc->cotd_queue);
126 		TAILQ_INIT(&sc->cotd_done);
127 		sc->cotd = cothread_create(vkd_io_thread, vkd_io_intr, sc,
128 					   "vkd");
129 
130 		bzero(&info, sizeof(info));
131 		info.d_media_blksize = DEV_BSIZE;
132 		info.d_media_blocks = st.st_size / info.d_media_blksize;
133 
134 		info.d_nheads = 1;
135 		info.d_ncylinders = 1;
136 		info.d_secpertrack = info.d_media_blocks;
137 		info.d_secpercyl = info.d_secpertrack * info.d_nheads;
138 
139 		disk_setdiskinfo(&sc->disk, &info);
140 	}
141 }
142 
143 SYSINIT(vkdisk, SI_SUB_DRIVERS, SI_ORDER_FIRST, vkdinit, NULL);
144 
145 static int
146 vkdopen(struct dev_open_args *ap)
147 {
148 	struct vkd_softc *sc;
149 	/* struct disk_info info; */
150 	struct stat st;
151 	cdev_t dev;
152 
153 	dev = ap->a_head.a_dev;
154 	sc = dev->si_drv1;
155 	if (fstat(sc->fd, &st) < 0)
156 		return(ENXIO);
157 
158 	/*
159 	 * Devices may return a st_size of 0, try to use
160 	 * lseek.
161 	 */
162 	if (st.st_size == 0) {
163 		st.st_size = lseek(sc->fd, 0L, SEEK_END);
164 		if (st.st_size == -1)
165 			st.st_size = 0;
166 	}
167 	if (st.st_size == 0)
168 		return(ENXIO);
169 
170 /*
171 	bzero(&info, sizeof(info));
172 	info.d_media_blksize = DEV_BSIZE;
173 	info.d_media_blocks = st.st_size / info.d_media_blksize;
174 
175 	info.d_nheads = 1;
176 	info.d_ncylinders = 1;
177 	info.d_secpertrack = info.d_media_blocks;
178 	info.d_secpercyl = info.d_secpertrack * info.d_nheads;
179 
180 	disk_setdiskinfo(&sc->disk, &info); */
181 	return(0);
182 }
183 
184 static int
185 vkdstrategy(struct dev_strategy_args *ap)
186 {
187 	struct bio *bio = ap->a_bio;
188 	struct vkd_softc *sc;
189 	cdev_t dev;
190 
191 	dev = ap->a_head.a_dev;
192 	sc = dev->si_drv1;
193 
194 	devstat_start_transaction(&sc->stats);
195 	cothread_lock(sc->cotd, 0);
196 	TAILQ_INSERT_TAIL(&sc->cotd_queue, bio, bio_act);
197 	cothread_signal(sc->cotd);
198 	cothread_unlock(sc->cotd, 0);
199 
200 	return(0);
201 }
202 
203 static
204 void
205 vkd_io_intr(cothread_t cotd)
206 {
207 	struct vkd_softc *sc;
208 	struct bio *bio;
209 	TAILQ_HEAD(, bio) tmpq;
210 
211 	sc = cotd->arg;
212 
213 	/*
214 	 * We can't call back into the kernel while holding cothread!
215 	 */
216 	TAILQ_INIT(&tmpq);
217 	cothread_lock(cotd, 0);
218 	while ((bio = TAILQ_FIRST(&sc->cotd_done)) != NULL) {
219 		TAILQ_REMOVE(&sc->cotd_done, bio, bio_act);
220 		TAILQ_INSERT_TAIL(&tmpq, bio, bio_act);
221 	}
222 	cothread_unlock(cotd, 0);
223 
224 	while ((bio = TAILQ_FIRST(&tmpq)) != NULL) {
225 		TAILQ_REMOVE(&tmpq, bio, bio_act);
226 		devstat_end_transaction_buf(&sc->stats, bio->bio_buf);
227 		biodone(bio);
228 	}
229 }
230 
231 /*
232  * WARNING!  This runs as a cothread and has no access to mycpu nor can it
233  * make vkernel-specific calls other then cothread_*() calls.
234  *
235  * WARNING!  A signal can occur and be discarded prior to our initial
236  * call to cothread_lock().  Process pending I/O before waiting.
237  */
238 static
239 void
240 vkd_io_thread(cothread_t cotd)
241 {
242 	struct bio *bio;
243 	struct vkd_softc *sc = cotd->arg;
244 	int count;
245 
246 	cothread_lock(cotd, 1);
247 	for (;;) {
248 		count = 0;
249 		while ((bio = TAILQ_FIRST(&sc->cotd_queue)) != NULL) {
250 			TAILQ_REMOVE(&sc->cotd_queue, bio, bio_act);
251 			cothread_unlock(cotd, 1);
252 			vkd_doio(sc, bio);
253 			cothread_lock(cotd, 1);
254 			TAILQ_INSERT_TAIL(&sc->cotd_done, bio, bio_act);
255 			if (++count == 8) {
256 				cothread_intr(cotd);
257 				count = 0;
258 			}
259 		}
260 		if (count)
261 			cothread_intr(cotd);
262 		cothread_wait(cotd);	/* interlocks cothread lock */
263 	}
264 	/* NOT REACHED */
265 	cothread_unlock(cotd, 1);
266 }
267 
268 static
269 void
270 vkd_doio(struct vkd_softc *sc, struct bio *bio)
271 {
272 	struct buf *bp = bio->bio_buf;
273 	int n;
274 
275 	switch(bp->b_cmd) {
276 	case BUF_CMD_READ:
277 		n = pread(sc->fd, bp->b_data, bp->b_bcount, bio->bio_offset);
278 		break;
279 	case BUF_CMD_WRITE:
280 		/* XXX HANDLE SHORT WRITE XXX */
281 		n = pwrite(sc->fd, bp->b_data, bp->b_bcount, bio->bio_offset);
282 		break;
283 	case BUF_CMD_FLUSH:
284 		if (fsync(sc->fd) < 0)
285 			n = -1;
286 		else
287 			n = bp->b_bcount;
288 		break;
289 	default:
290 		panic("vkd: bad b_cmd %d", bp->b_cmd);
291 		break; /* not reached */
292 	}
293 	if (n != bp->b_bcount) {
294 		bp->b_error = EIO;
295 		bp->b_flags |= B_ERROR;
296 	}
297 	bp->b_resid = bp->b_bcount - n;
298 }
299