xref: /freebsd/usr.sbin/bhyve/pci_virtio_block.c (revision a98ff317388a00b992f1bf8404dee596f9383f5e)
1 /*-
2  * Copyright (c) 2011 NetApp, Inc.
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 NETAPP, INC ``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 NETAPP, INC 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  * $FreeBSD$
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include <sys/param.h>
33 #include <sys/linker_set.h>
34 #include <sys/stat.h>
35 #include <sys/uio.h>
36 #include <sys/ioctl.h>
37 #include <sys/disk.h>
38 
39 #include <errno.h>
40 #include <fcntl.h>
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <stdint.h>
44 #include <string.h>
45 #include <strings.h>
46 #include <unistd.h>
47 #include <assert.h>
48 #include <pthread.h>
49 
50 #include "bhyverun.h"
51 #include "pci_emul.h"
52 #include "virtio.h"
53 
54 #define VTBLK_RINGSZ	64
55 
56 #define VTBLK_MAXSEGS	32
57 
58 #define VTBLK_S_OK	0
59 #define VTBLK_S_IOERR	1
60 
61 /*
62  * Host capabilities
63  */
64 #define VTBLK_S_HOSTCAPS      \
65   ( 0x00000004 |	/* host maximum request segments */ \
66     VIRTIO_RING_F_INDIRECT_DESC )	/* indirect descriptors */
67 
68 /*
69  * Config space "registers"
70  */
71 struct vtblk_config {
72 	uint64_t	vbc_capacity;
73 	uint32_t	vbc_size_max;
74 	uint32_t	vbc_seg_max;
75 	uint16_t	vbc_geom_c;
76 	uint8_t		vbc_geom_h;
77 	uint8_t		vbc_geom_s;
78 	uint32_t	vbc_blk_size;
79 	uint32_t	vbc_sectors_max;
80 } __packed;
81 
82 /*
83  * Fixed-size block header
84  */
85 struct virtio_blk_hdr {
86 #define	VBH_OP_READ		0
87 #define	VBH_OP_WRITE		1
88 #define	VBH_FLAG_BARRIER	0x80000000	/* OR'ed into vbh_type */
89 	uint32_t       	vbh_type;
90 	uint32_t	vbh_ioprio;
91 	uint64_t	vbh_sector;
92 } __packed;
93 
94 /*
95  * Debug printf
96  */
97 static int pci_vtblk_debug;
98 #define DPRINTF(params) if (pci_vtblk_debug) printf params
99 #define WPRINTF(params) printf params
100 
101 /*
102  * Per-device softc
103  */
104 struct pci_vtblk_softc {
105 	struct virtio_softc vbsc_vs;
106 	struct vqueue_info vbsc_vq;
107 	int		vbsc_fd;
108 	struct vtblk_config vbsc_cfg;
109 };
110 
111 static void pci_vtblk_reset(void *);
112 static void pci_vtblk_notify(void *, struct vqueue_info *);
113 static int pci_vtblk_cfgread(void *, int, int, uint32_t *);
114 static int pci_vtblk_cfgwrite(void *, int, int, uint32_t);
115 
116 static struct virtio_consts vtblk_vi_consts = {
117 	"vtblk",		/* our name */
118 	1,			/* we support 1 virtqueue */
119 	sizeof(struct vtblk_config), /* config reg size */
120 	pci_vtblk_reset,	/* reset */
121 	pci_vtblk_notify,	/* device-wide qnotify */
122 	pci_vtblk_cfgread,	/* read PCI config */
123 	pci_vtblk_cfgwrite,	/* write PCI config */
124 	VTBLK_S_HOSTCAPS,	/* our capabilities */
125 };
126 
127 static void
128 pci_vtblk_reset(void *vsc)
129 {
130 	struct pci_vtblk_softc *sc = vsc;
131 
132 	DPRINTF(("vtblk: device reset requested !\n"));
133 	vi_reset_dev(&sc->vbsc_vs);
134 }
135 
136 static void
137 pci_vtblk_proc(struct pci_vtblk_softc *sc, struct vqueue_info *vq)
138 {
139 	struct virtio_blk_hdr *vbh;
140 	uint8_t *status;
141 	int i, n;
142 	int err;
143 	int iolen;
144 	int writeop, type;
145 	off_t offset;
146 	struct iovec iov[VTBLK_MAXSEGS + 2];
147 	uint16_t flags[VTBLK_MAXSEGS + 2];
148 
149 	n = vq_getchain(vq, iov, VTBLK_MAXSEGS + 2, flags);
150 
151 	/*
152 	 * The first descriptor will be the read-only fixed header,
153 	 * and the last is for status (hence +2 above and below).
154 	 * The remaining iov's are the actual data I/O vectors.
155 	 *
156 	 * XXX - note - this fails on crash dump, which does a
157 	 * VIRTIO_BLK_T_FLUSH with a zero transfer length
158 	 */
159 	assert (n >= 3 && n < VTBLK_MAXSEGS + 2);
160 
161 	assert((flags[0] & VRING_DESC_F_WRITE) == 0);
162 	assert(iov[0].iov_len == sizeof(struct virtio_blk_hdr));
163 	vbh = iov[0].iov_base;
164 
165 	status = iov[--n].iov_base;
166 	assert(iov[n].iov_len == 1);
167 	assert(flags[n] & VRING_DESC_F_WRITE);
168 
169 	/*
170 	 * XXX
171 	 * The guest should not be setting the BARRIER flag because
172 	 * we don't advertise the capability.
173 	 */
174 	type = vbh->vbh_type & ~VBH_FLAG_BARRIER;
175 	writeop = (type == VBH_OP_WRITE);
176 
177 	offset = vbh->vbh_sector * DEV_BSIZE;
178 
179 	iolen = 0;
180 	for (i = 1; i < n; i++) {
181 		/*
182 		 * - write op implies read-only descriptor,
183 		 * - read op implies write-only descriptor,
184 		 * therefore test the inverse of the descriptor bit
185 		 * to the op.
186 		 */
187 		assert(((flags[i] & VRING_DESC_F_WRITE) == 0) == writeop);
188 		iolen += iov[i].iov_len;
189 	}
190 
191 	DPRINTF(("virtio-block: %s op, %d bytes, %d segs, offset %ld\n\r",
192 		 writeop ? "write" : "read", iolen, i - 1, offset));
193 
194 	if (writeop)
195 		err = pwritev(sc->vbsc_fd, iov + 1, i - 1, offset);
196 	else
197 		err = preadv(sc->vbsc_fd, iov + 1, i - 1, offset);
198 
199 	*status = err < 0 ? VTBLK_S_IOERR : VTBLK_S_OK;
200 
201 	/*
202 	 * Return the descriptor back to the host.
203 	 * We wrote 1 byte (our status) to host.
204 	 */
205 	vq_relchain(vq, 1);
206 }
207 
208 static void
209 pci_vtblk_notify(void *vsc, struct vqueue_info *vq)
210 {
211 	struct pci_vtblk_softc *sc = vsc;
212 
213 	vq_startchains(vq);
214 	while (vq_has_descs(vq))
215 		pci_vtblk_proc(sc, vq);
216 	vq_endchains(vq, 1);	/* Generate interrupt if appropriate. */
217 }
218 
219 static int
220 pci_vtblk_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts)
221 {
222 	struct stat sbuf;
223 	struct pci_vtblk_softc *sc;
224 	off_t size;
225 	int fd;
226 	int sectsz;
227 	int use_msix;
228 	const char *env_msi;
229 
230 	if (opts == NULL) {
231 		printf("virtio-block: backing device required\n");
232 		return (1);
233 	}
234 
235 	/*
236 	 * The supplied backing file has to exist
237 	 */
238 	fd = open(opts, O_RDWR);
239 	if (fd < 0) {
240 		perror("Could not open backing file");
241 		return (1);
242 	}
243 
244 	if (fstat(fd, &sbuf) < 0) {
245 		perror("Could not stat backing file");
246 		close(fd);
247 		return (1);
248 	}
249 
250 	/*
251 	 * Deal with raw devices
252 	 */
253 	size = sbuf.st_size;
254 	sectsz = DEV_BSIZE;
255 	if (S_ISCHR(sbuf.st_mode)) {
256 		if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 ||
257 		    ioctl(fd, DIOCGSECTORSIZE, &sectsz)) {
258 			perror("Could not fetch dev blk/sector size");
259 			close(fd);
260 			return (1);
261 		}
262 		assert(size != 0);
263 		assert(sectsz != 0);
264 	}
265 
266 	sc = malloc(sizeof(struct pci_vtblk_softc));
267 	memset(sc, 0, sizeof(struct pci_vtblk_softc));
268 
269 	/* record fd of storage device/file */
270 	sc->vbsc_fd = fd;
271 
272 	/* init virtio softc and virtqueues */
273 	vi_softc_linkup(&sc->vbsc_vs, &vtblk_vi_consts, sc, pi, &sc->vbsc_vq);
274 	sc->vbsc_vq.vq_qsize = VTBLK_RINGSZ;
275 	/* sc->vbsc_vq.vq_notify = we have no per-queue notify */
276 
277 	/* setup virtio block config space */
278 	sc->vbsc_cfg.vbc_capacity = size / sectsz;
279 	sc->vbsc_cfg.vbc_seg_max = VTBLK_MAXSEGS;
280 	sc->vbsc_cfg.vbc_blk_size = sectsz;
281 	sc->vbsc_cfg.vbc_size_max = 0;	/* not negotiated */
282 	sc->vbsc_cfg.vbc_geom_c = 0;	/* no geometry */
283 	sc->vbsc_cfg.vbc_geom_h = 0;
284 	sc->vbsc_cfg.vbc_geom_s = 0;
285 	sc->vbsc_cfg.vbc_sectors_max = 0;
286 
287 	/*
288 	 * Should we move some of this into virtio.c?  Could
289 	 * have the device, class, and subdev_0 as fields in
290 	 * the virtio constants structure.
291 	 */
292 	pci_set_cfgdata16(pi, PCIR_DEVICE, VIRTIO_DEV_BLOCK);
293 	pci_set_cfgdata16(pi, PCIR_VENDOR, VIRTIO_VENDOR);
294 	pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE);
295 	pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_BLOCK);
296 
297 	use_msix = 1;
298 	if ((env_msi = getenv("BHYVE_USE_MSI"))) {
299 		if (strcasecmp(env_msi, "yes") == 0)
300 			use_msix = 0;
301 	}
302 	if (vi_intr_init(&sc->vbsc_vs, 1, use_msix))
303 		return (1);
304 	vi_set_io_bar(&sc->vbsc_vs, 0);
305 	return (0);
306 }
307 
308 static int
309 pci_vtblk_cfgwrite(void *vsc, int offset, int size, uint32_t value)
310 {
311 
312 	DPRINTF(("vtblk: write to readonly reg %d\n\r", offset));
313 	return (1);
314 }
315 
316 static int
317 pci_vtblk_cfgread(void *vsc, int offset, int size, uint32_t *retval)
318 {
319 	struct pci_vtblk_softc *sc = vsc;
320 	void *ptr;
321 
322 	/* our caller has already verified offset and size */
323 	ptr = (uint8_t *)&sc->vbsc_cfg + offset;
324 	memcpy(retval, ptr, size);
325 	return (0);
326 }
327 
328 struct pci_devemu pci_de_vblk = {
329 	.pe_emu =	"virtio-blk",
330 	.pe_init =	pci_vtblk_init,
331 	.pe_barwrite =	vi_pci_write,
332 	.pe_barread =	vi_pci_read
333 };
334 PCI_EMUL_SET(pci_de_vblk);
335