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 #include <md5.h> 50 51 #include "bhyverun.h" 52 #include "pci_emul.h" 53 #include "virtio.h" 54 55 #ifndef min 56 #define min(a, b) ((a) < (b) ? (a) : (b)) 57 #endif 58 59 #define VTBLK_RINGSZ 64 60 61 #define VTBLK_MAXSEGS 32 62 63 #define VTBLK_S_OK 0 64 #define VTBLK_S_IOERR 1 65 #define VTBLK_S_UNSUPP 2 66 67 #define VTBLK_BLK_ID_BYTES 20 68 69 /* Capability bits */ 70 #define VTBLK_F_SEG_MAX (1 << 2) /* Maximum request segments */ 71 #define VTBLK_F_BLK_SIZE (1 << 6) /* cfg block size valid */ 72 73 /* 74 * Host capabilities 75 */ 76 #define VTBLK_S_HOSTCAPS \ 77 ( VTBLK_F_SEG_MAX | \ 78 VTBLK_F_BLK_SIZE | \ 79 VIRTIO_RING_F_INDIRECT_DESC ) /* indirect descriptors */ 80 81 /* 82 * Config space "registers" 83 */ 84 struct vtblk_config { 85 uint64_t vbc_capacity; 86 uint32_t vbc_size_max; 87 uint32_t vbc_seg_max; 88 uint16_t vbc_geom_c; 89 uint8_t vbc_geom_h; 90 uint8_t vbc_geom_s; 91 uint32_t vbc_blk_size; 92 uint32_t vbc_sectors_max; 93 } __packed; 94 95 /* 96 * Fixed-size block header 97 */ 98 struct virtio_blk_hdr { 99 #define VBH_OP_READ 0 100 #define VBH_OP_WRITE 1 101 #define VBH_OP_IDENT 8 102 #define VBH_FLAG_BARRIER 0x80000000 /* OR'ed into vbh_type */ 103 uint32_t vbh_type; 104 uint32_t vbh_ioprio; 105 uint64_t vbh_sector; 106 } __packed; 107 108 /* 109 * Debug printf 110 */ 111 static int pci_vtblk_debug; 112 #define DPRINTF(params) if (pci_vtblk_debug) printf params 113 #define WPRINTF(params) printf params 114 115 /* 116 * Per-device softc 117 */ 118 struct pci_vtblk_softc { 119 struct virtio_softc vbsc_vs; 120 struct vqueue_info vbsc_vq; 121 int vbsc_fd; 122 struct vtblk_config vbsc_cfg; 123 char vbsc_ident[VTBLK_BLK_ID_BYTES]; 124 }; 125 126 static void pci_vtblk_reset(void *); 127 static void pci_vtblk_notify(void *, struct vqueue_info *); 128 static int pci_vtblk_cfgread(void *, int, int, uint32_t *); 129 static int pci_vtblk_cfgwrite(void *, int, int, uint32_t); 130 131 static struct virtio_consts vtblk_vi_consts = { 132 "vtblk", /* our name */ 133 1, /* we support 1 virtqueue */ 134 sizeof(struct vtblk_config), /* config reg size */ 135 pci_vtblk_reset, /* reset */ 136 pci_vtblk_notify, /* device-wide qnotify */ 137 pci_vtblk_cfgread, /* read PCI config */ 138 pci_vtblk_cfgwrite, /* write PCI config */ 139 VTBLK_S_HOSTCAPS, /* our capabilities */ 140 }; 141 142 static void 143 pci_vtblk_reset(void *vsc) 144 { 145 struct pci_vtblk_softc *sc = vsc; 146 147 DPRINTF(("vtblk: device reset requested !\n")); 148 vi_reset_dev(&sc->vbsc_vs); 149 } 150 151 static void 152 pci_vtblk_proc(struct pci_vtblk_softc *sc, struct vqueue_info *vq) 153 { 154 struct virtio_blk_hdr *vbh; 155 uint8_t *status; 156 int i, n; 157 int err; 158 int iolen; 159 int writeop, type; 160 off_t offset; 161 struct iovec iov[VTBLK_MAXSEGS + 2]; 162 uint16_t flags[VTBLK_MAXSEGS + 2]; 163 164 n = vq_getchain(vq, iov, VTBLK_MAXSEGS + 2, flags); 165 166 /* 167 * The first descriptor will be the read-only fixed header, 168 * and the last is for status (hence +2 above and below). 169 * The remaining iov's are the actual data I/O vectors. 170 * 171 * XXX - note - this fails on crash dump, which does a 172 * VIRTIO_BLK_T_FLUSH with a zero transfer length 173 */ 174 assert(n >= 2 && n <= VTBLK_MAXSEGS + 2); 175 176 assert((flags[0] & VRING_DESC_F_WRITE) == 0); 177 assert(iov[0].iov_len == sizeof(struct virtio_blk_hdr)); 178 vbh = iov[0].iov_base; 179 180 status = iov[--n].iov_base; 181 assert(iov[n].iov_len == 1); 182 assert(flags[n] & VRING_DESC_F_WRITE); 183 184 /* 185 * XXX 186 * The guest should not be setting the BARRIER flag because 187 * we don't advertise the capability. 188 */ 189 type = vbh->vbh_type & ~VBH_FLAG_BARRIER; 190 writeop = (type == VBH_OP_WRITE); 191 192 offset = vbh->vbh_sector * DEV_BSIZE; 193 194 iolen = 0; 195 for (i = 1; i < n; i++) { 196 /* 197 * - write op implies read-only descriptor, 198 * - read/ident op implies write-only descriptor, 199 * therefore test the inverse of the descriptor bit 200 * to the op. 201 */ 202 assert(((flags[i] & VRING_DESC_F_WRITE) == 0) == writeop); 203 iolen += iov[i].iov_len; 204 } 205 206 DPRINTF(("virtio-block: %s op, %d bytes, %d segs, offset %ld\n\r", 207 writeop ? "write" : "read/ident", iolen, i - 1, offset)); 208 209 switch (type) { 210 case VBH_OP_WRITE: 211 err = pwritev(sc->vbsc_fd, iov + 1, i - 1, offset); 212 break; 213 case VBH_OP_READ: 214 err = preadv(sc->vbsc_fd, iov + 1, i - 1, offset); 215 break; 216 case VBH_OP_IDENT: 217 /* Assume a single buffer */ 218 strlcpy(iov[1].iov_base, sc->vbsc_ident, 219 min(iov[1].iov_len, sizeof(sc->vbsc_ident))); 220 err = 0; 221 break; 222 default: 223 err = -ENOSYS; 224 break; 225 } 226 227 /* convert errno into a virtio block error return */ 228 if (err < 0) { 229 if (err == -ENOSYS) 230 *status = VTBLK_S_UNSUPP; 231 else 232 *status = VTBLK_S_IOERR; 233 } else 234 *status = VTBLK_S_OK; 235 236 /* 237 * Return the descriptor back to the host. 238 * We wrote 1 byte (our status) to host. 239 */ 240 vq_relchain(vq, 1); 241 } 242 243 static void 244 pci_vtblk_notify(void *vsc, struct vqueue_info *vq) 245 { 246 struct pci_vtblk_softc *sc = vsc; 247 248 vq_startchains(vq); 249 while (vq_has_descs(vq)) 250 pci_vtblk_proc(sc, vq); 251 vq_endchains(vq, 1); /* Generate interrupt if appropriate. */ 252 } 253 254 static int 255 pci_vtblk_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) 256 { 257 struct stat sbuf; 258 MD5_CTX mdctx; 259 u_char digest[16]; 260 struct pci_vtblk_softc *sc; 261 off_t size; 262 int fd; 263 int sectsz; 264 265 if (opts == NULL) { 266 printf("virtio-block: backing device required\n"); 267 return (1); 268 } 269 270 /* 271 * The supplied backing file has to exist 272 */ 273 fd = open(opts, O_RDWR); 274 if (fd < 0) { 275 perror("Could not open backing file"); 276 return (1); 277 } 278 279 if (fstat(fd, &sbuf) < 0) { 280 perror("Could not stat backing file"); 281 close(fd); 282 return (1); 283 } 284 285 /* 286 * Deal with raw devices 287 */ 288 size = sbuf.st_size; 289 sectsz = DEV_BSIZE; 290 if (S_ISCHR(sbuf.st_mode)) { 291 if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 || 292 ioctl(fd, DIOCGSECTORSIZE, §sz)) { 293 perror("Could not fetch dev blk/sector size"); 294 close(fd); 295 return (1); 296 } 297 assert(size != 0); 298 assert(sectsz != 0); 299 } 300 301 sc = malloc(sizeof(struct pci_vtblk_softc)); 302 memset(sc, 0, sizeof(struct pci_vtblk_softc)); 303 304 /* record fd of storage device/file */ 305 sc->vbsc_fd = fd; 306 307 /* init virtio softc and virtqueues */ 308 vi_softc_linkup(&sc->vbsc_vs, &vtblk_vi_consts, sc, pi, &sc->vbsc_vq); 309 sc->vbsc_vq.vq_qsize = VTBLK_RINGSZ; 310 /* sc->vbsc_vq.vq_notify = we have no per-queue notify */ 311 312 /* 313 * Create an identifier for the backing file. Use parts of the 314 * md5 sum of the filename 315 */ 316 MD5Init(&mdctx); 317 MD5Update(&mdctx, opts, strlen(opts)); 318 MD5Final(digest, &mdctx); 319 sprintf(sc->vbsc_ident, "BHYVE-%02X%02X-%02X%02X-%02X%02X", 320 digest[0], digest[1], digest[2], digest[3], digest[4], digest[5]); 321 322 /* setup virtio block config space */ 323 sc->vbsc_cfg.vbc_capacity = size / DEV_BSIZE; /* 512-byte units */ 324 sc->vbsc_cfg.vbc_seg_max = VTBLK_MAXSEGS; 325 sc->vbsc_cfg.vbc_blk_size = sectsz; 326 sc->vbsc_cfg.vbc_size_max = 0; /* not negotiated */ 327 sc->vbsc_cfg.vbc_geom_c = 0; /* no geometry */ 328 sc->vbsc_cfg.vbc_geom_h = 0; 329 sc->vbsc_cfg.vbc_geom_s = 0; 330 sc->vbsc_cfg.vbc_sectors_max = 0; 331 332 /* 333 * Should we move some of this into virtio.c? Could 334 * have the device, class, and subdev_0 as fields in 335 * the virtio constants structure. 336 */ 337 pci_set_cfgdata16(pi, PCIR_DEVICE, VIRTIO_DEV_BLOCK); 338 pci_set_cfgdata16(pi, PCIR_VENDOR, VIRTIO_VENDOR); 339 pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE); 340 pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_BLOCK); 341 342 if (vi_intr_init(&sc->vbsc_vs, 1, fbsdrun_virtio_msix())) 343 return (1); 344 vi_set_io_bar(&sc->vbsc_vs, 0); 345 return (0); 346 } 347 348 static int 349 pci_vtblk_cfgwrite(void *vsc, int offset, int size, uint32_t value) 350 { 351 352 DPRINTF(("vtblk: write to readonly reg %d\n\r", offset)); 353 return (1); 354 } 355 356 static int 357 pci_vtblk_cfgread(void *vsc, int offset, int size, uint32_t *retval) 358 { 359 struct pci_vtblk_softc *sc = vsc; 360 void *ptr; 361 362 /* our caller has already verified offset and size */ 363 ptr = (uint8_t *)&sc->vbsc_cfg + offset; 364 memcpy(retval, ptr, size); 365 return (0); 366 } 367 368 struct pci_devemu pci_de_vblk = { 369 .pe_emu = "virtio-blk", 370 .pe_init = pci_vtblk_init, 371 .pe_barwrite = vi_pci_write, 372 .pe_barread = vi_pci_read 373 }; 374 PCI_EMUL_SET(pci_de_vblk); 375