1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2011 NetApp, Inc. 5 * All rights reserved. 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 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 * 28 * $FreeBSD$ 29 */ 30 31 #include <sys/cdefs.h> 32 __FBSDID("$FreeBSD$"); 33 34 #include <sys/param.h> 35 #include <sys/linker_set.h> 36 #include <sys/stat.h> 37 #include <sys/uio.h> 38 #include <sys/ioctl.h> 39 #include <sys/disk.h> 40 41 #include <errno.h> 42 #include <fcntl.h> 43 #include <stdio.h> 44 #include <stdlib.h> 45 #include <stdint.h> 46 #include <string.h> 47 #include <strings.h> 48 #include <unistd.h> 49 #include <assert.h> 50 #include <pthread.h> 51 #include <md5.h> 52 53 #include "bhyverun.h" 54 #include "pci_emul.h" 55 #include "virtio.h" 56 #include "block_if.h" 57 58 #define VTBLK_RINGSZ 64 59 60 #define VTBLK_S_OK 0 61 #define VTBLK_S_IOERR 1 62 #define VTBLK_S_UNSUPP 2 63 64 #define VTBLK_BLK_ID_BYTES 20 + 1 65 66 /* Capability bits */ 67 #define VTBLK_F_SEG_MAX (1 << 2) /* Maximum request segments */ 68 #define VTBLK_F_BLK_SIZE (1 << 6) /* cfg block size valid */ 69 #define VTBLK_F_FLUSH (1 << 9) /* Cache flush support */ 70 #define VTBLK_F_TOPOLOGY (1 << 10) /* Optimal I/O alignment */ 71 72 /* 73 * Host capabilities 74 */ 75 #define VTBLK_S_HOSTCAPS \ 76 ( VTBLK_F_SEG_MAX | \ 77 VTBLK_F_BLK_SIZE | \ 78 VTBLK_F_FLUSH | \ 79 VTBLK_F_TOPOLOGY | \ 80 VIRTIO_RING_F_INDIRECT_DESC ) /* indirect descriptors */ 81 82 /* 83 * Config space "registers" 84 */ 85 struct vtblk_config { 86 uint64_t vbc_capacity; 87 uint32_t vbc_size_max; 88 uint32_t vbc_seg_max; 89 struct { 90 uint16_t cylinders; 91 uint8_t heads; 92 uint8_t sectors; 93 } vbc_geometry; 94 uint32_t vbc_blk_size; 95 struct { 96 uint8_t physical_block_exp; 97 uint8_t alignment_offset; 98 uint16_t min_io_size; 99 uint32_t opt_io_size; 100 } vbc_topology; 101 uint8_t vbc_writeback; 102 } __packed; 103 104 /* 105 * Fixed-size block header 106 */ 107 struct virtio_blk_hdr { 108 #define VBH_OP_READ 0 109 #define VBH_OP_WRITE 1 110 #define VBH_OP_FLUSH 4 111 #define VBH_OP_FLUSH_OUT 5 112 #define VBH_OP_IDENT 8 113 #define VBH_FLAG_BARRIER 0x80000000 /* OR'ed into vbh_type */ 114 uint32_t vbh_type; 115 uint32_t vbh_ioprio; 116 uint64_t vbh_sector; 117 } __packed; 118 119 /* 120 * Debug printf 121 */ 122 static int pci_vtblk_debug; 123 #define DPRINTF(params) if (pci_vtblk_debug) printf params 124 #define WPRINTF(params) printf params 125 126 struct pci_vtblk_ioreq { 127 struct blockif_req io_req; 128 struct pci_vtblk_softc *io_sc; 129 uint8_t *io_status; 130 uint16_t io_idx; 131 }; 132 133 /* 134 * Per-device softc 135 */ 136 struct pci_vtblk_softc { 137 struct virtio_softc vbsc_vs; 138 pthread_mutex_t vsc_mtx; 139 struct vqueue_info vbsc_vq; 140 struct vtblk_config vbsc_cfg; 141 struct blockif_ctxt *bc; 142 char vbsc_ident[VTBLK_BLK_ID_BYTES]; 143 struct pci_vtblk_ioreq vbsc_ios[VTBLK_RINGSZ]; 144 }; 145 146 static void pci_vtblk_reset(void *); 147 static void pci_vtblk_notify(void *, struct vqueue_info *); 148 static int pci_vtblk_cfgread(void *, int, int, uint32_t *); 149 static int pci_vtblk_cfgwrite(void *, int, int, uint32_t); 150 151 static struct virtio_consts vtblk_vi_consts = { 152 "vtblk", /* our name */ 153 1, /* we support 1 virtqueue */ 154 sizeof(struct vtblk_config), /* config reg size */ 155 pci_vtblk_reset, /* reset */ 156 pci_vtblk_notify, /* device-wide qnotify */ 157 pci_vtblk_cfgread, /* read PCI config */ 158 pci_vtblk_cfgwrite, /* write PCI config */ 159 NULL, /* apply negotiated features */ 160 VTBLK_S_HOSTCAPS, /* our capabilities */ 161 }; 162 163 static void 164 pci_vtblk_reset(void *vsc) 165 { 166 struct pci_vtblk_softc *sc = vsc; 167 168 DPRINTF(("vtblk: device reset requested !\n")); 169 vi_reset_dev(&sc->vbsc_vs); 170 } 171 172 static void 173 pci_vtblk_done(struct blockif_req *br, int err) 174 { 175 struct pci_vtblk_ioreq *io = br->br_param; 176 struct pci_vtblk_softc *sc = io->io_sc; 177 178 /* convert errno into a virtio block error return */ 179 if (err == EOPNOTSUPP || err == ENOSYS) 180 *io->io_status = VTBLK_S_UNSUPP; 181 else if (err != 0) 182 *io->io_status = VTBLK_S_IOERR; 183 else 184 *io->io_status = VTBLK_S_OK; 185 186 /* 187 * Return the descriptor back to the host. 188 * We wrote 1 byte (our status) to host. 189 */ 190 pthread_mutex_lock(&sc->vsc_mtx); 191 vq_relchain(&sc->vbsc_vq, io->io_idx, 1); 192 vq_endchains(&sc->vbsc_vq, 0); 193 pthread_mutex_unlock(&sc->vsc_mtx); 194 } 195 196 static void 197 pci_vtblk_proc(struct pci_vtblk_softc *sc, struct vqueue_info *vq) 198 { 199 struct virtio_blk_hdr *vbh; 200 struct pci_vtblk_ioreq *io; 201 int i, n; 202 int err; 203 ssize_t iolen; 204 int writeop, type; 205 struct iovec iov[BLOCKIF_IOV_MAX + 2]; 206 uint16_t idx, flags[BLOCKIF_IOV_MAX + 2]; 207 208 n = vq_getchain(vq, &idx, iov, BLOCKIF_IOV_MAX + 2, flags); 209 210 /* 211 * The first descriptor will be the read-only fixed header, 212 * and the last is for status (hence +2 above and below). 213 * The remaining iov's are the actual data I/O vectors. 214 * 215 * XXX - note - this fails on crash dump, which does a 216 * VIRTIO_BLK_T_FLUSH with a zero transfer length 217 */ 218 assert(n >= 2 && n <= BLOCKIF_IOV_MAX + 2); 219 220 io = &sc->vbsc_ios[idx]; 221 assert((flags[0] & VRING_DESC_F_WRITE) == 0); 222 assert(iov[0].iov_len == sizeof(struct virtio_blk_hdr)); 223 vbh = iov[0].iov_base; 224 memcpy(&io->io_req.br_iov, &iov[1], sizeof(struct iovec) * (n - 2)); 225 io->io_req.br_iovcnt = n - 2; 226 io->io_req.br_offset = vbh->vbh_sector * DEV_BSIZE; 227 io->io_status = iov[--n].iov_base; 228 assert(iov[n].iov_len == 1); 229 assert(flags[n] & VRING_DESC_F_WRITE); 230 231 /* 232 * XXX 233 * The guest should not be setting the BARRIER flag because 234 * we don't advertise the capability. 235 */ 236 type = vbh->vbh_type & ~VBH_FLAG_BARRIER; 237 writeop = (type == VBH_OP_WRITE); 238 239 iolen = 0; 240 for (i = 1; i < n; i++) { 241 /* 242 * - write op implies read-only descriptor, 243 * - read/ident op implies write-only descriptor, 244 * therefore test the inverse of the descriptor bit 245 * to the op. 246 */ 247 assert(((flags[i] & VRING_DESC_F_WRITE) == 0) == writeop); 248 iolen += iov[i].iov_len; 249 } 250 io->io_req.br_resid = iolen; 251 252 DPRINTF(("virtio-block: %s op, %zd bytes, %d segs, offset %ld\n\r", 253 writeop ? "write" : "read/ident", iolen, i - 1, 254 io->io_req.br_offset)); 255 256 switch (type) { 257 case VBH_OP_READ: 258 err = blockif_read(sc->bc, &io->io_req); 259 break; 260 case VBH_OP_WRITE: 261 err = blockif_write(sc->bc, &io->io_req); 262 break; 263 case VBH_OP_FLUSH: 264 case VBH_OP_FLUSH_OUT: 265 err = blockif_flush(sc->bc, &io->io_req); 266 break; 267 case VBH_OP_IDENT: 268 /* Assume a single buffer */ 269 /* S/n equal to buffer is not zero-terminated. */ 270 memset(iov[1].iov_base, 0, iov[1].iov_len); 271 strncpy(iov[1].iov_base, sc->vbsc_ident, 272 MIN(iov[1].iov_len, sizeof(sc->vbsc_ident))); 273 pci_vtblk_done(&io->io_req, 0); 274 return; 275 default: 276 pci_vtblk_done(&io->io_req, EOPNOTSUPP); 277 return; 278 } 279 assert(err == 0); 280 } 281 282 static void 283 pci_vtblk_notify(void *vsc, struct vqueue_info *vq) 284 { 285 struct pci_vtblk_softc *sc = vsc; 286 287 while (vq_has_descs(vq)) 288 pci_vtblk_proc(sc, vq); 289 } 290 291 static int 292 pci_vtblk_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) 293 { 294 char bident[sizeof("XX:X:X")]; 295 struct blockif_ctxt *bctxt; 296 MD5_CTX mdctx; 297 u_char digest[16]; 298 struct pci_vtblk_softc *sc; 299 off_t size; 300 int i, sectsz, sts, sto; 301 302 if (opts == NULL) { 303 printf("virtio-block: backing device required\n"); 304 return (1); 305 } 306 307 /* 308 * The supplied backing file has to exist 309 */ 310 snprintf(bident, sizeof(bident), "%d:%d", pi->pi_slot, pi->pi_func); 311 bctxt = blockif_open(opts, bident); 312 if (bctxt == NULL) { 313 perror("Could not open backing file"); 314 return (1); 315 } 316 317 size = blockif_size(bctxt); 318 sectsz = blockif_sectsz(bctxt); 319 blockif_psectsz(bctxt, &sts, &sto); 320 321 sc = calloc(1, sizeof(struct pci_vtblk_softc)); 322 sc->bc = bctxt; 323 for (i = 0; i < VTBLK_RINGSZ; i++) { 324 struct pci_vtblk_ioreq *io = &sc->vbsc_ios[i]; 325 io->io_req.br_callback = pci_vtblk_done; 326 io->io_req.br_param = io; 327 io->io_sc = sc; 328 io->io_idx = i; 329 } 330 331 pthread_mutex_init(&sc->vsc_mtx, NULL); 332 333 /* init virtio softc and virtqueues */ 334 vi_softc_linkup(&sc->vbsc_vs, &vtblk_vi_consts, sc, pi, &sc->vbsc_vq); 335 sc->vbsc_vs.vs_mtx = &sc->vsc_mtx; 336 337 sc->vbsc_vq.vq_qsize = VTBLK_RINGSZ; 338 /* sc->vbsc_vq.vq_notify = we have no per-queue notify */ 339 340 /* 341 * Create an identifier for the backing file. Use parts of the 342 * md5 sum of the filename 343 */ 344 MD5Init(&mdctx); 345 MD5Update(&mdctx, opts, strlen(opts)); 346 MD5Final(digest, &mdctx); 347 snprintf(sc->vbsc_ident, VTBLK_BLK_ID_BYTES, 348 "BHYVE-%02X%02X-%02X%02X-%02X%02X", 349 digest[0], digest[1], digest[2], digest[3], digest[4], digest[5]); 350 351 /* setup virtio block config space */ 352 sc->vbsc_cfg.vbc_capacity = size / DEV_BSIZE; /* 512-byte units */ 353 sc->vbsc_cfg.vbc_size_max = 0; /* not negotiated */ 354 sc->vbsc_cfg.vbc_seg_max = BLOCKIF_IOV_MAX; 355 sc->vbsc_cfg.vbc_geometry.cylinders = 0; /* no geometry */ 356 sc->vbsc_cfg.vbc_geometry.heads = 0; 357 sc->vbsc_cfg.vbc_geometry.sectors = 0; 358 sc->vbsc_cfg.vbc_blk_size = sectsz; 359 sc->vbsc_cfg.vbc_topology.physical_block_exp = 360 (sts > sectsz) ? (ffsll(sts / sectsz) - 1) : 0; 361 sc->vbsc_cfg.vbc_topology.alignment_offset = 362 (sto != 0) ? ((sts - sto) / sectsz) : 0; 363 sc->vbsc_cfg.vbc_topology.min_io_size = 0; 364 sc->vbsc_cfg.vbc_topology.opt_io_size = 0; 365 sc->vbsc_cfg.vbc_writeback = 0; 366 367 /* 368 * Should we move some of this into virtio.c? Could 369 * have the device, class, and subdev_0 as fields in 370 * the virtio constants structure. 371 */ 372 pci_set_cfgdata16(pi, PCIR_DEVICE, VIRTIO_DEV_BLOCK); 373 pci_set_cfgdata16(pi, PCIR_VENDOR, VIRTIO_VENDOR); 374 pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE); 375 pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_BLOCK); 376 pci_set_cfgdata16(pi, PCIR_SUBVEND_0, VIRTIO_VENDOR); 377 378 if (vi_intr_init(&sc->vbsc_vs, 1, fbsdrun_virtio_msix())) { 379 blockif_close(sc->bc); 380 free(sc); 381 return (1); 382 } 383 vi_set_io_bar(&sc->vbsc_vs, 0); 384 return (0); 385 } 386 387 static int 388 pci_vtblk_cfgwrite(void *vsc, int offset, int size, uint32_t value) 389 { 390 391 DPRINTF(("vtblk: write to readonly reg %d\n\r", offset)); 392 return (1); 393 } 394 395 static int 396 pci_vtblk_cfgread(void *vsc, int offset, int size, uint32_t *retval) 397 { 398 struct pci_vtblk_softc *sc = vsc; 399 void *ptr; 400 401 /* our caller has already verified offset and size */ 402 ptr = (uint8_t *)&sc->vbsc_cfg + offset; 403 memcpy(retval, ptr, size); 404 return (0); 405 } 406 407 struct pci_devemu pci_de_vblk = { 408 .pe_emu = "virtio-blk", 409 .pe_init = pci_vtblk_init, 410 .pe_barwrite = vi_pci_write, 411 .pe_barread = vi_pci_read 412 }; 413 PCI_EMUL_SET(pci_de_vblk); 414