2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/linker_set.h>
36 #include <sys/ioctl.h>
54 #define VTBLK_RINGSZ 64
56 #define VTBLK_CFGSZ 28
58 #define VTBLK_R_CFG VTCFG_R_CFG1
59 #define VTBLK_R_CFG_END VTBLK_R_CFG + VTBLK_CFGSZ -1
60 #define VTBLK_R_MAX VTBLK_R_CFG_END
62 #define VTBLK_REGSZ VTBLK_R_MAX+1
64 #define VTBLK_MAXSEGS 32
67 #define VTBLK_S_IOERR 1
72 #define VTBLK_S_HOSTCAPS \
73 ( 0x00000004 | /* host maximum request segments */ \
74 0x10000000 ) /* supports indirect descriptors */
76 static int use_msix = 1;
81 uint16_t hq_cur_aidx; /* trails behind 'avail_idx' */
83 /* Host-context pointers to the queue */
84 struct virtio_desc *hq_dtable;
85 uint16_t *hq_avail_flags;
86 uint16_t *hq_avail_idx; /* monotonically increasing */
87 uint16_t *hq_avail_ring;
89 uint16_t *hq_used_flags;
90 uint16_t *hq_used_idx; /* monotonically increasing */
91 struct virtio_used *hq_used_ring;
98 uint64_t vbc_capacity;
99 uint32_t vbc_size_max;
100 uint32_t vbc_seg_max;
104 uint32_t vbc_blk_size;
105 uint32_t vbc_sectors_max;
107 CTASSERT(sizeof(struct vtblk_config) == VTBLK_CFGSZ);
110 * Fixed-size block header
112 struct virtio_blk_hdr {
113 #define VBH_OP_READ 0
114 #define VBH_OP_WRITE 1
123 static int pci_vtblk_debug;
124 #define DPRINTF(params) if (pci_vtblk_debug) printf params
125 #define WPRINTF(params) printf params
130 struct pci_vtblk_softc {
131 struct pci_devinst *vbsc_pi;
136 uint32_t vbsc_features;
138 struct vring_hqueue vbsc_q;
139 struct vtblk_config vbsc_cfg;
140 uint16_t msix_table_idx_req;
141 uint16_t msix_table_idx_cfg;
145 * Return the size of IO BAR that maps virtio header and device specific
146 * region. The size would vary depending on whether MSI-X is enabled or
150 pci_vtblk_iosize(struct pci_devinst *pi)
153 if (pci_msix_enabled(pi))
154 return (VTBLK_REGSZ);
156 return (VTBLK_REGSZ - (VTCFG_R_CFG1 - VTCFG_R_MSIX));
160 * Return the number of available descriptors in the vring taking care
161 * of the 16-bit index wraparound.
164 hq_num_avail(struct vring_hqueue *hq)
168 if (*hq->hq_avail_idx >= hq->hq_cur_aidx)
169 ndesc = *hq->hq_avail_idx - hq->hq_cur_aidx;
171 ndesc = UINT16_MAX - hq->hq_cur_aidx + *hq->hq_avail_idx + 1;
173 assert(ndesc >= 0 && ndesc <= hq->hq_size);
179 pci_vtblk_update_status(struct pci_vtblk_softc *sc, uint32_t value)
182 DPRINTF(("vtblk: device reset requested !\n"));
185 sc->vbsc_status = value;
189 pci_vtblk_proc(struct pci_vtblk_softc *sc, struct vring_hqueue *hq)
191 struct iovec iov[VTBLK_MAXSEGS];
192 struct virtio_blk_hdr *vbh;
193 struct virtio_desc *vd, *vid;
194 struct virtio_used *vu;
200 int uidx, aidx, didx;
204 uidx = *hq->hq_used_idx;
205 aidx = hq->hq_cur_aidx;
206 didx = hq->hq_avail_ring[aidx % hq->hq_size];
207 assert(didx >= 0 && didx < hq->hq_size);
209 vd = &hq->hq_dtable[didx];
212 * Verify that the descriptor is indirect, and obtain
213 * the pointer to the indirect descriptor.
214 * There has to be space for at least 3 descriptors
215 * in the indirect descriptor array: the block header,
216 * 1 or more data descriptors, and a status byte.
218 assert(vd->vd_flags & VRING_DESC_F_INDIRECT);
220 nsegs = vd->vd_len / sizeof(struct virtio_desc);
222 assert(nsegs < VTBLK_MAXSEGS + 2);
224 vid = paddr_guest2host(vd->vd_addr);
225 assert((vid->vd_flags & VRING_DESC_F_INDIRECT) == 0);
228 * The first descriptor will be the read-only fixed header
230 vbh = paddr_guest2host(vid[0].vd_addr);
231 assert(vid[0].vd_len == sizeof(struct virtio_blk_hdr));
232 assert(vid[0].vd_flags & VRING_DESC_F_NEXT);
233 assert((vid[0].vd_flags & VRING_DESC_F_WRITE) == 0);
235 writeop = (vbh->vbh_type == VBH_OP_WRITE);
237 offset = vbh->vbh_sector * DEV_BSIZE;
240 * Build up the iovec based on the guest's data descriptors
242 for (i = 1, iolen = 0; i < nsegs - 1; i++) {
243 iov[i-1].iov_base = paddr_guest2host(vid[i].vd_addr);
244 iov[i-1].iov_len = vid[i].vd_len;
245 iolen += vid[i].vd_len;
247 assert(vid[i].vd_flags & VRING_DESC_F_NEXT);
248 assert((vid[i].vd_flags & VRING_DESC_F_INDIRECT) == 0);
251 * - write op implies read-only descriptor,
252 * - read op implies write-only descriptor,
253 * therefore test the inverse of the descriptor bit
256 assert(((vid[i].vd_flags & VRING_DESC_F_WRITE) == 0) ==
260 /* Lastly, get the address of the status byte */
261 status = paddr_guest2host(vid[nsegs - 1].vd_addr);
262 assert(vid[nsegs - 1].vd_len == 1);
263 assert((vid[nsegs - 1].vd_flags & VRING_DESC_F_NEXT) == 0);
264 assert(vid[nsegs - 1].vd_flags & VRING_DESC_F_WRITE);
266 DPRINTF(("virtio-block: %s op, %d bytes, %d segs, offset %ld\n\r",
267 writeop ? "write" : "read", iolen, nsegs - 2, offset));
270 err = pwritev(sc->vbsc_fd, iov, nsegs - 2, offset);
272 err = preadv(sc->vbsc_fd, iov, nsegs - 2, offset);
275 *status = err < 0 ? VTBLK_S_IOERR : VTBLK_S_OK;
278 * Return the single indirect descriptor back to the host
280 vu = &hq->hq_used_ring[uidx % hq->hq_size];
284 *hq->hq_used_idx += 1;
288 pci_vtblk_qnotify(struct pci_vtblk_softc *sc)
290 struct vring_hqueue *hq = &sc->vbsc_q;
295 * Calculate number of ring entries to process
297 ndescs = hq_num_avail(hq);
303 * Run through all the entries, placing them into iovecs and
304 * sending when an end-of-packet is found
306 for (i = 0; i < ndescs; i++)
307 pci_vtblk_proc(sc, hq);
310 * Generate an interrupt if able
312 if ((*hq->hq_avail_flags & VRING_AVAIL_F_NO_INTERRUPT) == 0) {
314 pci_generate_msix(sc->vbsc_pi, sc->msix_table_idx_req);
315 } else if (sc->vbsc_isr == 0) {
317 pci_generate_msi(sc->vbsc_pi, 0);
324 pci_vtblk_ring_init(struct pci_vtblk_softc *sc, uint64_t pfn)
326 struct vring_hqueue *hq;
328 sc->vbsc_pfn = pfn << VRING_PFN;
331 * Set up host pointers to the various parts of the
335 hq->hq_size = VTBLK_RINGSZ;
337 hq->hq_dtable = paddr_guest2host(pfn << VRING_PFN);
338 hq->hq_avail_flags = (uint16_t *)(hq->hq_dtable + hq->hq_size);
339 hq->hq_avail_idx = hq->hq_avail_flags + 1;
340 hq->hq_avail_ring = hq->hq_avail_flags + 2;
341 hq->hq_used_flags = (uint16_t *)roundup2((uintptr_t)hq->hq_avail_ring,
343 hq->hq_used_idx = hq->hq_used_flags + 1;
344 hq->hq_used_ring = (struct virtio_used *)(hq->hq_used_flags + 2);
347 * Initialize queue indexes
353 pci_vtblk_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts)
356 struct pci_vtblk_softc *sc;
363 printf("virtio-block: backing device required\n");
368 * Access to guest memory is required. Fail if
371 if (paddr_guest2host(0) == NULL)
375 * The supplied backing file has to exist
377 fd = open(opts, O_RDWR);
379 perror("Could not open backing file");
383 if (fstat(fd, &sbuf) < 0) {
384 perror("Could not stat backing file");
390 * Deal with raw devices
394 if (S_ISCHR(sbuf.st_mode)) {
395 if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 ||
396 ioctl(fd, DIOCGSECTORSIZE, §sz)) {
397 perror("Could not fetch dev blk/sector size");
405 sc = malloc(sizeof(struct pci_vtblk_softc));
406 memset(sc, 0, sizeof(struct pci_vtblk_softc));
412 /* setup virtio block config space */
413 sc->vbsc_cfg.vbc_capacity = size / sectsz;
414 sc->vbsc_cfg.vbc_seg_max = VTBLK_MAXSEGS;
415 sc->vbsc_cfg.vbc_blk_size = sectsz;
416 sc->vbsc_cfg.vbc_size_max = 0; /* not negotiated */
417 sc->vbsc_cfg.vbc_geom_c = 0; /* no geometry */
418 sc->vbsc_cfg.vbc_geom_h = 0;
419 sc->vbsc_cfg.vbc_geom_s = 0;
420 sc->vbsc_cfg.vbc_sectors_max = 0;
422 /* initialize config space */
423 pci_set_cfgdata16(pi, PCIR_DEVICE, VIRTIO_DEV_BLOCK);
424 pci_set_cfgdata16(pi, PCIR_VENDOR, VIRTIO_VENDOR);
425 pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE);
426 pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_BLOCK);
428 if ((env_msi = getenv("BHYVE_USE_MSI"))) {
429 if (strcasecmp(env_msi, "yes") == 0)
435 sc->msix_table_idx_req = VIRTIO_MSI_NO_VECTOR;
436 sc->msix_table_idx_cfg = VIRTIO_MSI_NO_VECTOR;
438 if (pci_emul_add_msixcap(pi, 2, 1))
442 pci_emul_add_msicap(pi, 1);
445 pci_emul_alloc_bar(pi, 0, PCIBAR_IO, VTBLK_REGSZ);
451 vtblk_adjust_offset(struct pci_devinst *pi, uint64_t offset)
454 * Device specific offsets used by guest would change
455 * based on whether MSI-X capability is enabled or not
457 if (!pci_msix_enabled(pi)) {
458 if (offset >= VTCFG_R_MSIX)
459 return (offset + (VTCFG_R_CFG1 - VTCFG_R_MSIX));
466 pci_vtblk_write(struct vmctx *ctx, int vcpu, struct pci_devinst *pi,
467 int baridx, uint64_t offset, int size, uint64_t value)
469 struct pci_vtblk_softc *sc = pi->pi_arg;
472 if (baridx == pci_msix_table_bar(pi) ||
473 baridx == pci_msix_pba_bar(pi)) {
474 pci_emul_msix_twrite(pi, offset, size, value);
481 if (offset + size > pci_vtblk_iosize(pi)) {
482 DPRINTF(("vtblk_write: 2big, offset %ld size %d\n",
487 offset = vtblk_adjust_offset(pi, offset);
490 case VTCFG_R_GUESTCAP:
492 sc->vbsc_features = value & VTBLK_S_HOSTCAPS;
496 pci_vtblk_ring_init(sc, value);
500 sc->vbsc_lastq = value;
502 case VTCFG_R_QNOTIFY:
505 pci_vtblk_qnotify(sc);
509 pci_vtblk_update_status(sc, value);
513 sc->msix_table_idx_cfg = value;
517 sc->msix_table_idx_req = value;
519 case VTCFG_R_HOSTCAP:
522 case VTBLK_R_CFG ... VTBLK_R_CFG_END:
523 DPRINTF(("vtblk: write to readonly reg %ld\n\r", offset));
526 DPRINTF(("vtblk: unknown i/o write offset %ld\n\r", offset));
533 pci_vtblk_read(struct vmctx *ctx, int vcpu, struct pci_devinst *pi,
534 int baridx, uint64_t offset, int size)
536 struct pci_vtblk_softc *sc = pi->pi_arg;
541 if (baridx == pci_msix_table_bar(pi) ||
542 baridx == pci_msix_pba_bar(pi)) {
543 return (pci_emul_msix_tread(pi, offset, size));
549 if (offset + size > pci_vtblk_iosize(pi)) {
550 DPRINTF(("vtblk_read: 2big, offset %ld size %d\n",
555 offset = vtblk_adjust_offset(pi, offset);
558 case VTCFG_R_HOSTCAP:
560 value = VTBLK_S_HOSTCAPS;
562 case VTCFG_R_GUESTCAP:
564 value = sc->vbsc_features; /* XXX never read ? */
568 value = sc->vbsc_pfn >> VRING_PFN;
571 value = (sc->vbsc_lastq == 0) ? VTBLK_RINGSZ: 0;
575 value = sc->vbsc_lastq; /* XXX never read ? */
577 case VTCFG_R_QNOTIFY:
579 value = 0; /* XXX never read ? */
583 value = sc->vbsc_status;
587 value = sc->vbsc_isr;
588 sc->vbsc_isr = 0; /* a read clears this flag */
592 value = sc->msix_table_idx_cfg;
596 value = sc->msix_table_idx_req;
598 case VTBLK_R_CFG ... VTBLK_R_CFG_END:
599 assert(size + offset <= (VTBLK_R_CFG_END + 1));
600 ptr = (uint8_t *)&sc->vbsc_cfg + offset - VTBLK_R_CFG;
602 value = *(uint8_t *) ptr;
603 } else if (size == 2) {
604 value = *(uint16_t *) ptr;
606 value = *(uint32_t *) ptr;
610 DPRINTF(("vtblk: unknown i/o read offset %ld\n\r", offset));
618 struct pci_devemu pci_de_vblk = {
619 .pe_emu = "virtio-blk",
620 .pe_init = pci_vtblk_init,
621 .pe_barwrite = pci_vtblk_write,
622 .pe_barread = pci_vtblk_read
624 PCI_EMUL_SET(pci_de_vblk);