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>
56 #define VTBLK_RINGSZ 64
59 #define VTBLK_S_IOERR 1
60 #define VTBLK_S_UNSUPP 2
62 #define VTBLK_BLK_ID_BYTES 20
65 #define VTBLK_F_SEG_MAX (1 << 2) /* Maximum request segments */
66 #define VTBLK_F_BLK_SIZE (1 << 6) /* cfg block size valid */
67 #define VTBLK_F_TOPOLOGY (1 << 10) /* Optimal I/O alignment */
72 #define VTBLK_S_HOSTCAPS \
76 VIRTIO_RING_F_INDIRECT_DESC ) /* indirect descriptors */
79 * Config space "registers"
82 uint64_t vbc_capacity;
83 uint32_t vbc_size_max;
90 uint32_t vbc_blk_size;
92 uint8_t physical_block_exp;
93 uint8_t alignment_offset;
97 uint8_t vbc_writeback;
101 * Fixed-size block header
103 struct virtio_blk_hdr {
104 #define VBH_OP_READ 0
105 #define VBH_OP_WRITE 1
106 #define VBH_OP_FLUSH 4
107 #define VBH_OP_FLUSH_OUT 5
108 #define VBH_OP_IDENT 8
109 #define VBH_FLAG_BARRIER 0x80000000 /* OR'ed into vbh_type */
118 static int pci_vtblk_debug;
119 #define DPRINTF(params) if (pci_vtblk_debug) printf params
120 #define WPRINTF(params) printf params
122 struct pci_vtblk_ioreq {
123 struct blockif_req io_req;
124 struct pci_vtblk_softc *io_sc;
132 struct pci_vtblk_softc {
133 struct virtio_softc vbsc_vs;
134 pthread_mutex_t vsc_mtx;
135 struct vqueue_info vbsc_vq;
136 struct vtblk_config vbsc_cfg;
137 struct blockif_ctxt *bc;
138 char vbsc_ident[VTBLK_BLK_ID_BYTES];
139 struct pci_vtblk_ioreq vbsc_ios[VTBLK_RINGSZ];
142 static void pci_vtblk_reset(void *);
143 static void pci_vtblk_notify(void *, struct vqueue_info *);
144 static int pci_vtblk_cfgread(void *, int, int, uint32_t *);
145 static int pci_vtblk_cfgwrite(void *, int, int, uint32_t);
147 static struct virtio_consts vtblk_vi_consts = {
148 "vtblk", /* our name */
149 1, /* we support 1 virtqueue */
150 sizeof(struct vtblk_config), /* config reg size */
151 pci_vtblk_reset, /* reset */
152 pci_vtblk_notify, /* device-wide qnotify */
153 pci_vtblk_cfgread, /* read PCI config */
154 pci_vtblk_cfgwrite, /* write PCI config */
155 NULL, /* apply negotiated features */
156 VTBLK_S_HOSTCAPS, /* our capabilities */
160 pci_vtblk_reset(void *vsc)
162 struct pci_vtblk_softc *sc = vsc;
164 DPRINTF(("vtblk: device reset requested !\n"));
165 vi_reset_dev(&sc->vbsc_vs);
169 pci_vtblk_done(struct blockif_req *br, int err)
171 struct pci_vtblk_ioreq *io = br->br_param;
172 struct pci_vtblk_softc *sc = io->io_sc;
174 /* convert errno into a virtio block error return */
175 if (err == EOPNOTSUPP || err == ENOSYS)
176 *io->io_status = VTBLK_S_UNSUPP;
178 *io->io_status = VTBLK_S_IOERR;
180 *io->io_status = VTBLK_S_OK;
183 * Return the descriptor back to the host.
184 * We wrote 1 byte (our status) to host.
186 pthread_mutex_lock(&sc->vsc_mtx);
187 vq_relchain(&sc->vbsc_vq, io->io_idx, 1);
188 vq_endchains(&sc->vbsc_vq, 0);
189 pthread_mutex_unlock(&sc->vsc_mtx);
193 pci_vtblk_proc(struct pci_vtblk_softc *sc, struct vqueue_info *vq)
195 struct virtio_blk_hdr *vbh;
196 struct pci_vtblk_ioreq *io;
202 struct iovec iov[BLOCKIF_IOV_MAX + 2];
203 uint16_t idx, flags[BLOCKIF_IOV_MAX + 2];
205 n = vq_getchain(vq, &idx, iov, BLOCKIF_IOV_MAX + 2, flags);
208 * The first descriptor will be the read-only fixed header,
209 * and the last is for status (hence +2 above and below).
210 * The remaining iov's are the actual data I/O vectors.
212 * XXX - note - this fails on crash dump, which does a
213 * VIRTIO_BLK_T_FLUSH with a zero transfer length
215 assert(n >= 2 && n <= BLOCKIF_IOV_MAX + 2);
217 io = &sc->vbsc_ios[idx];
218 assert((flags[0] & VRING_DESC_F_WRITE) == 0);
219 assert(iov[0].iov_len == sizeof(struct virtio_blk_hdr));
220 vbh = iov[0].iov_base;
221 memcpy(&io->io_req.br_iov, &iov[1], sizeof(struct iovec) * (n - 2));
222 io->io_req.br_iovcnt = n - 2;
223 io->io_req.br_offset = vbh->vbh_sector * DEV_BSIZE;
224 io->io_status = iov[--n].iov_base;
225 assert(iov[n].iov_len == 1);
226 assert(flags[n] & VRING_DESC_F_WRITE);
230 * The guest should not be setting the BARRIER flag because
231 * we don't advertise the capability.
233 type = vbh->vbh_type & ~VBH_FLAG_BARRIER;
234 writeop = (type == VBH_OP_WRITE);
237 for (i = 1; i < n; i++) {
239 * - write op implies read-only descriptor,
240 * - read/ident op implies write-only descriptor,
241 * therefore test the inverse of the descriptor bit
244 assert(((flags[i] & VRING_DESC_F_WRITE) == 0) == writeop);
245 iolen += iov[i].iov_len;
248 DPRINTF(("virtio-block: %s op, %d bytes, %d segs, offset %ld\n\r",
249 writeop ? "write" : "read/ident", iolen, i - 1, offset));
253 err = blockif_read(sc->bc, &io->io_req);
256 err = blockif_write(sc->bc, &io->io_req);
259 case VBH_OP_FLUSH_OUT:
260 err = blockif_flush(sc->bc, &io->io_req);
263 /* Assume a single buffer */
264 /* S/n equal to buffer is not zero-terminated. */
265 memset(iov[1].iov_base, 0, iov[1].iov_len);
266 strncpy(iov[1].iov_base, sc->vbsc_ident,
267 MIN(iov[1].iov_len, sizeof(sc->vbsc_ident)));
268 pci_vtblk_done(&io->io_req, 0);
271 pci_vtblk_done(&io->io_req, EOPNOTSUPP);
278 pci_vtblk_notify(void *vsc, struct vqueue_info *vq)
280 struct pci_vtblk_softc *sc = vsc;
282 while (vq_has_descs(vq))
283 pci_vtblk_proc(sc, vq);
287 pci_vtblk_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts)
289 char bident[sizeof("XX:X:X")];
290 struct blockif_ctxt *bctxt;
293 struct pci_vtblk_softc *sc;
295 int i, sectsz, sts, sto;
298 printf("virtio-block: backing device required\n");
303 * The supplied backing file has to exist
305 snprintf(bident, sizeof(bident), "%d:%d", pi->pi_slot, pi->pi_func);
306 bctxt = blockif_open(opts, bident);
308 perror("Could not open backing file");
312 size = blockif_size(bctxt);
313 sectsz = blockif_sectsz(bctxt);
314 blockif_psectsz(bctxt, &sts, &sto);
316 sc = calloc(1, sizeof(struct pci_vtblk_softc));
318 for (i = 0; i < VTBLK_RINGSZ; i++) {
319 struct pci_vtblk_ioreq *io = &sc->vbsc_ios[i];
320 io->io_req.br_callback = pci_vtblk_done;
321 io->io_req.br_param = io;
326 pthread_mutex_init(&sc->vsc_mtx, NULL);
328 /* init virtio softc and virtqueues */
329 vi_softc_linkup(&sc->vbsc_vs, &vtblk_vi_consts, sc, pi, &sc->vbsc_vq);
330 sc->vbsc_vs.vs_mtx = &sc->vsc_mtx;
332 sc->vbsc_vq.vq_qsize = VTBLK_RINGSZ;
333 /* sc->vbsc_vq.vq_notify = we have no per-queue notify */
336 * Create an identifier for the backing file. Use parts of the
337 * md5 sum of the filename
340 MD5Update(&mdctx, opts, strlen(opts));
341 MD5Final(digest, &mdctx);
342 sprintf(sc->vbsc_ident, "BHYVE-%02X%02X-%02X%02X-%02X%02X",
343 digest[0], digest[1], digest[2], digest[3], digest[4], digest[5]);
345 /* setup virtio block config space */
346 sc->vbsc_cfg.vbc_capacity = size / DEV_BSIZE; /* 512-byte units */
347 sc->vbsc_cfg.vbc_size_max = 0; /* not negotiated */
348 sc->vbsc_cfg.vbc_seg_max = BLOCKIF_IOV_MAX;
349 sc->vbsc_cfg.vbc_geometry.cylinders = 0; /* no geometry */
350 sc->vbsc_cfg.vbc_geometry.heads = 0;
351 sc->vbsc_cfg.vbc_geometry.sectors = 0;
352 sc->vbsc_cfg.vbc_blk_size = sectsz;
353 sc->vbsc_cfg.vbc_topology.physical_block_exp =
354 (sts > sectsz) ? (ffsll(sts / sectsz) - 1) : 0;
355 sc->vbsc_cfg.vbc_topology.alignment_offset =
356 (sto != 0) ? ((sts - sto) / sectsz) : 0;
357 sc->vbsc_cfg.vbc_topology.min_io_size = 0;
358 sc->vbsc_cfg.vbc_topology.opt_io_size = 0;
359 sc->vbsc_cfg.vbc_writeback = 0;
362 * Should we move some of this into virtio.c? Could
363 * have the device, class, and subdev_0 as fields in
364 * the virtio constants structure.
366 pci_set_cfgdata16(pi, PCIR_DEVICE, VIRTIO_DEV_BLOCK);
367 pci_set_cfgdata16(pi, PCIR_VENDOR, VIRTIO_VENDOR);
368 pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE);
369 pci_set_cfgdata16(pi, PCIR_SUBDEV_0, VIRTIO_TYPE_BLOCK);
371 pci_lintr_request(pi);
373 if (vi_intr_init(&sc->vbsc_vs, 1, fbsdrun_virtio_msix())) {
374 blockif_close(sc->bc);
378 vi_set_io_bar(&sc->vbsc_vs, 0);
383 pci_vtblk_cfgwrite(void *vsc, int offset, int size, uint32_t value)
386 DPRINTF(("vtblk: write to readonly reg %d\n\r", offset));
391 pci_vtblk_cfgread(void *vsc, int offset, int size, uint32_t *retval)
393 struct pci_vtblk_softc *sc = vsc;
396 /* our caller has already verified offset and size */
397 ptr = (uint8_t *)&sc->vbsc_cfg + offset;
398 memcpy(retval, ptr, size);
402 struct pci_devemu pci_de_vblk = {
403 .pe_emu = "virtio-blk",
404 .pe_init = pci_vtblk_init,
405 .pe_barwrite = vi_pci_write,
406 .pe_barread = vi_pci_read
408 PCI_EMUL_SET(pci_de_vblk);