Index: head/usr.sbin/bhyve/bhyve.8 =================================================================== --- head/usr.sbin/bhyve/bhyve.8 (revision 282787) +++ head/usr.sbin/bhyve/bhyve.8 (revision 282788) @@ -1,324 +1,325 @@ .\" Copyright (c) 2013 Peter Grehan .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd September 17, 2014 .Dt BHYVE 8 .Os .Sh NAME .Nm bhyve .Nd "run a guest operating system inside a virtual machine" .Sh SYNOPSIS .Nm .Op Fl abehuwxACHPWY .Op Fl c Ar numcpus .Op Fl g Ar gdbport .Op Fl l Ar lpcdev Ns Op , Ns Ar conf .Op Fl m Ar size Ns Op Ar K|k|M|m|G|g|T|t .Op Fl p Ar vcpu:hostcpu .Op Fl s Ar slot,emulation Ns Op , Ns Ar conf .Op Fl U Ar uuid .Ar vmname .Sh DESCRIPTION .Nm is a hypervisor that runs guest operating systems inside a virtual machine. .Pp Parameters such as the number of virtual CPUs, amount of guest memory, and I/O connectivity can be specified with command-line parameters. .Pp The guest operating system must be loaded with .Xr bhyveload 4 or a similar boot loader before running .Nm . .Pp .Nm runs until the guest operating system reboots or an unhandled hypervisor exit is detected. .Sh OPTIONS .Bl -tag -width 10n .It Fl a The guest's local APIC is configured in xAPIC mode. The xAPIC mode is the default setting so this option is redundant. It will be deprecated in a future version. .It Fl A Generate ACPI tables. Required for .Fx Ns /amd64 guests. .It Fl b Enable a low-level console device supported by .Fx kernels compiled with .Cd "device bvmconsole" . This option will be deprecated in a future version. .It Fl c Ar numcpus Number of guest virtual CPUs. The default is 1 and the maximum is 16. .It Fl C Include guest memory in core file. .It Fl e Force .Nm to exit when a guest issues an access to an I/O port that is not emulated. This is intended for debug purposes. .It Fl g Ar gdbport For .Fx kernels compiled with .Cd "device bvmdebug" , allow a remote kernel kgdb to be relayed to the guest kernel gdb stub via a local IPv4 address and this port. This option will be deprecated in a future version. .It Fl h Print help message and exit. .It Fl H Yield the virtual CPU thread when a HLT instruction is detected. If this option is not specified, virtual CPUs will use 100% of a host CPU. .It Fl l Ar lpcdev Ns Op , Ns Ar conf Allow devices behind the LPC PCI-ISA bridge to be configured. The only supported devices are the TTY-class devices, .Li com1 and .Li com2 . .It Fl m Ar size Ns Op Ar K|k|M|m|G|g|T|t Guest physical memory size in bytes. This must be the same size that was given to .Xr bhyveload 8 . .Pp The size argument may be suffixed with one of K, M, G or T (either upper or lower case) to indicate a multiple of kilobytes, megabytes, gigabytes, or terabytes. If no suffix is given, the value is assumed to be in megabytes. .It Fl p Ar vcpu:hostcpu Pin guest's virtual CPU .Em vcpu to .Em hostcpu . .It Fl P Force the guest virtual CPU to exit when a PAUSE instruction is detected. .It Fl s Ar slot,emulation Ns Op , Ns Ar conf Configure a virtual PCI slot and function. .Pp .Nm bhyve provides PCI bus emulation and virtual devices that can be attached to slots on the bus. There are 32 available slots, with the option of providing up to 8 functions per slot. .Bl -tag -width 10n .It Ar slot .Ar pcislot[:function] .Ar bus:pcislot:function .Pp The .Ar pcislot value is 0 to 31. The optional function value is 0 to 7. The optional .Ar bus value is 0 to 255. If not specified, the function value defaults to 0. If not specified, the bus value defaults to 0. .It Ar emulation .Bl -tag -width 10n .It Li hostbridge | Li amd_hostbridge .Pp Provide a simple host bridge. This is usually configured at slot 0, and is required by most guest operating systems. The .Li amd_hostbridge emulation is identical but uses a PCI vendor ID of .Li AMD . .It Li passthru PCI pass-through device. .It Li virtio-net Virtio network interface. .It Li virtio-blk Virtio block storage interface. .It Li virtio-rnd Virtio RNG interface. .It Li ahci-cd AHCI controller attached to an ATAPI CD/DVD. .It Li ahci-hd AHCI controller attached to a SATA hard-drive. .It Li uart PCI 16550 serial device. .It Li lpc LPC PCI-ISA bridge with COM1 and COM2 16550 serial ports. The LPC bridge emulation can only be configured on bus 0. .El .It Op Ar conf This optional parameter describes the backend for device emulations. If .Ar conf is not specified, the device emulation has no backend and can be considered unconnected. .Pp Network devices: .Bl -tag -width 10n .It Ar tapN Ns Op , Ns Ar mac=xx:xx:xx:xx:xx:xx .It Ar vmnetN Ns Op , Ns Ar mac=xx:xx:xx:xx:xx:xx .Pp If .Ar mac is not specified, the MAC address is derived from a fixed OUI and the remaining bytes from an MD5 hash of the slot and function numbers and the device name. .Pp The MAC address is an ASCII string in .Xr ethers 5 format. .El .Pp Block storage devices: .Bl -tag -width 10n -.It Pa /filename Ns Oo , Ns Li nocache Oc Ns Oo , Ns Li direct Oc Ns Oo , Ns Li ro Oc -.It Pa /dev/xxx Ns Oo , Ns Ar nocache Oc Ns Oo , Ns Ar direct Oc Ns Oo , Ns Ar ro Oc +.It Pa /filename Ns Oo , Ns Ar block-device-options Oc +.It Pa /dev/xxx Ns Oo , Ns Ar block-device-options Oc +.El +.Pp +The +.Ar block-device-options +are: .Bl -tag -width 8n .It Li nocache Open the file with .Dv O_DIRECT . .It Li direct Open the file using .Dv O_SYNC . .It Li ro Force the file to be opened read-only. -.El -.Pp -The -.Li nocache , -.Li direct , -and -.Li ro -options are not available for virtio block devices. +.It Li sectorsize= Ns Ar logical Ns Oo / Ns Ar physical Oc +Specify the logical and physical sector sizes of the emulated disk. +The physical sector size is optional and is equal to the logical sector size +if not explicitly specified. .El .Pp TTY devices: .Bl -tag -width 10n .It Li stdio Connect the serial port to the standard input and output of the bhyve process. .It Pa /dev/xxx Use the host TTY device for serial port I/O. .El .Pp Pass-through devices: .Bl -tag -width 10n .It Ns Ar slot Ns / Ns Ar bus Ns / Ns Ar function Connect to a PCI device on the host at the selector described by .Ar slot , .Ar bus , and .Ar function numbers. .El .Pp The host device must have been reserved at boot-time using the .Va pptdev loader variable as described in .Xr vmm 4 . .El .It Fl u RTC keeps UTC time. .It Fl U Ar uuid Set the universally unique identifier .Pq UUID in the guest's System Management BIOS System Information structure. By default a UUID is generated from the host's hostname and .Ar vmname . .It Fl w Ignore accesses to unimplemented Model Specific Registers (MSRs). This is intended for debug purposes. .It Fl W Force virtio PCI device emulations to use MSI interrupts instead of MSI-X interrupts. .It Fl x The guest's local APIC is configured in x2APIC mode. .It Fl Y Disable MPtable generation. .It Ar vmname Alphanumeric name of the guest. This should be the same as that created by .Xr bhyveload 8 . .El .Sh EXAMPLES The guest operating system must have been loaded with .Xr bhyveload 4 or a similar boot loader before .Xr bhyve 4 can be run. .Pp To run a virtual machine with 1GB of memory, two virtual CPUs, a virtio block device backed by the .Pa /my/image filesystem image, and a serial port for the console: .Bd -literal -offset indent bhyve -c 2 -s 0,hostbridge -s 1,lpc -s 2,virtio-blk,/my/image \\ -l com1,stdio -A -H -P -m 1G vm1 .Ed .Pp Run a 24GB single-CPU virtual machine with three network ports, one of which has a MAC address specified: .Bd -literal -offset indent bhyve -s 0,hostbridge -s 1,lpc -s 2:0,virtio-net,tap0 \\ -s 2:1,virtio-net,tap1 \\ -s 2:2,virtio-net,tap2,mac=00:be:fa:76:45:00 \\ -s 3,virtio-blk,/my/image -l com1,stdio \\ -A -H -P -m 24G bigvm .Ed .Pp Run an 8GB quad-CPU virtual machine with 8 AHCI SATA disks, an AHCI ATAPI CD-ROM, a single virtio network port, an AMD hostbridge, and the console port connected to an .Xr nmdm 4 null-model device. .Bd -literal -offset indent bhyve -c 4 \e\ -s 0,amd_hostbridge -s 1,lpc \\ -s 1:0,ahci-hd,/images/disk.1 \\ -s 1:1,ahci-hd,/images/disk.2 \\ -s 1:2,ahci-hd,/images/disk.3 \\ -s 1:3,ahci-hd,/images/disk.4 \\ -s 1:4,ahci-hd,/images/disk.5 \\ -s 1:5,ahci-hd,/images/disk.6 \\ -s 1:6,ahci-hd,/images/disk.7 \\ -s 1:7,ahci-hd,/images/disk.8 \\ -s 2,ahci-cd,/images.install.iso \\ -s 3,virtio-net,tap0 \\ -l com1,/dev/nmdm0A \\ -A -H -P -m 8G .Ed .Sh SEE ALSO .Xr bhyve 4 , .Xr nmdm 4 , .Xr vmm 4 , .Xr ethers 5 , .Xr bhyvectl 8 , .Xr bhyveload 8 .Sh HISTORY .Nm first appeared in .Fx 10.0 . .Sh AUTHORS .An Neel Natu Aq Mt neel@freebsd.org .An Peter Grehan Aq Mt grehan@freebsd.org Index: head/usr.sbin/bhyve/block_if.c =================================================================== --- head/usr.sbin/bhyve/block_if.c (revision 282787) +++ head/usr.sbin/bhyve/block_if.c (revision 282788) @@ -1,780 +1,822 @@ /*- * Copyright (c) 2013 Peter Grehan * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "bhyverun.h" #include "mevent.h" #include "block_if.h" #define BLOCKIF_SIG 0xb109b109 #define BLOCKIF_NUMTHR 8 #define BLOCKIF_MAXREQ (64 + BLOCKIF_NUMTHR) enum blockop { BOP_READ, BOP_WRITE, BOP_FLUSH, BOP_DELETE }; enum blockstat { BST_FREE, BST_BLOCK, BST_PEND, BST_BUSY, BST_DONE }; struct blockif_elem { TAILQ_ENTRY(blockif_elem) be_link; struct blockif_req *be_req; enum blockop be_op; enum blockstat be_status; pthread_t be_tid; off_t be_block; }; struct blockif_ctxt { int bc_magic; int bc_fd; int bc_ischr; int bc_isgeom; int bc_candelete; int bc_rdonly; off_t bc_size; int bc_sectsz; int bc_psectsz; int bc_psectoff; int bc_closing; pthread_t bc_btid[BLOCKIF_NUMTHR]; pthread_mutex_t bc_mtx; pthread_cond_t bc_cond; /* Request elements and free/pending/busy queues */ TAILQ_HEAD(, blockif_elem) bc_freeq; TAILQ_HEAD(, blockif_elem) bc_pendq; TAILQ_HEAD(, blockif_elem) bc_busyq; struct blockif_elem bc_reqs[BLOCKIF_MAXREQ]; }; static pthread_once_t blockif_once = PTHREAD_ONCE_INIT; struct blockif_sig_elem { pthread_mutex_t bse_mtx; pthread_cond_t bse_cond; int bse_pending; struct blockif_sig_elem *bse_next; }; static struct blockif_sig_elem *blockif_bse_head; static int blockif_enqueue(struct blockif_ctxt *bc, struct blockif_req *breq, enum blockop op) { struct blockif_elem *be, *tbe; off_t off; int i; be = TAILQ_FIRST(&bc->bc_freeq); assert(be != NULL); assert(be->be_status == BST_FREE); TAILQ_REMOVE(&bc->bc_freeq, be, be_link); be->be_req = breq; be->be_op = op; switch (op) { case BOP_READ: case BOP_WRITE: case BOP_DELETE: off = breq->br_offset; for (i = 0; i < breq->br_iovcnt; i++) off += breq->br_iov[i].iov_len; break; default: off = OFF_MAX; } be->be_block = off; TAILQ_FOREACH(tbe, &bc->bc_pendq, be_link) { if (tbe->be_block == breq->br_offset) break; } if (tbe == NULL) { TAILQ_FOREACH(tbe, &bc->bc_busyq, be_link) { if (tbe->be_block == breq->br_offset) break; } } if (tbe == NULL) be->be_status = BST_PEND; else be->be_status = BST_BLOCK; TAILQ_INSERT_TAIL(&bc->bc_pendq, be, be_link); return (be->be_status == BST_PEND); } static int blockif_dequeue(struct blockif_ctxt *bc, pthread_t t, struct blockif_elem **bep) { struct blockif_elem *be; TAILQ_FOREACH(be, &bc->bc_pendq, be_link) { if (be->be_status == BST_PEND) break; assert(be->be_status == BST_BLOCK); } if (be == NULL) return (0); TAILQ_REMOVE(&bc->bc_pendq, be, be_link); be->be_status = BST_BUSY; be->be_tid = t; TAILQ_INSERT_TAIL(&bc->bc_busyq, be, be_link); *bep = be; return (1); } static void blockif_complete(struct blockif_ctxt *bc, struct blockif_elem *be) { struct blockif_elem *tbe; if (be->be_status == BST_DONE || be->be_status == BST_BUSY) TAILQ_REMOVE(&bc->bc_busyq, be, be_link); else TAILQ_REMOVE(&bc->bc_pendq, be, be_link); TAILQ_FOREACH(tbe, &bc->bc_pendq, be_link) { if (tbe->be_req->br_offset == be->be_block) tbe->be_status = BST_PEND; } be->be_tid = 0; be->be_status = BST_FREE; be->be_req = NULL; TAILQ_INSERT_TAIL(&bc->bc_freeq, be, be_link); } static void blockif_proc(struct blockif_ctxt *bc, struct blockif_elem *be, uint8_t *buf) { struct blockif_req *br; off_t arg[2]; ssize_t clen, len, off, boff, voff; int i, err; br = be->be_req; if (br->br_iovcnt <= 1) buf = NULL; err = 0; switch (be->be_op) { case BOP_READ: if (buf == NULL) { if ((len = preadv(bc->bc_fd, br->br_iov, br->br_iovcnt, br->br_offset)) < 0) err = errno; else br->br_resid -= len; break; } i = 0; off = voff = 0; while (br->br_resid > 0) { len = MIN(br->br_resid, MAXPHYS); if (pread(bc->bc_fd, buf, len, br->br_offset + off) < 0) { err = errno; break; } boff = 0; do { clen = MIN(len - boff, br->br_iov[i].iov_len - voff); memcpy(br->br_iov[i].iov_base + voff, buf + boff, clen); if (clen < br->br_iov[i].iov_len - voff) voff += clen; else { i++; voff = 0; } boff += clen; } while (boff < len); off += len; br->br_resid -= len; } break; case BOP_WRITE: if (bc->bc_rdonly) { err = EROFS; break; } if (buf == NULL) { if ((len = pwritev(bc->bc_fd, br->br_iov, br->br_iovcnt, br->br_offset)) < 0) err = errno; else br->br_resid -= len; break; } i = 0; off = voff = 0; while (br->br_resid > 0) { len = MIN(br->br_resid, MAXPHYS); boff = 0; do { clen = MIN(len - boff, br->br_iov[i].iov_len - voff); memcpy(buf + boff, br->br_iov[i].iov_base + voff, clen); if (clen < br->br_iov[i].iov_len - voff) voff += clen; else { i++; voff = 0; } boff += clen; } while (boff < len); if (pwrite(bc->bc_fd, buf, len, br->br_offset + off) < 0) { err = errno; break; } off += len; br->br_resid -= len; } break; case BOP_FLUSH: if (bc->bc_ischr) { if (ioctl(bc->bc_fd, DIOCGFLUSH)) err = errno; } else if (fsync(bc->bc_fd)) err = errno; break; case BOP_DELETE: if (!bc->bc_candelete) err = EOPNOTSUPP; else if (bc->bc_rdonly) err = EROFS; else if (bc->bc_ischr) { arg[0] = br->br_offset; arg[1] = br->br_resid; if (ioctl(bc->bc_fd, DIOCGDELETE, arg)) err = errno; else br->br_resid = 0; } else err = EOPNOTSUPP; break; default: err = EINVAL; break; } be->be_status = BST_DONE; (*br->br_callback)(br, err); } static void * blockif_thr(void *arg) { struct blockif_ctxt *bc; struct blockif_elem *be; pthread_t t; uint8_t *buf; bc = arg; if (bc->bc_isgeom) buf = malloc(MAXPHYS); else buf = NULL; t = pthread_self(); pthread_mutex_lock(&bc->bc_mtx); for (;;) { while (blockif_dequeue(bc, t, &be)) { pthread_mutex_unlock(&bc->bc_mtx); blockif_proc(bc, be, buf); pthread_mutex_lock(&bc->bc_mtx); blockif_complete(bc, be); } /* Check ctxt status here to see if exit requested */ if (bc->bc_closing) break; pthread_cond_wait(&bc->bc_cond, &bc->bc_mtx); } pthread_mutex_unlock(&bc->bc_mtx); if (buf) free(buf); pthread_exit(NULL); return (NULL); } static void blockif_sigcont_handler(int signal, enum ev_type type, void *arg) { struct blockif_sig_elem *bse; for (;;) { /* * Process the entire list even if not intended for * this thread. */ do { bse = blockif_bse_head; if (bse == NULL) return; } while (!atomic_cmpset_ptr((uintptr_t *)&blockif_bse_head, (uintptr_t)bse, (uintptr_t)bse->bse_next)); pthread_mutex_lock(&bse->bse_mtx); bse->bse_pending = 0; pthread_cond_signal(&bse->bse_cond); pthread_mutex_unlock(&bse->bse_mtx); } } static void blockif_init(void) { mevent_add(SIGCONT, EVF_SIGNAL, blockif_sigcont_handler, NULL); (void) signal(SIGCONT, SIG_IGN); } struct blockif_ctxt * blockif_open(const char *optstr, const char *ident) { char tname[MAXCOMLEN + 1]; char name[MAXPATHLEN]; - char *nopt, *xopts; + char *nopt, *xopts, *cp; struct blockif_ctxt *bc; struct stat sbuf; struct diocgattr_arg arg; off_t size, psectsz, psectoff; int extra, fd, i, sectsz; - int nocache, sync, ro, candelete, geom; + int nocache, sync, ro, candelete, geom, ssopt, pssopt; pthread_once(&blockif_once, blockif_init); + fd = -1; + ssopt = 0; nocache = 0; sync = 0; ro = 0; /* * The first element in the optstring is always a pathname. * Optional elements follow */ - nopt = strdup(optstr); - for (xopts = strtok(nopt, ","); - xopts != NULL; - xopts = strtok(NULL, ",")) { - if (!strcmp(xopts, "nocache")) + nopt = xopts = strdup(optstr); + while (xopts != NULL) { + cp = strsep(&xopts, ","); + if (cp == nopt) /* file or device pathname */ + continue; + else if (!strcmp(cp, "nocache")) nocache = 1; - else if (!strcmp(xopts, "sync")) + else if (!strcmp(cp, "sync") || !strcmp(cp, "direct")) sync = 1; - else if (!strcmp(xopts, "ro")) + else if (!strcmp(cp, "ro")) ro = 1; + else if (sscanf(cp, "sectorsize=%d/%d", &ssopt, &pssopt) == 2) + ; + else if (sscanf(cp, "sectorsize=%d", &ssopt) == 1) + pssopt = ssopt; + else { + fprintf(stderr, "Invalid device option \"%s\"\n", cp); + goto err; + } } extra = 0; if (nocache) extra |= O_DIRECT; if (sync) extra |= O_SYNC; fd = open(nopt, (ro ? O_RDONLY : O_RDWR) | extra); if (fd < 0 && !ro) { /* Attempt a r/w fail with a r/o open */ fd = open(nopt, O_RDONLY | extra); ro = 1; } if (fd < 0) { perror("Could not open backing file"); - return (NULL); + goto err; } if (fstat(fd, &sbuf) < 0) { perror("Could not stat backing file"); - close(fd); - return (NULL); + goto err; } /* * Deal with raw devices */ size = sbuf.st_size; sectsz = DEV_BSIZE; psectsz = psectoff = 0; candelete = geom = 0; if (S_ISCHR(sbuf.st_mode)) { if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 || ioctl(fd, DIOCGSECTORSIZE, §sz)) { perror("Could not fetch dev blk/sector size"); - close(fd); - return (NULL); + goto err; } assert(size != 0); assert(sectsz != 0); if (ioctl(fd, DIOCGSTRIPESIZE, &psectsz) == 0 && psectsz > 0) ioctl(fd, DIOCGSTRIPEOFFSET, &psectoff); strlcpy(arg.name, "GEOM::candelete", sizeof(arg.name)); arg.len = sizeof(arg.value.i); if (ioctl(fd, DIOCGATTR, &arg) == 0) candelete = arg.value.i; if (ioctl(fd, DIOCGPROVIDERNAME, name) == 0) geom = 1; } else psectsz = sbuf.st_blksize; + if (ssopt != 0) { + if (!powerof2(ssopt) || !powerof2(pssopt) || ssopt < 512 || + ssopt > pssopt) { + fprintf(stderr, "Invalid sector size %d/%d\n", + ssopt, pssopt); + goto err; + } + + /* + * Some backend drivers (e.g. cd0, ada0) require that the I/O + * size be a multiple of the device's sector size. + * + * Validate that the emulated sector size complies with this + * requirement. + */ + if (S_ISCHR(sbuf.st_mode)) { + if (ssopt < sectsz || (ssopt % sectsz) != 0) { + fprintf(stderr, "Sector size %d incompatible " + "with underlying device sector size %d\n", + ssopt, sectsz); + goto err; + } + } + + sectsz = ssopt; + psectsz = pssopt; + psectoff = 0; + } + bc = calloc(1, sizeof(struct blockif_ctxt)); if (bc == NULL) { - close(fd); - return (NULL); + perror("calloc"); + goto err; } bc->bc_magic = BLOCKIF_SIG; bc->bc_fd = fd; bc->bc_ischr = S_ISCHR(sbuf.st_mode); bc->bc_isgeom = geom; bc->bc_candelete = candelete; bc->bc_rdonly = ro; bc->bc_size = size; bc->bc_sectsz = sectsz; bc->bc_psectsz = psectsz; bc->bc_psectoff = psectoff; pthread_mutex_init(&bc->bc_mtx, NULL); pthread_cond_init(&bc->bc_cond, NULL); TAILQ_INIT(&bc->bc_freeq); TAILQ_INIT(&bc->bc_pendq); TAILQ_INIT(&bc->bc_busyq); for (i = 0; i < BLOCKIF_MAXREQ; i++) { bc->bc_reqs[i].be_status = BST_FREE; TAILQ_INSERT_HEAD(&bc->bc_freeq, &bc->bc_reqs[i], be_link); } for (i = 0; i < BLOCKIF_NUMTHR; i++) { pthread_create(&bc->bc_btid[i], NULL, blockif_thr, bc); snprintf(tname, sizeof(tname), "blk-%s-%d", ident, i); pthread_set_name_np(bc->bc_btid[i], tname); } return (bc); +err: + if (fd >= 0) + close(fd); + return (NULL); } static int blockif_request(struct blockif_ctxt *bc, struct blockif_req *breq, enum blockop op) { int err; err = 0; pthread_mutex_lock(&bc->bc_mtx); if (!TAILQ_EMPTY(&bc->bc_freeq)) { /* * Enqueue and inform the block i/o thread * that there is work available */ if (blockif_enqueue(bc, breq, op)) pthread_cond_signal(&bc->bc_cond); } else { /* * Callers are not allowed to enqueue more than * the specified blockif queue limit. Return an * error to indicate that the queue length has been * exceeded. */ err = E2BIG; } pthread_mutex_unlock(&bc->bc_mtx); return (err); } int blockif_read(struct blockif_ctxt *bc, struct blockif_req *breq) { assert(bc->bc_magic == BLOCKIF_SIG); return (blockif_request(bc, breq, BOP_READ)); } int blockif_write(struct blockif_ctxt *bc, struct blockif_req *breq) { assert(bc->bc_magic == BLOCKIF_SIG); return (blockif_request(bc, breq, BOP_WRITE)); } int blockif_flush(struct blockif_ctxt *bc, struct blockif_req *breq) { assert(bc->bc_magic == BLOCKIF_SIG); return (blockif_request(bc, breq, BOP_FLUSH)); } int blockif_delete(struct blockif_ctxt *bc, struct blockif_req *breq) { assert(bc->bc_magic == BLOCKIF_SIG); return (blockif_request(bc, breq, BOP_DELETE)); } int blockif_cancel(struct blockif_ctxt *bc, struct blockif_req *breq) { struct blockif_elem *be; assert(bc->bc_magic == BLOCKIF_SIG); pthread_mutex_lock(&bc->bc_mtx); /* * Check pending requests. */ TAILQ_FOREACH(be, &bc->bc_pendq, be_link) { if (be->be_req == breq) break; } if (be != NULL) { /* * Found it. */ blockif_complete(bc, be); pthread_mutex_unlock(&bc->bc_mtx); return (0); } /* * Check in-flight requests. */ TAILQ_FOREACH(be, &bc->bc_busyq, be_link) { if (be->be_req == breq) break; } if (be == NULL) { /* * Didn't find it. */ pthread_mutex_unlock(&bc->bc_mtx); return (EINVAL); } /* * Interrupt the processing thread to force it return * prematurely via it's normal callback path. */ while (be->be_status == BST_BUSY) { struct blockif_sig_elem bse, *old_head; pthread_mutex_init(&bse.bse_mtx, NULL); pthread_cond_init(&bse.bse_cond, NULL); bse.bse_pending = 1; do { old_head = blockif_bse_head; bse.bse_next = old_head; } while (!atomic_cmpset_ptr((uintptr_t *)&blockif_bse_head, (uintptr_t)old_head, (uintptr_t)&bse)); pthread_kill(be->be_tid, SIGCONT); pthread_mutex_lock(&bse.bse_mtx); while (bse.bse_pending) pthread_cond_wait(&bse.bse_cond, &bse.bse_mtx); pthread_mutex_unlock(&bse.bse_mtx); } pthread_mutex_unlock(&bc->bc_mtx); /* * The processing thread has been interrupted. Since it's not * clear if the callback has been invoked yet, return EBUSY. */ return (EBUSY); } int blockif_close(struct blockif_ctxt *bc) { void *jval; int err, i; err = 0; assert(bc->bc_magic == BLOCKIF_SIG); /* * Stop the block i/o thread */ pthread_mutex_lock(&bc->bc_mtx); bc->bc_closing = 1; pthread_mutex_unlock(&bc->bc_mtx); pthread_cond_broadcast(&bc->bc_cond); for (i = 0; i < BLOCKIF_NUMTHR; i++) pthread_join(bc->bc_btid[i], &jval); /* XXX Cancel queued i/o's ??? */ /* * Release resources */ bc->bc_magic = 0; close(bc->bc_fd); free(bc); return (0); } /* * Return virtual C/H/S values for a given block. Use the algorithm * outlined in the VHD specification to calculate values. */ void blockif_chs(struct blockif_ctxt *bc, uint16_t *c, uint8_t *h, uint8_t *s) { off_t sectors; /* total sectors of the block dev */ off_t hcyl; /* cylinders times heads */ uint16_t secpt; /* sectors per track */ uint8_t heads; assert(bc->bc_magic == BLOCKIF_SIG); sectors = bc->bc_size / bc->bc_sectsz; /* Clamp the size to the largest possible with CHS */ if (sectors > 65535UL*16*255) sectors = 65535UL*16*255; if (sectors >= 65536UL*16*63) { secpt = 255; heads = 16; hcyl = sectors / secpt; } else { secpt = 17; hcyl = sectors / secpt; heads = (hcyl + 1023) / 1024; if (heads < 4) heads = 4; if (hcyl >= (heads * 1024) || heads > 16) { secpt = 31; heads = 16; hcyl = sectors / secpt; } if (hcyl >= (heads * 1024)) { secpt = 63; heads = 16; hcyl = sectors / secpt; } } *c = hcyl / heads; *h = heads; *s = secpt; } /* * Accessors */ off_t blockif_size(struct blockif_ctxt *bc) { assert(bc->bc_magic == BLOCKIF_SIG); return (bc->bc_size); } int blockif_sectsz(struct blockif_ctxt *bc) { assert(bc->bc_magic == BLOCKIF_SIG); return (bc->bc_sectsz); } void blockif_psectsz(struct blockif_ctxt *bc, int *size, int *off) { assert(bc->bc_magic == BLOCKIF_SIG); *size = bc->bc_psectsz; *off = bc->bc_psectoff; } int blockif_queuesz(struct blockif_ctxt *bc) { assert(bc->bc_magic == BLOCKIF_SIG); return (BLOCKIF_MAXREQ - 1); } int blockif_is_ro(struct blockif_ctxt *bc) { assert(bc->bc_magic == BLOCKIF_SIG); return (bc->bc_rdonly); } int blockif_candelete(struct blockif_ctxt *bc) { assert(bc->bc_magic == BLOCKIF_SIG); return (bc->bc_candelete); }