2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2011 NetApp, Inc.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
34 #include <sys/param.h>
35 #include <sys/sysctl.h>
36 #include <sys/ioctl.h>
37 #include <sys/linker.h>
39 #include <sys/module.h>
40 #include <sys/_iovec.h>
41 #include <sys/cpuset.h>
43 #include <x86/segments.h>
44 #include <machine/specialreg.h>
58 #include <machine/vmm.h>
59 #include <machine/vmm_dev.h>
60 #include <machine/vmm_snapshot.h>
64 #define MB (1024 * 1024UL)
65 #define GB (1024 * 1024 * 1024UL)
68 * Size of the guard region before and after the virtual address space
69 * mapping the guest physical memory. This must be a multiple of the
70 * superpage size for performance reasons.
72 #define VM_MMAP_GUARD_SIZE (4 * MB)
74 #define PROT_RW (PROT_READ | PROT_WRITE)
75 #define PROT_ALL (PROT_READ | PROT_WRITE | PROT_EXEC)
79 uint32_t lowmem_limit;
87 #define CREATE(x) sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
88 #define DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
91 vm_device_open(const char *name)
96 len = strlen("/dev/vmm/") + strlen(name) + 1;
98 assert(vmfile != NULL);
99 snprintf(vmfile, len, "/dev/vmm/%s", name);
101 /* Open the device file */
102 fd = open(vmfile, O_RDWR, 0);
109 vm_create(const char *name)
111 /* Try to load vmm(4) module before creating a guest. */
112 if (modfind("vmm") < 0)
114 return (CREATE(name));
118 vm_open(const char *name)
123 vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
128 vm->lowmem_limit = 3 * GB;
129 vm->name = (char *)(vm + 1);
130 strcpy(vm->name, name);
132 if ((vm->fd = vm_device_open(vm->name)) < 0)
144 vm_destroy(struct vmctx *vm)
156 vm_parse_memsize(const char *opt, size_t *ret_memsize)
162 optval = strtoul(opt, &endptr, 0);
163 if (*opt != '\0' && *endptr == '\0') {
165 * For the sake of backward compatibility if the memory size
166 * specified on the command line is less than a megabyte then
167 * it is interpreted as being in units of MB.
171 *ret_memsize = optval;
174 error = expand_number(opt, ret_memsize);
180 vm_get_lowmem_limit(struct vmctx *ctx)
183 return (ctx->lowmem_limit);
187 vm_set_lowmem_limit(struct vmctx *ctx, uint32_t limit)
190 ctx->lowmem_limit = limit;
194 vm_set_memflags(struct vmctx *ctx, int flags)
197 ctx->memflags = flags;
201 vm_get_memflags(struct vmctx *ctx)
204 return (ctx->memflags);
208 * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len).
211 vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off,
212 size_t len, int prot)
214 struct vm_memmap memmap;
218 memmap.segid = segid;
224 if (ctx->memflags & VM_MEM_F_WIRED)
225 memmap.flags |= VM_MEMMAP_F_WIRED;
228 * If this mapping already exists then don't create it again. This
229 * is the common case for SYSMEM mappings created by bhyveload(8).
231 error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags);
232 if (error == 0 && gpa == memmap.gpa) {
233 if (segid != memmap.segid || off != memmap.segoff ||
234 prot != memmap.prot || flags != memmap.flags) {
242 error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap);
247 vm_get_guestmem_from_ctx(struct vmctx *ctx, char **guest_baseaddr,
248 size_t *lowmem_size, size_t *highmem_size)
251 *guest_baseaddr = ctx->baseaddr;
252 *lowmem_size = ctx->lowmem;
253 *highmem_size = ctx->highmem;
258 vm_munmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, size_t len)
260 struct vm_munmap munmap;
266 error = ioctl(ctx->fd, VM_MUNMAP_MEMSEG, &munmap);
271 vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid,
272 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags)
274 struct vm_memmap memmap;
277 bzero(&memmap, sizeof(struct vm_memmap));
279 error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap);
282 *segid = memmap.segid;
283 *segoff = memmap.segoff;
286 *flags = memmap.flags;
292 * Return 0 if the segments are identical and non-zero otherwise.
294 * This is slightly complicated by the fact that only device memory segments
298 cmpseg(size_t len, const char *str, size_t len2, const char *str2)
302 if ((!str && !str2) || (str && str2 && !strcmp(str, str2)))
309 vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name)
311 struct vm_memseg memseg;
316 * If the memory segment has already been created then just return.
317 * This is the usual case for the SYSMEM segment created by userspace
318 * loaders like bhyveload(8).
320 error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name,
321 sizeof(memseg.name));
325 if (memseg.len != 0) {
326 if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) {
334 bzero(&memseg, sizeof(struct vm_memseg));
335 memseg.segid = segid;
338 n = strlcpy(memseg.name, name, sizeof(memseg.name));
339 if (n >= sizeof(memseg.name)) {
340 errno = ENAMETOOLONG;
345 error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg);
350 vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf,
353 struct vm_memseg memseg;
357 memseg.segid = segid;
358 error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg);
361 n = strlcpy(namebuf, memseg.name, bufsize);
363 errno = ENAMETOOLONG;
371 setup_memory_segment(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char *base)
376 /* Map 'len' bytes starting at 'gpa' in the guest address space */
377 error = vm_mmap_memseg(ctx, gpa, VM_SYSMEM, gpa, len, PROT_ALL);
381 flags = MAP_SHARED | MAP_FIXED;
382 if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
385 /* mmap into the process address space on the host */
386 ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa);
387 if (ptr == MAP_FAILED)
394 vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms)
398 char *baseaddr, *ptr;
401 assert(vms == VM_MMAP_ALL);
404 * If 'memsize' cannot fit entirely in the 'lowmem' segment then
405 * create another 'highmem' segment above 4GB for the remainder.
407 if (memsize > ctx->lowmem_limit) {
408 ctx->lowmem = ctx->lowmem_limit;
409 ctx->highmem = memsize - ctx->lowmem_limit;
410 objsize = 4*GB + ctx->highmem;
412 ctx->lowmem = memsize;
414 objsize = ctx->lowmem;
417 error = vm_alloc_memseg(ctx, VM_SYSMEM, objsize, NULL);
422 * Stake out a contiguous region covering the guest physical memory
423 * and the adjoining guard regions.
425 len = VM_MMAP_GUARD_SIZE + objsize + VM_MMAP_GUARD_SIZE;
426 ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0);
427 if (ptr == MAP_FAILED)
430 baseaddr = ptr + VM_MMAP_GUARD_SIZE;
431 if (ctx->highmem > 0) {
434 error = setup_memory_segment(ctx, gpa, len, baseaddr);
439 if (ctx->lowmem > 0) {
442 error = setup_memory_segment(ctx, gpa, len, baseaddr);
447 ctx->baseaddr = baseaddr;
453 * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in
454 * the lowmem or highmem regions.
456 * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region.
457 * The instruction emulation code depends on this behavior.
460 vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len)
463 if (ctx->lowmem > 0) {
464 if (gaddr < ctx->lowmem && len <= ctx->lowmem &&
465 gaddr + len <= ctx->lowmem)
466 return (ctx->baseaddr + gaddr);
469 if (ctx->highmem > 0) {
471 if (gaddr < 4*GB + ctx->highmem &&
472 len <= ctx->highmem &&
473 gaddr + len <= 4*GB + ctx->highmem)
474 return (ctx->baseaddr + gaddr);
482 vm_rev_map_gpa(struct vmctx *ctx, void *addr)
486 offaddr = (char *)addr - ctx->baseaddr;
489 if (offaddr <= ctx->lowmem)
492 if (ctx->highmem > 0)
493 if (offaddr >= 4*GB && offaddr < 4*GB + ctx->highmem)
496 return ((vm_paddr_t)-1);
499 /* TODO: maximum size for vmname */
501 vm_get_name(struct vmctx *ctx, char *buf, size_t max_len)
504 if (strlcpy(buf, ctx->name, max_len) >= max_len)
510 vm_get_lowmem_size(struct vmctx *ctx)
513 return (ctx->lowmem);
517 vm_get_highmem_size(struct vmctx *ctx)
520 return (ctx->highmem);
524 vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len)
526 char pathname[MAXPATHLEN];
529 int fd, error, flags;
533 if (name == NULL || strlen(name) == 0) {
538 error = vm_alloc_memseg(ctx, segid, len, name);
542 strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname));
543 strlcat(pathname, ctx->name, sizeof(pathname));
544 strlcat(pathname, ".", sizeof(pathname));
545 strlcat(pathname, name, sizeof(pathname));
547 fd = open(pathname, O_RDWR);
552 * Stake out a contiguous region covering the device memory and the
553 * adjoining guard regions.
555 len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE;
556 base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1,
558 if (base == MAP_FAILED)
561 flags = MAP_SHARED | MAP_FIXED;
562 if ((ctx->memflags & VM_MEM_F_INCORE) == 0)
565 /* mmap the devmem region in the host address space */
566 ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0);
574 vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
575 uint64_t base, uint32_t limit, uint32_t access)
578 struct vm_seg_desc vmsegdesc;
580 bzero(&vmsegdesc, sizeof(vmsegdesc));
581 vmsegdesc.cpuid = vcpu;
582 vmsegdesc.regnum = reg;
583 vmsegdesc.desc.base = base;
584 vmsegdesc.desc.limit = limit;
585 vmsegdesc.desc.access = access;
587 error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
592 vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
593 uint64_t *base, uint32_t *limit, uint32_t *access)
596 struct vm_seg_desc vmsegdesc;
598 bzero(&vmsegdesc, sizeof(vmsegdesc));
599 vmsegdesc.cpuid = vcpu;
600 vmsegdesc.regnum = reg;
602 error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
604 *base = vmsegdesc.desc.base;
605 *limit = vmsegdesc.desc.limit;
606 *access = vmsegdesc.desc.access;
612 vm_get_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *seg_desc)
616 error = vm_get_desc(ctx, vcpu, reg, &seg_desc->base, &seg_desc->limit,
622 vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
625 struct vm_register vmreg;
627 bzero(&vmreg, sizeof(vmreg));
632 error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
637 vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
640 struct vm_register vmreg;
642 bzero(&vmreg, sizeof(vmreg));
646 error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
647 *ret_val = vmreg.regval;
652 vm_set_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
653 const int *regnums, uint64_t *regvals)
656 struct vm_register_set vmregset;
658 bzero(&vmregset, sizeof(vmregset));
659 vmregset.cpuid = vcpu;
660 vmregset.count = count;
661 vmregset.regnums = regnums;
662 vmregset.regvals = regvals;
664 error = ioctl(ctx->fd, VM_SET_REGISTER_SET, &vmregset);
669 vm_get_register_set(struct vmctx *ctx, int vcpu, unsigned int count,
670 const int *regnums, uint64_t *regvals)
673 struct vm_register_set vmregset;
675 bzero(&vmregset, sizeof(vmregset));
676 vmregset.cpuid = vcpu;
677 vmregset.count = count;
678 vmregset.regnums = regnums;
679 vmregset.regvals = regvals;
681 error = ioctl(ctx->fd, VM_GET_REGISTER_SET, &vmregset);
686 vm_run(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit)
691 bzero(&vmrun, sizeof(vmrun));
694 error = ioctl(ctx->fd, VM_RUN, &vmrun);
695 bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
700 vm_suspend(struct vmctx *ctx, enum vm_suspend_how how)
702 struct vm_suspend vmsuspend;
704 bzero(&vmsuspend, sizeof(vmsuspend));
706 return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend));
710 vm_reinit(struct vmctx *ctx)
713 return (ioctl(ctx->fd, VM_REINIT, 0));
717 vm_inject_exception(struct vmctx *ctx, int vcpu, int vector, int errcode_valid,
718 uint32_t errcode, int restart_instruction)
720 struct vm_exception exc;
724 exc.error_code = errcode;
725 exc.error_code_valid = errcode_valid;
726 exc.restart_instruction = restart_instruction;
728 return (ioctl(ctx->fd, VM_INJECT_EXCEPTION, &exc));
732 vm_apicid2vcpu(struct vmctx *ctx __unused, int apicid)
735 * The apic id associated with the 'vcpu' has the same numerical value
736 * as the 'vcpu' itself.
742 vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
744 struct vm_lapic_irq vmirq;
746 bzero(&vmirq, sizeof(vmirq));
748 vmirq.vector = vector;
750 return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
754 vm_lapic_local_irq(struct vmctx *ctx, int vcpu, int vector)
756 struct vm_lapic_irq vmirq;
758 bzero(&vmirq, sizeof(vmirq));
760 vmirq.vector = vector;
762 return (ioctl(ctx->fd, VM_LAPIC_LOCAL_IRQ, &vmirq));
766 vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg)
768 struct vm_lapic_msi vmmsi;
770 bzero(&vmmsi, sizeof(vmmsi));
774 return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi));
778 vm_ioapic_assert_irq(struct vmctx *ctx, int irq)
780 struct vm_ioapic_irq ioapic_irq;
782 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
783 ioapic_irq.irq = irq;
785 return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq));
789 vm_ioapic_deassert_irq(struct vmctx *ctx, int irq)
791 struct vm_ioapic_irq ioapic_irq;
793 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
794 ioapic_irq.irq = irq;
796 return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq));
800 vm_ioapic_pulse_irq(struct vmctx *ctx, int irq)
802 struct vm_ioapic_irq ioapic_irq;
804 bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq));
805 ioapic_irq.irq = irq;
807 return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq));
811 vm_ioapic_pincount(struct vmctx *ctx, int *pincount)
814 return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount));
818 vm_readwrite_kernemu_device(struct vmctx *ctx, int vcpu, vm_paddr_t gpa,
819 bool write, int size, uint64_t *value)
821 struct vm_readwrite_kernemu_device irp = {
823 .access_width = fls(size) - 1,
825 .value = write ? *value : ~0ul,
827 long cmd = (write ? VM_SET_KERNEMU_DEV : VM_GET_KERNEMU_DEV);
830 rc = ioctl(ctx->fd, cmd, &irp);
831 if (rc == 0 && !write)
837 vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
839 struct vm_isa_irq isa_irq;
841 bzero(&isa_irq, sizeof(struct vm_isa_irq));
842 isa_irq.atpic_irq = atpic_irq;
843 isa_irq.ioapic_irq = ioapic_irq;
845 return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq));
849 vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
851 struct vm_isa_irq isa_irq;
853 bzero(&isa_irq, sizeof(struct vm_isa_irq));
854 isa_irq.atpic_irq = atpic_irq;
855 isa_irq.ioapic_irq = ioapic_irq;
857 return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq));
861 vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq)
863 struct vm_isa_irq isa_irq;
865 bzero(&isa_irq, sizeof(struct vm_isa_irq));
866 isa_irq.atpic_irq = atpic_irq;
867 isa_irq.ioapic_irq = ioapic_irq;
869 return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq));
873 vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq,
874 enum vm_intr_trigger trigger)
876 struct vm_isa_irq_trigger isa_irq_trigger;
878 bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger));
879 isa_irq_trigger.atpic_irq = atpic_irq;
880 isa_irq_trigger.trigger = trigger;
882 return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger));
886 vm_inject_nmi(struct vmctx *ctx, int vcpu)
890 bzero(&vmnmi, sizeof(vmnmi));
893 return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
896 static const char *capstrmap[] = {
897 [VM_CAP_HALT_EXIT] = "hlt_exit",
898 [VM_CAP_MTRAP_EXIT] = "mtrap_exit",
899 [VM_CAP_PAUSE_EXIT] = "pause_exit",
900 [VM_CAP_UNRESTRICTED_GUEST] = "unrestricted_guest",
901 [VM_CAP_ENABLE_INVPCID] = "enable_invpcid",
902 [VM_CAP_BPT_EXIT] = "bpt_exit",
906 vm_capability_name2type(const char *capname)
910 for (i = 0; i < (int)nitems(capstrmap); i++) {
911 if (strcmp(capstrmap[i], capname) == 0)
919 vm_capability_type2name(int type)
921 if (type >= 0 && type < (int)nitems(capstrmap))
922 return (capstrmap[type]);
928 vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
932 struct vm_capability vmcap;
934 bzero(&vmcap, sizeof(vmcap));
938 error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
939 *retval = vmcap.capval;
944 vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
946 struct vm_capability vmcap;
948 bzero(&vmcap, sizeof(vmcap));
953 return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
957 vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
959 struct vm_pptdev pptdev;
961 bzero(&pptdev, sizeof(pptdev));
966 return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
970 vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
972 struct vm_pptdev pptdev;
974 bzero(&pptdev, sizeof(pptdev));
979 return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
983 vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
984 vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
986 struct vm_pptdev_mmio pptmmio;
988 bzero(&pptmmio, sizeof(pptmmio));
996 return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
1000 vm_unmap_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
1001 vm_paddr_t gpa, size_t len)
1003 struct vm_pptdev_mmio pptmmio;
1005 bzero(&pptmmio, sizeof(pptmmio));
1007 pptmmio.slot = slot;
1008 pptmmio.func = func;
1012 return (ioctl(ctx->fd, VM_UNMAP_PPTDEV_MMIO, &pptmmio));
1016 vm_setup_pptdev_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1017 uint64_t addr, uint64_t msg, int numvec)
1019 struct vm_pptdev_msi pptmsi;
1021 bzero(&pptmsi, sizeof(pptmsi));
1028 pptmsi.numvec = numvec;
1030 return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
1034 vm_setup_pptdev_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
1035 int idx, uint64_t addr, uint64_t msg, uint32_t vector_control)
1037 struct vm_pptdev_msix pptmsix;
1039 bzero(&pptmsix, sizeof(pptmsix));
1040 pptmsix.vcpu = vcpu;
1042 pptmsix.slot = slot;
1043 pptmsix.func = func;
1046 pptmsix.addr = addr;
1047 pptmsix.vector_control = vector_control;
1049 return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
1053 vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func)
1055 struct vm_pptdev ppt;
1057 bzero(&ppt, sizeof(ppt));
1062 return ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &ppt);
1066 vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
1069 static _Thread_local uint64_t *stats_buf;
1070 static _Thread_local u_int stats_count;
1071 uint64_t *new_stats;
1072 struct vm_stats vmstats;
1077 vmstats.cpuid = vcpu;
1079 for (index = 0;; index += nitems(vmstats.statbuf)) {
1080 vmstats.index = index;
1081 if (ioctl(ctx->fd, VM_STATS, &vmstats) != 0)
1083 if (stats_count < index + vmstats.num_entries) {
1084 new_stats = realloc(stats_buf,
1085 (index + vmstats.num_entries) * sizeof(uint64_t));
1086 if (new_stats == NULL) {
1090 stats_count = index + vmstats.num_entries;
1091 stats_buf = new_stats;
1093 memcpy(stats_buf + index, vmstats.statbuf,
1094 vmstats.num_entries * sizeof(uint64_t));
1095 count += vmstats.num_entries;
1098 if (vmstats.num_entries != nitems(vmstats.statbuf))
1103 *ret_entries = count;
1105 *ret_tv = vmstats.tv;
1112 vm_get_stat_desc(struct vmctx *ctx, int index)
1114 static struct vm_stat_desc statdesc;
1116 statdesc.index = index;
1117 if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
1118 return (statdesc.desc);
1124 vm_get_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state *state)
1127 struct vm_x2apic x2apic;
1129 bzero(&x2apic, sizeof(x2apic));
1130 x2apic.cpuid = vcpu;
1132 error = ioctl(ctx->fd, VM_GET_X2APIC_STATE, &x2apic);
1133 *state = x2apic.state;
1138 vm_set_x2apic_state(struct vmctx *ctx, int vcpu, enum x2apic_state state)
1141 struct vm_x2apic x2apic;
1143 bzero(&x2apic, sizeof(x2apic));
1144 x2apic.cpuid = vcpu;
1145 x2apic.state = state;
1147 error = ioctl(ctx->fd, VM_SET_X2APIC_STATE, &x2apic);
1153 * From Intel Vol 3a:
1154 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
1157 vcpu_reset(struct vmctx *vmctx, int vcpu)
1160 uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
1161 uint32_t desc_access, desc_limit;
1167 error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
1172 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
1176 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
1179 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
1183 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
1187 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
1189 desc_base = 0xffff0000;
1190 desc_limit = 0xffff;
1191 desc_access = 0x0093;
1192 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
1193 desc_base, desc_limit, desc_access);
1198 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
1202 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
1205 desc_limit = 0xffff;
1206 desc_access = 0x0093;
1207 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
1208 desc_base, desc_limit, desc_access);
1212 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
1213 desc_base, desc_limit, desc_access);
1217 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
1218 desc_base, desc_limit, desc_access);
1222 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
1223 desc_base, desc_limit, desc_access);
1227 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
1228 desc_base, desc_limit, desc_access);
1233 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
1235 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
1237 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
1239 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
1241 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
1244 /* General purpose registers */
1246 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
1248 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
1250 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
1252 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
1254 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
1256 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
1258 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
1260 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
1265 desc_limit = 0xffff;
1267 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
1268 desc_base, desc_limit, desc_access);
1272 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
1273 desc_base, desc_limit, desc_access);
1279 desc_limit = 0xffff;
1280 desc_access = 0x0000008b;
1281 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
1286 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
1291 desc_limit = 0xffff;
1292 desc_access = 0x00000082;
1293 error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
1294 desc_limit, desc_access);
1299 if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
1302 /* XXX cr2, debug registers */
1310 vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num)
1313 struct vm_gpa_pte gpapte;
1315 bzero(&gpapte, sizeof(gpapte));
1318 error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte);
1321 *num = gpapte.ptenum;
1322 for (i = 0; i < gpapte.ptenum; i++)
1323 pte[i] = gpapte.pte[i];
1330 vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities)
1333 struct vm_hpet_cap cap;
1335 bzero(&cap, sizeof(struct vm_hpet_cap));
1336 error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap);
1337 if (capabilities != NULL)
1338 *capabilities = cap.capabilities;
1343 vm_gla2gpa(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1344 uint64_t gla, int prot, uint64_t *gpa, int *fault)
1346 struct vm_gla2gpa gg;
1349 bzero(&gg, sizeof(struct vm_gla2gpa));
1353 gg.paging = *paging;
1355 error = ioctl(ctx->fd, VM_GLA2GPA, &gg);
1364 vm_gla2gpa_nofault(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1365 uint64_t gla, int prot, uint64_t *gpa, int *fault)
1367 struct vm_gla2gpa gg;
1370 bzero(&gg, sizeof(struct vm_gla2gpa));
1374 gg.paging = *paging;
1376 error = ioctl(ctx->fd, VM_GLA2GPA_NOFAULT, &gg);
1385 #define min(a,b) (((a) < (b)) ? (a) : (b))
1389 vm_copy_setup(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging,
1390 uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt,
1397 for (i = 0; i < iovcnt; i++) {
1398 iov[i].iov_base = 0;
1404 error = vm_gla2gpa(ctx, vcpu, paging, gla, prot, &gpa, fault);
1405 if (error || *fault)
1408 off = gpa & PAGE_MASK;
1409 n = MIN(len, PAGE_SIZE - off);
1411 va = vm_map_gpa(ctx, gpa, n);
1427 vm_copy_teardown(struct vmctx *ctx __unused, int vcpu __unused,
1428 struct iovec *iov __unused, int iovcnt __unused)
1433 vm_copyin(struct vmctx *ctx __unused, int vcpu __unused, struct iovec *iov,
1434 void *vp, size_t len)
1442 assert(iov->iov_len);
1443 n = min(len, iov->iov_len);
1444 src = iov->iov_base;
1454 vm_copyout(struct vmctx *ctx __unused, int vcpu __unused, const void *vp,
1455 struct iovec *iov, size_t len)
1463 assert(iov->iov_len);
1464 n = min(len, iov->iov_len);
1465 dst = iov->iov_base;
1475 vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus)
1477 struct vm_cpuset vm_cpuset;
1480 bzero(&vm_cpuset, sizeof(struct vm_cpuset));
1481 vm_cpuset.which = which;
1482 vm_cpuset.cpusetsize = sizeof(cpuset_t);
1483 vm_cpuset.cpus = cpus;
1485 error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset);
1490 vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus)
1493 return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus));
1497 vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus)
1500 return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus));
1504 vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus)
1507 return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus));
1511 vm_activate_cpu(struct vmctx *ctx, int vcpu)
1513 struct vm_activate_cpu ac;
1516 bzero(&ac, sizeof(struct vm_activate_cpu));
1518 error = ioctl(ctx->fd, VM_ACTIVATE_CPU, &ac);
1523 vm_suspend_cpu(struct vmctx *ctx, int vcpu)
1525 struct vm_activate_cpu ac;
1528 bzero(&ac, sizeof(struct vm_activate_cpu));
1530 error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac);
1535 vm_resume_cpu(struct vmctx *ctx, int vcpu)
1537 struct vm_activate_cpu ac;
1540 bzero(&ac, sizeof(struct vm_activate_cpu));
1542 error = ioctl(ctx->fd, VM_RESUME_CPU, &ac);
1547 vm_get_intinfo(struct vmctx *ctx, int vcpu, uint64_t *info1, uint64_t *info2)
1549 struct vm_intinfo vmii;
1552 bzero(&vmii, sizeof(struct vm_intinfo));
1554 error = ioctl(ctx->fd, VM_GET_INTINFO, &vmii);
1556 *info1 = vmii.info1;
1557 *info2 = vmii.info2;
1563 vm_set_intinfo(struct vmctx *ctx, int vcpu, uint64_t info1)
1565 struct vm_intinfo vmii;
1568 bzero(&vmii, sizeof(struct vm_intinfo));
1571 error = ioctl(ctx->fd, VM_SET_INTINFO, &vmii);
1576 vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value)
1578 struct vm_rtc_data rtcdata;
1581 bzero(&rtcdata, sizeof(struct vm_rtc_data));
1582 rtcdata.offset = offset;
1583 rtcdata.value = value;
1584 error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata);
1589 vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval)
1591 struct vm_rtc_data rtcdata;
1594 bzero(&rtcdata, sizeof(struct vm_rtc_data));
1595 rtcdata.offset = offset;
1596 error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata);
1598 *retval = rtcdata.value;
1603 vm_rtc_settime(struct vmctx *ctx, time_t secs)
1605 struct vm_rtc_time rtctime;
1608 bzero(&rtctime, sizeof(struct vm_rtc_time));
1609 rtctime.secs = secs;
1610 error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime);
1615 vm_rtc_gettime(struct vmctx *ctx, time_t *secs)
1617 struct vm_rtc_time rtctime;
1620 bzero(&rtctime, sizeof(struct vm_rtc_time));
1621 error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime);
1623 *secs = rtctime.secs;
1628 vm_restart_instruction(void *arg, int vcpu)
1630 struct vmctx *ctx = arg;
1632 return (ioctl(ctx->fd, VM_RESTART_INSTRUCTION, &vcpu));
1636 vm_snapshot_req(struct vm_snapshot_meta *meta)
1639 if (ioctl(meta->ctx->fd, VM_SNAPSHOT_REQ, meta) == -1) {
1640 #ifdef SNAPSHOT_DEBUG
1641 fprintf(stderr, "%s: snapshot failed for %s: %d\r\n",
1642 __func__, meta->dev_name, errno);
1650 vm_restore_time(struct vmctx *ctx)
1655 return (ioctl(ctx->fd, VM_RESTORE_TIME, &dummy));
1659 vm_set_topology(struct vmctx *ctx,
1660 uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus)
1662 struct vm_cpu_topology topology;
1664 bzero(&topology, sizeof (struct vm_cpu_topology));
1665 topology.sockets = sockets;
1666 topology.cores = cores;
1667 topology.threads = threads;
1668 topology.maxcpus = maxcpus;
1669 return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology));
1673 vm_get_topology(struct vmctx *ctx,
1674 uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus)
1676 struct vm_cpu_topology topology;
1679 bzero(&topology, sizeof (struct vm_cpu_topology));
1680 error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology);
1682 *sockets = topology.sockets;
1683 *cores = topology.cores;
1684 *threads = topology.threads;
1685 *maxcpus = topology.maxcpus;
1691 vm_get_device_fd(struct vmctx *ctx)
1698 vm_get_ioctls(size_t *len)
1701 /* keep in sync with machine/vmm_dev.h */
1702 static const cap_ioctl_t vm_ioctl_cmds[] = { VM_RUN, VM_SUSPEND, VM_REINIT,
1703 VM_ALLOC_MEMSEG, VM_GET_MEMSEG, VM_MMAP_MEMSEG, VM_MMAP_MEMSEG,
1704 VM_MMAP_GETNEXT, VM_MUNMAP_MEMSEG, VM_SET_REGISTER, VM_GET_REGISTER,
1705 VM_SET_SEGMENT_DESCRIPTOR, VM_GET_SEGMENT_DESCRIPTOR,
1706 VM_SET_REGISTER_SET, VM_GET_REGISTER_SET,
1707 VM_SET_KERNEMU_DEV, VM_GET_KERNEMU_DEV,
1708 VM_INJECT_EXCEPTION, VM_LAPIC_IRQ, VM_LAPIC_LOCAL_IRQ,
1709 VM_LAPIC_MSI, VM_IOAPIC_ASSERT_IRQ, VM_IOAPIC_DEASSERT_IRQ,
1710 VM_IOAPIC_PULSE_IRQ, VM_IOAPIC_PINCOUNT, VM_ISA_ASSERT_IRQ,
1711 VM_ISA_DEASSERT_IRQ, VM_ISA_PULSE_IRQ, VM_ISA_SET_IRQ_TRIGGER,
1712 VM_SET_CAPABILITY, VM_GET_CAPABILITY, VM_BIND_PPTDEV,
1713 VM_UNBIND_PPTDEV, VM_MAP_PPTDEV_MMIO, VM_PPTDEV_MSI,
1714 VM_PPTDEV_MSIX, VM_UNMAP_PPTDEV_MMIO, VM_PPTDEV_DISABLE_MSIX,
1715 VM_INJECT_NMI, VM_STATS, VM_STAT_DESC,
1716 VM_SET_X2APIC_STATE, VM_GET_X2APIC_STATE,
1717 VM_GET_HPET_CAPABILITIES, VM_GET_GPA_PMAP, VM_GLA2GPA,
1719 VM_ACTIVATE_CPU, VM_GET_CPUS, VM_SUSPEND_CPU, VM_RESUME_CPU,
1720 VM_SET_INTINFO, VM_GET_INTINFO,
1721 VM_RTC_WRITE, VM_RTC_READ, VM_RTC_SETTIME, VM_RTC_GETTIME,
1722 VM_RESTART_INSTRUCTION, VM_SET_TOPOLOGY, VM_GET_TOPOLOGY };
1725 cmds = malloc(sizeof(vm_ioctl_cmds));
1728 bcopy(vm_ioctl_cmds, cmds, sizeof(vm_ioctl_cmds));
1732 *len = nitems(vm_ioctl_cmds);