You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

413 lines
12 KiB

/*
* Check decoding of KVM_* commands of ioctl syscall using /dev/kvm API.
* Based on kvmtest.c from https://lwn.net/Articles/658512/
*
* kvmtest.c author: Josh Triplett <josh@joshtriplett.org>
* Copyright (c) 2015 Intel Corporation
* Copyright (c) 2017-2018 The strace developers.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*/
#include "tests.h"
#if defined HAVE_LINUX_KVM_H \
&& defined HAVE_STRUCT_KVM_CPUID2 \
&& defined HAVE_STRUCT_KVM_REGS \
&& defined HAVE_STRUCT_KVM_SREGS \
&& defined HAVE_STRUCT_KVM_USERSPACE_MEMORY_REGION \
&&(defined __x86_64__ || defined __i386__)
# include <fcntl.h>
# include <stdint.h>
# include <stdio.h>
# include <stdlib.h>
# include <string.h>
# include <sys/ioctl.h>
# include <sys/mman.h>
# include <unistd.h>
# include <linux/kvm.h>
# ifndef KVM_MAX_CPUID_ENTRIES
# define KVM_MAX_CPUID_ENTRIES 80
# endif
#include "xlat.h"
#include "xlat/kvm_cpuid_flags.h"
static int
kvm_ioctl(int fd, unsigned long cmd, const char *cmd_str, void *arg)
{
int rc = ioctl(fd, cmd, arg);
if (rc < 0)
perror_msg_and_skip("%s", cmd_str);
return rc;
}
#define KVM_IOCTL(fd_, cmd_, arg_) \
kvm_ioctl((fd_), (cmd_), #cmd_, (arg_))
static const char dev[] = "/dev/kvm";
static const char vm_dev[] = "anon_inode:kvm-vm";
static char vcpu_dev[] = "anon_inode:kvm-vcpu:0";
static size_t page_size;
extern const char code[];
extern const unsigned short code_size;
__asm__(
".type code, @object \n"
"code: \n"
" mov $0xd80003f8, %edx \n"
" mov $'\n', %al \n"
" out %al, (%dx) \n"
" hlt \n"
".size code, . - code \n"
".type code_size, @object \n"
"code_size: \n"
" .short . - code \n"
".size code_size, . - code_size \n"
);
static void
print_kvm_segment(const struct kvm_segment *seg)
{
printf("{base=%#jx, limit=%u, selector=%u, type=%u, present=%u, "
"dpl=%u, db=%u, s=%u, l=%u, g=%u, avl=%u}",
(uintmax_t) seg->base, seg->limit, seg->selector, seg->type,
seg->present, seg->dpl, seg->db, seg->s, seg->l, seg->g,
seg->avl);
}
static void
print_kvm_sregs(const struct kvm_sregs *sregs)
{
printf("{cs=");
print_kvm_segment(&sregs->cs);
#if VERBOSE
printf(", ds=");
print_kvm_segment(&sregs->ds);
printf(", es=");
print_kvm_segment(&sregs->es);
printf(", fs=");
print_kvm_segment(&sregs->fs);
printf(", gs=");
print_kvm_segment(&sregs->gs);
printf(", ss=");
print_kvm_segment(&sregs->ss);
printf(", tr=");
print_kvm_segment(&sregs->tr);
printf(", ldt=");
print_kvm_segment(&sregs->ldt);
printf(", gdt={base=%#jx, limit=%u}, idt={base=%#jx, limit=%u}, "
"cr0=%llu, cr2=%llu, cr3=%llu, cr4=%llu, cr8=%llu, efer=%llu, "
"apic_base=%#jx", (uintmax_t) sregs->gdt.base, sregs->gdt.limit,
(uintmax_t) sregs->idt.base, sregs->idt.limit, sregs->cr0,
sregs->cr2, sregs->cr3, sregs->cr4, sregs->cr8, sregs->efer,
(uintmax_t)sregs->apic_base);
printf(", interrupt_bitmap=[");
for (size_t i = 0; i < ARRAY_SIZE(sregs->interrupt_bitmap); i++) {
if (i)
printf(", ");
printf("%#jx", (uintmax_t) sregs->interrupt_bitmap[i]);
}
printf("]");
#else
printf(", ...");
#endif
printf("}");
}
static void
print_kvm_regs(const struct kvm_regs *regs)
{
printf("{rax=%#jx", (uintmax_t) regs->rax);
#if VERBOSE
printf(", rbx=%#jx, rcx=%#jx, rdx=%#jx, rsi=%#jx, rdi=%#jx",
(uintmax_t) regs->rbx, (uintmax_t) regs->rcx,
(uintmax_t) regs->rdx, (uintmax_t) regs->rsi,
(uintmax_t) regs->rdi);
#else
printf(", ...");
#endif
printf(", rsp=%#jx, rbp=%#jx", (uintmax_t) regs->rsp,
(uintmax_t) regs->rbp);
#if VERBOSE
printf(", r8=%#jx, r9=%#jx, r10=%#jx, r11=%#jx, r12=%#jx, r13=%#jx"
", r14=%#jx, r15=%#jx",
(uintmax_t) regs->r8, (uintmax_t) regs->r9,
(uintmax_t) regs->r10, (uintmax_t) regs->r11,
(uintmax_t) regs->r12, (uintmax_t) regs->r13,
(uintmax_t) regs->r14, (uintmax_t) regs->r15);
#else
printf(", ...");
#endif
printf(", rip=%#jx, rflags=%#jx}", (uintmax_t) regs->rip,
(uintmax_t) regs->rflags);
}
# define need_print_KVM_RUN 1
static void
print_KVM_RUN(const int fd, const char *const dev, const unsigned int reason);
static void
run_kvm(const int vcpu_fd, struct kvm_run *const run, const size_t mmap_size,
void *const mem)
{
/* Initialize CS to point at 0, via a read-modify-write of sregs. */
struct kvm_sregs sregs;
KVM_IOCTL(vcpu_fd, KVM_GET_SREGS, &sregs);
printf("ioctl(%d<%s>, KVM_GET_SREGS, ", vcpu_fd, vcpu_dev);
print_kvm_sregs(&sregs);
printf(") = 0\n");
sregs.cs.base = 0;
sregs.cs.selector = 0;
KVM_IOCTL(vcpu_fd, KVM_SET_SREGS, &sregs);
printf("ioctl(%d<%s>, KVM_SET_SREGS, ", vcpu_fd, vcpu_dev);
print_kvm_sregs(&sregs);
printf(") = 0\n");
/*
* Initialize registers: instruction pointer for our code, addends,
* and initial flags required by x86 architecture.
*/
struct kvm_regs regs = {
.rip = page_size,
.rax = 2,
.rbx = 2,
.rflags = 0x2,
};
KVM_IOCTL(vcpu_fd, KVM_SET_REGS, &regs);
printf("ioctl(%d<%s>, KVM_SET_REGS, ", vcpu_fd, vcpu_dev);
print_kvm_regs(&regs);
printf(") = 0\n");
/* Copy the code */
memcpy(mem, code, code_size);
const char *p = "\n";
/* Repeatedly run code and handle VM exits. */
for (;;) {
KVM_IOCTL(vcpu_fd, KVM_RUN, NULL);
print_KVM_RUN(vcpu_fd, vcpu_dev, run->exit_reason);
switch (run->exit_reason) {
case KVM_EXIT_HLT:
if (p)
error_msg_and_fail("premature KVM_EXIT_HLT");
return;
case KVM_EXIT_IO:
if (run->io.direction == KVM_EXIT_IO_OUT
&& run->io.size == 1
&& run->io.port == 0x03f8
&& run->io.count == 1
&& run->io.data_offset < mmap_size
&& p && *p == ((char *) run)[run->io.data_offset])
p = NULL;
else
error_msg_and_fail("unhandled KVM_EXIT_IO");
break;
case KVM_EXIT_MMIO:
error_msg_and_fail("Got an unexpected MMIO exit:"
" phys_addr %#llx,"
" data %02x %02x %02x %02x"
" %02x %02x %02x %02x,"
" len %u, is_write %hhu",
(unsigned long long) run->mmio.phys_addr,
run->mmio.data[0], run->mmio.data[1],
run->mmio.data[2], run->mmio.data[3],
run->mmio.data[4], run->mmio.data[5],
run->mmio.data[6], run->mmio.data[7],
run->mmio.len, run->mmio.is_write);
default:
error_msg_and_fail("exit_reason = %#x",
run->exit_reason);
}
}
}
static int
vcpu_dev_should_have_cpuid(int fd)
{
int r = 0;
char *filename = NULL;
char buf[sizeof(vcpu_dev)];
if (asprintf(&filename, "/proc/%d/fd/%d", getpid(), fd) < 0)
error_msg_and_fail("asprintf");
if (readlink(filename, buf, sizeof(buf)) == sizeof(buf) - 1
&& (memcmp(buf, vcpu_dev, sizeof(buf) - 1) == 0))
r = 1;
free(filename);
return r;
}
static void
print_cpuid_ioctl(int fd, const char *fd_dev,
const char *ioctl_name, const struct kvm_cpuid2 *cpuid)
{
printf("ioctl(%d<%s>, %s, {nent=%u, entries=[",
fd, fd_dev, ioctl_name, cpuid->nent);
#if VERBOSE
for (size_t i = 0; i < cpuid->nent; i++) {
if (i)
printf(", ");
printf("{function=%#x, index=%#x, flags=",
cpuid->entries[i].function, cpuid->entries[i].index);
printflags(kvm_cpuid_flags, cpuid->entries[i].flags,
"KVM_CPUID_FLAG_???");
printf(", eax=%#x, ebx=%#x, ecx=%#x, edx=%#x}",
cpuid->entries[i].eax, cpuid->entries[i].ebx,
cpuid->entries[i].ecx, cpuid->entries[i].edx);
}
#else
if (cpuid->nent)
printf("...");
#endif
printf("]}) = 0\n");
}
int
main(void)
{
skip_if_unavailable("/proc/self/fd/");
int kvm = open(dev, O_RDWR);
if (kvm < 0)
perror_msg_and_skip("open: %s", dev);
/* Make sure we have the stable version of the API */
int ret = KVM_IOCTL(kvm, KVM_GET_API_VERSION, 0);
if (ret != KVM_API_VERSION)
error_msg_and_skip("KVM_GET_API_VERSION returned %d"
", KVM_API_VERSION is %d",
kvm, KVM_API_VERSION);
printf("ioctl(%d<%s>, KVM_GET_API_VERSION, 0) = %d\n",
kvm, dev, ret);
ret = KVM_IOCTL(kvm, KVM_CHECK_EXTENSION,
(void *) (uintptr_t) KVM_CAP_USER_MEMORY);
printf("ioctl(%d<%s>, KVM_CHECK_EXTENSION, KVM_CAP_USER_MEMORY) = %d\n",
kvm, dev, ret);
int vm_fd = KVM_IOCTL(kvm, KVM_CREATE_VM, 0);
printf("ioctl(%d<%s>, KVM_CREATE_VM, 0) = %d<%s>\n",
kvm, dev, vm_fd, vm_dev);
/* Allocate one aligned page of guest memory to hold the code. */
page_size = get_page_size();
void *const mem = mmap(NULL, page_size, PROT_READ | PROT_WRITE,
MAP_SHARED | MAP_ANONYMOUS, -1, 0);
if (mem == MAP_FAILED)
perror_msg_and_fail("mmap page");
/* Map it to the second page frame (to avoid the real-mode IDT at 0). */
struct kvm_userspace_memory_region region = {
.slot = 0,
.guest_phys_addr = page_size,
.memory_size = page_size,
.userspace_addr = (uintptr_t) mem,
};
KVM_IOCTL(vm_fd, KVM_SET_USER_MEMORY_REGION, &region);
printf("ioctl(%d<%s>, KVM_SET_USER_MEMORY_REGION"
", {slot=0, flags=0, guest_phys_addr=%#lx, memory_size=%lu"
", userspace_addr=%p}) = 0\n", vm_fd, vm_dev,
(unsigned long) page_size, (unsigned long) page_size, mem);
int vcpu_fd = KVM_IOCTL(vm_fd, KVM_CREATE_VCPU, NULL);
if (!vcpu_dev_should_have_cpuid(vcpu_fd)) {
/*
* This is an older kernel that doesn't place a cpuid
* at the end of the dentry associated with vcpu_fd.
* Trim the cpuid part of vcpu_dev like:
* "anon_inode:kvm-vcpu:0" -> "anon_inode:kvm-vcpu"
*/
vcpu_dev[strlen (vcpu_dev) - 2] = '\0';
#ifdef KVM_NO_CPUID_CALLBACK
KVM_NO_CPUID_CALLBACK;
#endif
}
printf("ioctl(%d<%s>, KVM_CREATE_VCPU, 0) = %d<%s>\n",
vm_fd, vm_dev, vcpu_fd, vcpu_dev);
/* Map the shared kvm_run structure and following data. */
ret = KVM_IOCTL(kvm, KVM_GET_VCPU_MMAP_SIZE, NULL);
struct kvm_run *run;
if (ret < (int) sizeof(*run))
error_msg_and_fail("KVM_GET_VCPU_MMAP_SIZE returned %d < %d",
ret, (int) sizeof(*run));
printf("ioctl(%d<%s>, KVM_GET_VCPU_MMAP_SIZE, 0) = %d\n",
kvm, dev, ret);
const size_t mmap_size = (ret + page_size - 1) & -page_size;
run = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE,
MAP_SHARED, vcpu_fd, 0);
if (run == MAP_FAILED)
perror_msg_and_fail("mmap vcpu");
size_t cpuid_nent = KVM_MAX_CPUID_ENTRIES;
struct kvm_cpuid2 *cpuid = tail_alloc(sizeof(*cpuid) +
cpuid_nent *
sizeof(*cpuid->entries));
cpuid->nent = 0;
ioctl(kvm, KVM_GET_SUPPORTED_CPUID, cpuid);
printf("ioctl(%d<%s>, KVM_GET_SUPPORTED_CPUID, %p) = -1 E2BIG (%m)\n",
kvm, dev, cpuid);
cpuid->nent = cpuid_nent;
KVM_IOCTL(kvm, KVM_GET_SUPPORTED_CPUID, cpuid);
print_cpuid_ioctl(kvm, dev, "KVM_GET_SUPPORTED_CPUID", cpuid);
struct kvm_cpuid2 cpuid_tmp = { .nent = 0 };
KVM_IOCTL(vcpu_fd, KVM_SET_CPUID2, &cpuid_tmp);
printf("ioctl(%d<%s>, KVM_SET_CPUID2, {nent=%u, entries=[]}) = 0\n",
vcpu_fd, vcpu_dev, cpuid_tmp.nent);
KVM_IOCTL(vcpu_fd, KVM_SET_CPUID2, cpuid);
print_cpuid_ioctl(vcpu_fd, vcpu_dev, "KVM_SET_CPUID2", cpuid);
ioctl(vcpu_fd, KVM_SET_CPUID2, NULL);
printf("ioctl(%d<%s>, KVM_SET_CPUID2, NULL) = -1 EFAULT (%m)\n",
vcpu_fd, vcpu_dev);
run_kvm(vcpu_fd, run, mmap_size, mem);
puts("+++ exited with 0 +++");
return 0;
}
#else /* !HAVE_LINUX_KVM_H */
SKIP_MAIN_UNDEFINED("HAVE_LINUX_KVM_H && HAVE_STRUCT_KVM_CPUID2 && "
"HAVE_STRUCT_KVM_REGS && HAVE_STRUCT_KVM_SREGS && "
"HAVE_STRUCT_KVM_USERSPACE_MEMORY_REGION && "
"(__x86_64__ || __i386__)")
# define need_print_KVM_RUN 0
#endif