kvm-all.c 32.0 KB
Newer Older
A
aliguori 已提交
1 2 3 4
/*
 * QEMU KVM support
 *
 * Copyright IBM, Corp. 2008
5
 *           Red Hat, Inc. 2008
A
aliguori 已提交
6 7 8
 *
 * Authors:
 *  Anthony Liguori   <aliguori@us.ibm.com>
9
 *  Glauber Costa     <gcosta@redhat.com>
A
aliguori 已提交
10 11 12 13 14 15 16 17 18
 *
 * This work is licensed under the terms of the GNU GPL, version 2 or later.
 * See the COPYING file in the top-level directory.
 *
 */

#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
19
#include <stdarg.h>
A
aliguori 已提交
20 21 22 23

#include <linux/kvm.h>

#include "qemu-common.h"
24
#include "qemu-barrier.h"
A
aliguori 已提交
25
#include "sysemu.h"
J
Jan Kiszka 已提交
26
#include "hw/hw.h"
27
#include "gdbstub.h"
A
aliguori 已提交
28
#include "kvm.h"
29
#include "bswap.h"
A
aliguori 已提交
30

A
aliguori 已提交
31 32 33
/* KVM uses PAGE_SIZE in it's definition of COALESCED_MMIO_MAX */
#define PAGE_SIZE TARGET_PAGE_SIZE

A
aliguori 已提交
34 35 36
//#define DEBUG_KVM

#ifdef DEBUG_KVM
37
#define DPRINTF(fmt, ...) \
A
aliguori 已提交
38 39
    do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
#else
40
#define DPRINTF(fmt, ...) \
A
aliguori 已提交
41 42 43
    do { } while (0)
#endif

A
aliguori 已提交
44 45
typedef struct KVMSlot
{
A
Anthony Liguori 已提交
46 47 48
    target_phys_addr_t start_addr;
    ram_addr_t memory_size;
    ram_addr_t phys_offset;
A
aliguori 已提交
49 50 51
    int slot;
    int flags;
} KVMSlot;
A
aliguori 已提交
52

53 54
typedef struct kvm_dirty_log KVMDirtyLog;

A
aliguori 已提交
55 56 57 58 59
struct KVMState
{
    KVMSlot slots[32];
    int fd;
    int vmfd;
A
aliguori 已提交
60
    int coalesced_mmio;
61 62 63
#ifdef KVM_CAP_COALESCED_MMIO
    struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
#endif
64
    int broken_set_mem_region;
65
    int migration_log;
66
    int vcpu_events;
67
    int robust_singlestep;
68
    int debugregs;
69 70 71
#ifdef KVM_CAP_SET_GUEST_DEBUG
    struct kvm_sw_breakpoint_head kvm_sw_breakpoints;
#endif
72 73
    int irqchip_in_kernel;
    int pit_in_kernel;
74
    int xsave, xcrs;
A
aliguori 已提交
75 76 77 78 79 80 81 82 83
};

static KVMState *kvm_state;

static KVMSlot *kvm_alloc_slot(KVMState *s)
{
    int i;

    for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
A
aliguori 已提交
84 85 86
        /* KVM private memory slots */
        if (i >= 8 && i < 12)
            continue;
A
aliguori 已提交
87 88 89 90
        if (s->slots[i].memory_size == 0)
            return &s->slots[i];
    }

91 92 93 94 95
    fprintf(stderr, "%s: no free slot available\n", __func__);
    abort();
}

static KVMSlot *kvm_lookup_matching_slot(KVMState *s,
A
Anthony Liguori 已提交
96 97
                                         target_phys_addr_t start_addr,
                                         target_phys_addr_t end_addr)
98 99 100 101 102 103 104 105 106 107 108 109
{
    int i;

    for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
        KVMSlot *mem = &s->slots[i];

        if (start_addr == mem->start_addr &&
            end_addr == mem->start_addr + mem->memory_size) {
            return mem;
        }
    }

A
aliguori 已提交
110 111 112
    return NULL;
}

113 114 115 116
/*
 * Find overlapping slot with lowest start address
 */
static KVMSlot *kvm_lookup_overlapping_slot(KVMState *s,
A
Anthony Liguori 已提交
117 118
                                            target_phys_addr_t start_addr,
                                            target_phys_addr_t end_addr)
A
aliguori 已提交
119
{
120
    KVMSlot *found = NULL;
A
aliguori 已提交
121 122 123 124 125
    int i;

    for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
        KVMSlot *mem = &s->slots[i];

126 127 128 129 130 131 132 133 134
        if (mem->memory_size == 0 ||
            (found && found->start_addr < mem->start_addr)) {
            continue;
        }

        if (end_addr > mem->start_addr &&
            start_addr < mem->start_addr + mem->memory_size) {
            found = mem;
        }
A
aliguori 已提交
135 136
    }

137
    return found;
A
aliguori 已提交
138 139
}

140 141 142 143 144 145 146
static int kvm_set_user_memory_region(KVMState *s, KVMSlot *slot)
{
    struct kvm_userspace_memory_region mem;

    mem.slot = slot->slot;
    mem.guest_phys_addr = slot->start_addr;
    mem.memory_size = slot->memory_size;
P
pbrook 已提交
147
    mem.userspace_addr = (unsigned long)qemu_get_ram_ptr(slot->phys_offset);
148
    mem.flags = slot->flags;
149 150 151
    if (s->migration_log) {
        mem.flags |= KVM_MEM_LOG_DIRTY_PAGES;
    }
152 153 154
    return kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
}

J
Jan Kiszka 已提交
155 156 157 158
static void kvm_reset_vcpu(void *opaque)
{
    CPUState *env = opaque;

J
Jan Kiszka 已提交
159
    kvm_arch_reset_vcpu(env);
J
Jan Kiszka 已提交
160
}
161

162 163 164 165 166 167 168 169 170 171 172
int kvm_irqchip_in_kernel(void)
{
    return kvm_state->irqchip_in_kernel;
}

int kvm_pit_in_kernel(void)
{
    return kvm_state->pit_in_kernel;
}


A
aliguori 已提交
173 174 175 176 177 178
int kvm_init_vcpu(CPUState *env)
{
    KVMState *s = kvm_state;
    long mmap_size;
    int ret;

179
    DPRINTF("kvm_init_vcpu\n");
A
aliguori 已提交
180

181
    ret = kvm_vm_ioctl(s, KVM_CREATE_VCPU, env->cpu_index);
A
aliguori 已提交
182
    if (ret < 0) {
183
        DPRINTF("kvm_create_vcpu failed\n");
A
aliguori 已提交
184 185 186 187 188 189 190 191
        goto err;
    }

    env->kvm_fd = ret;
    env->kvm_state = s;

    mmap_size = kvm_ioctl(s, KVM_GET_VCPU_MMAP_SIZE, 0);
    if (mmap_size < 0) {
192
        DPRINTF("KVM_GET_VCPU_MMAP_SIZE failed\n");
A
aliguori 已提交
193 194 195 196 197 198 199
        goto err;
    }

    env->kvm_run = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE, MAP_SHARED,
                        env->kvm_fd, 0);
    if (env->kvm_run == MAP_FAILED) {
        ret = -errno;
200
        DPRINTF("mmap'ing vcpu state failed\n");
A
aliguori 已提交
201 202 203
        goto err;
    }

204 205 206 207 208 209
#ifdef KVM_CAP_COALESCED_MMIO
    if (s->coalesced_mmio && !s->coalesced_mmio_ring)
        s->coalesced_mmio_ring = (void *) env->kvm_run +
		s->coalesced_mmio * PAGE_SIZE;
#endif

A
aliguori 已提交
210
    ret = kvm_arch_init_vcpu(env);
J
Jan Kiszka 已提交
211
    if (ret == 0) {
212
        qemu_register_reset(kvm_reset_vcpu, env);
J
Jan Kiszka 已提交
213
        kvm_arch_reset_vcpu(env);
J
Jan Kiszka 已提交
214
    }
A
aliguori 已提交
215 216 217 218
err:
    return ret;
}

219 220 221
/*
 * dirty pages logging control
 */
A
Anthony Liguori 已提交
222 223
static int kvm_dirty_pages_log_change(target_phys_addr_t phys_addr,
                                      ram_addr_t size, int flags, int mask)
224 225
{
    KVMState *s = kvm_state;
226
    KVMSlot *mem = kvm_lookup_matching_slot(s, phys_addr, phys_addr + size);
227 228
    int old_flags;

229
    if (mem == NULL)  {
230 231
            fprintf(stderr, "BUG: %s: invalid parameters " TARGET_FMT_plx "-"
                    TARGET_FMT_plx "\n", __func__, phys_addr,
A
Anthony Liguori 已提交
232
                    (target_phys_addr_t)(phys_addr + size - 1));
233 234 235
            return -EINVAL;
    }

236
    old_flags = mem->flags;
237

238
    flags = (mem->flags & ~mask) | flags;
239 240
    mem->flags = flags;

241 242 243 244 245 246 247 248
    /* If nothing changed effectively, no need to issue ioctl */
    if (s->migration_log) {
        flags |= KVM_MEM_LOG_DIRTY_PAGES;
    }
    if (flags == old_flags) {
            return 0;
    }

249 250 251
    return kvm_set_user_memory_region(s, mem);
}

A
Anthony Liguori 已提交
252
int kvm_log_start(target_phys_addr_t phys_addr, ram_addr_t size)
253
{
254
        return kvm_dirty_pages_log_change(phys_addr, size,
255 256 257 258
                                          KVM_MEM_LOG_DIRTY_PAGES,
                                          KVM_MEM_LOG_DIRTY_PAGES);
}

A
Anthony Liguori 已提交
259
int kvm_log_stop(target_phys_addr_t phys_addr, ram_addr_t size)
260
{
261
        return kvm_dirty_pages_log_change(phys_addr, size,
262 263 264 265
                                          0,
                                          KVM_MEM_LOG_DIRTY_PAGES);
}

266
static int kvm_set_migration_log(int enable)
267 268 269 270 271 272 273 274 275 276
{
    KVMState *s = kvm_state;
    KVMSlot *mem;
    int i, err;

    s->migration_log = enable;

    for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
        mem = &s->slots[i];

277 278 279
        if (!mem->memory_size) {
            continue;
        }
280 281 282 283 284 285 286 287 288 289 290
        if (!!(mem->flags & KVM_MEM_LOG_DIRTY_PAGES) == enable) {
            continue;
        }
        err = kvm_set_user_memory_region(s, mem);
        if (err) {
            return err;
        }
    }
    return 0;
}

291 292 293 294 295
/* get kvm's dirty pages bitmap and update qemu's */
static int kvm_get_dirty_pages_log_range(unsigned long start_addr,
                                         unsigned long *bitmap,
                                         unsigned long offset,
                                         unsigned long mem_size)
A
Alexander Graf 已提交
296
{
297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321
    unsigned int i, j;
    unsigned long page_number, addr, addr1, c;
    ram_addr_t ram_addr;
    unsigned int len = ((mem_size / TARGET_PAGE_SIZE) + HOST_LONG_BITS - 1) /
        HOST_LONG_BITS;

    /*
     * bitmap-traveling is faster than memory-traveling (for addr...)
     * especially when most of the memory is not dirty.
     */
    for (i = 0; i < len; i++) {
        if (bitmap[i] != 0) {
            c = leul_to_cpu(bitmap[i]);
            do {
                j = ffsl(c) - 1;
                c &= ~(1ul << j);
                page_number = i * HOST_LONG_BITS + j;
                addr1 = page_number * TARGET_PAGE_SIZE;
                addr = offset + addr1;
                ram_addr = cpu_get_physical_page_desc(addr);
                cpu_physical_memory_set_dirty(ram_addr);
            } while (c != 0);
        }
    }
    return 0;
A
Alexander Graf 已提交
322 323
}

324 325
#define ALIGN(x, y)  (((x)+(y)-1) & ~((y)-1))

326 327 328 329 330
/**
 * kvm_physical_sync_dirty_bitmap - Grab dirty bitmap from kernel space
 * This function updates qemu's dirty bitmap using cpu_physical_memory_set_dirty().
 * This means all bits are set to dirty.
 *
331
 * @start_add: start of logged region.
332 333
 * @end_addr: end of logged region.
 */
334 335
static int kvm_physical_sync_dirty_bitmap(target_phys_addr_t start_addr,
					  target_phys_addr_t end_addr)
336 337
{
    KVMState *s = kvm_state;
338 339 340 341
    unsigned long size, allocated_size = 0;
    KVMDirtyLog d;
    KVMSlot *mem;
    int ret = 0;
342

343 344 345 346 347 348
    d.dirty_bitmap = NULL;
    while (start_addr < end_addr) {
        mem = kvm_lookup_overlapping_slot(s, start_addr, end_addr);
        if (mem == NULL) {
            break;
        }
349

350
        size = ALIGN(((mem->memory_size) >> TARGET_PAGE_BITS), HOST_LONG_BITS) / 8;
351 352 353 354 355 356 357
        if (!d.dirty_bitmap) {
            d.dirty_bitmap = qemu_malloc(size);
        } else if (size > allocated_size) {
            d.dirty_bitmap = qemu_realloc(d.dirty_bitmap, size);
        }
        allocated_size = size;
        memset(d.dirty_bitmap, 0, allocated_size);
358

359
        d.slot = mem->slot;
360

361
        if (kvm_vm_ioctl(s, KVM_GET_DIRTY_LOG, &d) == -1) {
362
            DPRINTF("ioctl failed %d\n", errno);
363 364 365
            ret = -1;
            break;
        }
366

367 368 369
        kvm_get_dirty_pages_log_range(mem->start_addr, d.dirty_bitmap,
                                      mem->start_addr, mem->memory_size);
        start_addr = mem->start_addr + mem->memory_size;
370 371
    }
    qemu_free(d.dirty_bitmap);
372 373

    return ret;
374 375
}

A
Anthony Liguori 已提交
376
int kvm_coalesce_mmio_region(target_phys_addr_t start, ram_addr_t size)
A
aliguori 已提交
377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
{
    int ret = -ENOSYS;
#ifdef KVM_CAP_COALESCED_MMIO
    KVMState *s = kvm_state;

    if (s->coalesced_mmio) {
        struct kvm_coalesced_mmio_zone zone;

        zone.addr = start;
        zone.size = size;

        ret = kvm_vm_ioctl(s, KVM_REGISTER_COALESCED_MMIO, &zone);
    }
#endif

    return ret;
}

A
Anthony Liguori 已提交
395
int kvm_uncoalesce_mmio_region(target_phys_addr_t start, ram_addr_t size)
A
aliguori 已提交
396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
{
    int ret = -ENOSYS;
#ifdef KVM_CAP_COALESCED_MMIO
    KVMState *s = kvm_state;

    if (s->coalesced_mmio) {
        struct kvm_coalesced_mmio_zone zone;

        zone.addr = start;
        zone.size = size;

        ret = kvm_vm_ioctl(s, KVM_UNREGISTER_COALESCED_MMIO, &zone);
    }
#endif

    return ret;
}

414 415 416 417 418 419 420 421 422 423 424 425
int kvm_check_extension(KVMState *s, unsigned int extension)
{
    int ret;

    ret = kvm_ioctl(s, KVM_CHECK_EXTENSION, extension);
    if (ret < 0) {
        ret = 0;
    }

    return ret;
}

426 427 428
static void kvm_set_phys_mem(target_phys_addr_t start_addr,
			     ram_addr_t size,
			     ram_addr_t phys_offset)
429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
{
    KVMState *s = kvm_state;
    ram_addr_t flags = phys_offset & ~TARGET_PAGE_MASK;
    KVMSlot *mem, old;
    int err;

    if (start_addr & ~TARGET_PAGE_MASK) {
        if (flags >= IO_MEM_UNASSIGNED) {
            if (!kvm_lookup_overlapping_slot(s, start_addr,
                                             start_addr + size)) {
                return;
            }
            fprintf(stderr, "Unaligned split of a KVM memory slot\n");
        } else {
            fprintf(stderr, "Only page-aligned memory slots supported\n");
        }
        abort();
    }

    /* KVM does not support read-only slots */
    phys_offset &= ~IO_MEM_ROM;

    while (1) {
        mem = kvm_lookup_overlapping_slot(s, start_addr, start_addr + size);
        if (!mem) {
            break;
        }

        if (flags < IO_MEM_UNASSIGNED && start_addr >= mem->start_addr &&
            (start_addr + size <= mem->start_addr + mem->memory_size) &&
            (phys_offset - start_addr == mem->phys_offset - mem->start_addr)) {
            /* The new slot fits into the existing one and comes with
             * identical parameters - nothing to be done. */
            return;
        }

        old = *mem;

        /* unregister the overlapping slot */
        mem->memory_size = 0;
        err = kvm_set_user_memory_region(s, mem);
        if (err) {
            fprintf(stderr, "%s: error unregistering overlapping slot: %s\n",
                    __func__, strerror(-err));
            abort();
        }

        /* Workaround for older KVM versions: we can't join slots, even not by
         * unregistering the previous ones and then registering the larger
         * slot. We have to maintain the existing fragmentation. Sigh.
         *
         * This workaround assumes that the new slot starts at the same
         * address as the first existing one. If not or if some overlapping
         * slot comes around later, we will fail (not seen in practice so far)
         * - and actually require a recent KVM version. */
        if (s->broken_set_mem_region &&
            old.start_addr == start_addr && old.memory_size < size &&
            flags < IO_MEM_UNASSIGNED) {
            mem = kvm_alloc_slot(s);
            mem->memory_size = old.memory_size;
            mem->start_addr = old.start_addr;
            mem->phys_offset = old.phys_offset;
            mem->flags = 0;

            err = kvm_set_user_memory_region(s, mem);
            if (err) {
                fprintf(stderr, "%s: error updating slot: %s\n", __func__,
                        strerror(-err));
                abort();
            }

            start_addr += old.memory_size;
            phys_offset += old.memory_size;
            size -= old.memory_size;
            continue;
        }

        /* register prefix slot */
        if (old.start_addr < start_addr) {
            mem = kvm_alloc_slot(s);
            mem->memory_size = start_addr - old.start_addr;
            mem->start_addr = old.start_addr;
            mem->phys_offset = old.phys_offset;
            mem->flags = 0;

            err = kvm_set_user_memory_region(s, mem);
            if (err) {
                fprintf(stderr, "%s: error registering prefix slot: %s\n",
                        __func__, strerror(-err));
                abort();
            }
        }

        /* register suffix slot */
        if (old.start_addr + old.memory_size > start_addr + size) {
            ram_addr_t size_delta;

            mem = kvm_alloc_slot(s);
            mem->start_addr = start_addr + size;
            size_delta = mem->start_addr - old.start_addr;
            mem->memory_size = old.memory_size - size_delta;
            mem->phys_offset = old.phys_offset + size_delta;
            mem->flags = 0;

            err = kvm_set_user_memory_region(s, mem);
            if (err) {
                fprintf(stderr, "%s: error registering suffix slot: %s\n",
                        __func__, strerror(-err));
                abort();
            }
        }
    }

    /* in case the KVM bug workaround already "consumed" the new slot */
    if (!size)
        return;

    /* KVM does not need to know about this memory */
    if (flags >= IO_MEM_UNASSIGNED)
        return;

    mem = kvm_alloc_slot(s);
    mem->memory_size = size;
    mem->start_addr = start_addr;
    mem->phys_offset = phys_offset;
    mem->flags = 0;

    err = kvm_set_user_memory_region(s, mem);
    if (err) {
        fprintf(stderr, "%s: error registering slot: %s\n", __func__,
                strerror(-err));
        abort();
    }
}

564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590
static void kvm_client_set_memory(struct CPUPhysMemoryClient *client,
				  target_phys_addr_t start_addr,
				  ram_addr_t size,
				  ram_addr_t phys_offset)
{
	kvm_set_phys_mem(start_addr, size, phys_offset);
}

static int kvm_client_sync_dirty_bitmap(struct CPUPhysMemoryClient *client,
					target_phys_addr_t start_addr,
					target_phys_addr_t end_addr)
{
	return kvm_physical_sync_dirty_bitmap(start_addr, end_addr);
}

static int kvm_client_migration_log(struct CPUPhysMemoryClient *client,
				    int enable)
{
	return kvm_set_migration_log(enable);
}

static CPUPhysMemoryClient kvm_cpu_phys_memory_client = {
	.set_memory = kvm_client_set_memory,
	.sync_dirty_bitmap = kvm_client_sync_dirty_bitmap,
	.migration_log = kvm_client_migration_log,
};

A
aliguori 已提交
591 592
int kvm_init(int smp_cpus)
{
593 594 595
    static const char upgrade_note[] =
        "Please upgrade to at least kernel 2.6.29 or recent kvm-kmod\n"
        "(see http://sourceforge.net/projects/kvm).\n";
A
aliguori 已提交
596 597 598 599 600 601
    KVMState *s;
    int ret;
    int i;

    s = qemu_mallocz(sizeof(KVMState));

602
#ifdef KVM_CAP_SET_GUEST_DEBUG
B
Blue Swirl 已提交
603
    QTAILQ_INIT(&s->kvm_sw_breakpoints);
604
#endif
A
aliguori 已提交
605 606 607 608
    for (i = 0; i < ARRAY_SIZE(s->slots); i++)
        s->slots[i].slot = i;

    s->vmfd = -1;
K
Kevin Wolf 已提交
609
    s->fd = qemu_open("/dev/kvm", O_RDWR);
A
aliguori 已提交
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
    if (s->fd == -1) {
        fprintf(stderr, "Could not access KVM kernel module: %m\n");
        ret = -errno;
        goto err;
    }

    ret = kvm_ioctl(s, KVM_GET_API_VERSION, 0);
    if (ret < KVM_API_VERSION) {
        if (ret > 0)
            ret = -EINVAL;
        fprintf(stderr, "kvm version too old\n");
        goto err;
    }

    if (ret > KVM_API_VERSION) {
        ret = -EINVAL;
        fprintf(stderr, "kvm version not supported\n");
        goto err;
    }

    s->vmfd = kvm_ioctl(s, KVM_CREATE_VM, 0);
631 632 633 634 635
    if (s->vmfd < 0) {
#ifdef TARGET_S390X
        fprintf(stderr, "Please add the 'switch_amode' kernel parameter to "
                        "your host kernel command line\n");
#endif
A
aliguori 已提交
636
        goto err;
637
    }
A
aliguori 已提交
638 639 640

    /* initially, KVM allocated its own memory and we had to jump through
     * hooks to make phys_ram_base point to this.  Modern versions of KVM
P
pbrook 已提交
641
     * just use a user allocated buffer so we can use regular pages
A
aliguori 已提交
642 643
     * unmodified.  Make sure we have a sufficiently modern version of KVM.
     */
644 645
    if (!kvm_check_extension(s, KVM_CAP_USER_MEMORY)) {
        ret = -EINVAL;
646 647
        fprintf(stderr, "kvm does not support KVM_CAP_USER_MEMORY\n%s",
                upgrade_note);
A
aliguori 已提交
648 649 650
        goto err;
    }

651 652 653
    /* There was a nasty bug in < kvm-80 that prevents memory slots from being
     * destroyed properly.  Since we rely on this capability, refuse to work
     * with any kernel without this capability. */
654 655
    if (!kvm_check_extension(s, KVM_CAP_DESTROY_MEMORY_REGION_WORKS)) {
        ret = -EINVAL;
656 657

        fprintf(stderr,
658 659
                "KVM kernel module broken (DESTROY_MEMORY_REGION).\n%s",
                upgrade_note);
660 661 662
        goto err;
    }

663
    s->coalesced_mmio = 0;
A
aliguori 已提交
664
#ifdef KVM_CAP_COALESCED_MMIO
665
    s->coalesced_mmio = kvm_check_extension(s, KVM_CAP_COALESCED_MMIO);
666
    s->coalesced_mmio_ring = NULL;
A
aliguori 已提交
667 668
#endif

669 670 671 672 673 674 675 676
    s->broken_set_mem_region = 1;
#ifdef KVM_CAP_JOIN_MEMORY_REGIONS_WORKS
    ret = kvm_ioctl(s, KVM_CHECK_EXTENSION, KVM_CAP_JOIN_MEMORY_REGIONS_WORKS);
    if (ret > 0) {
        s->broken_set_mem_region = 0;
    }
#endif

677 678 679 680 681
    s->vcpu_events = 0;
#ifdef KVM_CAP_VCPU_EVENTS
    s->vcpu_events = kvm_check_extension(s, KVM_CAP_VCPU_EVENTS);
#endif

682 683 684 685 686 687
    s->robust_singlestep = 0;
#ifdef KVM_CAP_X86_ROBUST_SINGLESTEP
    s->robust_singlestep =
        kvm_check_extension(s, KVM_CAP_X86_ROBUST_SINGLESTEP);
#endif

688 689 690 691 692
    s->debugregs = 0;
#ifdef KVM_CAP_DEBUGREGS
    s->debugregs = kvm_check_extension(s, KVM_CAP_DEBUGREGS);
#endif

693 694 695 696 697 698 699 700 701 702
    s->xsave = 0;
#ifdef KVM_CAP_XSAVE
    s->xsave = kvm_check_extension(s, KVM_CAP_XSAVE);
#endif

    s->xcrs = 0;
#ifdef KVM_CAP_XCRS
    s->xcrs = kvm_check_extension(s, KVM_CAP_XCRS);
#endif

A
aliguori 已提交
703 704 705 706 707
    ret = kvm_arch_init(s, smp_cpus);
    if (ret < 0)
        goto err;

    kvm_state = s;
708
    cpu_register_phys_memory_client(&kvm_cpu_phys_memory_client);
A
aliguori 已提交
709 710 711 712 713 714 715 716 717 718 719 720 721 722 723

    return 0;

err:
    if (s) {
        if (s->vmfd != -1)
            close(s->vmfd);
        if (s->fd != -1)
            close(s->fd);
    }
    qemu_free(s);

    return ret;
}

724 725
static int kvm_handle_io(uint16_t port, void *data, int direction, int size,
                         uint32_t count)
A
aliguori 已提交
726 727 728 729 730 731 732 733
{
    int i;
    uint8_t *ptr = data;

    for (i = 0; i < count; i++) {
        if (direction == KVM_EXIT_IO_IN) {
            switch (size) {
            case 1:
734
                stb_p(ptr, cpu_inb(port));
A
aliguori 已提交
735 736
                break;
            case 2:
737
                stw_p(ptr, cpu_inw(port));
A
aliguori 已提交
738 739
                break;
            case 4:
740
                stl_p(ptr, cpu_inl(port));
A
aliguori 已提交
741 742 743 744 745
                break;
            }
        } else {
            switch (size) {
            case 1:
746
                cpu_outb(port, ldub_p(ptr));
A
aliguori 已提交
747 748
                break;
            case 2:
749
                cpu_outw(port, lduw_p(ptr));
A
aliguori 已提交
750 751
                break;
            case 4:
752
                cpu_outl(port, ldl_p(ptr));
A
aliguori 已提交
753 754 755 756 757 758 759 760 761 762
                break;
            }
        }

        ptr += size;
    }

    return 1;
}

M
Marcelo Tosatti 已提交
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780
#ifdef KVM_CAP_INTERNAL_ERROR_DATA
static void kvm_handle_internal_error(CPUState *env, struct kvm_run *run)
{

    if (kvm_check_extension(kvm_state, KVM_CAP_INTERNAL_ERROR_DATA)) {
        int i;

        fprintf(stderr, "KVM internal error. Suberror: %d\n",
                run->internal.suberror);

        for (i = 0; i < run->internal.ndata; ++i) {
            fprintf(stderr, "extra data[%d]: %"PRIx64"\n",
                    i, (uint64_t)run->internal.data[i]);
        }
    }
    cpu_dump_state(env, stderr, fprintf, 0);
    if (run->internal.suberror == KVM_INTERNAL_ERROR_EMULATION) {
        fprintf(stderr, "emulation failure\n");
781 782
        if (!kvm_arch_stop_on_emulation_error(env))
		return;
M
Marcelo Tosatti 已提交
783 784 785 786 787 788 789 790
    }
    /* FIXME: Should trigger a qmp message to let management know
     * something went wrong.
     */
    vm_stop(0);
}
#endif

791
void kvm_flush_coalesced_mmio_buffer(void)
A
aliguori 已提交
792 793 794
{
#ifdef KVM_CAP_COALESCED_MMIO
    KVMState *s = kvm_state;
795 796
    if (s->coalesced_mmio_ring) {
        struct kvm_coalesced_mmio_ring *ring = s->coalesced_mmio_ring;
A
aliguori 已提交
797 798 799 800 801 802
        while (ring->first != ring->last) {
            struct kvm_coalesced_mmio *ent;

            ent = &ring->coalesced_mmio[ring->first];

            cpu_physical_memory_write(ent->phys_addr, ent->data, ent->len);
803
            smp_wmb();
A
aliguori 已提交
804 805 806 807 808 809
            ring->first = (ring->first + 1) % KVM_COALESCED_MMIO_MAX;
        }
    }
#endif
}

810
static void do_kvm_cpu_synchronize_state(void *_env)
811
{
812 813
    CPUState *env = _env;

J
Jan Kiszka 已提交
814
    if (!env->kvm_vcpu_dirty) {
815
        kvm_arch_get_registers(env);
J
Jan Kiszka 已提交
816
        env->kvm_vcpu_dirty = 1;
817 818 819
    }
}

820 821 822 823 824 825
void kvm_cpu_synchronize_state(CPUState *env)
{
    if (!env->kvm_vcpu_dirty)
        run_on_cpu(env, do_kvm_cpu_synchronize_state, env);
}

826 827 828 829 830 831 832 833 834 835 836 837
void kvm_cpu_synchronize_post_reset(CPUState *env)
{
    kvm_arch_put_registers(env, KVM_PUT_RESET_STATE);
    env->kvm_vcpu_dirty = 0;
}

void kvm_cpu_synchronize_post_init(CPUState *env)
{
    kvm_arch_put_registers(env, KVM_PUT_FULL_STATE);
    env->kvm_vcpu_dirty = 0;
}

A
aliguori 已提交
838 839 840 841 842
int kvm_cpu_exec(CPUState *env)
{
    struct kvm_run *run = env->kvm_run;
    int ret;

843
    DPRINTF("kvm_cpu_exec()\n");
A
aliguori 已提交
844 845

    do {
846
#ifndef CONFIG_IOTHREAD
847
        if (env->exit_request) {
848
            DPRINTF("interrupt exit requested\n");
A
aliguori 已提交
849 850 851
            ret = 0;
            break;
        }
852
#endif
A
aliguori 已提交
853

M
Marcelo Tosatti 已提交
854 855 856 857 858
        if (kvm_arch_process_irqchip_events(env)) {
            ret = 0;
            break;
        }

J
Jan Kiszka 已提交
859
        if (env->kvm_vcpu_dirty) {
860
            kvm_arch_put_registers(env, KVM_PUT_RUNTIME_STATE);
J
Jan Kiszka 已提交
861
            env->kvm_vcpu_dirty = 0;
862 863
        }

864
        kvm_arch_pre_run(env, run);
865
        cpu_single_env = NULL;
866
        qemu_mutex_unlock_iothread();
A
aliguori 已提交
867
        ret = kvm_vcpu_ioctl(env, KVM_RUN, 0);
868
        qemu_mutex_lock_iothread();
869
        cpu_single_env = env;
A
aliguori 已提交
870 871 872
        kvm_arch_post_run(env, run);

        if (ret == -EINTR || ret == -EAGAIN) {
873
            cpu_exit(env);
874
            DPRINTF("io window exit\n");
A
aliguori 已提交
875 876 877 878 879
            ret = 0;
            break;
        }

        if (ret < 0) {
880
            DPRINTF("kvm run failed %s\n", strerror(-ret));
A
aliguori 已提交
881 882 883
            abort();
        }

884
        kvm_flush_coalesced_mmio_buffer();
A
aliguori 已提交
885

A
aliguori 已提交
886 887 888
        ret = 0; /* exit loop */
        switch (run->exit_reason) {
        case KVM_EXIT_IO:
889
            DPRINTF("handle_io\n");
890
            ret = kvm_handle_io(run->io.port,
A
aliguori 已提交
891 892 893 894 895 896
                                (uint8_t *)run + run->io.data_offset,
                                run->io.direction,
                                run->io.size,
                                run->io.count);
            break;
        case KVM_EXIT_MMIO:
897
            DPRINTF("handle_mmio\n");
A
aliguori 已提交
898 899 900 901 902 903 904
            cpu_physical_memory_rw(run->mmio.phys_addr,
                                   run->mmio.data,
                                   run->mmio.len,
                                   run->mmio.is_write);
            ret = 1;
            break;
        case KVM_EXIT_IRQ_WINDOW_OPEN:
905
            DPRINTF("irq_window_open\n");
A
aliguori 已提交
906 907
            break;
        case KVM_EXIT_SHUTDOWN:
908
            DPRINTF("shutdown\n");
A
aliguori 已提交
909 910 911 912
            qemu_system_reset_request();
            ret = 1;
            break;
        case KVM_EXIT_UNKNOWN:
913
            DPRINTF("kvm_exit_unknown\n");
A
aliguori 已提交
914 915
            break;
        case KVM_EXIT_FAIL_ENTRY:
916
            DPRINTF("kvm_exit_fail_entry\n");
A
aliguori 已提交
917 918
            break;
        case KVM_EXIT_EXCEPTION:
919
            DPRINTF("kvm_exit_exception\n");
A
aliguori 已提交
920
            break;
M
Marcelo Tosatti 已提交
921 922 923 924 925
#ifdef KVM_CAP_INTERNAL_ERROR_DATA
        case KVM_EXIT_INTERNAL_ERROR:
            kvm_handle_internal_error(env, run);
            break;
#endif
A
aliguori 已提交
926
        case KVM_EXIT_DEBUG:
927
            DPRINTF("kvm_exit_debug\n");
928 929 930 931 932 933 934 935
#ifdef KVM_CAP_SET_GUEST_DEBUG
            if (kvm_arch_debug(&run->debug.arch)) {
                env->exception_index = EXCP_DEBUG;
                return 0;
            }
            /* re-enter, this exception was guest-internal */
            ret = 1;
#endif /* KVM_CAP_SET_GUEST_DEBUG */
A
aliguori 已提交
936 937
            break;
        default:
938
            DPRINTF("kvm_arch_handle_exit\n");
A
aliguori 已提交
939 940 941 942 943
            ret = kvm_arch_handle_exit(env, run);
            break;
        }
    } while (ret > 0);

944 945
    if (env->exit_request) {
        env->exit_request = 0;
A
aliguori 已提交
946 947 948
        env->exception_index = EXCP_INTERRUPT;
    }

A
aliguori 已提交
949 950 951
    return ret;
}

952
int kvm_ioctl(KVMState *s, int type, ...)
A
aliguori 已提交
953 954
{
    int ret;
955 956
    void *arg;
    va_list ap;
A
aliguori 已提交
957

958 959 960 961 962
    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);

    ret = ioctl(s->fd, type, arg);
A
aliguori 已提交
963 964 965 966 967 968
    if (ret == -1)
        ret = -errno;

    return ret;
}

969
int kvm_vm_ioctl(KVMState *s, int type, ...)
A
aliguori 已提交
970 971
{
    int ret;
972 973 974 975 976 977
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);
A
aliguori 已提交
978

979
    ret = ioctl(s->vmfd, type, arg);
A
aliguori 已提交
980 981 982 983 984 985
    if (ret == -1)
        ret = -errno;

    return ret;
}

986
int kvm_vcpu_ioctl(CPUState *env, int type, ...)
A
aliguori 已提交
987 988
{
    int ret;
989 990 991 992 993 994
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);
A
aliguori 已提交
995

996
    ret = ioctl(env->kvm_fd, type, arg);
A
aliguori 已提交
997 998 999 1000 1001
    if (ret == -1)
        ret = -errno;

    return ret;
}
A
aliguori 已提交
1002 1003 1004

int kvm_has_sync_mmu(void)
{
A
aurel32 已提交
1005
#ifdef KVM_CAP_SYNC_MMU
A
aliguori 已提交
1006 1007
    KVMState *s = kvm_state;

1008 1009
    return kvm_check_extension(s, KVM_CAP_SYNC_MMU);
#else
A
aliguori 已提交
1010
    return 0;
1011
#endif
A
aliguori 已提交
1012
}
1013

1014 1015 1016 1017 1018
int kvm_has_vcpu_events(void)
{
    return kvm_state->vcpu_events;
}

1019 1020 1021 1022 1023
int kvm_has_robust_singlestep(void)
{
    return kvm_state->robust_singlestep;
}

1024 1025 1026 1027 1028
int kvm_has_debugregs(void)
{
    return kvm_state->debugregs;
}

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
int kvm_has_xsave(void)
{
    return kvm_state->xsave;
}

int kvm_has_xcrs(void)
{
    return kvm_state->xcrs;
}

1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056
void kvm_setup_guest_memory(void *start, size_t size)
{
    if (!kvm_has_sync_mmu()) {
#ifdef MADV_DONTFORK
        int ret = madvise(start, size, MADV_DONTFORK);

        if (ret) {
            perror("madvice");
            exit(1);
        }
#else
        fprintf(stderr,
                "Need MADV_DONTFORK in absence of synchronous KVM MMU\n");
        exit(1);
#endif
    }
}

1057 1058 1059 1060 1061 1062
#ifdef KVM_CAP_SET_GUEST_DEBUG
struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *env,
                                                 target_ulong pc)
{
    struct kvm_sw_breakpoint *bp;

B
Blue Swirl 已提交
1063
    QTAILQ_FOREACH(bp, &env->kvm_state->kvm_sw_breakpoints, entry) {
1064 1065 1066 1067 1068 1069 1070 1071
        if (bp->pc == pc)
            return bp;
    }
    return NULL;
}

int kvm_sw_breakpoints_active(CPUState *env)
{
B
Blue Swirl 已提交
1072
    return !QTAILQ_EMPTY(&env->kvm_state->kvm_sw_breakpoints);
1073 1074
}

G
Glauber Costa 已提交
1075 1076 1077 1078 1079 1080 1081 1082 1083
struct kvm_set_guest_debug_data {
    struct kvm_guest_debug dbg;
    CPUState *env;
    int err;
};

static void kvm_invoke_set_guest_debug(void *data)
{
    struct kvm_set_guest_debug_data *dbg_data = data;
J
Jan Kiszka 已提交
1084 1085 1086
    CPUState *env = dbg_data->env;

    dbg_data->err = kvm_vcpu_ioctl(env, KVM_SET_GUEST_DEBUG, &dbg_data->dbg);
G
Glauber Costa 已提交
1087 1088
}

1089 1090
int kvm_update_guest_debug(CPUState *env, unsigned long reinject_trap)
{
G
Glauber Costa 已提交
1091
    struct kvm_set_guest_debug_data data;
1092

1093
    data.dbg.control = reinject_trap;
1094

1095 1096 1097
    if (env->singlestep_enabled) {
        data.dbg.control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_SINGLESTEP;
    }
G
Glauber Costa 已提交
1098 1099
    kvm_arch_update_guest_debug(env, &data.dbg);
    data.env = env;
1100

1101
    run_on_cpu(env, kvm_invoke_set_guest_debug, &data);
G
Glauber Costa 已提交
1102
    return data.err;
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130
}

int kvm_insert_breakpoint(CPUState *current_env, target_ulong addr,
                          target_ulong len, int type)
{
    struct kvm_sw_breakpoint *bp;
    CPUState *env;
    int err;

    if (type == GDB_BREAKPOINT_SW) {
        bp = kvm_find_sw_breakpoint(current_env, addr);
        if (bp) {
            bp->use_count++;
            return 0;
        }

        bp = qemu_malloc(sizeof(struct kvm_sw_breakpoint));
        if (!bp)
            return -ENOMEM;

        bp->pc = addr;
        bp->use_count = 1;
        err = kvm_arch_insert_sw_breakpoint(current_env, bp);
        if (err) {
            free(bp);
            return err;
        }

B
Blue Swirl 已提交
1131
        QTAILQ_INSERT_HEAD(&current_env->kvm_state->kvm_sw_breakpoints,
1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
                          bp, entry);
    } else {
        err = kvm_arch_insert_hw_breakpoint(addr, len, type);
        if (err)
            return err;
    }

    for (env = first_cpu; env != NULL; env = env->next_cpu) {
        err = kvm_update_guest_debug(env, 0);
        if (err)
            return err;
    }
    return 0;
}

int kvm_remove_breakpoint(CPUState *current_env, target_ulong addr,
                          target_ulong len, int type)
{
    struct kvm_sw_breakpoint *bp;
    CPUState *env;
    int err;

    if (type == GDB_BREAKPOINT_SW) {
        bp = kvm_find_sw_breakpoint(current_env, addr);
        if (!bp)
            return -ENOENT;

        if (bp->use_count > 1) {
            bp->use_count--;
            return 0;
        }

        err = kvm_arch_remove_sw_breakpoint(current_env, bp);
        if (err)
            return err;

B
Blue Swirl 已提交
1168
        QTAILQ_REMOVE(&current_env->kvm_state->kvm_sw_breakpoints, bp, entry);
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
        qemu_free(bp);
    } else {
        err = kvm_arch_remove_hw_breakpoint(addr, len, type);
        if (err)
            return err;
    }

    for (env = first_cpu; env != NULL; env = env->next_cpu) {
        err = kvm_update_guest_debug(env, 0);
        if (err)
            return err;
    }
    return 0;
}

void kvm_remove_all_breakpoints(CPUState *current_env)
{
    struct kvm_sw_breakpoint *bp, *next;
    KVMState *s = current_env->kvm_state;
    CPUState *env;

B
Blue Swirl 已提交
1190
    QTAILQ_FOREACH_SAFE(bp, &s->kvm_sw_breakpoints, entry, next) {
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
        if (kvm_arch_remove_sw_breakpoint(current_env, bp) != 0) {
            /* Try harder to find a CPU that currently sees the breakpoint. */
            for (env = first_cpu; env != NULL; env = env->next_cpu) {
                if (kvm_arch_remove_sw_breakpoint(env, bp) == 0)
                    break;
            }
        }
    }
    kvm_arch_remove_all_hw_breakpoints();

    for (env = first_cpu; env != NULL; env = env->next_cpu)
        kvm_update_guest_debug(env, 0);
}

#else /* !KVM_CAP_SET_GUEST_DEBUG */

int kvm_update_guest_debug(CPUState *env, unsigned long reinject_trap)
{
    return -EINVAL;
}

int kvm_insert_breakpoint(CPUState *current_env, target_ulong addr,
                          target_ulong len, int type)
{
    return -EINVAL;
}

int kvm_remove_breakpoint(CPUState *current_env, target_ulong addr,
                          target_ulong len, int type)
{
    return -EINVAL;
}

void kvm_remove_all_breakpoints(CPUState *current_env)
{
}
#endif /* !KVM_CAP_SET_GUEST_DEBUG */
1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245

int kvm_set_signal_mask(CPUState *env, const sigset_t *sigset)
{
    struct kvm_signal_mask *sigmask;
    int r;

    if (!sigset)
        return kvm_vcpu_ioctl(env, KVM_SET_SIGNAL_MASK, NULL);

    sigmask = qemu_malloc(sizeof(*sigmask) + sizeof(*sigset));

    sigmask->len = 8;
    memcpy(sigmask->sigset, sigset, sizeof(*sigset));
    r = kvm_vcpu_ioctl(env, KVM_SET_SIGNAL_MASK, sigmask);
    free(sigmask);

    return r;
}
1246 1247 1248

int kvm_set_ioeventfd_pio_word(int fd, uint16_t addr, uint16_t val, bool assign)
{
1249
#ifdef KVM_IOEVENTFD
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
    struct kvm_ioeventfd kick = {
        .datamatch = val,
        .addr = addr,
        .len = 2,
        .flags = KVM_IOEVENTFD_FLAG_DATAMATCH | KVM_IOEVENTFD_FLAG_PIO,
        .fd = fd,
    };
    int r;
    if (!kvm_enabled())
        return -ENOSYS;
    if (!assign)
        kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
    r = kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
    if (r < 0)
        return r;
    return 0;
1266 1267
#else
    return -ENOSYS;
1268
#endif
1269
}